diff options
Diffstat (limited to 'servers')
95 files changed, 8592 insertions, 5058 deletions
diff --git a/servers/audio/audio_driver_dummy.cpp b/servers/audio/audio_driver_dummy.cpp index 12cadb9301..faddced155 100644 --- a/servers/audio/audio_driver_dummy.cpp +++ b/servers/audio/audio_driver_dummy.cpp @@ -48,7 +48,7 @@ Error AudioDriverDummy::init() { samples_in = memnew_arr(int32_t, buffer_frames * channels); - thread = Thread::create(AudioDriverDummy::thread_func, this); + thread.start(AudioDriverDummy::thread_func, this); return OK; }; @@ -86,31 +86,18 @@ AudioDriver::SpeakerMode AudioDriverDummy::get_speaker_mode() const { }; void AudioDriverDummy::lock() { - if (!thread) { - return; - } mutex.lock(); }; void AudioDriverDummy::unlock() { - if (!thread) { - return; - } mutex.unlock(); }; void AudioDriverDummy::finish() { - if (!thread) { - return; - } - exit_thread = true; - Thread::wait_to_finish(thread); + thread.wait_to_finish(); if (samples_in) { memdelete_arr(samples_in); }; - - memdelete(thread); - thread = nullptr; }; diff --git a/servers/audio/audio_driver_dummy.h b/servers/audio/audio_driver_dummy.h index 617ffb2c79..7d84e7ffc8 100644 --- a/servers/audio/audio_driver_dummy.h +++ b/servers/audio/audio_driver_dummy.h @@ -37,7 +37,7 @@ #include "core/os/thread.h" class AudioDriverDummy : public AudioDriver { - Thread *thread = nullptr; + Thread thread; Mutex mutex; int32_t *samples_in; diff --git a/servers/audio/audio_filter_sw.cpp b/servers/audio/audio_filter_sw.cpp index 580e061496..bcfa4c4c37 100644 --- a/servers/audio/audio_filter_sw.cpp +++ b/servers/audio/audio_filter_sw.cpp @@ -58,7 +58,7 @@ void AudioFilterSW::prepare_coefficients(Coeffs *p_coeffs) { final_cutoff = 1; //don't allow less than this } - double omega = 2.0 * Math_PI * final_cutoff / sampling_rate; + double omega = Math_TAU * final_cutoff / sampling_rate; double sin_v = Math::sin(omega); double cos_v = Math::cos(omega); @@ -132,7 +132,7 @@ void AudioFilterSW::prepare_coefficients(Coeffs *p_coeffs) { double hicutoff = resonance; double centercutoff = (cutoff + resonance) / 2.0; double bandwidth = (Math::log(centercutoff) - Math::log(hicutoff)) / Math::log((double)2); - omega = 2.0 * Math_PI * centercutoff / sampling_rate; + omega = Math_TAU * centercutoff / sampling_rate; alpha = Math::sin(omega) * Math::sinh(Math::log((double)2) / 2 * bandwidth * omega / Math::sin(omega)); a0 = 1 + alpha; @@ -197,7 +197,7 @@ void AudioFilterSW::set_stages(int p_stages) { //adjust for multiple stages /* Fouriertransform kernel to obtain response */ float AudioFilterSW::get_response(float p_freq, Coeffs *p_coeffs) { - float freq = p_freq / sampling_rate * Math_PI * 2.0f; + float freq = p_freq / sampling_rate * Math_TAU; float cx = p_coeffs->b0, cy = 0.0; diff --git a/servers/audio/effects/audio_effect_capture.cpp b/servers/audio/effects/audio_effect_capture.cpp new file mode 100644 index 0000000000..f37938eec8 --- /dev/null +++ b/servers/audio/effects/audio_effect_capture.cpp @@ -0,0 +1,140 @@ +/*************************************************************************/ +/* audio_effect_capture.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "audio_effect_capture.h" + +bool AudioEffectCapture::can_get_buffer(int p_frames) const { + return buffer.data_left() >= p_frames; +} + +PackedVector2Array AudioEffectCapture::get_buffer(int p_frames) { + ERR_FAIL_COND_V(!buffer_initialized, PackedVector2Array()); + ERR_FAIL_INDEX_V(p_frames, buffer.size(), PackedVector2Array()); + int data_left = buffer.data_left(); + if (data_left < p_frames || p_frames == 0) { + return PackedVector2Array(); + } + + PackedVector2Array ret; + ret.resize(p_frames); + + Vector<AudioFrame> streaming_data; + streaming_data.resize(p_frames); + buffer.read(streaming_data.ptrw(), p_frames); + for (int32_t i = 0; i < p_frames; i++) { + ret.write[i] = Vector2(streaming_data[i].l, streaming_data[i].r); + } + return ret; +} + +void AudioEffectCapture::clear_buffer() { + const int32_t data_left = buffer.data_left(); + buffer.advance_read(data_left); +} + +void AudioEffectCapture::_bind_methods() { + ClassDB::bind_method(D_METHOD("can_get_buffer", "frames"), &AudioEffectCapture::can_get_buffer); + ClassDB::bind_method(D_METHOD("get_buffer", "frames"), &AudioEffectCapture::get_buffer); + ClassDB::bind_method(D_METHOD("clear_buffer"), &AudioEffectCapture::clear_buffer); + ClassDB::bind_method(D_METHOD("set_buffer_length", "buffer_length_seconds"), &AudioEffectCapture::set_buffer_length); + ClassDB::bind_method(D_METHOD("get_buffer_length"), &AudioEffectCapture::get_buffer_length); + ClassDB::bind_method(D_METHOD("get_frames_available"), &AudioEffectCapture::get_frames_available); + ClassDB::bind_method(D_METHOD("get_discarded_frames"), &AudioEffectCapture::get_discarded_frames); + ClassDB::bind_method(D_METHOD("get_buffer_length_frames"), &AudioEffectCapture::get_buffer_length_frames); + ClassDB::bind_method(D_METHOD("get_pushed_frames"), &AudioEffectCapture::get_pushed_frames); + + ADD_PROPERTY(PropertyInfo(Variant::FLOAT, "buffer_length", PROPERTY_HINT_RANGE, "0.01,10,0.01"), "set_buffer_length", "get_buffer_length"); +} + +Ref<AudioEffectInstance> AudioEffectCapture::instance() { + if (!buffer_initialized) { + float target_buffer_size = AudioServer::get_singleton()->get_mix_rate() * buffer_length_seconds; + ERR_FAIL_COND_V(target_buffer_size <= 0 || target_buffer_size >= (1 << 27), Ref<AudioEffectInstance>()); + buffer.resize(nearest_shift((int)target_buffer_size)); + buffer_initialized = true; + } + + clear_buffer(); + + Ref<AudioEffectCaptureInstance> ins; + ins.instance(); + ins->base = Ref<AudioEffectCapture>(this); + + return ins; +} + +void AudioEffectCapture::set_buffer_length(float p_buffer_length_seconds) { + ERR_FAIL_COND(buffer_initialized); + + buffer_length_seconds = p_buffer_length_seconds; +} + +float AudioEffectCapture::get_buffer_length() { + return buffer_length_seconds; +} + +int AudioEffectCapture::get_frames_available() const { + ERR_FAIL_COND_V(!buffer_initialized, 0); + return buffer.data_left(); +} + +int64_t AudioEffectCapture::get_discarded_frames() const { + return discarded_frames; +} + +int AudioEffectCapture::get_buffer_length_frames() const { + ERR_FAIL_COND_V(!buffer_initialized, 0); + return buffer.size(); +} + +int64_t AudioEffectCapture::get_pushed_frames() const { + return pushed_frames; +} + +void AudioEffectCaptureInstance::process(const AudioFrame *p_src_frames, AudioFrame *p_dst_frames, int p_frame_count) { + RingBuffer<AudioFrame> &buffer = base->buffer; + + for (int i = 0; i < p_frame_count; i++) { + p_dst_frames[i] = p_src_frames[i]; + } + + if (buffer.space_left() >= p_frame_count) { + // Add incoming audio frames to the IO ring buffer + int32_t ret = buffer.write(p_src_frames, p_frame_count); + ERR_FAIL_COND_MSG(ret != p_frame_count, "Failed to add data to effect capture ring buffer despite sufficient space."); + atomic_add(&base->pushed_frames, p_frame_count); + } else { + atomic_add(&base->discarded_frames, p_frame_count); + } +} + +bool AudioEffectCaptureInstance::process_silence() const { + return true; +} diff --git a/servers/audio/effects/audio_effect_capture.h b/servers/audio/effects/audio_effect_capture.h new file mode 100644 index 0000000000..b154be85de --- /dev/null +++ b/servers/audio/effects/audio_effect_capture.h @@ -0,0 +1,82 @@ +/*************************************************************************/ +/* audio_effect_capture.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef AUDIO_EFFECT_CAPTURE_H +#define AUDIO_EFFECT_CAPTURE_H + +#include "core/config/engine.h" +#include "core/math/audio_frame.h" +#include "core/object/reference.h" +#include "core/templates/vector.h" +#include "servers/audio/audio_effect.h" +#include "servers/audio_server.h" + +class AudioEffectCapture; + +class AudioEffectCaptureInstance : public AudioEffectInstance { + GDCLASS(AudioEffectCaptureInstance, AudioEffectInstance); + friend class AudioEffectCapture; + Ref<AudioEffectCapture> base; + +public: + virtual void process(const AudioFrame *p_src_frames, AudioFrame *p_dst_frames, int p_frame_count) override; + virtual bool process_silence() const override; +}; + +class AudioEffectCapture : public AudioEffect { + GDCLASS(AudioEffectCapture, AudioEffect) + friend class AudioEffectCaptureInstance; + + RingBuffer<AudioFrame> buffer; + uint64_t discarded_frames = 0; + uint64_t pushed_frames = 0; + float buffer_length_seconds = 0.1f; + bool buffer_initialized = false; + +protected: + static void _bind_methods(); + +public: + virtual Ref<AudioEffectInstance> instance() override; + + void set_buffer_length(float p_buffer_length_seconds); + float get_buffer_length(); + + bool can_get_buffer(int p_frames) const; + PackedVector2Array get_buffer(int p_len); + void clear_buffer(); + + int get_frames_available() const; + int64_t get_discarded_frames() const; + int get_buffer_length_frames() const; + int64_t get_pushed_frames() const; +}; + +#endif // AUDIO_EFFECT_CAPTURE_H diff --git a/servers/audio/effects/audio_effect_chorus.cpp b/servers/audio/effects/audio_effect_chorus.cpp index 1542273a24..76a995eb37 100644 --- a/servers/audio/effects/audio_effect_chorus.cpp +++ b/servers/audio/effects/audio_effect_chorus.cpp @@ -84,7 +84,7 @@ void AudioEffectChorusInstance::_process_chunk(const AudioFrame *p_src_frames, A if (v.cutoff == 0) { continue; } - float auxlp = expf(-2.0 * Math_PI * v.cutoff / mix_rate); + float auxlp = expf(-Math_TAU * v.cutoff / mix_rate); float c1 = 1.0 - auxlp; float c2 = auxlp; AudioFrame h = filter_h[vc]; @@ -104,7 +104,7 @@ void AudioEffectChorusInstance::_process_chunk(const AudioFrame *p_src_frames, A float phase = (float)(local_cycles & AudioEffectChorus::CYCLES_MASK) / (float)(1 << AudioEffectChorus::CYCLES_FRAC); - float wave_delay = sinf(phase * 2.0 * Math_PI) * max_depth_frames; + float wave_delay = sinf(phase * Math_TAU) * max_depth_frames; int wave_delay_frames = lrint(floor(wave_delay)); float wave_delay_frac = wave_delay - (float)wave_delay_frames; diff --git a/servers/audio/effects/audio_effect_delay.cpp b/servers/audio/effects/audio_effect_delay.cpp index f04ab45ec1..ba50eeb0a3 100644 --- a/servers/audio/effects/audio_effect_delay.cpp +++ b/servers/audio/effects/audio_effect_delay.cpp @@ -75,7 +75,7 @@ void AudioEffectDelayInstance::_process_chunk(const AudioFrame *p_src_frames, Au tap2_vol.r *= CLAMP(1.0 + base->tap_2_pan, 0, 1); // feedback lowpass here - float lpf_c = expf(-2.0 * Math_PI * base->feedback_lowpass / mix_rate); // 0 .. 10khz + float lpf_c = expf(-Math_TAU * base->feedback_lowpass / mix_rate); // 0 .. 10khz float lpf_ic = 1.0 - lpf_c; const AudioFrame *src = p_src_frames; diff --git a/servers/audio/effects/audio_effect_distortion.cpp b/servers/audio/effects/audio_effect_distortion.cpp index 8f713ace22..b79434e7c2 100644 --- a/servers/audio/effects/audio_effect_distortion.cpp +++ b/servers/audio/effects/audio_effect_distortion.cpp @@ -36,8 +36,8 @@ void AudioEffectDistortionInstance::process(const AudioFrame *p_src_frames, Audi const float *src = (const float *)p_src_frames; float *dst = (float *)p_dst_frames; - //float lpf_c=expf(-2.0*Math_PI*keep_hf_hz.get()/(mix_rate*(float)OVERSAMPLE)); - float lpf_c = expf(-2.0 * Math_PI * base->keep_hf_hz / (AudioServer::get_singleton()->get_mix_rate())); + //float lpf_c=expf(-Math_TAU*keep_hf_hz.get()/(mix_rate*(float)OVERSAMPLE)); + float lpf_c = expf(-Math_TAU * base->keep_hf_hz / (AudioServer::get_singleton()->get_mix_rate())); float lpf_ic = 1.0 - lpf_c; float drive_f = base->drive; diff --git a/servers/audio/effects/audio_effect_phaser.cpp b/servers/audio/effects/audio_effect_phaser.cpp index 5e4e183ccf..9b70f03a19 100644 --- a/servers/audio/effects/audio_effect_phaser.cpp +++ b/servers/audio/effects/audio_effect_phaser.cpp @@ -38,13 +38,13 @@ void AudioEffectPhaserInstance::process(const AudioFrame *p_src_frames, AudioFra float dmin = base->range_min / (sampling_rate / 2.0); float dmax = base->range_max / (sampling_rate / 2.0); - float increment = 2.f * Math_PI * (base->rate / sampling_rate); + float increment = Math_TAU * (base->rate / sampling_rate); for (int i = 0; i < p_frame_count; i++) { phase += increment; - while (phase >= Math_PI * 2.f) { - phase -= Math_PI * 2.f; + while (phase >= Math_TAU) { + phase -= Math_TAU; } float d = dmin + (dmax - dmin) * ((sin(phase) + 1.f) / 2.f); diff --git a/servers/audio/effects/audio_effect_record.cpp b/servers/audio/effects/audio_effect_record.cpp index e8832c92a3..2015ede81f 100644 --- a/servers/audio/effects/audio_effect_record.cpp +++ b/servers/audio/effects/audio_effect_record.cpp @@ -118,7 +118,7 @@ void AudioEffectRecordInstance::init() { #ifdef NO_THREADS AudioServer::get_singleton()->add_update_callback(&AudioEffectRecordInstance::_update, this); #else - io_thread = Thread::create(_thread_callback, this); + io_thread.start(_thread_callback, this); #endif } @@ -126,9 +126,7 @@ void AudioEffectRecordInstance::finish() { #ifdef NO_THREADS AudioServer::get_singleton()->remove_update_callback(&AudioEffectRecordInstance::_update, this); #else - if (thread_active) { - Thread::wait_to_finish(io_thread); - } + io_thread.wait_to_finish(); #endif } diff --git a/servers/audio/effects/audio_effect_record.h b/servers/audio/effects/audio_effect_record.h index 14e646e29d..b97ec43946 100644 --- a/servers/audio/effects/audio_effect_record.h +++ b/servers/audio/effects/audio_effect_record.h @@ -48,7 +48,7 @@ class AudioEffectRecordInstance : public AudioEffectInstance { Ref<AudioEffectRecord> base; bool is_recording; - Thread *io_thread; + Thread io_thread; bool thread_active = false; Vector<AudioFrame> ring_buffer; diff --git a/servers/audio/effects/audio_effect_spectrum_analyzer.cpp b/servers/audio/effects/audio_effect_spectrum_analyzer.cpp index 7f73f2e880..3f7ab74a74 100644 --- a/servers/audio/effects/audio_effect_spectrum_analyzer.cpp +++ b/servers/audio/effects/audio_effect_spectrum_analyzer.cpp @@ -110,10 +110,11 @@ void AudioEffectSpectrumAnalyzerInstance::process(const AudioFrame *p_src_frames while (p_frame_count) { int to_fill = fft_size * 2 - temporal_fft_pos; to_fill = MIN(to_fill, p_frame_count); + const double to_fill_step = Math_TAU / (double)fft_size; float *fftw = temporal_fft.ptrw(); for (int i = 0; i < to_fill; i++) { //left and right buffers - float window = -0.5 * Math::cos(2.0 * Math_PI * (double)temporal_fft_pos / (double)fft_size) + 0.5; + float window = -0.5 * Math::cos(to_fill_step * (double)temporal_fft_pos) + 0.5; fftw[temporal_fft_pos * 2] = window * p_src_frames->l; fftw[temporal_fft_pos * 2 + 1] = 0; fftw[(temporal_fft_pos + fft_size * 2) * 2] = window * p_src_frames->r; diff --git a/servers/audio/effects/eq.cpp b/servers/audio/effects/eq.cpp index 2181411b9e..e0c3eb6d3a 100644 --- a/servers/audio/effects/eq.cpp +++ b/servers/audio/effects/eq.cpp @@ -89,8 +89,8 @@ void EQ::recalculate_band_coefficients() { double frq_l = round(frq / pow(2.0, octave_size / 2.0)); double side_gain2 = POW2(Math_SQRT12); - double th = 2.0 * Math_PI * frq / mix_rate; - double th_l = 2.0 * Math_PI * frq_l / mix_rate; + double th = Math_TAU * frq / mix_rate; + double th_l = Math_TAU * frq_l / mix_rate; double c2a = side_gain2 * POW2(cos(th)) - 2.0 * side_gain2 * cos(th_l) * cos(th) + side_gain2 - POW2(sin(th_l)); diff --git a/servers/audio/effects/reverb.cpp b/servers/audio/effects/reverb.cpp index eb96e21659..7df2f99f67 100644 --- a/servers/audio/effects/reverb.cpp +++ b/servers/audio/effects/reverb.cpp @@ -91,7 +91,7 @@ void Reverb::process(float *p_src, float *p_dst, int p_frames) { } if (params.hpf > 0) { - float hpaux = expf(-2.0 * Math_PI * params.hpf * 6000 / params.mix_rate); + float hpaux = expf(-Math_TAU * params.hpf * 6000 / params.mix_rate); float hp_a1 = (1.0 + hpaux) / 2.0; float hp_a2 = -(1.0 + hpaux) / 2.0; float hp_b1 = hpaux; @@ -293,7 +293,7 @@ void Reverb::update_parameters() { float auxdmp = params.damp / 2.0 + 0.5; //only half the range (0.5 .. 1.0 is enough) auxdmp *= auxdmp; - c.damp = expf(-2.0 * Math_PI * auxdmp * 10000 / params.mix_rate); // 0 .. 10khz + c.damp = expf(-Math_TAU * auxdmp * 10000 / params.mix_rate); // 0 .. 10khz } } diff --git a/servers/audio_server.cpp b/servers/audio_server.cpp index 295059a1ea..16c6a26595 100644 --- a/servers/audio_server.cpp +++ b/servers/audio_server.cpp @@ -71,13 +71,20 @@ void AudioDriver::update_mix_time(int p_frames) { } } -double AudioDriver::get_time_since_last_mix() const { - return (OS::get_singleton()->get_ticks_usec() - _last_mix_time) / 1000000.0; +double AudioDriver::get_time_since_last_mix() { + lock(); + uint64_t last_mix_time = _last_mix_time; + unlock(); + return (OS::get_singleton()->get_ticks_usec() - last_mix_time) / 1000000.0; } -double AudioDriver::get_time_to_next_mix() const { - double total = (OS::get_singleton()->get_ticks_usec() - _last_mix_time) / 1000000.0; - double mix_buffer = _last_mix_frames / (double)get_mix_rate(); +double AudioDriver::get_time_to_next_mix() { + lock(); + uint64_t last_mix_time = _last_mix_time; + uint64_t last_mix_frames = _last_mix_frames; + unlock(); + double total = (OS::get_singleton()->get_ticks_usec() - last_mix_time) / 1000000.0; + double mix_buffer = last_mix_frames / (double)get_mix_rate(); return mix_buffer - total; } @@ -394,6 +401,7 @@ void AudioServer::_mix_step() { for (int k = 0; k < bus->channels.size(); k++) { if (!bus->channels[k].active) { + bus->channels.write[k].peak_volume = AudioFrame(AUDIO_MIN_PEAK_DB, AUDIO_MIN_PEAK_DB); continue; } @@ -427,7 +435,7 @@ void AudioServer::_mix_step() { } } - bus->channels.write[k].peak_volume = AudioFrame(Math::linear2db(peak.l + 0.0000000001), Math::linear2db(peak.r + 0.0000000001)); + bus->channels.write[k].peak_volume = AudioFrame(Math::linear2db(peak.l + AUDIO_PEAK_OFFSET), Math::linear2db(peak.r + AUDIO_PEAK_OFFSET)); if (!bus->channels[k].used) { //see if any audio is contained, because channel was not used diff --git a/servers/audio_server.h b/servers/audio_server.h index 631fef0f65..a1a373e1ca 100644 --- a/servers/audio_server.h +++ b/servers/audio_server.h @@ -70,8 +70,8 @@ protected: #endif public: - double get_time_since_last_mix() const; //useful for video -> audio sync - double get_time_to_next_mix() const; + double get_time_since_last_mix(); //useful for video -> audio sync + double get_time_to_next_mix(); enum SpeakerMode { SPEAKER_MODE_STEREO, @@ -199,7 +199,7 @@ private: last_mix_with_audio = 0; used = false; active = false; - peak_volume = AudioFrame(0, 0); + peak_volume = AudioFrame(AUDIO_MIN_PEAK_DB, AUDIO_MIN_PEAK_DB); } }; diff --git a/servers/physics_2d/body_pair_2d_sw.cpp b/servers/physics_2d/body_pair_2d_sw.cpp index 6a13453f9f..feced36a2b 100644 --- a/servers/physics_2d/body_pair_2d_sw.cpp +++ b/servers/physics_2d/body_pair_2d_sw.cpp @@ -288,21 +288,17 @@ bool BodyPair2DSW::setup(real_t p_step) { if (A->is_shape_set_as_one_way_collision(shape_A)) { Vector2 direction = xform_A.get_axis(1).normalized(); bool valid = false; - if (B->get_linear_velocity().dot(direction) >= 0) { - for (int i = 0; i < contact_count; i++) { - Contact &c = contacts[i]; - if (!c.reused) { - continue; - } - if (c.normal.dot(direction) > 0) { //greater (normal inverted) - continue; - } - - valid = true; - break; + for (int i = 0; i < contact_count; i++) { + Contact &c = contacts[i]; + if (!c.reused) { + continue; } + if (c.normal.dot(direction) > -CMP_EPSILON) { //greater (normal inverted) + continue; + } + valid = true; + break; } - if (!valid) { collided = false; oneway_disabled = true; @@ -313,19 +309,16 @@ bool BodyPair2DSW::setup(real_t p_step) { if (B->is_shape_set_as_one_way_collision(shape_B)) { Vector2 direction = xform_B.get_axis(1).normalized(); bool valid = false; - if (A->get_linear_velocity().dot(direction) >= 0) { - for (int i = 0; i < contact_count; i++) { - Contact &c = contacts[i]; - if (!c.reused) { - continue; - } - if (c.normal.dot(direction) < 0) { //less (normal ok) - continue; - } - - valid = true; - break; + for (int i = 0; i < contact_count; i++) { + Contact &c = contacts[i]; + if (!c.reused) { + continue; + } + if (c.normal.dot(direction) < CMP_EPSILON) { //less (normal ok) + continue; } + valid = true; + break; } if (!valid) { collided = false; diff --git a/servers/physics_2d/collision_object_2d_sw.h b/servers/physics_2d/collision_object_2d_sw.h index 2939b4b99f..2db3961f41 100644 --- a/servers/physics_2d/collision_object_2d_sw.h +++ b/servers/physics_2d/collision_object_2d_sw.h @@ -61,7 +61,7 @@ private: Variant metadata; bool disabled; bool one_way_collision; - float one_way_collision_margin; + real_t one_way_collision_margin; Shape() { disabled = false; one_way_collision = false; @@ -153,7 +153,7 @@ public: return shapes[p_idx].disabled; } - _FORCE_INLINE_ void set_shape_as_one_way_collision(int p_idx, bool p_one_way_collision, float p_margin) { + _FORCE_INLINE_ void set_shape_as_one_way_collision(int p_idx, bool p_one_way_collision, real_t p_margin) { CRASH_BAD_INDEX(p_idx, shapes.size()); shapes.write[p_idx].one_way_collision = p_one_way_collision; shapes.write[p_idx].one_way_collision_margin = p_margin; @@ -163,7 +163,7 @@ public: return shapes[p_idx].one_way_collision; } - _FORCE_INLINE_ float get_shape_one_way_collision_margin(int p_idx) const { + _FORCE_INLINE_ real_t get_shape_one_way_collision_margin(int p_idx) const { CRASH_BAD_INDEX(p_idx, shapes.size()); return shapes[p_idx].one_way_collision_margin; } diff --git a/servers/physics_2d/physics_server_2d_sw.cpp b/servers/physics_2d/physics_server_2d_sw.cpp index 85e24ca537..14fcf1520a 100644 --- a/servers/physics_2d/physics_server_2d_sw.cpp +++ b/servers/physics_2d/physics_server_2d_sw.cpp @@ -149,24 +149,19 @@ void PhysicsServer2DSW::_shape_col_cbk(const Vector2 &p_point_A, const Vector2 & return; } + Vector2 rel_dir = (p_point_A - p_point_B); + real_t rel_length2 = rel_dir.length_squared(); if (cbk->valid_dir != Vector2()) { - if (p_point_A.distance_squared_to(p_point_B) > cbk->valid_depth * cbk->valid_depth) { - cbk->invalid_by_dir++; - return; - } - Vector2 rel_dir = (p_point_A - p_point_B).normalized(); - - if (cbk->valid_dir.dot(rel_dir) < Math_SQRT12) { //sqrt(2)/2.0 - 45 degrees - cbk->invalid_by_dir++; - - /* - print_line("A: "+p_point_A); - print_line("B: "+p_point_B); - print_line("discard too angled "+rtos(cbk->valid_dir.dot((p_point_A-p_point_B)))); - print_line("resnorm: "+(p_point_A-p_point_B).normalized()); - print_line("distance: "+rtos(p_point_A.distance_to(p_point_B))); - */ - return; + if (cbk->valid_depth < 10e20) { + if (rel_length2 > cbk->valid_depth * cbk->valid_depth || + (rel_length2 > CMP_EPSILON && cbk->valid_dir.dot(rel_dir.normalized()) < CMP_EPSILON)) { + cbk->invalid_by_dir++; + return; + } + } else { + if (rel_length2 > 0 && cbk->valid_dir.dot(rel_dir.normalized()) < CMP_EPSILON) { + return; + } } } @@ -182,8 +177,7 @@ void PhysicsServer2DSW::_shape_col_cbk(const Vector2 &p_point_A, const Vector2 & } } - real_t d = p_point_A.distance_squared_to(p_point_B); - if (d < min_depth) { + if (rel_length2 < min_depth) { return; } cbk->ptr[min_depth_idx * 2 + 0] = p_point_A; @@ -673,7 +667,7 @@ void PhysicsServer2DSW::body_set_shape_disabled(RID p_body, int p_shape_idx, boo body->set_shape_as_disabled(p_shape_idx, p_disabled); } -void PhysicsServer2DSW::body_set_shape_as_one_way_collision(RID p_body, int p_shape_idx, bool p_enable, float p_margin) { +void PhysicsServer2DSW::body_set_shape_as_one_way_collision(RID p_body, int p_shape_idx, bool p_enable, real_t p_margin) { Body2DSW *body = body_owner.getornull(p_body); ERR_FAIL_COND(!body); ERR_FAIL_INDEX(p_shape_idx, body->get_shape_count()); @@ -964,7 +958,7 @@ bool PhysicsServer2DSW::body_test_motion(RID p_body, const Transform2D &p_from, return body->get_space()->test_body_motion(body, p_from, p_motion, p_infinite_inertia, p_margin, r_result, p_exclude_raycast_shapes); } -int PhysicsServer2DSW::body_test_ray_separation(RID p_body, const Transform2D &p_transform, bool p_infinite_inertia, Vector2 &r_recover_motion, SeparationResult *r_results, int p_result_max, float p_margin) { +int PhysicsServer2DSW::body_test_ray_separation(RID p_body, const Transform2D &p_transform, bool p_infinite_inertia, Vector2 &r_recover_motion, SeparationResult *r_results, int p_result_max, real_t p_margin) { Body2DSW *body = body_owner.getornull(p_body); ERR_FAIL_COND_V(!body, false); ERR_FAIL_COND_V(!body->get_space(), false); diff --git a/servers/physics_2d/physics_server_2d_sw.h b/servers/physics_2d/physics_server_2d_sw.h index 3305c0bd3d..62ea30b3f6 100644 --- a/servers/physics_2d/physics_server_2d_sw.h +++ b/servers/physics_2d/physics_server_2d_sw.h @@ -191,7 +191,7 @@ public: virtual void body_clear_shapes(RID p_body) override; virtual void body_set_shape_disabled(RID p_body, int p_shape_idx, bool p_disabled) override; - virtual void body_set_shape_as_one_way_collision(RID p_body, int p_shape_idx, bool p_enable, float p_margin) override; + virtual void body_set_shape_as_one_way_collision(RID p_body, int p_shape_idx, bool p_enable, real_t p_margin) override; virtual void body_attach_object_instance_id(RID p_body, ObjectID p_id) override; virtual ObjectID body_get_object_instance_id(RID p_body) const override; @@ -248,7 +248,7 @@ public: virtual void body_set_pickable(RID p_body, bool p_pickable) override; virtual bool body_test_motion(RID p_body, const Transform2D &p_from, const Vector2 &p_motion, bool p_infinite_inertia, real_t p_margin = 0.001, MotionResult *r_result = nullptr, bool p_exclude_raycast_shapes = true) override; - virtual int body_test_ray_separation(RID p_body, const Transform2D &p_transform, bool p_infinite_inertia, Vector2 &r_recover_motion, SeparationResult *r_results, int p_result_max, float p_margin = 0.001) override; + virtual int body_test_ray_separation(RID p_body, const Transform2D &p_transform, bool p_infinite_inertia, Vector2 &r_recover_motion, SeparationResult *r_results, int p_result_max, real_t p_margin = 0.001) override; // this function only works on physics process, errors and returns null otherwise virtual PhysicsDirectBodyState2D *body_get_direct_state(RID p_body) override; diff --git a/servers/physics_2d/physics_server_2d_wrap_mt.cpp b/servers/physics_2d/physics_server_2d_wrap_mt.cpp index 15d875b3b7..897724fe6f 100644 --- a/servers/physics_2d/physics_server_2d_wrap_mt.cpp +++ b/servers/physics_2d/physics_server_2d_wrap_mt.cpp @@ -76,7 +76,7 @@ void PhysicsServer2DWrapMT::step(real_t p_step) { } void PhysicsServer2DWrapMT::sync() { - if (thread) { + if (create_thread) { if (first_frame) { first_frame = false; } else { @@ -97,7 +97,7 @@ void PhysicsServer2DWrapMT::end_sync() { void PhysicsServer2DWrapMT::init() { if (create_thread) { //OS::get_singleton()->release_rendering_thread(); - thread = Thread::create(_thread_callback, this); + thread.start(_thread_callback, this); while (!step_thread_up) { OS::get_singleton()->delay_usec(1000); } @@ -107,12 +107,9 @@ void PhysicsServer2DWrapMT::init() { } void PhysicsServer2DWrapMT::finish() { - if (thread) { + if (thread.is_started()) { command_queue.push(this, &PhysicsServer2DWrapMT::thread_exit); - Thread::wait_to_finish(thread); - memdelete(thread); - - thread = nullptr; + thread.wait_to_finish(); } else { physics_2d_server->finish(); } @@ -135,7 +132,6 @@ PhysicsServer2DWrapMT::PhysicsServer2DWrapMT(PhysicsServer2D *p_contained, bool command_queue(p_create_thread) { physics_2d_server = p_contained; create_thread = p_create_thread; - thread = nullptr; step_pending = 0; step_thread_up = false; diff --git a/servers/physics_2d/physics_server_2d_wrap_mt.h b/servers/physics_2d/physics_server_2d_wrap_mt.h index 9207081a51..fbc5b1eaa1 100644 --- a/servers/physics_2d/physics_server_2d_wrap_mt.h +++ b/servers/physics_2d/physics_server_2d_wrap_mt.h @@ -53,7 +53,7 @@ class PhysicsServer2DWrapMT : public PhysicsServer2D { Thread::ID server_thread; Thread::ID main_thread; volatile bool exit; - Thread *thread; + Thread thread; volatile bool step_thread_up; bool create_thread; @@ -189,7 +189,7 @@ public: FUNC2RC(RID, body_get_shape, RID, int); FUNC3(body_set_shape_disabled, RID, int, bool); - FUNC4(body_set_shape_as_one_way_collision, RID, int, bool, float); + FUNC4(body_set_shape_as_one_way_collision, RID, int, bool, real_t); FUNC2(body_remove_shape, RID, int); FUNC1(body_clear_shapes, RID); @@ -255,7 +255,7 @@ public: return physics_2d_server->body_test_motion(p_body, p_from, p_motion, p_infinite_inertia, p_margin, r_result, p_exclude_raycast_shapes); } - int body_test_ray_separation(RID p_body, const Transform2D &p_transform, bool p_infinite_inertia, Vector2 &r_recover_motion, SeparationResult *r_results, int p_result_max, float p_margin = 0.001) { + int body_test_ray_separation(RID p_body, const Transform2D &p_transform, bool p_infinite_inertia, Vector2 &r_recover_motion, SeparationResult *r_results, int p_result_max, real_t p_margin = 0.001) { ERR_FAIL_COND_V(main_thread != Thread::get_caller_id(), false); return physics_2d_server->body_test_ray_separation(p_body, p_transform, p_infinite_inertia, r_recover_motion, r_results, p_result_max, p_margin); } diff --git a/servers/physics_2d/shape_2d_sw.cpp b/servers/physics_2d/shape_2d_sw.cpp index 24c73314d8..6e7e802a8b 100644 --- a/servers/physics_2d/shape_2d_sw.cpp +++ b/servers/physics_2d/shape_2d_sw.cpp @@ -339,10 +339,10 @@ void RectangleShape2DSW::get_supports(const Vector2 &p_normal, Vector2 *r_suppor } bool RectangleShape2DSW::contains_point(const Vector2 &p_point) const { - float x = p_point.x; - float y = p_point.y; - float edge_x = half_extents.x; - float edge_y = half_extents.y; + real_t x = p_point.x; + real_t y = p_point.y; + real_t edge_x = half_extents.x; + real_t edge_y = half_extents.y; return (x >= -edge_x) && (x < edge_x) && (y >= -edge_y) && (y < edge_y); } @@ -590,7 +590,11 @@ real_t ConvexPolygonShape2DSW::get_moment_of_inertia(real_t p_mass, const Size2 } void ConvexPolygonShape2DSW::set_data(const Variant &p_data) { +#ifdef REAL_T_IS_DOUBLE + ERR_FAIL_COND(p_data.get_type() != Variant::PACKED_VECTOR2_ARRAY && p_data.get_type() != Variant::PACKED_FLOAT64_ARRAY); +#else ERR_FAIL_COND(p_data.get_type() != Variant::PACKED_VECTOR2_ARRAY && p_data.get_type() != Variant::PACKED_FLOAT32_ARRAY); +#endif if (points) { memdelete_arr(points); @@ -829,7 +833,11 @@ int ConcavePolygonShape2DSW::_generate_bvh(BVH *p_bvh, int p_len, int p_depth) { } void ConcavePolygonShape2DSW::set_data(const Variant &p_data) { +#ifdef REAL_T_IS_DOUBLE + ERR_FAIL_COND(p_data.get_type() != Variant::PACKED_VECTOR2_ARRAY && p_data.get_type() != Variant::PACKED_FLOAT64_ARRAY); +#else ERR_FAIL_COND(p_data.get_type() != Variant::PACKED_VECTOR2_ARRAY && p_data.get_type() != Variant::PACKED_FLOAT32_ARRAY); +#endif Rect2 aabb; diff --git a/servers/physics_2d/space_2d_sw.cpp b/servers/physics_2d/space_2d_sw.cpp index c2a6dc828e..05136e2501 100644 --- a/servers/physics_2d/space_2d_sw.cpp +++ b/servers/physics_2d/space_2d_sw.cpp @@ -278,9 +278,9 @@ bool PhysicsDirectSpaceState2DSW::cast_motion(const RID &p_shape, const Transfor continue; } - //test initial overlap + //test initial overlap, ignore objects it's inside of. if (CollisionSolver2DSW::solve(shape, p_xform, Vector2(), col_obj->get_shape(shape_idx), col_obj_xform, Vector2(), nullptr, nullptr, nullptr, p_margin)) { - return false; + continue; } //just do kinematic solving @@ -376,25 +376,25 @@ struct _RestCallbackData2D { Vector2 best_normal; real_t best_len; Vector2 valid_dir; - real_t valid_depth; real_t min_allowed_depth; }; static void _rest_cbk_result(const Vector2 &p_point_A, const Vector2 &p_point_B, void *p_userdata) { _RestCallbackData2D *rd = (_RestCallbackData2D *)p_userdata; - if (rd->valid_dir != Vector2()) { - if (p_point_A.distance_squared_to(p_point_B) > rd->valid_depth * rd->valid_depth) { - return; - } - if (rd->valid_dir.dot((p_point_A - p_point_B).normalized()) < Math_PI * 0.25) { - return; - } - } - Vector2 contact_rel = p_point_B - p_point_A; real_t len = contact_rel.length(); + if (len == 0) { + return; + } + + Vector2 normal = contact_rel / len; + + if (rd->valid_dir != Vector2() && rd->valid_dir.dot(normal) > -CMP_EPSILON) { + return; + } + if (len < rd->min_allowed_depth) { return; } @@ -405,7 +405,7 @@ static void _rest_cbk_result(const Vector2 &p_point_A, const Vector2 &p_point_B, rd->best_len = len; rd->best_contact = p_point_B; - rd->best_normal = contact_rel / len; + rd->best_normal = normal; rd->best_object = rd->object; rd->best_shape = rd->shape; rd->best_local_shape = rd->local_shape; @@ -440,7 +440,6 @@ bool PhysicsDirectSpaceState2DSW::rest_info(RID p_shape, const Transform2D &p_sh } rcd.valid_dir = Vector2(); - rcd.valid_depth = 0; rcd.object = col_obj; rcd.shape = shape_idx; rcd.local_shape = 0; @@ -643,9 +642,9 @@ int Space2DSW::test_body_ray_separation(Body2DSW *p_body, const Transform2D &p_t Vector2 a = sr[k * 2 + 0]; Vector2 b = sr[k * 2 + 1]; - recover_motion += (b - a) * 0.4; + recover_motion += (b - a) / cbk.amount; - float depth = a.distance_to(b); + real_t depth = a.distance_to(b); if (depth > result.collision_depth) { result.collision_depth = depth; result.collision_point = b; @@ -740,7 +739,7 @@ bool Space2DSW::test_body_motion(Body2DSW *p_body, const Transform2D &p_from, co ExcludedShapeSW excluded_shape_pairs[max_excluded_shape_pairs]; int excluded_shape_pair_count = 0; - float separation_margin = MIN(p_margin, MAX(0.0, p_motion.length() - CMP_EPSILON)); //don't separate by more than the intended motion + real_t separation_margin = MIN(p_margin, MAX(0.0, p_motion.length() - CMP_EPSILON)); //don't separate by more than the intended motion Transform2D body_transform = p_from; @@ -794,7 +793,7 @@ bool Space2DSW::test_body_motion(Body2DSW *p_body, const Transform2D &p_from, co if (col_obj->is_shape_set_as_one_way_collision(shape_idx)) { cbk.valid_dir = col_obj_shape_xform.get_axis(1).normalized(); - float owc_margin = col_obj->get_shape_one_way_collision_margin(shape_idx); + real_t owc_margin = col_obj->get_shape_one_way_collision_margin(shape_idx); cbk.valid_depth = MAX(owc_margin, p_margin); //user specified, but never less than actual margin or it won't work cbk.invalid_by_dir = 0; @@ -805,7 +804,7 @@ bool Space2DSW::test_body_motion(Body2DSW *p_body, const Transform2D &p_from, co Vector2 lv = b->get_linear_velocity(); //compute displacement from linear velocity Vector2 motion = lv * PhysicsDirectBodyState2DSW::singleton->step; - float motion_len = motion.length(); + real_t motion_len = motion.length(); motion.normalize(); cbk.valid_depth += motion_len * MAX(motion.dot(-cbk.valid_dir), 0.0); } @@ -850,7 +849,7 @@ bool Space2DSW::test_body_motion(Body2DSW *p_body, const Transform2D &p_from, co for (int i = 0; i < cbk.amount; i++) { Vector2 a = sr[i * 2 + 0]; Vector2 b = sr[i * 2 + 1]; - recover_motion += (b - a) * 0.4; + recover_motion += (b - a) / cbk.amount; } if (recover_motion == Vector2()) { @@ -1002,7 +1001,7 @@ bool Space2DSW::test_body_motion(Body2DSW *p_body, const Transform2D &p_from, co best_shape = -1; //no best shape with cast, reset to -1 } - { + if (safe < 1) { //it collided, let's get the rest info in unsafe advance Transform2D ugt = body_transform; ugt.elements[2] += p_motion * unsafe; @@ -1061,10 +1060,8 @@ bool Space2DSW::test_body_motion(Body2DSW *p_body, const Transform2D &p_from, co if (col_obj->is_shape_set_as_one_way_collision(shape_idx)) { rcd.valid_dir = col_obj_shape_xform.get_axis(1).normalized(); - rcd.valid_depth = 10e20; } else { rcd.valid_dir = Vector2(); - rcd.valid_depth = 0; } rcd.object = col_obj; @@ -1331,7 +1328,7 @@ Space2DSW::Space2DSW() { constraint_bias = 0.2; body_linear_velocity_sleep_threshold = GLOBAL_DEF("physics/2d/sleep_threshold_linear", 2.0); - body_angular_velocity_sleep_threshold = GLOBAL_DEF("physics/2d/sleep_threshold_angular", (8.0 / 180.0 * Math_PI)); + body_angular_velocity_sleep_threshold = GLOBAL_DEF("physics/2d/sleep_threshold_angular", Math::deg2rad(8.0)); body_time_to_sleep = GLOBAL_DEF("physics/2d/time_before_sleep", 0.5); ProjectSettings::get_singleton()->set_custom_property_info("physics/2d/time_before_sleep", PropertyInfo(Variant::FLOAT, "physics/2d/time_before_sleep", PROPERTY_HINT_RANGE, "0,5,0.01,or_greater")); diff --git a/servers/physics_3d/body_3d_sw.h b/servers/physics_3d/body_3d_sw.h index 41578778f6..8e21003a5f 100644 --- a/servers/physics_3d/body_3d_sw.h +++ b/servers/physics_3d/body_3d_sw.h @@ -426,7 +426,7 @@ public: ERR_FAIL_INDEX_V(p_contact_idx, body->contact_count, Vector3()); return body->contacts[p_contact_idx].local_normal; } - virtual float get_contact_impulse(int p_contact_idx) const override { + virtual real_t get_contact_impulse(int p_contact_idx) const override { return 0.0f; // Only implemented for bullet } virtual int get_contact_local_shape(int p_contact_idx) const override { diff --git a/servers/physics_3d/joints/cone_twist_joint_3d_sw.cpp b/servers/physics_3d/joints/cone_twist_joint_3d_sw.cpp index 7b10257157..9c4493f4a2 100644 --- a/servers/physics_3d/joints/cone_twist_joint_3d_sw.cpp +++ b/servers/physics_3d/joints/cone_twist_joint_3d_sw.cpp @@ -92,9 +92,9 @@ ConeTwistJoint3DSW::ConeTwistJoint3DSW(Body3DSW *rbA, Body3DSW *rbB, const Trans m_rbAFrame = rbAFrame; m_rbBFrame = rbBFrame; - m_swingSpan1 = Math_PI / 4.0; - m_swingSpan2 = Math_PI / 4.0; - m_twistSpan = Math_PI * 2; + m_swingSpan1 = Math_TAU / 8.0; + m_swingSpan2 = Math_TAU / 8.0; + m_twistSpan = Math_TAU; m_biasFactor = 0.3f; m_relaxationFactor = 1.0f; diff --git a/servers/physics_3d/joints/generic_6dof_joint_3d_sw.cpp b/servers/physics_3d/joints/generic_6dof_joint_3d_sw.cpp index 7eb49e657b..13b389251f 100644 --- a/servers/physics_3d/joints/generic_6dof_joint_3d_sw.cpp +++ b/servers/physics_3d/joints/generic_6dof_joint_3d_sw.cpp @@ -132,7 +132,7 @@ real_t G6DOFRotationalLimitMotor3DSW::solveAngularLimits( real_t oldaccumImpulse = m_accumulatedImpulse; real_t sum = oldaccumImpulse + clippedMotorImpulse; - m_accumulatedImpulse = sum > hi ? real_t(0.) : sum < lo ? real_t(0.) : sum; + m_accumulatedImpulse = sum > hi ? real_t(0.) : (sum < lo ? real_t(0.) : sum); clippedMotorImpulse = m_accumulatedImpulse - oldaccumImpulse; @@ -201,7 +201,7 @@ real_t G6DOFTranslationalLimitMotor3DSW::solveLinearAxis( real_t oldNormalImpulse = m_accumulatedImpulse[limit_index]; real_t sum = oldNormalImpulse + normalImpulse; - m_accumulatedImpulse[limit_index] = sum > hi ? real_t(0.) : sum < lo ? real_t(0.) : sum; + m_accumulatedImpulse[limit_index] = sum > hi ? real_t(0.) : (sum < lo ? real_t(0.) : sum); normalImpulse = m_accumulatedImpulse[limit_index] - oldNormalImpulse; Vector3 impulse_vector = axis_normal_on_a * normalImpulse; diff --git a/servers/physics_3d/physics_server_3d_sw.cpp b/servers/physics_3d/physics_server_3d_sw.cpp index 274de8411c..b554a23bf2 100644 --- a/servers/physics_3d/physics_server_3d_sw.cpp +++ b/servers/physics_3d/physics_server_3d_sw.cpp @@ -874,7 +874,7 @@ bool PhysicsServer3DSW::body_test_motion(RID p_body, const Transform &p_from, co return body->get_space()->test_body_motion(body, p_from, p_motion, p_infinite_inertia, body->get_kinematic_margin(), r_result, p_exclude_raycast_shapes); } -int PhysicsServer3DSW::body_test_ray_separation(RID p_body, const Transform &p_transform, bool p_infinite_inertia, Vector3 &r_recover_motion, SeparationResult *r_results, int p_result_max, float p_margin) { +int PhysicsServer3DSW::body_test_ray_separation(RID p_body, const Transform &p_transform, bool p_infinite_inertia, Vector3 &r_recover_motion, SeparationResult *r_results, int p_result_max, real_t p_margin) { Body3DSW *body = body_owner.getornull(p_body); ERR_FAIL_COND_V(!body, false); ERR_FAIL_COND_V(!body->get_space(), false); diff --git a/servers/physics_3d/physics_server_3d_sw.h b/servers/physics_3d/physics_server_3d_sw.h index 9b6b113677..c48db81d97 100644 --- a/servers/physics_3d/physics_server_3d_sw.h +++ b/servers/physics_3d/physics_server_3d_sw.h @@ -235,7 +235,7 @@ public: virtual bool body_is_ray_pickable(RID p_body) const override; virtual bool body_test_motion(RID p_body, const Transform &p_from, const Vector3 &p_motion, bool p_infinite_inertia, MotionResult *r_result = nullptr, bool p_exclude_raycast_shapes = true) override; - virtual int body_test_ray_separation(RID p_body, const Transform &p_transform, bool p_infinite_inertia, Vector3 &r_recover_motion, SeparationResult *r_results, int p_result_max, float p_margin = 0.001) override; + virtual int body_test_ray_separation(RID p_body, const Transform &p_transform, bool p_infinite_inertia, Vector3 &r_recover_motion, SeparationResult *r_results, int p_result_max, real_t p_margin = 0.001) override; // this function only works on physics process, errors and returns null otherwise virtual PhysicsDirectBodyState3D *body_get_direct_state(RID p_body) override; diff --git a/servers/physics_3d/shape_3d_sw.cpp b/servers/physics_3d/shape_3d_sw.cpp index f2adcc1072..9c37060bea 100644 --- a/servers/physics_3d/shape_3d_sw.cpp +++ b/servers/physics_3d/shape_3d_sw.cpp @@ -261,7 +261,7 @@ bool SphereShape3DSW::intersect_point(const Vector3 &p_point) const { Vector3 SphereShape3DSW::get_closest_point_to(const Vector3 &p_point) const { Vector3 p = p_point; - float l = p.length(); + real_t l = p.length(); if (l < radius) { return p_point; } @@ -429,7 +429,7 @@ Vector3 BoxShape3DSW::get_closest_point_to(const Vector3 &p_point) const { } //check segments - float min_distance = 1e20; + real_t min_distance = 1e20; Vector3 closest_vertex = half_extents * p_point.sign(); Vector3 s[2] = { closest_vertex, @@ -442,7 +442,7 @@ Vector3 BoxShape3DSW::get_closest_point_to(const Vector3 &p_point) const { Vector3 closest_edge = Geometry3D::get_closest_point_to_segment(p_point, s); - float d = p_point.distance_to(closest_edge); + real_t d = p_point.distance_to(closest_edge); if (d < min_distance) { min_point = closest_edge; min_distance = d; @@ -839,7 +839,7 @@ Vector3 ConvexPolygonShape3DSW::get_closest_point_to(const Vector3 &p_point) con return p_point; } - float min_distance = 1e20; + real_t min_distance = 1e20; Vector3 min_point; //check edges @@ -852,7 +852,7 @@ Vector3 ConvexPolygonShape3DSW::get_closest_point_to(const Vector3 &p_point) con }; Vector3 closest = Geometry3D::get_closest_point_to_segment(p_point, s); - float d = closest.distance_to(p_point); + real_t d = closest.distance_to(p_point); if (d < min_distance) { min_distance = d; min_point = closest; diff --git a/servers/physics_3d/space_3d_sw.cpp b/servers/physics_3d/space_3d_sw.cpp index 2b2b5122da..43cc032120 100644 --- a/servers/physics_3d/space_3d_sw.cpp +++ b/servers/physics_3d/space_3d_sw.cpp @@ -274,11 +274,11 @@ bool PhysicsDirectSpaceState3DSW::cast_motion(const RID &p_shape, const Transfor continue; } - //test initial overlap + //test initial overlap, ignore objects it's inside of. sep_axis = p_motion.normalized(); if (!CollisionSolver3DSW::solve_distance(shape, p_xform, col_obj->get_shape(shape_idx), col_obj_xform, point_A, point_B, aabb, &sep_axis)) { - return false; + continue; } //just do kinematic solving @@ -476,7 +476,7 @@ Vector3 PhysicsDirectSpaceState3DSW::get_closest_point_to_object_volume(RID p_ob ERR_FAIL_COND_V(obj->get_space() != space, Vector3()); - float min_distance = 1e20; + real_t min_distance = 1e20; Vector3 min_point; bool shapes_found = false; @@ -492,7 +492,7 @@ Vector3 PhysicsDirectSpaceState3DSW::get_closest_point_to_object_volume(RID p_ob Vector3 point = shape->get_closest_point_to(shape_xform.affine_inverse().xform(p_point)); point = shape_xform.xform(point); - float dist = point.distance_to(p_point); + real_t dist = point.distance_to(p_point); if (dist < min_distance) { min_distance = dist; min_point = point; @@ -649,9 +649,9 @@ int Space3DSW::test_body_ray_separation(Body3DSW *p_body, const Transform &p_tra Vector3 a = sr[k * 2 + 0]; Vector3 b = sr[k * 2 + 1]; - recover_motion += (b - a) * 0.4; + recover_motion += (b - a) / cbk.amount; - float depth = a.distance_to(b); + real_t depth = a.distance_to(b); if (depth > result.collision_depth) { result.collision_depth = depth; result.collision_point = b; @@ -791,7 +791,7 @@ bool Space3DSW::test_body_motion(Body3DSW *p_body, const Transform &p_from, cons for (int i = 0; i < cbk.amount; i++) { Vector3 a = sr[i * 2 + 0]; Vector3 b = sr[i * 2 + 1]; - recover_motion += (b - a) * 0.4; + recover_motion += (b - a) / cbk.amount; } if (recover_motion == Vector3()) { @@ -1211,7 +1211,7 @@ Space3DSW::Space3DSW() { constraint_bias = 0.01; body_linear_velocity_sleep_threshold = GLOBAL_DEF("physics/3d/sleep_threshold_linear", 0.1); - body_angular_velocity_sleep_threshold = GLOBAL_DEF("physics/3d/sleep_threshold_angular", (8.0 / 180.0 * Math_PI)); + body_angular_velocity_sleep_threshold = GLOBAL_DEF("physics/3d/sleep_threshold_angular", Math::deg2rad(8.0)); body_time_to_sleep = GLOBAL_DEF("physics/3d/time_before_sleep", 0.5); ProjectSettings::get_singleton()->set_custom_property_info("physics/3d/time_before_sleep", PropertyInfo(Variant::FLOAT, "physics/3d/time_before_sleep", PROPERTY_HINT_RANGE, "0,5,0.01,or_greater")); body_angular_velocity_damp_ratio = 10; diff --git a/servers/physics_server_2d.cpp b/servers/physics_server_2d.cpp index a6f64f5848..7c36229e9f 100644 --- a/servers/physics_server_2d.cpp +++ b/servers/physics_server_2d.cpp @@ -42,7 +42,7 @@ void PhysicsDirectBodyState2D::integrate_forces() { real_t av = get_angular_velocity(); - float damp = 1.0 - step * get_total_linear_damp(); + real_t damp = 1.0 - step * get_total_linear_damp(); if (damp < 0) { // reached zero in the given time damp = 0; @@ -168,11 +168,11 @@ Vector2 PhysicsShapeQueryParameters2D::get_motion() const { return motion; } -void PhysicsShapeQueryParameters2D::set_margin(float p_margin) { +void PhysicsShapeQueryParameters2D::set_margin(real_t p_margin) { margin = p_margin; } -float PhysicsShapeQueryParameters2D::get_margin() const { +real_t PhysicsShapeQueryParameters2D::get_margin() const { return margin; } @@ -311,7 +311,7 @@ Array PhysicsDirectSpaceState2D::_intersect_shape(const Ref<PhysicsShapeQueryPar Array PhysicsDirectSpaceState2D::_cast_motion(const Ref<PhysicsShapeQueryParameters2D> &p_shape_query) { ERR_FAIL_COND_V(!p_shape_query.is_valid(), Array()); - float closest_safe, closest_unsafe; + real_t closest_safe, closest_unsafe; bool res = cast_motion(p_shape_query->shape, p_shape_query->transform, p_shape_query->motion, p_shape_query->margin, closest_safe, closest_unsafe, p_shape_query->exclude, p_shape_query->collision_mask, p_shape_query->collide_with_bodies, p_shape_query->collide_with_areas); if (!res) { return Array(); @@ -517,7 +517,7 @@ PhysicsTestMotionResult2D::PhysicsTestMotionResult2D() { /////////////////////////////////////// -bool PhysicsServer2D::_body_test_motion(RID p_body, const Transform2D &p_from, const Vector2 &p_motion, bool p_infinite_inertia, float p_margin, const Ref<PhysicsTestMotionResult2D> &p_result) { +bool PhysicsServer2D::_body_test_motion(RID p_body, const Transform2D &p_from, const Vector2 &p_motion, bool p_infinite_inertia, real_t p_margin, const Ref<PhysicsTestMotionResult2D> &p_result) { MotionResult *r = nullptr; if (p_result.is_valid()) { r = p_result->get_result_ptr(); diff --git a/servers/physics_server_2d.h b/servers/physics_server_2d.h index dd38855199..710eecfdec 100644 --- a/servers/physics_server_2d.h +++ b/servers/physics_server_2d.h @@ -45,10 +45,10 @@ protected: public: virtual Vector2 get_total_gravity() const = 0; // get gravity vector working on this body space/area - virtual float get_total_linear_damp() const = 0; // get density of this body space/area - virtual float get_total_angular_damp() const = 0; // get density of this body space/area + virtual real_t get_total_linear_damp() const = 0; // get density of this body space/area + virtual real_t get_total_angular_damp() const = 0; // get density of this body space/area - virtual float get_inverse_mass() const = 0; // get the mass + virtual real_t get_inverse_mass() const = 0; // get the mass virtual real_t get_inverse_inertia() const = 0; // get density of this body space virtual void set_linear_velocity(const Vector2 &p_velocity) = 0; @@ -103,7 +103,7 @@ class PhysicsShapeQueryParameters2D : public Reference { RID shape; Transform2D transform; Vector2 motion; - float margin; + real_t margin; Set<RID> exclude; uint32_t collision_mask; @@ -125,8 +125,8 @@ public: void set_motion(const Vector2 &p_motion); Vector2 get_motion() const; - void set_margin(float p_margin); - float get_margin() const; + void set_margin(real_t p_margin); + real_t get_margin() const; void set_collision_mask(int p_collision_mask); int get_collision_mask() const; @@ -182,11 +182,11 @@ public: virtual int intersect_point(const Vector2 &p_point, ShapeResult *r_results, int p_result_max, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false, bool p_pick_point = false) = 0; virtual int intersect_point_on_canvas(const Vector2 &p_point, ObjectID p_canvas_instance_id, ShapeResult *r_results, int p_result_max, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false, bool p_pick_point = false) = 0; - virtual int intersect_shape(const RID &p_shape, const Transform2D &p_xform, const Vector2 &p_motion, float p_margin, ShapeResult *r_results, int p_result_max, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; + virtual int intersect_shape(const RID &p_shape, const Transform2D &p_xform, const Vector2 &p_motion, real_t p_margin, ShapeResult *r_results, int p_result_max, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; - virtual bool cast_motion(const RID &p_shape, const Transform2D &p_xform, const Vector2 &p_motion, float p_margin, float &p_closest_safe, float &p_closest_unsafe, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; + virtual bool cast_motion(const RID &p_shape, const Transform2D &p_xform, const Vector2 &p_motion, real_t p_margin, real_t &p_closest_safe, real_t &p_closest_unsafe, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; - virtual bool collide_shape(RID p_shape, const Transform2D &p_shape_xform, const Vector2 &p_motion, float p_margin, Vector2 *r_results, int p_result_max, int &r_result_count, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; + virtual bool collide_shape(RID p_shape, const Transform2D &p_shape_xform, const Vector2 &p_motion, real_t p_margin, Vector2 *r_results, int p_result_max, int &r_result_count, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; struct ShapeRestInfo { Vector2 point; @@ -198,7 +198,7 @@ public: Variant metadata; }; - virtual bool rest_info(RID p_shape, const Transform2D &p_shape_xform, const Vector2 &p_motion, float p_margin, ShapeRestInfo *r_info, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; + virtual bool rest_info(RID p_shape, const Transform2D &p_shape_xform, const Vector2 &p_motion, real_t p_margin, ShapeRestInfo *r_info, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_layer = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; PhysicsDirectSpaceState2D(); }; @@ -230,7 +230,7 @@ class PhysicsServer2D : public Object { static PhysicsServer2D *singleton; - virtual bool _body_test_motion(RID p_body, const Transform2D &p_from, const Vector2 &p_motion, bool p_infinite_inertia, float p_margin = 0.08, const Ref<PhysicsTestMotionResult2D> &p_result = Ref<PhysicsTestMotionResult2D>()); + virtual bool _body_test_motion(RID p_body, const Transform2D &p_from, const Vector2 &p_motion, bool p_infinite_inertia, real_t p_margin = 0.08, const Ref<PhysicsTestMotionResult2D> &p_result = Ref<PhysicsTestMotionResult2D>()); protected: static void _bind_methods(); @@ -393,7 +393,7 @@ public: virtual Variant body_get_shape_metadata(RID p_body, int p_shape_idx) const = 0; virtual void body_set_shape_disabled(RID p_body, int p_shape, bool p_disabled) = 0; - virtual void body_set_shape_as_one_way_collision(RID p_body, int p_shape, bool p_enabled, float p_margin = 0) = 0; + virtual void body_set_shape_as_one_way_collision(RID p_body, int p_shape, bool p_enabled, real_t p_margin = 0) = 0; virtual void body_remove_shape(RID p_body, int p_shape_idx) = 0; virtual void body_clear_shapes(RID p_body) = 0; @@ -431,8 +431,8 @@ public: BODY_PARAM_MAX, }; - virtual void body_set_param(RID p_body, BodyParameter p_param, float p_value) = 0; - virtual float body_get_param(RID p_body, BodyParameter p_param) const = 0; + virtual void body_set_param(RID p_body, BodyParameter p_param, real_t p_value) = 0; + virtual real_t body_get_param(RID p_body, BodyParameter p_param) const = 0; //state enum BodyState { @@ -450,15 +450,15 @@ public: virtual void body_set_applied_force(RID p_body, const Vector2 &p_force) = 0; virtual Vector2 body_get_applied_force(RID p_body) const = 0; - virtual void body_set_applied_torque(RID p_body, float p_torque) = 0; - virtual float body_get_applied_torque(RID p_body) const = 0; + virtual void body_set_applied_torque(RID p_body, real_t p_torque) = 0; + virtual real_t body_get_applied_torque(RID p_body) const = 0; virtual void body_add_central_force(RID p_body, const Vector2 &p_force) = 0; virtual void body_add_force(RID p_body, const Vector2 &p_force, const Vector2 &p_position = Vector2()) = 0; - virtual void body_add_torque(RID p_body, float p_torque) = 0; + virtual void body_add_torque(RID p_body, real_t p_torque) = 0; virtual void body_apply_central_impulse(RID p_body, const Vector2 &p_impulse) = 0; - virtual void body_apply_torque_impulse(RID p_body, float p_torque) = 0; + virtual void body_apply_torque_impulse(RID p_body, real_t p_torque) = 0; virtual void body_apply_impulse(RID p_body, const Vector2 &p_impulse, const Vector2 &p_position = Vector2()) = 0; virtual void body_set_axis_velocity(RID p_body, const Vector2 &p_axis_velocity) = 0; @@ -471,8 +471,8 @@ public: virtual int body_get_max_contacts_reported(RID p_body) const = 0; //missing remove - virtual void body_set_contacts_reported_depth_threshold(RID p_body, float p_threshold) = 0; - virtual float body_get_contacts_reported_depth_threshold(RID p_body) const = 0; + virtual void body_set_contacts_reported_depth_threshold(RID p_body, real_t p_threshold) = 0; + virtual real_t body_get_contacts_reported_depth_threshold(RID p_body) const = 0; virtual void body_set_omit_force_integration(RID p_body, bool p_omit) = 0; virtual bool body_is_omitting_force_integration(RID p_body) const = 0; @@ -506,10 +506,10 @@ public: } }; - virtual bool body_test_motion(RID p_body, const Transform2D &p_from, const Vector2 &p_motion, bool p_infinite_inertia, float p_margin = 0.001, MotionResult *r_result = nullptr, bool p_exclude_raycast_shapes = true) = 0; + virtual bool body_test_motion(RID p_body, const Transform2D &p_from, const Vector2 &p_motion, bool p_infinite_inertia, real_t p_margin = 0.001, MotionResult *r_result = nullptr, bool p_exclude_raycast_shapes = true) = 0; struct SeparationResult { - float collision_depth; + real_t collision_depth; Vector2 collision_point; Vector2 collision_normal; Vector2 collider_velocity; @@ -520,7 +520,7 @@ public: Variant collider_metadata; }; - virtual int body_test_ray_separation(RID p_body, const Transform2D &p_transform, bool p_infinite_inertia, Vector2 &r_recover_motion, SeparationResult *r_results, int p_result_max, float p_margin = 0.001) = 0; + virtual int body_test_ray_separation(RID p_body, const Transform2D &p_transform, bool p_infinite_inertia, Vector2 &r_recover_motion, SeparationResult *r_results, int p_result_max, real_t p_margin = 0.001) = 0; /* JOINT API */ @@ -576,7 +576,7 @@ public: virtual void set_active(bool p_active) = 0; virtual void init() = 0; - virtual void step(float p_step) = 0; + virtual void step(real_t p_step) = 0; virtual void sync() = 0; virtual void flush_queries() = 0; virtual void end_sync() = 0; diff --git a/servers/physics_server_3d.cpp b/servers/physics_server_3d.cpp index 27ebe32e50..a4dc80b0a6 100644 --- a/servers/physics_server_3d.cpp +++ b/servers/physics_server_3d.cpp @@ -42,13 +42,13 @@ void PhysicsDirectBodyState3D::integrate_forces() { Vector3 av = get_angular_velocity(); - float linear_damp = 1.0 - step * get_total_linear_damp(); + real_t linear_damp = 1.0 - step * get_total_linear_damp(); if (linear_damp < 0) { // reached zero in the given time linear_damp = 0; } - float angular_damp = 1.0 - step * get_total_angular_damp(); + real_t angular_damp = 1.0 - step * get_total_angular_damp(); if (angular_damp < 0) { // reached zero in the given time angular_damp = 0; @@ -128,7 +128,7 @@ void PhysicsDirectBodyState3D::_bind_methods() { ADD_PROPERTY(PropertyInfo(Variant::VECTOR3, "angular_velocity"), "set_angular_velocity", "get_angular_velocity"); ADD_PROPERTY(PropertyInfo(Variant::VECTOR3, "linear_velocity"), "set_linear_velocity", "get_linear_velocity"); ADD_PROPERTY(PropertyInfo(Variant::BOOL, "sleeping"), "set_sleep_state", "is_sleeping"); - ADD_PROPERTY(PropertyInfo(Variant::TRANSFORM2D, "transform"), "set_transform", "get_transform"); + ADD_PROPERTY(PropertyInfo(Variant::TRANSFORM, "transform"), "set_transform", "get_transform"); } PhysicsDirectBodyState3D::PhysicsDirectBodyState3D() {} @@ -164,11 +164,11 @@ Transform PhysicsShapeQueryParameters3D::get_transform() const { return transform; } -void PhysicsShapeQueryParameters3D::set_margin(float p_margin) { +void PhysicsShapeQueryParameters3D::set_margin(real_t p_margin) { margin = p_margin; } -float PhysicsShapeQueryParameters3D::get_margin() const { +real_t PhysicsShapeQueryParameters3D::get_margin() const { return margin; } @@ -303,7 +303,7 @@ Array PhysicsDirectSpaceState3D::_intersect_shape(const Ref<PhysicsShapeQueryPar Array PhysicsDirectSpaceState3D::_cast_motion(const Ref<PhysicsShapeQueryParameters3D> &p_shape_query, const Vector3 &p_motion) { ERR_FAIL_COND_V(!p_shape_query.is_valid(), Array()); - float closest_safe = 1.0f, closest_unsafe = 1.0f; + real_t closest_safe = 1.0f, closest_unsafe = 1.0f; bool res = cast_motion(p_shape_query->shape, p_shape_query->transform, p_motion, p_shape_query->margin, closest_safe, closest_unsafe, p_shape_query->exclude, p_shape_query->collision_mask, p_shape_query->collide_with_bodies, p_shape_query->collide_with_areas); if (!res) { return Array(); diff --git a/servers/physics_server_3d.h b/servers/physics_server_3d.h index 303825f37c..1349e0e033 100644 --- a/servers/physics_server_3d.h +++ b/servers/physics_server_3d.h @@ -44,12 +44,12 @@ protected: public: virtual Vector3 get_total_gravity() const = 0; - virtual float get_total_angular_damp() const = 0; - virtual float get_total_linear_damp() const = 0; + virtual real_t get_total_angular_damp() const = 0; + virtual real_t get_total_linear_damp() const = 0; virtual Vector3 get_center_of_mass() const = 0; virtual Basis get_principal_inertia_axes() const = 0; - virtual float get_inverse_mass() const = 0; // get the mass + virtual real_t get_inverse_mass() const = 0; // get the mass virtual Vector3 get_inverse_inertia() const = 0; // get density of this body space virtual Basis get_inverse_inertia_tensor() const = 0; // get density of this body space @@ -76,7 +76,7 @@ public: virtual Vector3 get_contact_local_position(int p_contact_idx) const = 0; virtual Vector3 get_contact_local_normal(int p_contact_idx) const = 0; - virtual float get_contact_impulse(int p_contact_idx) const = 0; + virtual real_t get_contact_impulse(int p_contact_idx) const = 0; virtual int get_contact_local_shape(int p_contact_idx) const = 0; virtual RID get_contact_collider(int p_contact_idx) const = 0; @@ -103,7 +103,7 @@ class PhysicsShapeQueryParameters3D : public Reference { RES shape_ref; RID shape; Transform transform; - float margin; + real_t margin; Set<RID> exclude; uint32_t collision_mask; @@ -122,8 +122,8 @@ public: void set_transform(const Transform &p_transform); Transform get_transform() const; - void set_margin(float p_margin); - float get_margin() const; + void set_margin(real_t p_margin); + real_t get_margin() const; void set_collision_mask(int p_collision_mask); int get_collision_mask() const; @@ -174,7 +174,7 @@ public: virtual bool intersect_ray(const Vector3 &p_from, const Vector3 &p_to, RayResult &r_result, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_mask = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false, bool p_pick_ray = false) = 0; - virtual int intersect_shape(const RID &p_shape, const Transform &p_xform, float p_margin, ShapeResult *r_results, int p_result_max, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_mask = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; + virtual int intersect_shape(const RID &p_shape, const Transform &p_xform, real_t p_margin, ShapeResult *r_results, int p_result_max, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_mask = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; struct ShapeRestInfo { Vector3 point; @@ -185,11 +185,11 @@ public: Vector3 linear_velocity; //velocity at contact point }; - virtual bool cast_motion(const RID &p_shape, const Transform &p_xform, const Vector3 &p_motion, float p_margin, float &p_closest_safe, float &p_closest_unsafe, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_mask = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false, ShapeRestInfo *r_info = nullptr) = 0; + virtual bool cast_motion(const RID &p_shape, const Transform &p_xform, const Vector3 &p_motion, real_t p_margin, real_t &p_closest_safe, real_t &p_closest_unsafe, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_mask = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false, ShapeRestInfo *r_info = nullptr) = 0; - virtual bool collide_shape(RID p_shape, const Transform &p_shape_xform, float p_margin, Vector3 *r_results, int p_result_max, int &r_result_count, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_mask = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; + virtual bool collide_shape(RID p_shape, const Transform &p_shape_xform, real_t p_margin, Vector3 *r_results, int p_result_max, int &r_result_count, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_mask = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; - virtual bool rest_info(RID p_shape, const Transform &p_shape_xform, float p_margin, ShapeRestInfo *r_info, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_mask = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; + virtual bool rest_info(RID p_shape, const Transform &p_shape_xform, real_t p_margin, ShapeRestInfo *r_info, const Set<RID> &p_exclude = Set<RID>(), uint32_t p_collision_mask = 0xFFFFFFFF, bool p_collide_with_bodies = true, bool p_collide_with_areas = false) = 0; virtual Vector3 get_closest_point_to_object_volume(RID p_object, const Vector3 p_point) const = 0; @@ -404,8 +404,8 @@ public: BODY_PARAM_MAX, }; - virtual void body_set_param(RID p_body, BodyParameter p_param, float p_value) = 0; - virtual float body_get_param(RID p_body, BodyParameter p_param) const = 0; + virtual void body_set_param(RID p_body, BodyParameter p_param, real_t p_value) = 0; + virtual real_t body_get_param(RID p_body, BodyParameter p_param) const = 0; virtual void body_set_kinematic_safe_margin(RID p_body, real_t p_margin) = 0; virtual real_t body_get_kinematic_safe_margin(RID p_body) const = 0; @@ -459,8 +459,8 @@ public: virtual int body_get_max_contacts_reported(RID p_body) const = 0; //missing remove - virtual void body_set_contacts_reported_depth_threshold(RID p_body, float p_threshold) = 0; - virtual float body_get_contacts_reported_depth_threshold(RID p_body) const = 0; + virtual void body_set_contacts_reported_depth_threshold(RID p_body, real_t p_threshold) = 0; + virtual real_t body_get_contacts_reported_depth_threshold(RID p_body) const = 0; virtual void body_set_omit_force_integration(RID p_body, bool p_omit) = 0; virtual bool body_is_omitting_force_integration(RID p_body) const = 0; @@ -495,7 +495,7 @@ public: virtual bool body_test_motion(RID p_body, const Transform &p_from, const Vector3 &p_motion, bool p_infinite_inertia, MotionResult *r_result = nullptr, bool p_exclude_raycast_shapes = true) = 0; struct SeparationResult { - float collision_depth; + real_t collision_depth; Vector3 collision_point; Vector3 collision_normal; Vector3 collider_velocity; @@ -506,7 +506,7 @@ public: Variant collider_metadata; }; - virtual int body_test_ray_separation(RID p_body, const Transform &p_transform, bool p_infinite_inertia, Vector3 &r_recover_motion, SeparationResult *r_results, int p_result_max, float p_margin = 0.001) = 0; + virtual int body_test_ray_separation(RID p_body, const Transform &p_transform, bool p_infinite_inertia, Vector3 &r_recover_motion, SeparationResult *r_results, int p_result_max, real_t p_margin = 0.001) = 0; /* SOFT BODY */ @@ -601,8 +601,8 @@ public: PIN_JOINT_IMPULSE_CLAMP }; - virtual void pin_joint_set_param(RID p_joint, PinJointParam p_param, float p_value) = 0; - virtual float pin_joint_get_param(RID p_joint, PinJointParam p_param) const = 0; + virtual void pin_joint_set_param(RID p_joint, PinJointParam p_param, real_t p_value) = 0; + virtual real_t pin_joint_get_param(RID p_joint, PinJointParam p_param) const = 0; virtual void pin_joint_set_local_a(RID p_joint, const Vector3 &p_A) = 0; virtual Vector3 pin_joint_get_local_a(RID p_joint) const = 0; @@ -631,8 +631,8 @@ public: virtual RID joint_create_hinge(RID p_body_A, const Transform &p_hinge_A, RID p_body_B, const Transform &p_hinge_B) = 0; virtual RID joint_create_hinge_simple(RID p_body_A, const Vector3 &p_pivot_A, const Vector3 &p_axis_A, RID p_body_B, const Vector3 &p_pivot_B, const Vector3 &p_axis_B) = 0; - virtual void hinge_joint_set_param(RID p_joint, HingeJointParam p_param, float p_value) = 0; - virtual float hinge_joint_get_param(RID p_joint, HingeJointParam p_param) const = 0; + virtual void hinge_joint_set_param(RID p_joint, HingeJointParam p_param, real_t p_value) = 0; + virtual real_t hinge_joint_get_param(RID p_joint, HingeJointParam p_param) const = 0; virtual void hinge_joint_set_flag(RID p_joint, HingeJointFlag p_flag, bool p_value) = 0; virtual bool hinge_joint_get_flag(RID p_joint, HingeJointFlag p_flag) const = 0; @@ -667,8 +667,8 @@ public: virtual RID joint_create_slider(RID p_body_A, const Transform &p_local_frame_A, RID p_body_B, const Transform &p_local_frame_B) = 0; //reference frame is A - virtual void slider_joint_set_param(RID p_joint, SliderJointParam p_param, float p_value) = 0; - virtual float slider_joint_get_param(RID p_joint, SliderJointParam p_param) const = 0; + virtual void slider_joint_set_param(RID p_joint, SliderJointParam p_param, real_t p_value) = 0; + virtual real_t slider_joint_get_param(RID p_joint, SliderJointParam p_param) const = 0; enum ConeTwistJointParam { CONE_TWIST_JOINT_SWING_SPAN, @@ -681,8 +681,8 @@ public: virtual RID joint_create_cone_twist(RID p_body_A, const Transform &p_local_frame_A, RID p_body_B, const Transform &p_local_frame_B) = 0; //reference frame is A - virtual void cone_twist_joint_set_param(RID p_joint, ConeTwistJointParam p_param, float p_value) = 0; - virtual float cone_twist_joint_get_param(RID p_joint, ConeTwistJointParam p_param) const = 0; + virtual void cone_twist_joint_set_param(RID p_joint, ConeTwistJointParam p_param, real_t p_value) = 0; + virtual real_t cone_twist_joint_get_param(RID p_joint, ConeTwistJointParam p_param) const = 0; enum G6DOFJointAxisParam { G6DOF_JOINT_LINEAR_LOWER_LIMIT, @@ -722,8 +722,8 @@ public: virtual RID joint_create_generic_6dof(RID p_body_A, const Transform &p_local_frame_A, RID p_body_B, const Transform &p_local_frame_B) = 0; //reference frame is A - virtual void generic_6dof_joint_set_param(RID p_joint, Vector3::Axis, G6DOFJointAxisParam p_param, float p_value) = 0; - virtual float generic_6dof_joint_get_param(RID p_joint, Vector3::Axis, G6DOFJointAxisParam p_param) = 0; + virtual void generic_6dof_joint_set_param(RID p_joint, Vector3::Axis, G6DOFJointAxisParam p_param, real_t p_value) = 0; + virtual real_t generic_6dof_joint_get_param(RID p_joint, Vector3::Axis, G6DOFJointAxisParam p_param) = 0; virtual void generic_6dof_joint_set_flag(RID p_joint, Vector3::Axis, G6DOFJointAxisFlag p_flag, bool p_enable) = 0; virtual bool generic_6dof_joint_get_flag(RID p_joint, Vector3::Axis, G6DOFJointAxisFlag p_flag) = 0; @@ -741,7 +741,7 @@ public: virtual void set_active(bool p_active) = 0; virtual void init() = 0; - virtual void step(float p_step) = 0; + virtual void step(real_t p_step) = 0; virtual void flush_queries() = 0; virtual void finish() = 0; diff --git a/servers/register_server_types.cpp b/servers/register_server_types.cpp index 58bcdf5802..50efd7c554 100644 --- a/servers/register_server_types.cpp +++ b/servers/register_server_types.cpp @@ -36,6 +36,7 @@ #include "audio/audio_effect.h" #include "audio/audio_stream.h" #include "audio/effects/audio_effect_amplify.h" +#include "audio/effects/audio_effect_capture.h" #include "audio/effects/audio_effect_chorus.h" #include "audio/effects/audio_effect_compressor.h" #include "audio/effects/audio_effect_delay.h" @@ -166,6 +167,8 @@ void register_server_types() { ClassDB::register_class<AudioEffectRecord>(); ClassDB::register_class<AudioEffectSpectrumAnalyzer>(); ClassDB::register_virtual_class<AudioEffectSpectrumAnalyzerInstance>(); + + ClassDB::register_class<AudioEffectCapture>(); } ClassDB::register_virtual_class<RenderingDevice>(); diff --git a/servers/rendering/renderer_canvas_cull.cpp b/servers/rendering/renderer_canvas_cull.cpp index 2d2847e6ca..e1ce52661c 100644 --- a/servers/rendering/renderer_canvas_cull.cpp +++ b/servers/rendering/renderer_canvas_cull.cpp @@ -721,8 +721,10 @@ void RendererCanvasCull::canvas_item_add_circle(RID p_item, const Point2 &p_pos, static const int circle_points = 64; points.resize(circle_points); + const real_t circle_point_step = Math_TAU / circle_points; + for (int i = 0; i < circle_points; i++) { - float angle = (i / float(circle_points)) * 2 * Math_PI; + float angle = i * circle_point_step; points.write[i].x = Math::cos(angle) * p_radius; points.write[i].y = Math::sin(angle) * p_radius; points.write[i] += p_pos; diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.cpp b/servers/rendering/renderer_rd/cluster_builder_rd.cpp new file mode 100644 index 0000000000..0fdd864d47 --- /dev/null +++ b/servers/rendering/renderer_rd/cluster_builder_rd.cpp @@ -0,0 +1,555 @@ +/*************************************************************************/ +/* cluster_builder_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "cluster_builder_rd.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/rendering_server_globals.h" + +ClusterBuilderSharedDataRD::ClusterBuilderSharedDataRD() { + RD::VertexFormatID vertex_format; + + { + Vector<RD::VertexAttribute> attributes; + { + RD::VertexAttribute va; + va.format = RD::DATA_FORMAT_R32G32B32_SFLOAT; + va.stride = sizeof(float) * 3; + attributes.push_back(va); + } + vertex_format = RD::get_singleton()->vertex_format_create(attributes); + } + + { + Vector<String> versions; + versions.push_back(""); + cluster_render.cluster_render_shader.initialize(versions); + cluster_render.shader_version = cluster_render.cluster_render_shader.version_create(); + cluster_render.shader = cluster_render.cluster_render_shader.version_get_shader(cluster_render.shader_version, 0); + cluster_render.shader_pipelines[ClusterRender::PIPELINE_NORMAL] = RD::get_singleton()->render_pipeline_create(cluster_render.shader, RD::get_singleton()->framebuffer_format_create_empty(), vertex_format, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState(), 0); + RD::PipelineMultisampleState ms; + ms.sample_count = RD::TEXTURE_SAMPLES_4; + cluster_render.shader_pipelines[ClusterRender::PIPELINE_MSAA] = RD::get_singleton()->render_pipeline_create(cluster_render.shader, RD::get_singleton()->framebuffer_format_create_empty(), vertex_format, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), ms, RD::PipelineDepthStencilState(), RD::PipelineColorBlendState(), 0); + } + { + Vector<String> versions; + versions.push_back(""); + cluster_store.cluster_store_shader.initialize(versions); + cluster_store.shader_version = cluster_store.cluster_store_shader.version_create(); + cluster_store.shader = cluster_store.cluster_store_shader.version_get_shader(cluster_store.shader_version, 0); + cluster_store.shader_pipeline = RD::get_singleton()->compute_pipeline_create(cluster_store.shader); + } + { + Vector<String> versions; + versions.push_back(""); + cluster_debug.cluster_debug_shader.initialize(versions); + cluster_debug.shader_version = cluster_debug.cluster_debug_shader.version_create(); + cluster_debug.shader = cluster_debug.cluster_debug_shader.version_get_shader(cluster_debug.shader_version, 0); + cluster_debug.shader_pipeline = RD::get_singleton()->compute_pipeline_create(cluster_debug.shader); + } + + { // SPHERE + static const uint32_t icosphere_vertex_count = 42; + static const float icosphere_vertices[icosphere_vertex_count * 3] = { + 0, 0, -1, 0.7236073, -0.5257253, -0.4472195, -0.276388, -0.8506492, -0.4472199, -0.8944262, 0, -0.4472156, -0.276388, 0.8506492, -0.4472199, 0.7236073, 0.5257253, -0.4472195, 0.276388, -0.8506492, 0.4472199, -0.7236073, -0.5257253, 0.4472195, -0.7236073, 0.5257253, 0.4472195, 0.276388, 0.8506492, 0.4472199, 0.8944262, 0, 0.4472156, 0, 0, 1, -0.1624555, -0.4999952, -0.8506544, 0.4253227, -0.3090114, -0.8506542, 0.2628688, -0.8090116, -0.5257377, 0.8506479, 0, -0.5257359, 0.4253227, 0.3090114, -0.8506542, -0.5257298, 0, -0.8506517, -0.6881894, -0.4999969, -0.5257362, -0.1624555, 0.4999952, -0.8506544, -0.6881894, 0.4999969, -0.5257362, 0.2628688, 0.8090116, -0.5257377, 0.9510579, -0.3090126, 0, 0.9510579, 0.3090126, 0, 0, -1, 0, 0.5877856, -0.8090167, 0, -0.9510579, -0.3090126, 0, -0.5877856, -0.8090167, 0, -0.5877856, 0.8090167, 0, -0.9510579, 0.3090126, 0, 0.5877856, 0.8090167, 0, 0, 1, 0, 0.6881894, -0.4999969, 0.5257362, -0.2628688, -0.8090116, 0.5257377, -0.8506479, 0, 0.5257359, -0.2628688, 0.8090116, 0.5257377, 0.6881894, 0.4999969, 0.5257362, 0.1624555, -0.4999952, 0.8506544, 0.5257298, 0, 0.8506517, -0.4253227, -0.3090114, 0.8506542, -0.4253227, 0.3090114, 0.8506542, 0.1624555, 0.4999952, 0.8506544 + }; + static const uint32_t icosphere_triangle_count = 80; + static const uint32_t icosphere_triangle_indices[icosphere_triangle_count * 3] = { + 0, 13, 12, 1, 13, 15, 0, 12, 17, 0, 17, 19, 0, 19, 16, 1, 15, 22, 2, 14, 24, 3, 18, 26, 4, 20, 28, 5, 21, 30, 1, 22, 25, 2, 24, 27, 3, 26, 29, 4, 28, 31, 5, 30, 23, 6, 32, 37, 7, 33, 39, 8, 34, 40, 9, 35, 41, 10, 36, 38, 38, 41, 11, 38, 36, 41, 36, 9, 41, 41, 40, 11, 41, 35, 40, 35, 8, 40, 40, 39, 11, 40, 34, 39, 34, 7, 39, 39, 37, 11, 39, 33, 37, 33, 6, 37, 37, 38, 11, 37, 32, 38, 32, 10, 38, 23, 36, 10, 23, 30, 36, 30, 9, 36, 31, 35, 9, 31, 28, 35, 28, 8, 35, 29, 34, 8, 29, 26, 34, 26, 7, 34, 27, 33, 7, 27, 24, 33, 24, 6, 33, 25, 32, 6, 25, 22, 32, 22, 10, 32, 30, 31, 9, 30, 21, 31, 21, 4, 31, 28, 29, 8, 28, 20, 29, 20, 3, 29, 26, 27, 7, 26, 18, 27, 18, 2, 27, 24, 25, 6, 24, 14, 25, 14, 1, 25, 22, 23, 10, 22, 15, 23, 15, 5, 23, 16, 21, 5, 16, 19, 21, 19, 4, 21, 19, 20, 4, 19, 17, 20, 17, 3, 20, 17, 18, 3, 17, 12, 18, 12, 2, 18, 15, 16, 5, 15, 13, 16, 13, 0, 16, 12, 14, 2, 12, 13, 14, 13, 1, 14 + }; + + Vector<uint8_t> vertex_data; + vertex_data.resize(sizeof(float) * icosphere_vertex_count * 3); + copymem(vertex_data.ptrw(), icosphere_vertices, vertex_data.size()); + + sphere_vertex_buffer = RD::get_singleton()->vertex_buffer_create(vertex_data.size(), vertex_data); + + Vector<uint8_t> index_data; + index_data.resize(sizeof(uint32_t) * icosphere_triangle_count * 3); + copymem(index_data.ptrw(), icosphere_triangle_indices, index_data.size()); + + sphere_index_buffer = RD::get_singleton()->index_buffer_create(icosphere_triangle_count * 3, RD::INDEX_BUFFER_FORMAT_UINT32, index_data); + + Vector<RID> buffers; + buffers.push_back(sphere_vertex_buffer); + + sphere_vertex_array = RD::get_singleton()->vertex_array_create(icosphere_vertex_count, vertex_format, buffers); + + sphere_index_array = RD::get_singleton()->index_array_create(sphere_index_buffer, 0, icosphere_triangle_count * 3); + + float min_d = 1e20; + for (uint32_t i = 0; i < icosphere_triangle_count; i++) { + Vector3 vertices[3]; + for (uint32_t j = 0; j < 3; j++) { + uint32_t index = icosphere_triangle_indices[i * 3 + j]; + for (uint32_t k = 0; k < 3; k++) { + vertices[j][k] = icosphere_vertices[index * 3 + k]; + } + } + Plane p(vertices[0], vertices[1], vertices[2]); + min_d = MIN(Math::abs(p.d), min_d); + } + sphere_overfit = 1.0 / min_d; + } + + { // CONE + static const uint32_t cone_vertex_count = 99; + static const float cone_vertices[cone_vertex_count * 3] = { + 0, 1, -1, 0.1950903, 0.9807853, -1, 0.3826835, 0.9238795, -1, 0.5555703, 0.8314696, -1, 0.7071068, 0.7071068, -1, 0.8314697, 0.5555702, -1, 0.9238795, 0.3826834, -1, 0.9807853, 0.1950903, -1, 1, 0, -1, 0.9807853, -0.1950902, -1, 0.9238796, -0.3826833, -1, 0.8314697, -0.5555702, -1, 0.7071068, -0.7071068, -1, 0.5555702, -0.8314697, -1, 0.3826833, -0.9238796, -1, 0.1950901, -0.9807853, -1, -3.25841e-7, -1, -1, -0.1950907, -0.9807852, -1, -0.3826839, -0.9238793, -1, -0.5555707, -0.8314693, -1, -0.7071073, -0.7071063, -1, -0.83147, -0.5555697, -1, -0.9238799, -0.3826827, -1, 0, 0, 0, -0.9807854, -0.1950894, -1, -1, 9.65599e-7, -1, -0.9807851, 0.1950913, -1, -0.9238791, 0.3826845, -1, -0.8314689, 0.5555713, -1, -0.7071059, 0.7071077, -1, -0.5555691, 0.8314704, -1, -0.3826821, 0.9238801, -1, -0.1950888, 0.9807856, -1 + }; + static const uint32_t cone_triangle_count = 62; + static const uint32_t cone_triangle_indices[cone_triangle_count * 3] = { + 0, 23, 1, 1, 23, 2, 2, 23, 3, 3, 23, 4, 4, 23, 5, 5, 23, 6, 6, 23, 7, 7, 23, 8, 8, 23, 9, 9, 23, 10, 10, 23, 11, 11, 23, 12, 12, 23, 13, 13, 23, 14, 14, 23, 15, 15, 23, 16, 16, 23, 17, 17, 23, 18, 18, 23, 19, 19, 23, 20, 20, 23, 21, 21, 23, 22, 22, 23, 24, 24, 23, 25, 25, 23, 26, 26, 23, 27, 27, 23, 28, 28, 23, 29, 29, 23, 30, 30, 23, 31, 31, 23, 32, 32, 23, 0, 7, 15, 24, 32, 0, 1, 1, 2, 3, 3, 4, 5, 5, 6, 3, 6, 7, 3, 7, 8, 9, 9, 10, 7, 10, 11, 7, 11, 12, 15, 12, 13, 15, 13, 14, 15, 15, 16, 17, 17, 18, 19, 19, 20, 24, 20, 21, 24, 21, 22, 24, 24, 25, 26, 26, 27, 28, 28, 29, 30, 30, 31, 32, 32, 1, 3, 15, 17, 24, 17, 19, 24, 24, 26, 32, 26, 28, 32, 28, 30, 32, 32, 3, 7, 7, 11, 15, 32, 7, 24 + }; + + Vector<uint8_t> vertex_data; + vertex_data.resize(sizeof(float) * cone_vertex_count * 3); + copymem(vertex_data.ptrw(), cone_vertices, vertex_data.size()); + + cone_vertex_buffer = RD::get_singleton()->vertex_buffer_create(vertex_data.size(), vertex_data); + + Vector<uint8_t> index_data; + index_data.resize(sizeof(uint32_t) * cone_triangle_count * 3); + copymem(index_data.ptrw(), cone_triangle_indices, index_data.size()); + + cone_index_buffer = RD::get_singleton()->index_buffer_create(cone_triangle_count * 3, RD::INDEX_BUFFER_FORMAT_UINT32, index_data); + + Vector<RID> buffers; + buffers.push_back(cone_vertex_buffer); + + cone_vertex_array = RD::get_singleton()->vertex_array_create(cone_vertex_count, vertex_format, buffers); + + cone_index_array = RD::get_singleton()->index_array_create(cone_index_buffer, 0, cone_triangle_count * 3); + + float min_d = 1e20; + for (uint32_t i = 0; i < cone_triangle_count; i++) { + Vector3 vertices[3]; + int32_t zero_index = -1; + for (uint32_t j = 0; j < 3; j++) { + uint32_t index = cone_triangle_indices[i * 3 + j]; + for (uint32_t k = 0; k < 3; k++) { + vertices[j][k] = cone_vertices[index * 3 + k]; + } + if (vertices[j] == Vector3()) { + zero_index = j; + } + } + + if (zero_index != -1) { + Vector3 a = vertices[(zero_index + 1) % 3]; + Vector3 b = vertices[(zero_index + 2) % 3]; + Vector3 c = a + Vector3(0, 0, 1); + Plane p(a, b, c); + min_d = MIN(Math::abs(p.d), min_d); + } + } + cone_overfit = 1.0 / min_d; + } + + { // BOX + static const uint32_t box_vertex_count = 8; + static const float box_vertices[box_vertex_count * 3] = { + -1, -1, -1, -1, -1, 1, -1, 1, -1, -1, 1, 1, 1, -1, -1, 1, -1, 1, 1, 1, -1, 1, 1, 1 + }; + static const uint32_t box_triangle_count = 12; + static const uint32_t box_triangle_indices[box_triangle_count * 3] = { + 1, 2, 0, 3, 6, 2, 7, 4, 6, 5, 0, 4, 6, 0, 2, 3, 5, 7, 1, 3, 2, 3, 7, 6, 7, 5, 4, 5, 1, 0, 6, 4, 0, 3, 1, 5 + }; + + Vector<uint8_t> vertex_data; + vertex_data.resize(sizeof(float) * box_vertex_count * 3); + copymem(vertex_data.ptrw(), box_vertices, vertex_data.size()); + + box_vertex_buffer = RD::get_singleton()->vertex_buffer_create(vertex_data.size(), vertex_data); + + Vector<uint8_t> index_data; + index_data.resize(sizeof(uint32_t) * box_triangle_count * 3); + copymem(index_data.ptrw(), box_triangle_indices, index_data.size()); + + box_index_buffer = RD::get_singleton()->index_buffer_create(box_triangle_count * 3, RD::INDEX_BUFFER_FORMAT_UINT32, index_data); + + Vector<RID> buffers; + buffers.push_back(box_vertex_buffer); + + box_vertex_array = RD::get_singleton()->vertex_array_create(box_vertex_count, vertex_format, buffers); + + box_index_array = RD::get_singleton()->index_array_create(box_index_buffer, 0, box_triangle_count * 3); + } +} +ClusterBuilderSharedDataRD::~ClusterBuilderSharedDataRD() { + RD::get_singleton()->free(sphere_vertex_buffer); + RD::get_singleton()->free(sphere_index_buffer); + RD::get_singleton()->free(cone_vertex_buffer); + RD::get_singleton()->free(cone_index_buffer); + RD::get_singleton()->free(box_vertex_buffer); + RD::get_singleton()->free(box_index_buffer); + + cluster_render.cluster_render_shader.version_free(cluster_render.shader_version); + cluster_store.cluster_store_shader.version_free(cluster_store.shader_version); + cluster_debug.cluster_debug_shader.version_free(cluster_debug.shader_version); +} + +///////////////////////////// + +void ClusterBuilderRD::_clear() { + if (cluster_buffer.is_null()) { + return; //nothing to clear + } + RD::get_singleton()->free(cluster_buffer); + RD::get_singleton()->free(cluster_render_buffer); + RD::get_singleton()->free(element_buffer); + cluster_buffer = RID(); + cluster_render_buffer = RID(); + element_buffer = RID(); + + memfree(render_elements); + + render_elements = nullptr; + render_element_max = 0; + render_element_count = 0; + + RD::get_singleton()->free(framebuffer); + framebuffer = RID(); + + cluster_render_uniform_set = RID(); + cluster_store_uniform_set = RID(); +} + +void ClusterBuilderRD::setup(Size2i p_screen_size, uint32_t p_max_elements, RID p_depth_buffer, RID p_depth_buffer_sampler, RID p_color_buffer) { + ERR_FAIL_COND(p_max_elements == 0); + ERR_FAIL_COND(p_screen_size.x < 1); + ERR_FAIL_COND(p_screen_size.y < 1); + + _clear(); + + screen_size = p_screen_size; + + cluster_screen_size.width = (p_screen_size.width - 1) / cluster_size + 1; + cluster_screen_size.height = (p_screen_size.height - 1) / cluster_size + 1; + + max_elements_by_type = p_max_elements; + if (max_elements_by_type % 32) { //need to be 32 aligned + max_elements_by_type += 32 - (max_elements_by_type % 32); + } + + cluster_buffer_size = cluster_screen_size.x * cluster_screen_size.y * (max_elements_by_type / 32 + 32) * ELEMENT_TYPE_MAX * 4; + + render_element_max = max_elements_by_type * ELEMENT_TYPE_MAX; + + uint32_t element_tag_bits_size = render_element_max / 32; + uint32_t element_tag_depth_bits_size = render_element_max; + cluster_render_buffer_size = cluster_screen_size.x * cluster_screen_size.y * (element_tag_bits_size + element_tag_depth_bits_size) * 4; // tag bits (element was used) and tag depth (depth range in which it was used) + + cluster_render_buffer = RD::get_singleton()->storage_buffer_create(cluster_render_buffer_size); + cluster_buffer = RD::get_singleton()->storage_buffer_create(cluster_buffer_size); + + render_elements = (RenderElementData *)memalloc(sizeof(RenderElementData *) * render_element_max); + render_element_count = 0; + + element_buffer = RD::get_singleton()->storage_buffer_create(sizeof(RenderElementData) * render_element_max); + + uint32_t div_value = 1 << divisor; + if (use_msaa) { + framebuffer = RD::get_singleton()->framebuffer_create_empty(p_screen_size / div_value, RD::TEXTURE_SAMPLES_4); + } else { + framebuffer = RD::get_singleton()->framebuffer_create_empty(p_screen_size / div_value); + } + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 1; + u.ids.push_back(state_uniform); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.ids.push_back(element_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 3; + u.ids.push_back(cluster_render_buffer); + uniforms.push_back(u); + } + + cluster_render_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shared->cluster_render.shader, 0); + } + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.ids.push_back(cluster_render_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.ids.push_back(cluster_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 3; + u.ids.push_back(element_buffer); + uniforms.push_back(u); + } + + cluster_store_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shared->cluster_store.shader, 0); + } + + if (p_color_buffer.is_valid()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.ids.push_back(cluster_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.ids.push_back(p_color_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 3; + u.ids.push_back(p_depth_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 4; + u.ids.push_back(p_depth_buffer_sampler); + uniforms.push_back(u); + } + + debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shared->cluster_debug.shader, 0); + } else { + debug_uniform_set = RID(); + } +} + +void ClusterBuilderRD::begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection, bool p_flip_y) { + view_xform = p_view_transform.affine_inverse(); + projection = p_cam_projection; + z_near = projection.get_z_near(); + z_far = projection.get_z_far(); + orthogonal = p_cam_projection.is_orthogonal(); + adjusted_projection = projection; + if (!orthogonal) { + adjusted_projection.adjust_perspective_znear(0.0001); + } + + CameraMatrix correction; + correction.set_depth_correction(p_flip_y); + projection = correction * projection; + adjusted_projection = correction * adjusted_projection; + + //reset counts + render_element_count = 0; + for (uint32_t i = 0; i < ELEMENT_TYPE_MAX; i++) { + cluster_count_by_type[i] = 0; + } +} + +void ClusterBuilderRD::bake_cluster() { + RENDER_TIMESTAMP(">Bake Cluster"); + + RD::get_singleton()->draw_command_begin_label("Bake Light Cluster"); + + //clear cluster buffer + RD::get_singleton()->buffer_clear(cluster_buffer, 0, cluster_buffer_size, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + + if (render_element_count > 0) { + //clear render buffer + RD::get_singleton()->buffer_clear(cluster_render_buffer, 0, cluster_render_buffer_size, RD::BARRIER_MASK_RASTER); + + { //fill state uniform + + StateUniform state; + + RendererStorageRD::store_camera(adjusted_projection, state.projection); + state.inv_z_far = 1.0 / z_far; + state.screen_to_clusters_shift = get_shift_from_power_of_2(cluster_size); + state.screen_to_clusters_shift -= divisor; //screen is smaller, shift one less + + state.cluster_screen_width = cluster_screen_size.x; + state.cluster_depth_offset = (render_element_max / 32); + state.cluster_data_size = state.cluster_depth_offset + render_element_max; + + RD::get_singleton()->buffer_update(state_uniform, 0, sizeof(StateUniform), &state, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + + //update instances + + RD::get_singleton()->buffer_update(element_buffer, 0, sizeof(RenderElementData) * render_element_count, render_elements, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + + RENDER_TIMESTAMP("Render Elements"); + + //render elements + { + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD); + ClusterBuilderSharedDataRD::ClusterRender::PushConstant push_constant = {}; + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, shared->cluster_render.shader_pipelines[use_msaa ? ClusterBuilderSharedDataRD::ClusterRender::PIPELINE_MSAA : ClusterBuilderSharedDataRD::ClusterRender::PIPELINE_NORMAL]); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, cluster_render_uniform_set, 0); + + for (uint32_t i = 0; i < render_element_count;) { + push_constant.base_index = i; + switch (render_elements[i].type) { + case ELEMENT_TYPE_OMNI_LIGHT: { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, shared->sphere_vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->sphere_index_array); + } break; + case ELEMENT_TYPE_SPOT_LIGHT: { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, shared->cone_vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->cone_index_array); + } break; + case ELEMENT_TYPE_DECAL: + case ELEMENT_TYPE_REFLECTION_PROBE: { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, shared->box_vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->box_index_array); + } break; + } + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(ClusterBuilderSharedDataRD::ClusterRender::PushConstant)); + + uint32_t instances = 1; +#if 0 + for (uint32_t j = i+1; j < element_count; j++) { + if (elements[i].type!=elements[j].type) { + break; + } + instances++; + } +#endif + RD::get_singleton()->draw_list_draw(draw_list, true, instances); + i += instances; + } + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_COMPUTE); + } + //store elements + RENDER_TIMESTAMP("Pack Elements"); + + { + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shared->cluster_store.shader_pipeline); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cluster_store_uniform_set, 0); + + ClusterBuilderSharedDataRD::ClusterStore::PushConstant push_constant; + push_constant.cluster_render_data_size = render_element_max / 32 + render_element_max; + push_constant.max_render_element_count_div_32 = render_element_max / 32; + push_constant.cluster_screen_size[0] = cluster_screen_size.x; + push_constant.cluster_screen_size[1] = cluster_screen_size.y; + push_constant.render_element_count_div_32 = render_element_count > 0 ? (render_element_count - 1) / 32 + 1 : 0; + push_constant.max_cluster_element_count_div_32 = max_elements_by_type / 32; + push_constant.pad1 = 0; + push_constant.pad2 = 0; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ClusterBuilderSharedDataRD::ClusterStore::PushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cluster_screen_size.x, cluster_screen_size.y, 1); + + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + } else { + RD::get_singleton()->barrier(RD::BARRIER_MASK_TRANSFER, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + RENDER_TIMESTAMP("<Bake Cluster"); + RD::get_singleton()->draw_command_end_label(); +} + +void ClusterBuilderRD::debug(ElementType p_element) { + ERR_FAIL_COND(debug_uniform_set.is_null()); + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shared->cluster_debug.shader_pipeline); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, debug_uniform_set, 0); + + ClusterBuilderSharedDataRD::ClusterDebug::PushConstant push_constant; + push_constant.screen_size[0] = screen_size.x; + push_constant.screen_size[1] = screen_size.y; + push_constant.cluster_screen_size[0] = cluster_screen_size.x; + push_constant.cluster_screen_size[1] = cluster_screen_size.y; + push_constant.cluster_shift = get_shift_from_power_of_2(cluster_size); + push_constant.cluster_type = p_element; + push_constant.orthogonal = orthogonal; + push_constant.z_far = z_far; + push_constant.z_near = z_near; + push_constant.max_cluster_element_count_div_32 = max_elements_by_type / 32; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ClusterBuilderSharedDataRD::ClusterDebug::PushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, screen_size.x, screen_size.y, 1); + + RD::get_singleton()->compute_list_end(); +} + +RID ClusterBuilderRD::get_cluster_buffer() const { + return cluster_buffer; +} + +uint32_t ClusterBuilderRD::get_cluster_size() const { + return cluster_size; +} + +uint32_t ClusterBuilderRD::get_max_cluster_elements() const { + return max_elements_by_type; +} + +void ClusterBuilderRD::set_shared(ClusterBuilderSharedDataRD *p_shared) { + shared = p_shared; +} + +ClusterBuilderRD::ClusterBuilderRD() { + state_uniform = RD::get_singleton()->uniform_buffer_create(sizeof(StateUniform)); +} + +ClusterBuilderRD::~ClusterBuilderRD() { + _clear(); + RD::get_singleton()->free(state_uniform); +} diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.h b/servers/rendering/renderer_rd/cluster_builder_rd.h new file mode 100644 index 0000000000..dc1707b534 --- /dev/null +++ b/servers/rendering/renderer_rd/cluster_builder_rd.h @@ -0,0 +1,378 @@ +/*************************************************************************/ +/* cluster_builder_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef CLUSTER_BUILDER_RD_H +#define CLUSTER_BUILDER_RD_H + +#include "servers/rendering/renderer_rd/renderer_storage_rd.h" +#include "servers/rendering/renderer_rd/shaders/cluster_debug.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cluster_render.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cluster_store.glsl.gen.h" + +class ClusterBuilderSharedDataRD { + friend class ClusterBuilderRD; + + RID sphere_vertex_buffer; + RID sphere_vertex_array; + RID sphere_index_buffer; + RID sphere_index_array; + float sphere_overfit = 0.0; //because an icosphere is not a perfect sphere, we need to enlarge it to cover the sphere area + + RID cone_vertex_buffer; + RID cone_vertex_array; + RID cone_index_buffer; + RID cone_index_array; + float cone_overfit = 0.0; //because an cone mesh is not a perfect sphere, we need to enlarge it to cover the actual cone area + + RID box_vertex_buffer; + RID box_vertex_array; + RID box_index_buffer; + RID box_index_array; + + enum Divisor { + DIVISOR_1, + DIVISOR_2, + DIVISOR_4, + }; + + struct ClusterRender { + struct PushConstant { + uint32_t base_index; + uint32_t pad0; + uint32_t pad1; + uint32_t pad2; + }; + + ClusterRenderShaderRD cluster_render_shader; + RID shader_version; + RID shader; + enum PipelineVersion { + PIPELINE_NORMAL, + PIPELINE_MSAA, + PIPELINE_MAX + }; + + RID shader_pipelines[PIPELINE_MAX]; + } cluster_render; + + struct ClusterStore { + struct PushConstant { + uint32_t cluster_render_data_size; // how much data for a single cluster takes + uint32_t max_render_element_count_div_32; //divided by 32 + uint32_t cluster_screen_size[2]; + uint32_t render_element_count_div_32; //divided by 32 + uint32_t max_cluster_element_count_div_32; //divided by 32 + uint32_t pad1; + uint32_t pad2; + }; + + ClusterStoreShaderRD cluster_store_shader; + RID shader_version; + RID shader; + RID shader_pipeline; + } cluster_store; + + struct ClusterDebug { + struct PushConstant { + uint32_t screen_size[2]; + uint32_t cluster_screen_size[2]; + + uint32_t cluster_shift; + uint32_t cluster_type; + float z_near; + float z_far; + + uint32_t orthogonal; + uint32_t max_cluster_element_count_div_32; + uint32_t pad1; + uint32_t pad2; + }; + + ClusterDebugShaderRD cluster_debug_shader; + RID shader_version; + RID shader; + RID shader_pipeline; + } cluster_debug; + +public: + ClusterBuilderSharedDataRD(); + ~ClusterBuilderSharedDataRD(); +}; + +class ClusterBuilderRD { +public: + enum LightType { + LIGHT_TYPE_OMNI, + LIGHT_TYPE_SPOT + }; + + enum BoxType { + BOX_TYPE_REFLECTION_PROBE, + BOX_TYPE_DECAL, + }; + + enum ElementType { + ELEMENT_TYPE_OMNI_LIGHT, + ELEMENT_TYPE_SPOT_LIGHT, + ELEMENT_TYPE_DECAL, + ELEMENT_TYPE_REFLECTION_PROBE, + ELEMENT_TYPE_MAX, + + }; + +private: + ClusterBuilderSharedDataRD *shared = nullptr; + + struct RenderElementData { + uint32_t type; //0-4 + uint32_t touches_near; + uint32_t touches_far; + uint32_t original_index; + float transform_inv[12]; //transposed transform for less space + float scale[3]; + uint32_t pad; + }; + + uint32_t cluster_count_by_type[ELEMENT_TYPE_MAX] = {}; + uint32_t max_elements_by_type = 0; + + RenderElementData *render_elements = nullptr; + uint32_t render_element_count = 0; + uint32_t render_element_max = 0; + + Transform view_xform; + CameraMatrix adjusted_projection; + CameraMatrix projection; + float z_far = 0; + float z_near = 0; + bool orthogonal = false; + + enum Divisor { + DIVISOR_1, + DIVISOR_2, + DIVISOR_4, + }; + + uint32_t cluster_size = 32; + bool use_msaa = true; + Divisor divisor = DIVISOR_4; + + Size2i screen_size; + Size2i cluster_screen_size; + + RID framebuffer; + RID cluster_render_buffer; //used for creating + RID cluster_buffer; //used for rendering + RID element_buffer; //used for storing, to hint element touches far plane or near plane + uint32_t cluster_render_buffer_size = 0; + uint32_t cluster_buffer_size = 0; + + RID cluster_render_uniform_set; + RID cluster_store_uniform_set; + + //persistent data + + void _clear(); + + struct StateUniform { + float projection[16]; + float inv_z_far; + uint32_t screen_to_clusters_shift; // shift to obtain coordinates in block indices + uint32_t cluster_screen_width; // + uint32_t cluster_data_size; // how much data for a single cluster takes + uint32_t cluster_depth_offset; + uint32_t pad0; + uint32_t pad1; + uint32_t pad2; + }; + + RID state_uniform; + + RID debug_uniform_set; + +public: + void setup(Size2i p_screen_size, uint32_t p_max_elements, RID p_depth_buffer, RID p_depth_buffer_sampler, RID p_color_buffer); + + void begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection, bool p_flip_y); + + _FORCE_INLINE_ void add_light(LightType p_type, const Transform &p_transform, float p_radius, float p_spot_aperture) { + if (p_type == LIGHT_TYPE_OMNI && cluster_count_by_type[ELEMENT_TYPE_OMNI_LIGHT] == max_elements_by_type) { + return; //max number elements reached + } + if (p_type == LIGHT_TYPE_SPOT && cluster_count_by_type[ELEMENT_TYPE_SPOT_LIGHT] == max_elements_by_type) { + return; //max number elements reached + } + + RenderElementData &e = render_elements[render_element_count]; + + Transform xform = view_xform * p_transform; + + float radius = xform.basis.get_uniform_scale(); + if (radius > 0.98 || radius < 1.02) { + xform.basis.orthonormalize(); + } + + radius *= p_radius; + + if (p_type == LIGHT_TYPE_OMNI) { + radius *= shared->sphere_overfit; // overfit icosphere + + //omni + float depth = -xform.origin.z; + if (orthogonal) { + e.touches_near = (depth - radius) < z_near; + } else { + //contains camera inside light + float radius2 = radius * shared->sphere_overfit; // overfit again for outer size (camera may be outside actual sphere but behind an icosphere vertex) + e.touches_near = xform.origin.length_squared() < radius2 * radius2; + } + + e.touches_far = (depth + radius) > z_far; + e.scale[0] = radius; + e.scale[1] = radius; + e.scale[2] = radius; + e.type = ELEMENT_TYPE_OMNI_LIGHT; + e.original_index = cluster_count_by_type[ELEMENT_TYPE_OMNI_LIGHT]; + + RendererStorageRD::store_transform_transposed_3x4(xform, e.transform_inv); + + cluster_count_by_type[ELEMENT_TYPE_OMNI_LIGHT]++; + + } else { + //spot + radius *= shared->cone_overfit; // overfit icosphere + + real_t len = Math::tan(Math::deg2rad(p_spot_aperture)) * radius; + //approximate, probably better to use a cone support function + float max_d = -1e20; + float min_d = 1e20; +#define CONE_MINMAX(m_x, m_y) \ + { \ + float d = -xform.xform(Vector3(len * m_x, len * m_y, -radius)).z; \ + min_d = MIN(d, min_d); \ + max_d = MAX(d, max_d); \ + } + + CONE_MINMAX(1, 1); + CONE_MINMAX(-1, 1); + CONE_MINMAX(-1, -1); + CONE_MINMAX(1, -1); + + if (orthogonal) { + e.touches_near = min_d < z_near; + } else { + //contains camera inside light + Plane base_plane(xform.origin, -xform.basis.get_axis(Vector3::AXIS_Z)); + float dist = base_plane.distance_to(Vector3()); + if (dist >= 0 && dist < radius) { + //inside, check angle + float angle = Math::rad2deg(Math::acos((-xform.origin.normalized()).dot(-xform.basis.get_axis(Vector3::AXIS_Z)))); + e.touches_near = angle < p_spot_aperture * 1.05; //overfit aperture a little due to cone overfit + } else { + e.touches_near = false; + } + } + + e.touches_far = max_d > z_far; + + e.scale[0] = len * shared->cone_overfit; + e.scale[1] = len * shared->cone_overfit; + e.scale[2] = radius; + + e.type = ELEMENT_TYPE_SPOT_LIGHT; + e.original_index = cluster_count_by_type[ELEMENT_TYPE_SPOT_LIGHT]; //use omni since they share index + + RendererStorageRD::store_transform_transposed_3x4(xform, e.transform_inv); + + cluster_count_by_type[ELEMENT_TYPE_SPOT_LIGHT]++; + } + + render_element_count++; + } + + _FORCE_INLINE_ void add_box(BoxType p_box_type, const Transform &p_transform, const Vector3 &p_half_extents) { + if (p_box_type == BOX_TYPE_DECAL && cluster_count_by_type[ELEMENT_TYPE_DECAL] == max_elements_by_type) { + return; //max number elements reached + } + if (p_box_type == BOX_TYPE_REFLECTION_PROBE && cluster_count_by_type[ELEMENT_TYPE_REFLECTION_PROBE] == max_elements_by_type) { + return; //max number elements reached + } + + RenderElementData &e = render_elements[render_element_count]; + Transform xform = view_xform * p_transform; + + //extract scale and scale the matrix by it, makes things simpler + Vector3 scale = p_half_extents; + for (uint32_t i = 0; i < 3; i++) { + float s = xform.basis.elements[i].length(); + scale[i] *= s; + xform.basis.elements[i] /= s; + }; + + float box_depth = Math::abs(xform.basis.xform_inv(Vector3(0, 0, -1)).dot(scale)); + float depth = -xform.origin.z; + + if (orthogonal) { + e.touches_near = depth - box_depth < z_near; + } else { + //contains camera inside box + Vector3 inside = xform.xform_inv(Vector3(0, 0, 0)).abs(); + e.touches_near = inside.x < scale.x && inside.y < scale.y && inside.z < scale.z; + } + + e.touches_far = depth + box_depth > z_far; + + e.scale[0] = scale.x; + e.scale[1] = scale.y; + e.scale[2] = scale.z; + + e.type = (p_box_type == BOX_TYPE_DECAL) ? ELEMENT_TYPE_DECAL : ELEMENT_TYPE_REFLECTION_PROBE; + e.original_index = cluster_count_by_type[e.type]; + + RendererStorageRD::store_transform_transposed_3x4(xform, e.transform_inv); + + cluster_count_by_type[e.type]++; + render_element_count++; + } + + void bake_cluster(); + void debug(ElementType p_element); + + RID get_cluster_buffer() const; + uint32_t get_cluster_size() const; + uint32_t get_max_cluster_elements() const; + + void set_shared(ClusterBuilderSharedDataRD *p_shared); + + ClusterBuilderRD(); + ~ClusterBuilderRD(); +}; + +#endif // CLUSTER_BUILDER_H diff --git a/servers/rendering/renderer_rd/effects_rd.cpp b/servers/rendering/renderer_rd/effects_rd.cpp index a73eb3782c..a9cadb40df 100644 --- a/servers/rendering/renderer_rd/effects_rd.cpp +++ b/servers/rendering/renderer_rd/effects_rd.cpp @@ -299,15 +299,12 @@ void EffectsRD::copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const copy.push_constant.target[0] = p_rect.position.x; copy.push_constant.target[1] = p_rect.position.y; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_8_bit_dst ? COPY_MODE_SIMPLY_COPY_8BIT : COPY_MODE_SIMPLY_COPY]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -322,15 +319,12 @@ void EffectsRD::copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, copy.push_constant.target[1] = 0; copy.push_constant.camera_z_far = p_lod; - int32_t x_groups = (p_panorama_size.width - 1) / 8 + 1; - int32_t y_groups = (p_panorama_size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_is_array ? COPY_MODE_CUBE_ARRAY_TO_PANORAMA : COPY_MODE_CUBE_TO_PANORAMA]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_cube), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_panorama), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_panorama_size.width, p_panorama_size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -349,15 +343,12 @@ void EffectsRD::copy_depth_to_rect_and_linearize(RID p_source_rd_texture, RID p_ copy.push_constant.camera_z_far = p_z_far; copy.push_constant.camera_z_near = p_z_near; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[COPY_MODE_LINEARIZE_DEPTH]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -374,15 +365,12 @@ void EffectsRD::copy_depth_to_rect(RID p_source_rd_texture, RID p_dest_texture, copy.push_constant.target[0] = p_rect.position.x; copy.push_constant.target[1] = p_rect.position.y; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[COPY_MODE_SIMPLY_COPY_DEPTH]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -400,14 +388,11 @@ void EffectsRD::set_color(RID p_dest_texture, const Color &p_color, const Rect2i copy.push_constant.set_color[2] = p_color.b; copy.push_constant.set_color[3] = p_color.a; - int32_t x_groups = (p_region.size.width - 1) / 8 + 1; - int32_t y_groups = (p_region.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_8bit_dst ? COPY_MODE_SET_COLOR_8BIT : COPY_MODE_SET_COLOR]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -420,8 +405,6 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back copy.push_constant.section[2] = p_region.size.width; copy.push_constant.section[3] = p_region.size.height; - int32_t x_groups = (p_region.size.width - 1) / 8 + 1; - int32_t y_groups = (p_region.size.height - 1) / 8 + 1; //HORIZONTAL RD::DrawListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_8bit_dst ? COPY_MODE_GAUSSIAN_COPY_8BIT : COPY_MODE_GAUSSIAN_COPY]); @@ -431,7 +414,7 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back copy.push_constant.flags = base_flags | COPY_FLAG_HORIZONTAL; RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -442,7 +425,7 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back copy.push_constant.flags = base_flags; RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -452,9 +435,6 @@ void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const CopyMode copy_mode = p_first_pass && p_auto_exposure.is_valid() ? COPY_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE : COPY_MODE_GAUSSIAN_GLOW; uint32_t base_flags = 0; - int32_t x_groups = (p_size.width + 7) / 8; - int32_t y_groups = (p_size.height + 7) / 8; - copy.push_constant.section[2] = p_size.x; copy.push_constant.section[3] = p_size.y; @@ -479,16 +459,13 @@ void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const copy.push_constant.flags = base_flags | (p_first_pass ? COPY_FLAG_GLOW_FIRST_PASS : 0) | (p_high_quality ? COPY_FLAG_HIGH_QUALITY_GLOW : 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.width, p_size.height, 1); RD::get_singleton()->compute_list_end(); } void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, RenderingServer::EnvironmentSSRRoughnessQuality p_roughness_quality, RID p_blur_radius, RID p_blur_radius2, RID p_metallic, const Color &p_metallic_mask, RID p_depth, RID p_scale_depth, RID p_scale_normal, RID p_output, RID p_output_blur, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const CameraMatrix &p_camera) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - int32_t x_groups = (p_screen_size.width - 1) / 8 + 1; - int32_t y_groups = (p_screen_size.height - 1) / 8 + 1; - { //scale color and depth to half ssr_scale.push_constant.camera_z_far = p_camera.get_z_far(); ssr_scale.push_constant.camera_z_near = p_camera.get_z_near(); @@ -506,7 +483,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssr_scale.push_constant, sizeof(ScreenSpaceReflectionScalePushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); } @@ -547,7 +524,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R } RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_scale_normal), 2); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); } if (p_roughness_quality != RS::ENV_SSR_ROUGNESS_QUALITY_DISABLED) { @@ -585,7 +562,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssr_filter.push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -600,7 +577,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssr_filter.push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); } RD::get_singleton()->compute_list_end(); @@ -609,9 +586,6 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R void EffectsRD::sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_depth, const CameraMatrix &p_camera, const Size2i &p_screen_size, float p_scale, float p_depth_scale, RenderingServer::SubSurfaceScatteringQuality p_quality) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - int32_t x_groups = (p_screen_size.width - 1) / 8 + 1; - int32_t y_groups = (p_screen_size.height - 1) / 8 + 1; - Plane p = p_camera.xform4(Plane(1, 0, -1, 1)); p.normal /= p.d; float unit_size = p.normal.x; @@ -635,7 +609,7 @@ void EffectsRD::sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_dept RD::get_singleton()->compute_list_set_push_constant(compute_list, &sss.push_constant, sizeof(SubSurfaceScatteringPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -646,7 +620,7 @@ void EffectsRD::sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_dept sss.push_constant.vertical = true; RD::get_singleton()->compute_list_set_push_constant(compute_list, &sss.push_constant, sizeof(SubSurfaceScatteringPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -690,39 +664,33 @@ void EffectsRD::make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const S copy.push_constant.section[2] = p_size.width; copy.push_constant.section[3] = p_size.height; - int32_t x_groups = (p_size.width - 1) / 8 + 1; - int32_t y_groups = (p_size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[COPY_MODE_MIPMAP]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.width, p_size.height, 1); RD::get_singleton()->compute_list_end(); } -void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, float p_z_near, float p_z_far, float p_bias, bool p_dp_flip) { +void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip) { CopyToDPPushConstant push_constant; - push_constant.screen_size[0] = p_rect.size.x; - push_constant.screen_size[1] = p_rect.size.y; - push_constant.dest_offset[0] = p_rect.position.x; - push_constant.dest_offset[1] = p_rect.position.y; - push_constant.bias = p_bias; + push_constant.screen_rect[0] = p_rect.position.x; + push_constant.screen_rect[1] = p_rect.position.y; + push_constant.screen_rect[2] = p_rect.size.width; + push_constant.screen_rect[3] = p_rect.size.height; push_constant.z_far = p_z_far; push_constant.z_near = p_z_near; push_constant.z_flip = p_dp_flip; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dst_framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, cube_to_dp.pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, cube_to_dp.pipeline); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 1); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(CopyToDPPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(CopyToDPPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_TRANSFER); } void EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings) { @@ -807,10 +775,7 @@ void EffectsRD::luminance_reduction(RID p_source_texture, const Size2i p_source_ RD::get_singleton()->compute_list_set_push_constant(compute_list, &luminance_reduce.push_constant, sizeof(LuminanceReducePushConstant)); - int32_t x_groups = (luminance_reduce.push_constant.source_size[0] - 1) / 8 + 1; - int32_t y_groups = (luminance_reduce.push_constant.source_size[1] - 1) / 8 + 1; - - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, luminance_reduce.push_constant.source_size[0], luminance_reduce.push_constant.source_size[1], 1); luminance_reduce.push_constant.source_size[0] = MAX(luminance_reduce.push_constant.source_size[0] / 8, 1); luminance_reduce.push_constant.source_size[1] = MAX(luminance_reduce.push_constant.source_size[1] / 8, 1); @@ -851,14 +816,12 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_texture), 1); - int32_t x_groups = (p_base_texture_size.x - 1) / 8 + 1; - int32_t y_groups = (p_base_texture_size.y - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x; bokeh.push_constant.size[1] = p_base_texture_size.y; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); if (p_bokeh_shape == RS::DOF_BOKEH_BOX || p_bokeh_shape == RS::DOF_BOKEH_HEXAGON) { @@ -875,8 +838,6 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); - x_groups = ((p_base_texture_size.x >> 1) - 1) / 8 + 1; - y_groups = ((p_base_texture_size.y >> 1) - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; bokeh.push_constant.size[1] = p_base_texture_size.y >> 1; bokeh.push_constant.half_size = true; @@ -890,7 +851,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, bokeh.push_constant.size[0], bokeh.push_constant.size[1], 1); RD::get_singleton()->compute_list_add_barrier(compute_list); //third pass @@ -906,7 +867,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, bokeh.push_constant.size[0], bokeh.push_constant.size[1], 1); RD::get_singleton()->compute_list_add_barrier(compute_list); if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { @@ -917,8 +878,6 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture2), 1); - x_groups = (p_base_texture_size.x - 1) / 8 + 1; - y_groups = (p_base_texture_size.y - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x; bokeh.push_constant.size[1] = p_base_texture_size.y; bokeh.push_constant.half_size = false; @@ -926,7 +885,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); } } else { //circle @@ -944,15 +903,13 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); - x_groups = ((p_base_texture_size.x >> 1) - 1) / 8 + 1; - y_groups = ((p_base_texture_size.y >> 1) - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; bokeh.push_constant.size[1] = p_base_texture_size.y >> 1; bokeh.push_constant.half_size = true; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, bokeh.push_constant.size[0], bokeh.push_constant.size[1], 1); RD::get_singleton()->compute_list_add_barrier(compute_list); //circle is just one pass, then upscale @@ -964,8 +921,6 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture1), 1); - x_groups = (p_base_texture_size.x - 1) / 8 + 1; - y_groups = (p_base_texture_size.y - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x; bokeh.push_constant.size[1] = p_base_texture_size.y; bokeh.push_constant.half_size = false; @@ -973,7 +928,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); } RD::get_singleton()->compute_list_end(); @@ -992,47 +947,47 @@ void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> ssao.gather_push_constant.pass_coord_offset[0] = i % 2; ssao.gather_push_constant.pass_coord_offset[1] = i / 2; - ssao.gather_push_constant.pass_uv_offset[0] = ((i % 2) - 0.0) / p_settings.screen_size.x; - ssao.gather_push_constant.pass_uv_offset[1] = ((i / 2) - 0.0) / p_settings.screen_size.y; + ssao.gather_push_constant.pass_uv_offset[0] = ((i % 2) - 0.0) / p_settings.full_screen_size.x; + ssao.gather_push_constant.pass_uv_offset[1] = ((i / 2) - 0.0) / p_settings.full_screen_size.y; ssao.gather_push_constant.pass = i; RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, _get_uniform_set_from_image(p_ao_slices[i]), 2); RD::get_singleton()->compute_list_set_push_constant(p_compute_list, &ssao.gather_push_constant, sizeof(SSAOGatherPushConstant)); - int x_groups = ((p_settings.screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - int y_groups = ((p_settings.screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; + Size2i size = Size2i(p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1), p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)); - RD::get_singleton()->compute_list_dispatch(p_compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(p_compute_list, size.x, size.y, 1); } RD::get_singleton()->compute_list_add_barrier(p_compute_list); } -void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets) { +void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &p_depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - + RD::get_singleton()->draw_command_begin_label("SSAO"); /* FIRST PASS */ // Downsample and deinterleave the depth buffer. { + RD::get_singleton()->draw_command_begin_label("Downsample Depth"); if (p_invalidate_uniform_sets) { Vector<RD::Uniform> uniforms; { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; u.binding = 0; - u.ids.push_back(depth_mipmaps[1]); + u.ids.push_back(p_depth_mipmaps[1]); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; u.binding = 1; - u.ids.push_back(depth_mipmaps[2]); + u.ids.push_back(p_depth_mipmaps[2]); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; u.binding = 2; - u.ids.push_back(depth_mipmaps[3]); + u.ids.push_back(p_depth_mipmaps[3]); uniforms.push_back(u); } ssao.downsample_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.downsample_shader.version_get_shader(ssao.downsample_shader_version, 2), 2); @@ -1051,8 +1006,8 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep ssao.downsample_push_constant.z_near = p_projection.get_z_near(); ssao.downsample_push_constant.z_far = p_projection.get_z_far(); } - ssao.downsample_push_constant.pixel_size[0] = 1.0 / p_settings.screen_size.x; - ssao.downsample_push_constant.pixel_size[1] = 1.0 / p_settings.screen_size.y; + ssao.downsample_push_constant.pixel_size[0] = 1.0 / p_settings.full_screen_size.x; + ssao.downsample_push_constant.pixel_size[1] = 1.0 / p_settings.full_screen_size.y; ssao.downsample_push_constant.radius_sq = p_settings.radius * p_settings.radius; int downsample_pipeline = SSAO_DOWNSAMPLE; @@ -1068,24 +1023,25 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[downsample_pipeline]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_buffer), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(depth_mipmaps[0]), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_depth_mipmaps[0]), 1); if (p_settings.quality > RS::ENV_SSAO_QUALITY_MEDIUM) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssao.downsample_uniform_set, 2); } RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.downsample_push_constant, sizeof(SSAODownsamplePushConstant)); - int x_groups = (MAX(1, p_settings.screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - int y_groups = (MAX(1, p_settings.screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; + Size2i size(MAX(1, p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1)), MAX(1, p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1))); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, size.x, size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); // Downsample SSAO } /* SECOND PASS */ // Sample SSAO { - ssao.gather_push_constant.screen_size[0] = p_settings.screen_size.x; - ssao.gather_push_constant.screen_size[1] = p_settings.screen_size.y; + RD::get_singleton()->draw_command_begin_label("Gather Samples"); + ssao.gather_push_constant.screen_size[0] = p_settings.full_screen_size.x; + ssao.gather_push_constant.screen_size[1] = p_settings.full_screen_size.y; ssao.gather_push_constant.half_screen_pixel_size[0] = 1.0 / p_settings.half_screen_size.x; ssao.gather_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y; @@ -1122,7 +1078,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep ssao.gather_push_constant.inv_radius_near_limit = 1.0f / radius_near_limit; ssao.gather_push_constant.neg_inv_radius = -1.0 / ssao.gather_push_constant.radius; - ssao.gather_push_constant.load_counter_avg_div = 9.0 / float((p_settings.quarter_size.x) * (p_settings.quarter_size.y) * 255); + ssao.gather_push_constant.load_counter_avg_div = 9.0 / float((p_settings.quarter_screen_size.x) * (p_settings.quarter_screen_size.y) * 255); ssao.gather_push_constant.adaptive_sample_limit = p_settings.adaptive_target; ssao.gather_push_constant.detail_intensity = p_settings.detail; @@ -1184,6 +1140,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep } if (p_settings.quality == RS::ENV_SSAO_QUALITY_ULTRA) { + RD::get_singleton()->draw_command_begin_label("Generate Importance Map"); ssao.importance_map_push_constant.half_screen_pixel_size[0] = 1.0 / p_settings.half_screen_size.x; ssao.importance_map_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y; ssao.importance_map_push_constant.intensity = p_settings.intensity; @@ -1192,21 +1149,19 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_BASE]); gather_ssao(compute_list, p_ao_pong_slices, p_settings, true); //generate importance map - int x_groups = (p_settings.quarter_size.x - 1) / 8 + 1; - int y_groups = (p_settings.quarter_size.y - 1) / 8 + 1; RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GENERATE_IMPORTANCE_MAP]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao_pong), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_importance_map), 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.quarter_screen_size.x, p_settings.quarter_screen_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); //process importance map A RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPA]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_importance_map), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_importance_map_pong), 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.quarter_screen_size.x, p_settings.quarter_screen_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); //process Importance Map B RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPB]); @@ -1214,21 +1169,24 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_importance_map), 1); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssao.counter_uniform_set, 2); RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.quarter_screen_size.x, p_settings.quarter_screen_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_ADAPTIVE]); + RD::get_singleton()->draw_command_end_label(); // Importance Map } else { RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER]); } gather_ssao(compute_list, p_ao_slices, p_settings, false); + RD::get_singleton()->draw_command_end_label(); // Gather SSAO } // /* THIRD PASS */ // // Blur // { + RD::get_singleton()->draw_command_begin_label("Edge Aware Blur"); ssao.blur_push_constant.edge_sharpness = 1.0 - p_settings.sharpness; ssao.blur_push_constant.half_screen_pixel_size[0] = 1.0 / p_settings.half_screen_size.x; ssao.blur_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y; @@ -1268,25 +1226,25 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep } RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.blur_push_constant, sizeof(SSAOBlurPushConstant)); - int x_groups = ((p_settings.screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - int y_groups = ((p_settings.screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + Size2i size(p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1), p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, size.x, size.y, 1); } if (p_settings.quality > RS::ENV_SSAO_QUALITY_VERY_LOW) { RD::get_singleton()->compute_list_add_barrier(compute_list); } } + RD::get_singleton()->draw_command_end_label(); // Blur } /* FOURTH PASS */ // Interleave buffers // back to full size { + RD::get_singleton()->draw_command_begin_label("Interleave Buffers"); ssao.interleave_push_constant.inv_sharpness = 1.0 - p_settings.sharpness; - ssao.interleave_push_constant.pixel_size[0] = 1.0 / p_settings.screen_size.x; - ssao.interleave_push_constant.pixel_size[1] = 1.0 / p_settings.screen_size.y; + ssao.interleave_push_constant.pixel_size[0] = 1.0 / p_settings.full_screen_size.x; + ssao.interleave_push_constant.pixel_size[1] = 1.0 / p_settings.full_screen_size.y; ssao.interleave_push_constant.size_modifier = uint32_t(p_settings.half_size ? 4 : 2); int interleave_pipeline = SSAO_INTERLEAVE_HALF; @@ -1307,17 +1265,15 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.interleave_push_constant, sizeof(SSAOInterleavePushConstant)); - int x_groups = (p_settings.screen_size.x - 1) / 8 + 1; - int y_groups = (p_settings.screen_size.y - 1) / 8 + 1; - - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.full_screen_size.x, p_settings.full_screen_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); // Interleave } - - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->draw_command_end_label(); //SSAO + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_TRANSFER); //wait for upcoming transfer int zero[1] = { 0 }; - RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, false); + RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, 0); //no barrier } void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve) { @@ -1330,12 +1286,9 @@ void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_normal), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_roughness), 1); - int x_groups = (p_size.x - 1) / 8 + 1; - int y_groups = (p_size.y - 1) / 8 + 1; - RD::get_singleton()->compute_list_set_push_constant(compute_list, &roughness_limiter.push_constant, sizeof(RoughnessLimiterPushConstant)); //not used but set anyway - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.x, p_size.y, 1); RD::get_singleton()->compute_list_end(); } @@ -1448,7 +1401,7 @@ void EffectsRD::render_sky(RD::DrawListID p_list, float p_time, RID p_fb, RID p_ RD::get_singleton()->draw_list_draw(draw_list, true); } -void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples) { +void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { ResolvePushConstant push_constant; push_constant.screen_size[0] = p_screen_size.x; push_constant.screen_size[1] = p_screen_size.y; @@ -1465,54 +1418,9 @@ void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RI RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ResolvePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1); - RD::get_singleton()->compute_list_end(); -} - -void EffectsRD::reduce_shadow(RID p_source_shadow, RID p_dest_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, int p_shrink_limit, RD::ComputeListID compute_list) { - uint32_t push_constant[8] = { (uint32_t)p_source_size.x, (uint32_t)p_source_size.y, (uint32_t)p_source_rect.position.x, (uint32_t)p_source_rect.position.y, (uint32_t)p_shrink_limit, 0, 0, 0 }; - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shadow_reduce.pipelines[SHADOW_REDUCE_REDUCE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_source_shadow, p_dest_shadow), 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(uint32_t) * 8); - - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_source_rect.size.width, p_source_rect.size.height, 1, 8, 8, 1); -} -void EffectsRD::filter_shadow(RID p_shadow, RID p_backing_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, RenderingServer::EnvVolumetricFogShadowFilter p_filter, RD::ComputeListID compute_list, bool p_vertical, bool p_horizontal) { - uint32_t push_constant[8] = { (uint32_t)p_source_size.x, (uint32_t)p_source_size.y, (uint32_t)p_source_rect.position.x, (uint32_t)p_source_rect.position.y, 0, 0, 0, 0 }; - - switch (p_filter) { - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED: - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_LOW: { - push_constant[5] = 0; - } break; - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_MEDIUM: { - push_constant[5] = 9; - } break; - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_HIGH: { - push_constant[5] = 18; - } break; - } - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shadow_reduce.pipelines[SHADOW_REDUCE_FILTER]); - if (p_vertical) { - push_constant[6] = 1; - push_constant[7] = 0; - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_shadow, p_backing_shadow), 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(uint32_t) * 8); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_source_rect.size.width, p_source_rect.size.height, 1, 8, 8, 1); - } - if (p_vertical && p_horizontal) { - RD::get_singleton()->compute_list_add_barrier(compute_list); - } - if (p_horizontal) { - push_constant[6] = 0; - push_constant[7] = 1; - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_backing_shadow, p_shadow), 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(uint32_t) * 8); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_source_rect.size.width, p_source_rect.size.height, 1, 8, 8, 1); - } + RD::get_singleton()->compute_list_end(p_barrier); } void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { @@ -1678,8 +1586,12 @@ EffectsRD::EffectsRD() { cube_to_dp.shader.initialize(copy_modes); cube_to_dp.shader_version = cube_to_dp.shader.version_create(); - - cube_to_dp.pipeline = RD::get_singleton()->compute_pipeline_create(cube_to_dp.shader.version_get_shader(cube_to_dp.shader_version, 0)); + RID shader = cube_to_dp.shader.version_get_shader(cube_to_dp.shader_version, 0); + RD::PipelineDepthStencilState dss; + dss.enable_depth_test = true; + dss.depth_compare_operator = RD::COMPARE_OP_ALWAYS; + dss.enable_depth_write = true; + cube_to_dp.pipeline.setup(shader, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), dss, RD::PipelineColorBlendState(), 0); } { @@ -1776,7 +1688,7 @@ EffectsRD::EffectsRD() { } } - RD::get_singleton()->buffer_update(ssao.gather_constants_buffer, 0, sizeof(SSAOGatherConstants), &gather_constants, false); + RD::get_singleton()->buffer_update(ssao.gather_constants_buffer, 0, sizeof(SSAOGatherConstants), &gather_constants); } { Vector<String> ssao_modes; @@ -1795,7 +1707,8 @@ EffectsRD::EffectsRD() { } ssao.importance_map_load_counter = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t)); int zero[1] = { 0 }; - RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, false); + RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero); + RD::get_singleton()->set_resource_name(ssao.importance_map_load_counter, "Importance Map Load Counter"); Vector<RD::Uniform> uniforms; { @@ -1806,6 +1719,7 @@ EffectsRD::EffectsRD() { uniforms.push_back(u); } ssao.counter_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.importance_map_shader.version_get_shader(ssao.importance_map_shader_version, 2), 2); + RD::get_singleton()->set_resource_name(ssao.counter_uniform_set, "Load Counter Uniform Set"); } { Vector<String> ssao_modes; @@ -1834,7 +1748,7 @@ EffectsRD::EffectsRD() { ssao.interleave_shader_version = ssao.interleave_shader.version_create(); for (int i = SSAO_INTERLEAVE; i <= SSAO_INTERLEAVE_HALF; i++) { ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.interleave_shader.version_get_shader(ssao.interleave_shader_version, i - SSAO_INTERLEAVE)); - + RD::get_singleton()->set_resource_name(ssao.pipelines[pipeline], "Interleave Pipeline " + itos(i)); pipeline++; } } @@ -1883,10 +1797,10 @@ EffectsRD::EffectsRD() { if (filter.use_high_quality) { filter.coefficient_buffer = RD::get_singleton()->storage_buffer_create(sizeof(high_quality_coeffs)); - RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(high_quality_coeffs), &high_quality_coeffs[0], false); + RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(high_quality_coeffs), &high_quality_coeffs[0]); } else { filter.coefficient_buffer = RD::get_singleton()->storage_buffer_create(sizeof(low_quality_coeffs)); - RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(low_quality_coeffs), &low_quality_coeffs[0], false); + RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(low_quality_coeffs), &low_quality_coeffs[0]); } Vector<RD::Uniform> uniforms; @@ -2005,20 +1919,6 @@ EffectsRD::EffectsRD() { } { - Vector<String> shadow_reduce_modes; - shadow_reduce_modes.push_back("\n#define MODE_REDUCE\n"); - shadow_reduce_modes.push_back("\n#define MODE_FILTER\n"); - - shadow_reduce.shader.initialize(shadow_reduce_modes); - - shadow_reduce.shader_version = shadow_reduce.shader.version_create(); - - for (int i = 0; i < SHADOW_REDUCE_MAX; i++) { - shadow_reduce.pipelines[i] = RD::get_singleton()->compute_pipeline_create(shadow_reduce.shader.version_get_shader(shadow_reduce.shader_version, i)); - } - } - - { Vector<String> sort_modes; sort_modes.push_back("\n#define MODE_SORT_BLOCK\n"); sort_modes.push_back("\n#define MODE_SORT_STEP\n"); @@ -2039,12 +1939,14 @@ EffectsRD::EffectsRD() { sampler.max_lod = 0; default_sampler = RD::get_singleton()->sampler_create(sampler); + RD::get_singleton()->set_resource_name(default_sampler, "Default Linear Sampler"); sampler.min_filter = RD::SAMPLER_FILTER_LINEAR; sampler.mip_filter = RD::SAMPLER_FILTER_LINEAR; sampler.max_lod = 1e20; default_mipmap_sampler = RD::get_singleton()->sampler_create(sampler); + RD::get_singleton()->set_resource_name(default_mipmap_sampler, "Default MipMap Sampler"); { //create index array for copy shaders Vector<uint8_t> pv; @@ -2104,5 +2006,4 @@ EffectsRD::~EffectsRD() { ssr_scale.shader.version_free(ssr_scale.shader_version); sss.shader.version_free(sss.shader_version); tonemap.shader.version_free(tonemap.shader_version); - shadow_reduce.shader.version_free(shadow_reduce.shader_version); } diff --git a/servers/rendering/renderer_rd/effects_rd.h b/servers/rendering/renderer_rd/effects_rd.h index ad4a660944..1ba25e301b 100644 --- a/servers/rendering/renderer_rd/effects_rd.h +++ b/servers/rendering/renderer_rd/effects_rd.h @@ -46,7 +46,6 @@ #include "servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/shadow_reduce.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/sort.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/specular_merge.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/ssao.glsl.gen.h" @@ -234,18 +233,17 @@ class EffectsRD { } luminance_reduce; struct CopyToDPPushConstant { - int32_t screen_size[2]; - int32_t dest_offset[2]; - float bias; float z_far; float z_near; uint32_t z_flip; + uint32_t pad; + float screen_rect[4]; }; struct CoptToDP { CubeToDpShaderRD shader; RID shader_version; - RID pipeline; + PipelineCacheRD pipeline; } cube_to_dp; struct BokehPushConstant { @@ -598,18 +596,6 @@ class EffectsRD { RID pipelines[RESOLVE_MODE_MAX]; //3 quality levels } resolve; - enum ShadowReduceMode { - SHADOW_REDUCE_REDUCE, - SHADOW_REDUCE_FILTER, - SHADOW_REDUCE_MAX - }; - - struct ShadowReduce { - ShadowReduceShaderRD shader; - RID shader_version; - RID pipelines[SHADOW_REDUCE_MAX]; - } shadow_reduce; - enum SortMode { SORT_MODE_BLOCK, SORT_MODE_STEP, @@ -687,7 +673,7 @@ public: void cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); void make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size); - void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, float p_z_near, float p_z_far, float p_bias, bool p_dp_flip); + void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip); void luminance_reduction(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set = false); void bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_secondary_texture, RID p_bokeh_texture1, RID p_bokeh_texture2, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RS::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); @@ -745,9 +731,9 @@ public: float fadeout_from = 50.0; float fadeout_to = 300.0; - Size2i screen_size = Size2i(); + Size2i full_screen_size = Size2i(); Size2i half_screen_size = Size2i(); - Size2i quarter_size = Size2i(); + Size2i quarter_screen_size = Size2i(); }; void tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings); @@ -764,10 +750,7 @@ public: void merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_base, RID p_reflection); void sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_depth, const CameraMatrix &p_camera, const Size2i &p_screen_size, float p_scale, float p_depth_scale, RS::SubSurfaceScatteringQuality p_quality); - void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples); - - void reduce_shadow(RID p_source_shadow, RID p_dest_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, int p_shrink_limit, RenderingDevice::ComputeListID compute_list); - void filter_shadow(RID p_shadow, RID p_backing_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, RS::EnvVolumetricFogShadowFilter p_filter, RenderingDevice::ComputeListID compute_list, bool p_vertical = true, bool p_horizontal = true); + void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); void sort_buffer(RID p_uniform_set, int p_size); diff --git a/servers/rendering/renderer_rd/light_cluster_builder.cpp b/servers/rendering/renderer_rd/light_cluster_builder.cpp deleted file mode 100644 index bb807ca4ca..0000000000 --- a/servers/rendering/renderer_rd/light_cluster_builder.cpp +++ /dev/null @@ -1,252 +0,0 @@ -/*************************************************************************/ -/* light_cluster_builder.cpp */ -/*************************************************************************/ -/* This file is part of: */ -/* GODOT ENGINE */ -/* https://godotengine.org */ -/*************************************************************************/ -/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ -/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ -/* */ -/* Permission is hereby granted, free of charge, to any person obtaining */ -/* a copy of this software and associated documentation files (the */ -/* "Software"), to deal in the Software without restriction, including */ -/* without limitation the rights to use, copy, modify, merge, publish, */ -/* distribute, sublicense, and/or sell copies of the Software, and to */ -/* permit persons to whom the Software is furnished to do so, subject to */ -/* the following conditions: */ -/* */ -/* The above copyright notice and this permission notice shall be */ -/* included in all copies or substantial portions of the Software. */ -/* */ -/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ -/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ -/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ -/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ -/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ -/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ -/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/*************************************************************************/ - -#include "light_cluster_builder.h" - -void LightClusterBuilder::begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection) { - view_xform = p_view_transform; - projection = p_cam_projection; - z_near = -projection.get_z_near(); - z_far = -projection.get_z_far(); - - //reset counts - light_count = 0; - refprobe_count = 0; - decal_count = 0; - item_count = 0; - sort_id_count = 0; -} - -void LightClusterBuilder::bake_cluster() { - float slice_depth = (z_near - z_far) / depth; - - uint8_t *cluster_dataw = cluster_data.ptrw(); - Cell *cluster_data_ptr = (Cell *)cluster_dataw; - //clear the cluster - zeromem(cluster_data_ptr, (width * height * depth * sizeof(Cell))); - - /* Step 1, create cell positions and count them */ - - for (uint32_t i = 0; i < item_count; i++) { - const Item &item = items[i]; - - int from_slice = Math::floor((z_near - (item.aabb.position.z + item.aabb.size.z)) / slice_depth); - int to_slice = Math::floor((z_near - item.aabb.position.z) / slice_depth); - - if (from_slice >= (int)depth || to_slice < 0) { - continue; //sorry no go - } - - from_slice = MAX(0, from_slice); - to_slice = MIN((int)depth - 1, to_slice); - - for (int j = from_slice; j <= to_slice; j++) { - Vector3 min = item.aabb.position; - Vector3 max = item.aabb.position + item.aabb.size; - - float limit_near = MIN((z_near - slice_depth * j), max.z); - float limit_far = MAX((z_near - slice_depth * (j + 1)), min.z); - - max.z = limit_near; - min.z = limit_near; - - Vector3 proj_min = projection.xform(min); - Vector3 proj_max = projection.xform(max); - - int near_from_x = int(Math::floor((proj_min.x * 0.5 + 0.5) * width)); - int near_from_y = int(Math::floor((-proj_max.y * 0.5 + 0.5) * height)); - int near_to_x = int(Math::floor((proj_max.x * 0.5 + 0.5) * width)); - int near_to_y = int(Math::floor((-proj_min.y * 0.5 + 0.5) * height)); - - max.z = limit_far; - min.z = limit_far; - - proj_min = projection.xform(min); - proj_max = projection.xform(max); - - int far_from_x = int(Math::floor((proj_min.x * 0.5 + 0.5) * width)); - int far_from_y = int(Math::floor((-proj_max.y * 0.5 + 0.5) * height)); - int far_to_x = int(Math::floor((proj_max.x * 0.5 + 0.5) * width)); - int far_to_y = int(Math::floor((-proj_min.y * 0.5 + 0.5) * height)); - - //print_line(itos(j) + " near - " + Vector2i(near_from_x, near_from_y) + " -> " + Vector2i(near_to_x, near_to_y)); - //print_line(itos(j) + " far - " + Vector2i(far_from_x, far_from_y) + " -> " + Vector2i(far_to_x, far_to_y)); - - int from_x = MIN(near_from_x, far_from_x); - int from_y = MIN(near_from_y, far_from_y); - int to_x = MAX(near_to_x, far_to_x); - int to_y = MAX(near_to_y, far_to_y); - - if (from_x >= (int)width || to_x < 0 || from_y >= (int)height || to_y < 0) { - continue; - } - - int sx = MAX(0, from_x); - int sy = MAX(0, from_y); - int dx = MIN((int)width - 1, to_x); - int dy = MIN((int)height - 1, to_y); - - //print_line(itos(j) + " - " + Vector2i(sx, sy) + " -> " + Vector2i(dx, dy)); - - for (int x = sx; x <= dx; x++) { - for (int y = sy; y <= dy; y++) { - uint32_t offset = j * (width * height) + y * width + x; - - if (unlikely(sort_id_count == sort_id_max)) { - sort_id_max = nearest_power_of_2_templated(sort_id_max + 1); - sort_ids = (SortID *)memrealloc(sort_ids, sizeof(SortID) * sort_id_max); - if (ids.size()) { - ids.resize(sort_id_max); - RD::get_singleton()->free(items_buffer); - items_buffer = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t) * sort_id_max); - } - } - - sort_ids[sort_id_count].cell_index = offset; - sort_ids[sort_id_count].item_index = item.index; - sort_ids[sort_id_count].item_type = item.type; - - sort_id_count++; - - //for now, only count - cluster_data_ptr[offset].item_pointers[item.type]++; - //print_line("at offset " + itos(offset) + " value: " + itos(cluster_data_ptr[offset].item_pointers[item.type])); - } - } - } - } - - /* Step 2, Assign pointers (and reset counters) */ - - uint32_t offset = 0; - for (uint32_t i = 0; i < (width * height * depth); i++) { - for (int j = 0; j < ITEM_TYPE_MAX; j++) { - uint32_t count = cluster_data_ptr[i].item_pointers[j]; //save count - cluster_data_ptr[i].item_pointers[j] = offset; //replace count by pointer - offset += count; //increase offset by count; - } - } - - //print_line("offset: " + itos(offset)); - /* Step 3, Place item lists */ - - uint32_t *ids_ptr = ids.ptrw(); - - for (uint32_t i = 0; i < sort_id_count; i++) { - const SortID &id = sort_ids[i]; - Cell &cell = cluster_data_ptr[id.cell_index]; - uint32_t pointer = cell.item_pointers[id.item_type] & POINTER_MASK; - uint32_t counter = cell.item_pointers[id.item_type] >> COUNTER_SHIFT; - ids_ptr[pointer + counter] = id.item_index; - - cell.item_pointers[id.item_type] = pointer | ((counter + 1) << COUNTER_SHIFT); - } - - RD::get_singleton()->texture_update(cluster_texture, 0, cluster_data, true); - RD::get_singleton()->buffer_update(items_buffer, 0, offset * sizeof(uint32_t), ids_ptr, true); -} - -void LightClusterBuilder::setup(uint32_t p_width, uint32_t p_height, uint32_t p_depth) { - if (width == p_width && height == p_height && depth == p_depth) { - return; - } - if (cluster_texture.is_valid()) { - RD::get_singleton()->free(cluster_texture); - } - - width = p_width; - height = p_height; - depth = p_depth; - - cluster_data.resize(width * height * depth * sizeof(Cell)); - - { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32G32B32A32_UINT; - tf.texture_type = RD::TEXTURE_TYPE_3D; - tf.width = width; - tf.height = height; - tf.depth = depth; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; - - cluster_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } -} - -RID LightClusterBuilder::get_cluster_texture() const { - return cluster_texture; -} - -RID LightClusterBuilder::get_cluster_indices_buffer() const { - return items_buffer; -} - -LightClusterBuilder::LightClusterBuilder() { - //initialize accumulators to something - lights = (LightData *)memalloc(sizeof(LightData) * 1024); - light_max = 1024; - - refprobes = (OrientedBoxData *)memalloc(sizeof(OrientedBoxData) * 1024); - refprobe_max = 1024; - - decals = (OrientedBoxData *)memalloc(sizeof(OrientedBoxData) * 1024); - decal_max = 1024; - - items = (Item *)memalloc(sizeof(Item) * 1024); - item_max = 1024; - - sort_ids = (SortID *)memalloc(sizeof(SortID) * 1024); - ids.resize(2014); - items_buffer = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t) * 1024); - item_max = 1024; -} - -LightClusterBuilder::~LightClusterBuilder() { - if (cluster_data.size()) { - RD::get_singleton()->free(cluster_texture); - } - - if (lights) { - memfree(lights); - } - if (refprobes) { - memfree(refprobes); - } - if (decals) { - memfree(decals); - } - if (items) { - memfree(items); - } - if (sort_ids) { - memfree(sort_ids); - RD::get_singleton()->free(items_buffer); - } -} diff --git a/servers/rendering/renderer_rd/light_cluster_builder.h b/servers/rendering/renderer_rd/light_cluster_builder.h deleted file mode 100644 index 8f77ece6f5..0000000000 --- a/servers/rendering/renderer_rd/light_cluster_builder.h +++ /dev/null @@ -1,290 +0,0 @@ -/*************************************************************************/ -/* light_cluster_builder.h */ -/*************************************************************************/ -/* This file is part of: */ -/* GODOT ENGINE */ -/* https://godotengine.org */ -/*************************************************************************/ -/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ -/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ -/* */ -/* Permission is hereby granted, free of charge, to any person obtaining */ -/* a copy of this software and associated documentation files (the */ -/* "Software"), to deal in the Software without restriction, including */ -/* without limitation the rights to use, copy, modify, merge, publish, */ -/* distribute, sublicense, and/or sell copies of the Software, and to */ -/* permit persons to whom the Software is furnished to do so, subject to */ -/* the following conditions: */ -/* */ -/* The above copyright notice and this permission notice shall be */ -/* included in all copies or substantial portions of the Software. */ -/* */ -/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ -/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ -/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ -/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ -/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ -/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ -/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/*************************************************************************/ - -#ifndef LIGHT_CLUSTER_BUILDER_H -#define LIGHT_CLUSTER_BUILDER_H - -#include "servers/rendering/renderer_rd/renderer_storage_rd.h" - -class LightClusterBuilder { -public: - enum LightType { - LIGHT_TYPE_OMNI, - LIGHT_TYPE_SPOT - }; - - enum ItemType { - ITEM_TYPE_OMNI_LIGHT, - ITEM_TYPE_SPOT_LIGHT, - ITEM_TYPE_REFLECTION_PROBE, - ITEM_TYPE_DECAL, - ITEM_TYPE_MAX //should always be 4 - }; - - enum { - COUNTER_SHIFT = 20, //one million total ids - POINTER_MASK = (1 << COUNTER_SHIFT) - 1, - COUNTER_MASK = 0xfff // 4096 items per cell - }; - -private: - struct LightData { - float position[3]; - uint32_t type; - float radius; - float spot_aperture; - uint32_t pad[2]; - }; - - uint32_t light_count = 0; - uint32_t light_max = 0; - LightData *lights = nullptr; - - struct OrientedBoxData { - float position[3]; - uint32_t pad; - float x_axis[3]; - uint32_t pad2; - float y_axis[3]; - uint32_t pad3; - float z_axis[3]; - uint32_t pad4; - }; - - uint32_t refprobe_count = 0; - uint32_t refprobe_max = 0; - OrientedBoxData *refprobes = nullptr; - - uint32_t decal_count = 0; - uint32_t decal_max = 0; - OrientedBoxData *decals = nullptr; - - struct Item { - AABB aabb; - ItemType type; - uint32_t index; - }; - - Item *items = nullptr; - uint32_t item_count = 0; - uint32_t item_max = 0; - - uint32_t width = 0; - uint32_t height = 0; - uint32_t depth = 0; - - struct Cell { - uint32_t item_pointers[ITEM_TYPE_MAX]; - }; - - Vector<uint8_t> cluster_data; - RID cluster_texture; - - struct SortID { - uint32_t cell_index; - uint32_t item_index; - ItemType item_type; - }; - - SortID *sort_ids = nullptr; - Vector<uint32_t> ids; - uint32_t sort_id_count = 0; - uint32_t sort_id_max = 0; - RID items_buffer; - - Transform view_xform; - CameraMatrix projection; - float z_far = 0; - float z_near = 0; - - _FORCE_INLINE_ void _add_item(const AABB &p_aabb, ItemType p_type, uint32_t p_index) { - if (unlikely(item_count == item_max)) { - item_max = nearest_power_of_2_templated(item_max + 1); - items = (Item *)memrealloc(items, sizeof(Item) * item_max); - } - - Item &item = items[item_count]; - item.aabb = p_aabb; - item.index = p_index; - item.type = p_type; - item_count++; - } - -public: - void begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection); - - _FORCE_INLINE_ void add_light(LightType p_type, const Transform &p_transform, float p_radius, float p_spot_aperture) { - if (unlikely(light_count == light_max)) { - light_max = nearest_power_of_2_templated(light_max + 1); - lights = (LightData *)memrealloc(lights, sizeof(LightData) * light_max); - } - - LightData &ld = lights[light_count]; - ld.type = p_type; - ld.position[0] = p_transform.origin.x; - ld.position[1] = p_transform.origin.y; - ld.position[2] = p_transform.origin.z; - ld.radius = p_radius; - ld.spot_aperture = p_spot_aperture; - - Transform xform = view_xform * p_transform; - - ld.radius *= xform.basis.get_uniform_scale(); - - AABB aabb; - - switch (p_type) { - case LIGHT_TYPE_OMNI: { - aabb.position = xform.origin; - aabb.size = Vector3(ld.radius, ld.radius, ld.radius); - aabb.position -= aabb.size; - aabb.size *= 2.0; - - _add_item(aabb, ITEM_TYPE_OMNI_LIGHT, light_count); - } break; - case LIGHT_TYPE_SPOT: { - float r = ld.radius; - real_t len = Math::tan(Math::deg2rad(ld.spot_aperture)) * r; - - aabb.position = xform.origin; - aabb.expand_to(xform.xform(Vector3(len, len, -r))); - aabb.expand_to(xform.xform(Vector3(-len, len, -r))); - aabb.expand_to(xform.xform(Vector3(-len, -len, -r))); - aabb.expand_to(xform.xform(Vector3(len, -len, -r))); - _add_item(aabb, ITEM_TYPE_SPOT_LIGHT, light_count); - } break; - } - - light_count++; - } - - _FORCE_INLINE_ void add_reflection_probe(const Transform &p_transform, const Vector3 &p_half_extents) { - if (unlikely(refprobe_count == refprobe_max)) { - refprobe_max = nearest_power_of_2_templated(refprobe_max + 1); - refprobes = (OrientedBoxData *)memrealloc(refprobes, sizeof(OrientedBoxData) * refprobe_max); - } - - Transform xform = view_xform * p_transform; - - OrientedBoxData &rp = refprobes[refprobe_count]; - Vector3 origin = xform.origin; - rp.position[0] = origin.x; - rp.position[1] = origin.y; - rp.position[2] = origin.z; - - Vector3 x_axis = xform.basis.get_axis(0) * p_half_extents.x; - rp.x_axis[0] = x_axis.x; - rp.x_axis[1] = x_axis.y; - rp.x_axis[2] = x_axis.z; - - Vector3 y_axis = xform.basis.get_axis(1) * p_half_extents.y; - rp.y_axis[0] = y_axis.x; - rp.y_axis[1] = y_axis.y; - rp.y_axis[2] = y_axis.z; - - Vector3 z_axis = xform.basis.get_axis(2) * p_half_extents.z; - rp.z_axis[0] = z_axis.x; - rp.z_axis[1] = z_axis.y; - rp.z_axis[2] = z_axis.z; - - AABB aabb; - - aabb.position = origin + x_axis + y_axis + z_axis; - aabb.expand_to(origin + x_axis + y_axis - z_axis); - aabb.expand_to(origin + x_axis - y_axis + z_axis); - aabb.expand_to(origin + x_axis - y_axis - z_axis); - aabb.expand_to(origin - x_axis + y_axis + z_axis); - aabb.expand_to(origin - x_axis + y_axis - z_axis); - aabb.expand_to(origin - x_axis - y_axis + z_axis); - aabb.expand_to(origin - x_axis - y_axis - z_axis); - - _add_item(aabb, ITEM_TYPE_REFLECTION_PROBE, refprobe_count); - - refprobe_count++; - } - - _FORCE_INLINE_ void add_decal(const Transform &p_transform, const Vector3 &p_half_extents) { - if (unlikely(decal_count == decal_max)) { - decal_max = nearest_power_of_2_templated(decal_max + 1); - decals = (OrientedBoxData *)memrealloc(decals, sizeof(OrientedBoxData) * decal_max); - } - - Transform xform = view_xform * p_transform; - - OrientedBoxData &dc = decals[decal_count]; - - Vector3 origin = xform.origin; - dc.position[0] = origin.x; - dc.position[1] = origin.y; - dc.position[2] = origin.z; - - Vector3 x_axis = xform.basis.get_axis(0) * p_half_extents.x; - dc.x_axis[0] = x_axis.x; - dc.x_axis[1] = x_axis.y; - dc.x_axis[2] = x_axis.z; - - Vector3 y_axis = xform.basis.get_axis(1) * p_half_extents.y; - dc.y_axis[0] = y_axis.x; - dc.y_axis[1] = y_axis.y; - dc.y_axis[2] = y_axis.z; - - Vector3 z_axis = xform.basis.get_axis(2) * p_half_extents.z; - dc.z_axis[0] = z_axis.x; - dc.z_axis[1] = z_axis.y; - dc.z_axis[2] = z_axis.z; - - AABB aabb; - - aabb.position = origin + x_axis + y_axis + z_axis; - aabb.expand_to(origin + x_axis + y_axis - z_axis); - aabb.expand_to(origin + x_axis - y_axis + z_axis); - aabb.expand_to(origin + x_axis - y_axis - z_axis); - aabb.expand_to(origin - x_axis + y_axis + z_axis); - aabb.expand_to(origin - x_axis + y_axis - z_axis); - aabb.expand_to(origin - x_axis - y_axis + z_axis); - aabb.expand_to(origin - x_axis - y_axis - z_axis); - - _add_item(aabb, ITEM_TYPE_DECAL, decal_count); - - decal_count++; - } - - void bake_cluster(); - - void setup(uint32_t p_width, uint32_t p_height, uint32_t p_depth); - - RID get_cluster_texture() const; - RID get_cluster_indices_buffer() const; - - LightClusterBuilder(); - ~LightClusterBuilder(); -}; - -#endif // LIGHT_CLUSTER_BUILDER_H diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp index 508d56cfab..2a1a4efe48 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp @@ -1367,7 +1367,7 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p } if (light_count > 0) { - RD::get_singleton()->buffer_update(state.lights_uniform_buffer, 0, sizeof(LightUniform) * light_count, &state.light_uniforms[0], true); + RD::get_singleton()->buffer_update(state.lights_uniform_buffer, 0, sizeof(LightUniform) * light_count, &state.light_uniforms[0]); } { @@ -1421,7 +1421,7 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p //print_line("w: " + itos(ssize.width) + " s: " + rtos(canvas_scale)); state_buffer.tex_to_sdf = 1.0 / ((canvas_scale.x + canvas_scale.y) * 0.5); - RD::get_singleton()->buffer_update(state.canvas_state_buffer, 0, sizeof(State::Buffer), &state_buffer, true); + RD::get_singleton()->buffer_update(state.canvas_state_buffer, 0, sizeof(State::Buffer), &state_buffer); } { //default filter/repeat @@ -1622,7 +1622,7 @@ void RendererCanvasRenderRD::light_update_shadow(RID p_rid, int p_shadow_index, projection.set_frustum(xmin, xmax, ymin, ymax, nearp, farp); } - Vector3 cam_target = Basis(Vector3(0, 0, Math_PI * 2 * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); + Vector3 cam_target = Basis(Vector3(0, 0, Math_TAU * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); projection = projection * CameraMatrix(Transform().looking_at(cam_target, Vector3(0, 0, -1)).affine_inverse()); ShadowRenderPushConstant push_constant; @@ -2239,6 +2239,11 @@ Variant RendererCanvasRenderRD::ShaderData::get_default_parameter(const StringNa return Variant(); } +RS::ShaderNativeSourceCode RendererCanvasRenderRD::ShaderData::get_native_source_code() const { + RendererCanvasRenderRD *canvas_singleton = (RendererCanvasRenderRD *)RendererCanvasRender::singleton; + return canvas_singleton->shader.canvas_shader.version_get_native_source_code(version); +} + RendererCanvasRenderRD::ShaderData::ShaderData() { valid = false; uses_screen_texture = false; @@ -2489,7 +2494,7 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { actions.renames["COLOR"] = "color"; actions.renames["NORMAL"] = "normal"; actions.renames["NORMAL_MAP"] = "normal_map"; - actions.renames["NORMAL_MAP_DEPTH"] = "normal_depth"; + actions.renames["NORMAL_MAP_DEPTH"] = "normal_map_depth"; actions.renames["TEXTURE"] = "color_texture"; actions.renames["TEXTURE_PIXEL_SIZE"] = "draw_data.color_texture_pixel_size"; actions.renames["NORMAL_TEXTURE"] = "normal_texture"; @@ -2501,7 +2506,7 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { actions.renames["FRAGCOORD"] = "gl_FragCoord"; actions.renames["POINT_COORD"] = "gl_PointCoord"; - actions.renames["LIGHT_POSITION"] = "light_pos"; + actions.renames["LIGHT_POSITION"] = "light_position"; actions.renames["LIGHT_COLOR"] = "light_color"; actions.renames["LIGHT_ENERGY"] = "light_energy"; actions.renames["LIGHT"] = "light"; diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h index 545eeaa106..cb947d7180 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h @@ -188,6 +188,8 @@ class RendererCanvasRenderRD : public RendererCanvasRender { virtual bool is_animated() const; virtual bool casts_shadows() const; virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + ShaderData(); virtual ~ShaderData(); }; diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp index fb9c114ade..be2552bd32 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp @@ -154,12 +154,9 @@ void RendererCompositorRD::initialize() { } } -ThreadWorkPool RendererCompositorRD::thread_work_pool; uint64_t RendererCompositorRD::frame = 1; void RendererCompositorRD::finalize() { - thread_work_pool.finish(); - memdelete(scene); memdelete(canvas); memdelete(storage); @@ -174,7 +171,6 @@ RendererCompositorRD *RendererCompositorRD::singleton = nullptr; RendererCompositorRD::RendererCompositorRD() { singleton = this; - thread_work_pool.init(); time = 0; storage = memnew(RendererStorageRD); diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.h b/servers/rendering/renderer_rd/renderer_compositor_rd.h index e1995872af..cb85fc79e0 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.h +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.h @@ -90,8 +90,6 @@ public: virtual bool is_low_end() const { return false; } - static ThreadWorkPool thread_work_pool; - static RendererCompositorRD *singleton; RendererCompositorRD(); ~RendererCompositorRD() {} diff --git a/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp b/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp index 6881d7913f..509495680a 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp @@ -394,6 +394,12 @@ Variant RendererSceneRenderForward::ShaderData::get_default_parameter(const Stri return Variant(); } +RS::ShaderNativeSourceCode RendererSceneRenderForward::ShaderData::get_native_source_code() const { + RendererSceneRenderForward *scene_singleton = (RendererSceneRenderForward *)RendererSceneRenderForward::singleton; + + return scene_singleton->shader.scene_shader.version_get_native_source_code(version); +} + RendererSceneRenderForward::ShaderData::ShaderData() { valid = false; uses_screen_texture = false; @@ -447,7 +453,7 @@ void RendererSceneRenderForward::MaterialData::update_parameters(const Map<Strin //check whether buffer changed if (p_uniform_dirty && ubo_data.size()) { update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); + RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw(), RD::BARRIER_MASK_RASTER); } uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); @@ -577,19 +583,6 @@ void RendererSceneRenderForward::RenderBufferDataForward::ensure_specular() { } } -void RendererSceneRenderForward::RenderBufferDataForward::ensure_gi() { - if (!reflection_buffer.is_valid()) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - tf.width = width; - tf.height = height; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - - reflection_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); - ambient_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } -} - void RendererSceneRenderForward::RenderBufferDataForward::ensure_giprobe() { if (!giprobe_buffer.is_valid()) { RD::TextureFormat tf; @@ -627,16 +620,6 @@ void RendererSceneRenderForward::RenderBufferDataForward::ensure_giprobe() { } void RendererSceneRenderForward::RenderBufferDataForward::clear() { - if (ambient_buffer != RID() && ambient_buffer != color) { - RD::get_singleton()->free(ambient_buffer); - ambient_buffer = RID(); - } - - if (reflection_buffer != RID() && reflection_buffer != specular) { - RD::get_singleton()->free(reflection_buffer); - reflection_buffer = RID(); - } - if (giprobe_buffer != RID()) { RD::get_singleton()->free(giprobe_buffer); giprobe_buffer = RID(); @@ -806,252 +789,92 @@ bool RendererSceneRenderForward::free(RID p_rid) { return false; } -void RendererSceneRenderForward::_fill_instances(RenderList::Element **p_elements, int p_element_count, bool p_for_depth, bool p_has_sdfgi, bool p_has_opaque_gi) { - uint32_t lightmap_captures_used = 0; - - for (int i = 0; i < p_element_count; i++) { - const RenderList::Element *e = p_elements[i]; - InstanceData &id = scene_state.instances[i]; - bool store_transform = true; - id.flags = 0; - id.mask = e->instance->layer_mask; - id.instance_uniforms_ofs = e->instance->instance_allocated_shader_parameters_offset >= 0 ? e->instance->instance_allocated_shader_parameters_offset : 0; - - if (e->instance->base_type == RS::INSTANCE_MULTIMESH) { - id.flags |= INSTANCE_DATA_FLAG_MULTIMESH; - uint32_t stride; - if (storage->multimesh_get_transform_format(e->instance->base) == RS::MULTIMESH_TRANSFORM_2D) { - id.flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; - stride = 2; - } else { - stride = 3; - } - if (storage->multimesh_uses_colors(e->instance->base)) { - id.flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; - stride += 1; - } - if (storage->multimesh_uses_custom_data(e->instance->base)) { - id.flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; - stride += 1; - } - - id.flags |= (stride << INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_SHIFT); - } else if (e->instance->base_type == RS::INSTANCE_PARTICLES) { - id.flags |= INSTANCE_DATA_FLAG_MULTIMESH; - uint32_t stride; - if (false) { // 2D particles - id.flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; - stride = 2; - } else { - stride = 3; - } - - id.flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; - stride += 1; - - id.flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; - stride += 1; - - id.flags |= (stride << INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_SHIFT); - - if (!storage->particles_is_using_local_coords(e->instance->base)) { - store_transform = false; - } - - } else if (e->instance->base_type == RS::INSTANCE_MESH) { - if (e->instance->skeleton.is_valid()) { - id.flags |= INSTANCE_DATA_FLAG_SKELETON; - } - } - - if (store_transform) { - RendererStorageRD::store_transform(e->instance->transform, id.transform); - RendererStorageRD::store_transform(Transform(e->instance->transform.basis.inverse().transposed()), id.normal_transform); - } else { - RendererStorageRD::store_transform(Transform(), id.transform); - RendererStorageRD::store_transform(Transform(), id.normal_transform); - } - - if (p_for_depth) { - id.gi_offset = 0xFFFFFFFF; - continue; - } - - if (e->instance->lightmap) { - int32_t lightmap_index = storage->lightmap_get_array_index(e->instance->lightmap->base); - if (lightmap_index >= 0) { - id.gi_offset = lightmap_index; - id.gi_offset |= e->instance->lightmap_slice_index << 12; - id.gi_offset |= e->instance->lightmap_cull_index << 20; - id.lightmap_uv_scale[0] = e->instance->lightmap_uv_scale.position.x; - id.lightmap_uv_scale[1] = e->instance->lightmap_uv_scale.position.y; - id.lightmap_uv_scale[2] = e->instance->lightmap_uv_scale.size.width; - id.lightmap_uv_scale[3] = e->instance->lightmap_uv_scale.size.height; - id.flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP; - if (storage->lightmap_uses_spherical_harmonics(e->instance->lightmap->base)) { - id.flags |= INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP; - } - } else { - id.gi_offset = 0xFFFFFFFF; - } - } else if (!e->instance->lightmap_sh.is_empty()) { - if (lightmap_captures_used < scene_state.max_lightmap_captures) { - const Color *src_capture = e->instance->lightmap_sh.ptr(); - LightmapCaptureData &lcd = scene_state.lightmap_captures[lightmap_captures_used]; - for (int j = 0; j < 9; j++) { - lcd.sh[j * 4 + 0] = src_capture[j].r; - lcd.sh[j * 4 + 1] = src_capture[j].g; - lcd.sh[j * 4 + 2] = src_capture[j].b; - lcd.sh[j * 4 + 3] = src_capture[j].a; - } - id.flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE; - id.gi_offset = lightmap_captures_used; - lightmap_captures_used++; - } - - } else { - if (p_has_opaque_gi) { - id.flags |= INSTANCE_DATA_FLAG_USE_GI_BUFFERS; - } - - if (!low_end && !e->instance->gi_probe_instances.is_empty()) { - uint32_t written = 0; - for (int j = 0; j < e->instance->gi_probe_instances.size(); j++) { - RID probe = e->instance->gi_probe_instances[j]; - - uint32_t index = gi_probe_instance_get_render_index(probe); - - if (written == 0) { - id.gi_offset = index; - id.flags |= INSTANCE_DATA_FLAG_USE_GIPROBE; - written = 1; - } else { - id.gi_offset = index << 16; - written = 2; - break; - } - } - if (written == 0) { - id.gi_offset = 0xFFFFFFFF; - } else if (written == 1) { - id.gi_offset |= 0xFFFF0000; - } - } else { - if (p_has_sdfgi && (e->instance->baked_light || e->instance->dynamic_gi)) { - id.flags |= INSTANCE_DATA_FLAG_USE_SDFGI; - } - id.gi_offset = 0xFFFFFFFF; - } - } - } - - RD::get_singleton()->buffer_update(scene_state.instance_buffer, 0, sizeof(InstanceData) * p_element_count, scene_state.instances, true); - if (lightmap_captures_used) { - RD::get_singleton()->buffer_update(scene_state.lightmap_capture_buffer, 0, sizeof(LightmapCaptureData) * lightmap_captures_used, scene_state.lightmap_captures, true); - } -} - /// RENDERING /// -void RendererSceneRenderForward::_render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderList::Element **p_elements, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe, const Vector2 &p_uv_offset, const Plane &p_lod_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) { +template <RendererSceneRenderForward::PassMode p_pass_mode> +void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element) { RD::DrawListID draw_list = p_draw_list; RD::FramebufferFormatID framebuffer_format = p_framebuffer_Format; //global scope bindings RD::get_singleton()->draw_list_bind_uniform_set(draw_list, render_base_uniform_set, SCENE_UNIFORM_SET); - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_render_pass_uniform_set, RENDER_PASS_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_params->render_pass_uniform_set, RENDER_PASS_UNIFORM_SET); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, default_vec4_xform_uniform_set, TRANSFORMS_UNIFORM_SET); - MaterialData *prev_material = nullptr; + RID prev_material_uniform_set; RID prev_vertex_array_rd; RID prev_index_array_rd; RID prev_pipeline_rd; RID prev_xforms_uniform_set; - PushConstant push_constant; - zeromem(&push_constant, sizeof(PushConstant)); - push_constant.bake_uv2_offset[0] = p_uv_offset.x; - push_constant.bake_uv2_offset[1] = p_uv_offset.y; + bool shadow_pass = (p_params->pass_mode == PASS_MODE_SHADOW) || (p_params->pass_mode == PASS_MODE_SHADOW_DP); + + SceneState::PushConstant push_constant; + + if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { + push_constant.uv_offset = Math::make_half_float(p_params->uv_offset.y) << 16; + push_constant.uv_offset |= Math::make_half_float(p_params->uv_offset.x); + } else { + push_constant.uv_offset = 0; + } + + for (uint32_t i = p_from_element; i < p_to_element; i++) { + const GeometryInstanceSurfaceDataCache *surf = p_params->elements[i]; + const RenderElementInfo &element_info = p_params->element_info[i]; + + push_constant.base_index = i + p_params->element_offset; + + RID material_uniform_set; + ShaderData *shader; + void *mesh_surface; + + if (shadow_pass || p_params->pass_mode == PASS_MODE_DEPTH) { //regular depth pass can use these too + material_uniform_set = surf->material_uniform_set_shadow; + shader = surf->shader_shadow; + mesh_surface = surf->surface_shadow; - for (int i = 0; i < p_element_count; i++) { - const RenderList::Element *e = p_elements[i]; + } else { + material_uniform_set = surf->material_uniform_set; + shader = surf->shader; + mesh_surface = surf->surface; + } - MaterialData *material = e->material; - ShaderData *shader = material->shader_data; - RID xforms_uniform_set; + if (!mesh_surface) { + continue; + } //find cull variant ShaderData::CullVariant cull_variant; - if (p_pass_mode == PASS_MODE_DEPTH_MATERIAL || p_pass_mode == PASS_MODE_SDF || ((p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) && e->instance->cast_shadows == RS::SHADOW_CASTING_SETTING_DOUBLE_SIDED)) { + if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL || p_params->pass_mode == PASS_MODE_SDF || ((p_params->pass_mode == PASS_MODE_SHADOW || p_params->pass_mode == PASS_MODE_SHADOW_DP) && surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS)) { cull_variant = ShaderData::CULL_VARIANT_DOUBLE_SIDED; } else { - bool mirror = e->instance->mirror; - if (p_reverse_cull) { + bool mirror = surf->owner->mirror; + if (p_params->reverse_cull) { mirror = !mirror; } cull_variant = mirror ? ShaderData::CULL_VARIANT_REVERSED : ShaderData::CULL_VARIANT_NORMAL; } - //find primitive and vertex format - RS::PrimitiveType primitive; - void *mesh_surface = nullptr; - - switch (e->instance->base_type) { - case RS::INSTANCE_MESH: { - mesh_surface = storage->mesh_get_surface(e->instance->base, e->surface_index); - - primitive = storage->mesh_surface_get_primitive(mesh_surface); - if (e->instance->skeleton.is_valid()) { - xforms_uniform_set = storage->skeleton_get_3d_uniform_set(e->instance->skeleton, default_shader_rd, TRANSFORMS_UNIFORM_SET); - } - } break; - case RS::INSTANCE_MULTIMESH: { - RID mesh = storage->multimesh_get_mesh(e->instance->base); - ERR_CONTINUE(!mesh.is_valid()); //should be a bug - - mesh_surface = storage->mesh_get_surface(e->instance->base, e->surface_index); - - primitive = storage->mesh_surface_get_primitive(mesh_surface); - - xforms_uniform_set = storage->multimesh_get_3d_uniform_set(e->instance->base, default_shader_rd, TRANSFORMS_UNIFORM_SET); - - } break; - case RS::INSTANCE_IMMEDIATE: { - ERR_CONTINUE(true); //should be a bug - } break; - case RS::INSTANCE_PARTICLES: { - RID mesh = storage->particles_get_draw_pass_mesh(e->instance->base, e->surface_index >> 16); - ERR_CONTINUE(!mesh.is_valid()); //should be a bug - - mesh_surface = storage->mesh_get_surface(e->instance->base, e->surface_index & 0xFFFF); - - primitive = storage->mesh_surface_get_primitive(mesh_surface); - - xforms_uniform_set = storage->particles_get_instance_buffer_uniform_set(e->instance->base, default_shader_rd, TRANSFORMS_UNIFORM_SET); - - } break; - default: { - ERR_CONTINUE(true); //should be a bug - } - } + RS::PrimitiveType primitive = surf->primitive; + RID xforms_uniform_set = surf->owner->transforms_uniform_set; ShaderVersion shader_version = SHADER_VERSION_MAX; // Assigned to silence wrong -Wmaybe-initialized. - switch (p_pass_mode) { + switch (p_params->pass_mode) { case PASS_MODE_COLOR: case PASS_MODE_COLOR_TRANSPARENT: { - if (e->uses_lightmap) { + if (element_info.uses_lightmap) { shader_version = SHADER_VERSION_LIGHTMAP_COLOR_PASS; - } else if (e->uses_forward_gi) { + } else if (element_info.uses_forward_gi) { shader_version = SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI; } else { shader_version = SHADER_VERSION_COLOR_PASS; } } break; case PASS_MODE_COLOR_SPECULAR: { - if (e->uses_lightmap) { + if (element_info.uses_lightmap) { shader_version = SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR; } else { shader_version = SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR; @@ -1086,42 +909,15 @@ void RendererSceneRenderForward::_render_list(RenderingDevice::DrawListID p_draw RID vertex_array_rd; RID index_array_rd; - if (mesh_surface) { - if (e->instance->mesh_instance.is_valid()) { //skeleton and blend shape - storage->mesh_instance_surface_get_vertex_arrays_and_format(e->instance->mesh_instance, e->surface_index, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); - } else { - storage->mesh_surface_get_vertex_arrays_and_format(mesh_surface, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); - } - - if (p_screen_lod_threshold > 0.0 && storage->mesh_surface_has_lod(mesh_surface)) { - Vector3 support_min = e->instance->transformed_aabb.get_support(-p_lod_plane.normal); - Vector3 support_max = e->instance->transformed_aabb.get_support(p_lod_plane.normal); - - float distance_min = p_lod_plane.distance_to(support_min); - float distance_max = p_lod_plane.distance_to(support_max); - - float distance = 0.0; - - if (distance_min * distance_max < 0.0) { - //crossing plane - distance = 0.0; - } else if (distance_min >= 0.0) { - distance = distance_min; - } else if (distance_max <= 0.0) { - distance = -distance_max; - } - - Vector3 model_scale_vec = e->instance->transform.basis.get_scale_abs(); - - float model_scale = MAX(model_scale_vec.x, MAX(model_scale_vec.y, model_scale_vec.z)); - - index_array_rd = storage->mesh_surface_get_index_array_with_lod(mesh_surface, model_scale * e->instance->lod_bias, distance * p_lod_distance_multiplier, p_screen_lod_threshold); - - } else { - index_array_rd = storage->mesh_surface_get_index_array(mesh_surface); - } + //skeleton and blend shape + if (surf->owner->mesh_instance.is_valid()) { + storage->mesh_instance_surface_get_vertex_arrays_and_format(surf->owner->mesh_instance, surf->surface_index, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); + } else { + storage->mesh_surface_get_vertex_arrays_and_format(mesh_surface, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); } + index_array_rd = storage->mesh_surface_get_index_array(mesh_surface, element_info.lod_index); + if (prev_vertex_array_rd != vertex_array_rd) { RD::get_singleton()->draw_list_bind_vertex_array(draw_list, vertex_array_rd); prev_vertex_array_rd = vertex_array_rd; @@ -1134,7 +930,7 @@ void RendererSceneRenderForward::_render_list(RenderingDevice::DrawListID p_draw prev_index_array_rd = index_array_rd; } - RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_force_wireframe); + RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe); if (pipeline_rd != prev_pipeline_rd) { // checking with prev shader does not make so much sense, as @@ -1148,40 +944,87 @@ void RendererSceneRenderForward::_render_list(RenderingDevice::DrawListID p_draw prev_xforms_uniform_set = xforms_uniform_set; } - if (material != prev_material) { + if (material_uniform_set != prev_material_uniform_set) { //update uniform set - if (material->uniform_set.is_valid()) { - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material->uniform_set, MATERIAL_UNIFORM_SET); + if (material_uniform_set.is_valid()) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_uniform_set, MATERIAL_UNIFORM_SET); } - prev_material = material; + prev_material_uniform_set = material_uniform_set; } - push_constant.index = i; - RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(PushConstant)); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(SceneState::PushConstant)); - switch (e->instance->base_type) { - case RS::INSTANCE_MESH: { - RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid()); - } break; - case RS::INSTANCE_MULTIMESH: { - uint32_t instances = storage->multimesh_get_instances_to_draw(e->instance->base); - RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), instances); - } break; - case RS::INSTANCE_IMMEDIATE: { - } break; - case RS::INSTANCE_PARTICLES: { - uint32_t instances = storage->particles_get_amount(e->instance->base); - RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), instances); - } break; - default: { - ERR_CONTINUE(true); //should be a bug - } - } + uint32_t instance_count = surf->owner->instance_count > 1 ? surf->owner->instance_count : element_info.repeat; + RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), instance_count); + i += element_info.repeat - 1; //skip equal elements } } -void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2 &p_screen_pixel_size, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers, bool p_pancake_shadows) { +void RendererSceneRenderForward::_render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element) { + //use template for faster performance (pass mode comparisons are inlined) + + switch (p_params->pass_mode) { + case PASS_MODE_COLOR: { + _render_list_template<PASS_MODE_COLOR>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_COLOR_SPECULAR: { + _render_list_template<PASS_MODE_COLOR_SPECULAR>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_COLOR_TRANSPARENT: { + _render_list_template<PASS_MODE_COLOR_TRANSPARENT>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_SHADOW: { + _render_list_template<PASS_MODE_SHADOW>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_SHADOW_DP: { + _render_list_template<PASS_MODE_SHADOW_DP>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH: { + _render_list_template<PASS_MODE_DEPTH>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS: { + _render_list_template<PASS_MODE_DEPTH_NORMAL_ROUGHNESS>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE: { + _render_list_template<PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH_MATERIAL: { + _render_list_template<PASS_MODE_DEPTH_MATERIAL>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_SDF: { + _render_list_template<PASS_MODE_SDF>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + } +} + +void RendererSceneRenderForward::_render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params) { + uint32_t render_total = p_params->element_count; + uint32_t total_threads = RendererThreadPool::singleton->thread_work_pool.get_thread_count(); + uint32_t render_from = p_thread * render_total / total_threads; + uint32_t render_to = (p_thread + 1 == total_threads) ? render_total : ((p_thread + 1) * render_total / total_threads); + _render_list(thread_draw_lists[p_thread], p_params->framebuffer_format, p_params, render_from, render_to); +} + +void RendererSceneRenderForward::_render_list_with_threads(RenderListParameters *p_params, RID p_framebuffer, RD::InitialAction p_initial_color_action, RD::FinalAction p_final_color_action, RD::InitialAction p_initial_depth_action, RD::FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values, float p_clear_depth, uint32_t p_clear_stencil, const Rect2 &p_region, const Vector<RID> &p_storage_textures) { + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(p_framebuffer); + p_params->framebuffer_format = fb_format; + + if ((uint32_t)p_params->element_count > render_list_thread_threshold && false) { // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); + RD::get_singleton()->draw_list_begin_split(p_framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); + RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RendererSceneRenderForward::_render_list_thread_function, p_params); + RD::get_singleton()->draw_list_end(p_params->barrier); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); + _render_list(draw_list, fb_format, p_params, 0, p_params->element_count); + RD::get_singleton()->draw_list_end(p_params->barrier); + } +} + +void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers, bool p_pancake_shadows, int p_index) { //CameraMatrix projection = p_cam_projection; //projection.flip_y(); // Vulkan and modern APIs use Y-Down CameraMatrix correction; @@ -1209,8 +1052,18 @@ void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_ren scene_state.ubo.penumbra_shadow_samples = penumbra_shadow_samples_get(); scene_state.ubo.soft_shadow_samples = soft_shadow_samples_get(); - scene_state.ubo.screen_pixel_size[0] = p_screen_pixel_size.x; - scene_state.ubo.screen_pixel_size[1] = p_screen_pixel_size.y; + Size2 screen_pixel_size = Vector2(1.0, 1.0) / Size2(p_screen_size); + scene_state.ubo.screen_pixel_size[0] = screen_pixel_size.x; + scene_state.ubo.screen_pixel_size[1] = screen_pixel_size.y; + + scene_state.ubo.cluster_shift = get_shift_from_power_of_2(p_cluster_size); + scene_state.ubo.max_cluster_element_count_div_32 = p_max_cluster_elements / 32; + { + uint32_t cluster_screen_width = (p_screen_size.width - 1) / p_cluster_size + 1; + uint32_t cluster_screen_height = (p_screen_size.height - 1) / p_cluster_size + 1; + scene_state.ubo.cluster_type_size = cluster_screen_width * cluster_screen_height * (scene_state.ubo.max_cluster_element_count_div_32 + 32); + scene_state.ubo.cluster_width = cluster_screen_width; + } if (p_shadow_atlas.is_valid()) { Vector2 sas = shadow_atlas_get_size(p_shadow_atlas); @@ -1410,276 +1263,332 @@ void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_ren scene_state.ubo.roughness_limiter_amount = screen_space_roughness_limiter_get_amount(); scene_state.ubo.roughness_limiter_limit = screen_space_roughness_limiter_get_limit(); - RD::get_singleton()->buffer_update(scene_state.uniform_buffer, 0, sizeof(SceneState::UBO), &scene_state.ubo, true); -} - -void RendererSceneRenderForward::_add_geometry(InstanceBase *p_instance, uint32_t p_surface, RID p_material, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi) { - RID m_src; - - m_src = p_instance->material_override.is_valid() ? p_instance->material_override : p_material; - - if (unlikely(get_debug_draw_mode() != RS::VIEWPORT_DEBUG_DRAW_DISABLED)) { - if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW) { - m_src = overdraw_material; - } else if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_LIGHTING) { - m_src = default_material; - } - } - - MaterialData *material = nullptr; - - if (m_src.is_valid()) { - material = (MaterialData *)storage->material_get_data(m_src, RendererStorageRD::SHADER_TYPE_3D); - if (!material || !material->shader_data->valid) { - material = nullptr; + if (p_index >= (int)scene_state.uniform_buffers.size()) { + uint32_t from = scene_state.uniform_buffers.size(); + scene_state.uniform_buffers.resize(p_index + 1); + render_pass_uniform_sets.resize(p_index + 1); + for (uint32_t i = from; i < scene_state.uniform_buffers.size(); i++) { + scene_state.uniform_buffers[i] = RD::get_singleton()->uniform_buffer_create(sizeof(SceneState::UBO)); } } + RD::get_singleton()->buffer_update(scene_state.uniform_buffers[p_index], 0, sizeof(SceneState::UBO), &scene_state.ubo, RD::BARRIER_MASK_RASTER); +} - if (!material) { - material = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); - m_src = default_material; - } - - ERR_FAIL_COND(!material); - - _add_geometry_with_material(p_instance, p_surface, material, m_src, p_pass_mode, p_geometry_index, p_using_sdfgi); - - while (material->next_pass.is_valid()) { - material = (MaterialData *)storage->material_get_data(material->next_pass, RendererStorageRD::SHADER_TYPE_3D); - if (!material || !material->shader_data->valid) { - break; +void RendererSceneRenderForward::_update_instance_data_buffer(RenderListType p_render_list) { + if (scene_state.instance_data[p_render_list].size() > 0) { + if (scene_state.instance_buffer[p_render_list] == RID() || scene_state.instance_buffer_size[p_render_list] < scene_state.instance_data[p_render_list].size()) { + if (scene_state.instance_buffer[p_render_list] != RID()) { + RD::get_singleton()->free(scene_state.instance_buffer[p_render_list]); + } + uint32_t new_size = nearest_power_of_2_templated(MAX(uint64_t(INSTANCE_DATA_BUFFER_MIN_SIZE), scene_state.instance_data[p_render_list].size())); + scene_state.instance_buffer[p_render_list] = RD::get_singleton()->storage_buffer_create(new_size * sizeof(SceneState::InstanceData)); + scene_state.instance_buffer_size[p_render_list] = new_size; } - _add_geometry_with_material(p_instance, p_surface, material, material->next_pass, p_pass_mode, p_geometry_index, p_using_sdfgi); + RD::get_singleton()->buffer_update(scene_state.instance_buffer[p_render_list], 0, sizeof(SceneState::InstanceData) * scene_state.instance_data[p_render_list].size(), scene_state.instance_data[p_render_list].ptr(), RD::BARRIER_MASK_RASTER); } } +void RendererSceneRenderForward::_fill_instance_data(RenderListType p_render_list, uint32_t p_offset, int32_t p_max_elements, bool p_update_buffer) { + RenderList *rl = &render_list[p_render_list]; + uint32_t element_total = p_max_elements >= 0 ? uint32_t(p_max_elements) : rl->elements.size(); -void RendererSceneRenderForward::_add_geometry_with_material(InstanceBase *p_instance, uint32_t p_surface, MaterialData *p_material, RID p_material_rid, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi) { - bool has_read_screen_alpha = p_material->shader_data->uses_screen_texture || p_material->shader_data->uses_depth_texture || p_material->shader_data->uses_normal_texture; - bool has_base_alpha = (p_material->shader_data->uses_alpha || has_read_screen_alpha); - bool has_blend_alpha = p_material->shader_data->uses_blend_alpha; - bool has_alpha = has_base_alpha || has_blend_alpha; + scene_state.instance_data[p_render_list].resize(p_offset + element_total); + rl->element_info.resize(p_offset + element_total); - if (p_material->shader_data->uses_sss) { - scene_state.used_sss = true; - } + uint32_t repeats = 0; + GeometryInstanceSurfaceDataCache *prev_surface = nullptr; + for (uint32_t i = 0; i < element_total; i++) { + GeometryInstanceSurfaceDataCache *surface = rl->elements[i + p_offset]; + GeometryInstanceForward *inst = surface->owner; - if (p_material->shader_data->uses_screen_texture) { - scene_state.used_screen_texture = true; - } + SceneState::InstanceData &instance_data = scene_state.instance_data[p_render_list][i + p_offset]; - if (p_material->shader_data->uses_depth_texture) { - scene_state.used_depth_texture = true; - } + if (inst->store_transform_cache) { + RendererStorageRD::store_transform(inst->transform, instance_data.transform); + } else { + RendererStorageRD::store_transform(Transform(), instance_data.transform); + } - if (p_material->shader_data->uses_normal_texture) { - scene_state.used_normal_texture = true; - } + instance_data.flags = inst->flags_cache; + instance_data.gi_offset = inst->gi_offset_cache; + instance_data.layer_mask = inst->layer_mask; + instance_data.instance_uniforms_ofs = uint32_t(inst->shader_parameters_offset); + instance_data.lightmap_uv_scale[0] = inst->lightmap_uv_scale.position.x; + instance_data.lightmap_uv_scale[1] = inst->lightmap_uv_scale.position.y; + instance_data.lightmap_uv_scale[2] = inst->lightmap_uv_scale.size.x; + instance_data.lightmap_uv_scale[3] = inst->lightmap_uv_scale.size.y; - if (p_pass_mode != PASS_MODE_COLOR && p_pass_mode != PASS_MODE_COLOR_SPECULAR) { - if (has_blend_alpha || has_read_screen_alpha || (has_base_alpha && !p_material->shader_data->uses_depth_pre_pass) || p_material->shader_data->depth_draw == ShaderData::DEPTH_DRAW_DISABLED || p_material->shader_data->depth_test == ShaderData::DEPTH_TEST_DISABLED || p_instance->cast_shadows == RS::SHADOW_CASTING_SETTING_OFF) { - //conditions in which no depth pass should be processed - return; - } + bool cant_repeat = instance_data.flags & INSTANCE_DATA_FLAG_MULTIMESH || inst->mesh_instance.is_valid(); - if ((p_pass_mode != PASS_MODE_DEPTH_MATERIAL && p_pass_mode != PASS_MODE_SDF) && !p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_pre_pass) { - //shader does not use discard and does not write a vertex position, use generic material - if (p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_DEPTH) { - p_material = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); - } else if ((p_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || p_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE) && !p_material->shader_data->uses_normal && !p_material->shader_data->uses_roughness) { - p_material = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); + if (prev_surface != nullptr && !cant_repeat && prev_surface->sort.sort_key1 == surface->sort.sort_key1 && prev_surface->sort.sort_key2 == surface->sort.sort_key2) { + //this element is the same as the previous one, count repeats to draw it using instancing + repeats++; + } else { + if (repeats > 0) { + for (uint32_t j = 1; j <= repeats; j++) { + rl->element_info[p_offset + i - j].repeat = j; + } } + repeats = 1; } - has_alpha = false; - } - - has_alpha = has_alpha || p_material->shader_data->depth_test == ShaderData::DEPTH_TEST_DISABLED; + RenderElementInfo &element_info = rl->element_info[p_offset + i]; - RenderList::Element *e = has_alpha ? render_list.add_alpha_element() : render_list.add_element(); + element_info.lod_index = surface->sort.lod_index; + element_info.uses_forward_gi = surface->sort.uses_forward_gi; + element_info.uses_lightmap = surface->sort.uses_lightmap; - if (!e) { - return; + if (cant_repeat) { + prev_surface = nullptr; + } else { + prev_surface = surface; + } } - e->instance = p_instance; - e->material = p_material; - e->surface_index = p_surface; - e->sort_key = 0; - - if (e->material->last_pass != render_pass) { - if (!RD::get_singleton()->uniform_set_is_valid(e->material->uniform_set)) { - //uniform set no longer valid, probably a texture changed - storage->material_force_update_textures(p_material_rid, RendererStorageRD::SHADER_TYPE_3D); - } - e->material->last_pass = render_pass; - e->material->index = scene_state.current_material_index++; - if (e->material->shader_data->last_pass != render_pass) { - e->material->shader_data->last_pass = scene_state.current_material_index++; - e->material->shader_data->index = scene_state.current_shader_index++; + if (repeats > 0) { + for (uint32_t j = 1; j <= repeats; j++) { + rl->element_info[p_offset + element_total - j].repeat = j; } } - e->geometry_index = p_geometry_index; - e->material_index = e->material->index; - e->uses_instancing = e->instance->base_type == RS::INSTANCE_MULTIMESH; - e->uses_lightmap = e->instance->lightmap != nullptr || !e->instance->lightmap_sh.is_empty(); - e->uses_forward_gi = has_alpha && (e->instance->gi_probe_instances.size() || p_using_sdfgi); - e->shader_index = e->shader_index; - e->depth_layer = e->instance->depth_layer; - e->priority = p_material->priority; - if (p_material->shader_data->uses_time) { - RenderingServerDefault::redraw_request(); + if (p_update_buffer) { + _update_instance_data_buffer(p_render_list); } } -void RendererSceneRenderForward::_fill_render_list(const PagedArray<InstanceBase *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi) { - scene_state.current_shader_index = 0; - scene_state.current_material_index = 0; - scene_state.used_sss = false; - scene_state.used_screen_texture = false; - scene_state.used_normal_texture = false; - scene_state.used_depth_texture = false; +void RendererSceneRenderForward::_fill_render_list(RenderListType p_render_list, const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi, bool p_using_opaque_gi, const Plane &p_lod_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_append) { + if (p_render_list == RENDER_LIST_OPAQUE) { + scene_state.used_sss = false; + scene_state.used_screen_texture = false; + scene_state.used_normal_texture = false; + scene_state.used_depth_texture = false; + } + uint32_t lightmap_captures_used = 0; Plane near_plane(p_cam_transform.origin, -p_cam_transform.basis.get_axis(Vector3::AXIS_Z)); near_plane.d += p_cam_projection.get_z_near(); float z_max = p_cam_projection.get_z_far() - p_cam_projection.get_z_near(); - uint32_t geometry_index = 0; + RenderList *rl = &render_list[p_render_list]; + _update_dirty_geometry_instances(); + + if (!p_append) { + rl->clear(); + if (p_render_list == RENDER_LIST_OPAQUE) { + render_list[RENDER_LIST_ALPHA].clear(); //opaque fills alpha too + } + } //fill list for (int i = 0; i < (int)p_instances.size(); i++) { - InstanceBase *inst = p_instances[i]; + GeometryInstanceForward *inst = static_cast<GeometryInstanceForward *>(p_instances[i]); - inst->depth = near_plane.distance_to(inst->transform.origin); - inst->depth_layer = CLAMP(int(inst->depth * 16 / z_max), 0, 15); + Vector3 support_min = inst->transformed_aabb.get_support(-near_plane.normal); + inst->depth = near_plane.distance_to(support_min); + uint32_t depth_layer = CLAMP(int(inst->depth * 16 / z_max), 0, 15); - //add geometry for drawing - switch (inst->base_type) { - case RS::INSTANCE_MESH: { - const RID *materials = nullptr; - uint32_t surface_count; + uint32_t flags = inst->base_flags; //fill flags if appropriate - materials = storage->mesh_get_surface_count_and_materials(inst->base, surface_count); - if (!materials) { - continue; //nothing to do - } + bool uses_lightmap = false; + bool uses_gi = false; - const RID *inst_materials = inst->materials.ptr(); + if (p_render_list == RENDER_LIST_OPAQUE) { + //setup GI - for (uint32_t j = 0; j < surface_count; j++) { - RID material = inst_materials[j].is_valid() ? inst_materials[j] : materials[j]; + if (inst->lightmap_instance.is_valid()) { + int32_t lightmap_cull_index = -1; + for (uint32_t j = 0; j < scene_state.lightmaps_used; j++) { + if (scene_state.lightmap_ids[j] == inst->lightmap_instance) { + lightmap_cull_index = j; + break; + } + } + if (lightmap_cull_index >= 0) { + inst->gi_offset_cache = inst->lightmap_slice_index << 16; + inst->gi_offset_cache |= lightmap_cull_index; + flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP; + if (scene_state.lightmap_has_sh[lightmap_cull_index]) { + flags |= INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP; + } + uses_lightmap = true; + } else { + inst->gi_offset_cache = 0xFFFFFFFF; + } - uint32_t surface_index = storage->mesh_surface_get_render_pass_index(inst->base, j, render_pass, &geometry_index); - _add_geometry(inst, j, material, p_pass_mode, surface_index, p_using_sdfgi); + } else if (inst->lightmap_sh) { + if (lightmap_captures_used < scene_state.max_lightmap_captures) { + const Color *src_capture = inst->lightmap_sh->sh; + LightmapCaptureData &lcd = scene_state.lightmap_captures[lightmap_captures_used]; + for (int j = 0; j < 9; j++) { + lcd.sh[j * 4 + 0] = src_capture[j].r; + lcd.sh[j * 4 + 1] = src_capture[j].g; + lcd.sh[j * 4 + 2] = src_capture[j].b; + lcd.sh[j * 4 + 3] = src_capture[j].a; + } + flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE; + inst->gi_offset_cache = lightmap_captures_used; + lightmap_captures_used++; + uses_lightmap = true; } - //mesh->last_pass=frame; + } else if (!low_end) { + if (p_using_opaque_gi) { + flags |= INSTANCE_DATA_FLAG_USE_GI_BUFFERS; + } - } break; + if (inst->gi_probes[0].is_valid()) { + uint32_t probe0_index = 0xFFFF; + uint32_t probe1_index = 0xFFFF; - case RS::INSTANCE_MULTIMESH: { - if (storage->multimesh_get_instances_to_draw(inst->base) == 0) { - //not visible, 0 instances - continue; - } + for (uint32_t j = 0; j < scene_state.giprobes_used; j++) { + if (scene_state.giprobe_ids[j] == inst->gi_probes[0]) { + probe0_index = j; + } else if (scene_state.giprobe_ids[j] == inst->gi_probes[1]) { + probe1_index = j; + } + } - RID mesh = storage->multimesh_get_mesh(inst->base); - if (!mesh.is_valid()) { - continue; + if (probe0_index == 0xFFFF && probe1_index != 0xFFFF) { + //0 must always exist if a probe exists + SWAP(probe0_index, probe1_index); + } + + inst->gi_offset_cache = probe0_index | (probe1_index << 16); + flags |= INSTANCE_DATA_FLAG_USE_GIPROBE; + uses_gi = true; + } else { + if (p_using_sdfgi && inst->can_sdfgi) { + flags |= INSTANCE_DATA_FLAG_USE_SDFGI; + uses_gi = true; + } + inst->gi_offset_cache = 0xFFFFFFFF; } + } + } + inst->flags_cache = flags; - const RID *materials = nullptr; - uint32_t surface_count; + GeometryInstanceSurfaceDataCache *surf = inst->surface_caches; - materials = storage->mesh_get_surface_count_and_materials(mesh, surface_count); - if (!materials) { - continue; //nothing to do - } + while (surf) { + surf->sort.uses_forward_gi = 0; + surf->sort.uses_lightmap = 0; - for (uint32_t j = 0; j < surface_count; j++) { - uint32_t surface_index = storage->mesh_surface_get_multimesh_render_pass_index(mesh, j, render_pass, &geometry_index); - _add_geometry(inst, j, materials[j], p_pass_mode, surface_index, p_using_sdfgi); - } + // LOD - } break; -#if 0 - case RS::INSTANCE_IMMEDIATE: { - RasterizerStorageGLES3::Immediate *immediate = storage->immediate_owner.getornull(inst->base); - ERR_CONTINUE(!immediate); + if (p_screen_lod_threshold > 0.0 && storage->mesh_surface_has_lod(surf->surface)) { + //lod + Vector3 lod_support_min = inst->transformed_aabb.get_support(-p_lod_plane.normal); + Vector3 lod_support_max = inst->transformed_aabb.get_support(p_lod_plane.normal); - _add_geometry(immediate, inst, nullptr, -1, p_depth_pass, p_shadow_pass); + float distance_min = p_lod_plane.distance_to(lod_support_min); + float distance_max = p_lod_plane.distance_to(lod_support_max); - } break; -#endif - case RS::INSTANCE_PARTICLES: { - int draw_passes = storage->particles_get_draw_passes(inst->base); + float distance = 0.0; - for (int j = 0; j < draw_passes; j++) { - RID mesh = storage->particles_get_draw_pass_mesh(inst->base, j); - if (!mesh.is_valid()) - continue; + if (distance_min * distance_max < 0.0) { + //crossing plane + distance = 0.0; + } else if (distance_min >= 0.0) { + distance = distance_min; + } else if (distance_max <= 0.0) { + distance = -distance_max; + } - const RID *materials = nullptr; - uint32_t surface_count; + surf->sort.lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_lod_distance_multiplier, p_screen_lod_threshold); + } else { + surf->sort.lod_index = 0; + } - materials = storage->mesh_get_surface_count_and_materials(mesh, surface_count); - if (!materials) { - continue; //nothing to do + // ADD Element + if (p_pass_mode == PASS_MODE_COLOR) { + if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { + rl->add_element(surf); + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA) { + render_list[RENDER_LIST_ALPHA].add_element(surf); + if (uses_gi) { + surf->sort.uses_forward_gi = 1; } + } - for (uint32_t k = 0; k < surface_count; k++) { - uint32_t surface_index = storage->mesh_surface_get_particles_render_pass_index(mesh, j, render_pass, &geometry_index); - _add_geometry(inst, (j << 16) | k, materials[j], p_pass_mode, surface_index, p_using_sdfgi); - } + if (uses_lightmap) { + surf->sort.uses_lightmap = 1; } - } break; + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_SUBSURFACE_SCATTERING) { + scene_state.used_sss = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_SCREEN_TEXTURE) { + scene_state.used_screen_texture = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_NORMAL_TEXTURE) { + scene_state.used_normal_texture = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DEPTH_TEXTURE) { + scene_state.used_depth_texture = true; + } - default: { + } else if (p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) { + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW) { + rl->add_element(surf); + } + } else { + if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { + rl->add_element(surf); + } } + + surf->sort.depth_layer = depth_layer; + + surf = surf->next; } } + + if (p_render_list == RENDER_LIST_OPAQUE && lightmap_captures_used) { + RD::get_singleton()->buffer_update(scene_state.lightmap_capture_buffer, 0, sizeof(LightmapCaptureData) * lightmap_captures_used, scene_state.lightmap_captures, RD::BARRIER_MASK_RASTER); + } +} + +void RendererSceneRenderForward::_setup_giprobes(const PagedArray<RID> &p_giprobes) { + scene_state.giprobes_used = MIN(p_giprobes.size(), uint32_t(MAX_GI_PROBES)); + for (uint32_t i = 0; i < scene_state.giprobes_used; i++) { + scene_state.giprobe_ids[i] = p_giprobes[i]; + } } -void RendererSceneRenderForward::_setup_lightmaps(const PagedArray<InstanceBase *> &p_lightmaps, const Transform &p_cam_transform) { - uint32_t lightmaps_used = 0; +void RendererSceneRenderForward::_setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform &p_cam_transform) { + scene_state.lightmaps_used = 0; for (int i = 0; i < (int)p_lightmaps.size(); i++) { if (i >= (int)scene_state.max_lightmaps) { break; } - InstanceBase *lm = p_lightmaps[i]; - Basis to_lm = lm->transform.basis.inverse() * p_cam_transform.basis; + RID lightmap = lightmap_instance_get_lightmap(p_lightmaps[i]); + + Basis to_lm = lightmap_instance_get_transform(p_lightmaps[i]).basis.inverse() * p_cam_transform.basis; to_lm = to_lm.inverse().transposed(); //will transform normals RendererStorageRD::store_transform_3x3(to_lm, scene_state.lightmaps[i].normal_xform); - lm->lightmap_cull_index = i; - lightmaps_used++; + scene_state.lightmap_ids[i] = p_lightmaps[i]; + scene_state.lightmap_has_sh[i] = storage->lightmap_uses_spherical_harmonics(lightmap); + + scene_state.lightmaps_used++; } - if (lightmaps_used > 0) { - RD::get_singleton()->buffer_update(scene_state.lightmap_buffer, 0, sizeof(LightmapData) * lightmaps_used, scene_state.lightmaps, true); + if (scene_state.lightmaps_used > 0) { + RD::get_singleton()->buffer_update(scene_state.lightmap_buffer, 0, sizeof(LightmapData) * scene_state.lightmaps_used, scene_state.lightmaps, RD::BARRIER_MASK_RASTER); } } -void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_screen_lod_threshold) { +void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_screen_lod_threshold) { RenderBufferDataForward *render_buffer = nullptr; if (p_render_buffer.is_valid()) { render_buffer = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffer); } //first of all, make a new render pass - render_pass++; - //fill up ubo RENDER_TIMESTAMP("Setup 3D Scene"); - if (p_reflection_probe.is_valid()) { - scene_state.ubo.reflection_multiplier = 0.0; - } else { - scene_state.ubo.reflection_multiplier = 1.0; - } - float lod_distance_multiplier = p_cam_projection.get_lod_multiplier(); Plane lod_camera_plane(p_cam_transform.get_origin(), -p_cam_transform.basis.get_axis(Vector3::AXIS_Z)); @@ -1692,9 +1601,8 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf Vector2 vp_he = p_cam_projection.get_viewport_half_extents(); scene_state.ubo.viewport_size[0] = vp_he.x; scene_state.ubo.viewport_size[1] = vp_he.y; - scene_state.ubo.directional_light_count = p_directional_light_count; + scene_state.ubo.directional_light_count = 0; - Size2 screen_pixel_size; Size2i screen_size; RID opaque_framebuffer; RID opaque_specular_framebuffer; @@ -1709,8 +1617,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf bool using_giprobe = false; if (render_buffer) { - screen_pixel_size.width = 1.0 / render_buffer->width; - screen_pixel_size.height = 1.0 / render_buffer->height; screen_size.x = render_buffer->width; screen_size.y = render_buffer->height; @@ -1718,7 +1624,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf if (!low_end && p_gi_probes.size() > 0) { using_giprobe = true; - render_buffer->ensure_gi(); } if (!p_environment.is_valid() && using_giprobe) { @@ -1728,7 +1633,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf if (environment_is_sdfgi_enabled(p_environment)) { depth_pass_mode = using_giprobe ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; // also giprobe using_sdfgi = true; - render_buffer->ensure_gi(); } else { depth_pass_mode = using_giprobe ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; } @@ -1767,8 +1671,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf alpha_framebuffer = opaque_framebuffer; } else if (p_reflection_probe.is_valid()) { uint32_t resolution = reflection_probe_instance_get_resolution(p_reflection_probe); - screen_pixel_size.width = 1.0 / resolution; - screen_pixel_size.height = 1.0 / resolution; screen_size.x = resolution; screen_size.y = resolution; @@ -1783,13 +1685,21 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf ERR_FAIL(); //bug? } + RD::get_singleton()->draw_command_begin_label("Render Setup"); + _setup_lightmaps(p_lightmaps, p_cam_transform); - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_pixel_size, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); + _setup_giprobes(p_gi_probes); + _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) - render_list.clear(); - _fill_render_list(p_instances, PASS_MODE_COLOR, p_cam_projection, p_cam_transform, using_sdfgi); + _fill_render_list(RENDER_LIST_OPAQUE, p_instances, PASS_MODE_COLOR, p_cam_projection, p_cam_transform, using_sdfgi, using_sdfgi || using_giprobe, lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + render_list[RENDER_LIST_OPAQUE].sort_by_key(); + render_list[RENDER_LIST_ALPHA].sort_by_depth(); + _fill_instance_data(RENDER_LIST_OPAQUE); + _fill_instance_data(RENDER_LIST_ALPHA); + + RD::get_singleton()->draw_command_end_label(); bool using_sss = !low_end && render_buffer && scene_state.used_sss && sub_surface_scattering_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED; @@ -1849,6 +1759,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf // setup sky if used for ambient, reflections, or background if (draw_sky || draw_sky_fog_only || environment_get_reflection_source(p_environment) == RS::ENV_REFLECTION_SOURCE_SKY || environment_get_ambient_source(p_environment) == RS::ENV_AMBIENT_SOURCE_SKY) { RENDER_TIMESTAMP("Setup Sky"); + RD::get_singleton()->draw_command_begin_label("Setup Sky"); CameraMatrix projection = p_cam_projection; if (p_reflection_probe.is_valid()) { CameraMatrix correction; @@ -1866,58 +1777,77 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf // do not try to draw sky if invalid draw_sky = false; } + RD::get_singleton()->draw_command_end_label(); } } else { clear_color = p_default_bg_color; } - render_list.sort_by_key(false); - - _fill_instances(render_list.elements, render_list.element_count, false, false, using_sdfgi || using_giprobe); - bool debug_giprobes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_ALBEDO || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION; bool debug_sdfgi_probes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_SDFGI_PROBES; - bool depth_pre_pass = !low_end && depth_framebuffer.is_valid(); bool using_ssao = depth_pre_pass && p_render_buffer.is_valid() && p_environment.is_valid() && environment_is_ssao_enabled(p_environment); bool continue_depth = false; if (depth_pre_pass) { //depth pre pass - RENDER_TIMESTAMP("Render Depth Pre-Pass"); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>()); + bool needs_pre_resolve = _needs_post_prepass_render(using_sdfgi || using_giprobe); + if (needs_pre_resolve) { + RENDER_TIMESTAMP("GI + Render Depth Pre-Pass (parallel)"); + } else { + RENDER_TIMESTAMP("Render Depth Pre-Pass"); + } + if (needs_pre_resolve) { + //pre clear the depth framebuffer, as AMD (and maybe others?) use compute for it, and barrier other compute shaders. + RD::get_singleton()->draw_list_begin(depth_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE, depth_pass_clear); + RD::get_singleton()->draw_list_end(); + //start compute processes here, so they run at the same time as depth pre-pass + _post_prepass_render(using_sdfgi || using_giprobe); + } + + RD::get_singleton()->draw_command_begin_label("Render Depth Pre-Pass"); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); bool finish_depth = using_ssao || using_sdfgi || using_giprobe; - RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(depth_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, finish_depth ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE, depth_pass_clear); - _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(depth_framebuffer), render_list.elements, render_list.element_count, false, depth_pass_mode, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); - RD::get_singleton()->draw_list_end(); + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), false, depth_pass_mode, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + _render_list_with_threads(&render_list_params, depth_framebuffer, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, finish_depth ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE, needs_pre_resolve ? Vector<Color>() : depth_pass_clear); + + RD::get_singleton()->draw_command_end_label(); + + if (needs_pre_resolve) { + _pre_resolve_render(using_sdfgi || using_giprobe); + } if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { RENDER_TIMESTAMP("Resolve Depth Pre-Pass"); + RD::get_singleton()->draw_command_begin_label("Resolve Depth Pre-Pass"); if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE) { + if (needs_pre_resolve) { + RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, RD::BARRIER_MASK_COMPUTE); + } static int texture_samples[RS::VIEWPORT_MSAA_MAX] = { 1, 2, 4, 8, 16 }; storage->get_effects()->resolve_gi(render_buffer->depth_msaa, render_buffer->normal_roughness_buffer_msaa, using_giprobe ? render_buffer->giprobe_buffer_msaa : RID(), render_buffer->depth, render_buffer->normal_roughness_buffer, using_giprobe ? render_buffer->giprobe_buffer : RID(), Vector2i(render_buffer->width, render_buffer->height), texture_samples[render_buffer->msaa]); } else if (finish_depth) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); } + RD::get_singleton()->draw_command_end_label(); } continue_depth = !finish_depth; } - if (using_ssao) { - _process_ssao(p_render_buffer, p_environment, render_buffer->normal_roughness_buffer, p_cam_projection); - } + _pre_opaque_render(using_ssao, using_sdfgi || using_giprobe, render_buffer ? render_buffer->normal_roughness_buffer : RID(), render_buffer ? render_buffer->giprobe_buffer : RID()); - if (using_sdfgi || using_giprobe) { - _process_gi(p_render_buffer, render_buffer->normal_roughness_buffer, render_buffer->ambient_buffer, render_buffer->reflection_buffer, render_buffer->giprobe_buffer, p_environment, p_cam_projection, p_cam_transform, p_gi_probes); - } + RD::get_singleton()->draw_command_begin_label("Render Opaque Pass"); + + scene_state.ubo.directional_light_count = _get_render_state_directional_light_count(); - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_pixel_size, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), p_render_buffer.is_valid()); + _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), p_render_buffer.is_valid()); RENDER_TIMESTAMP("Render Opaque Pass"); - RID rp_uniform_set = _setup_render_pass_uniform_set(p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_gi_probes); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps, true); bool can_continue_color = !scene_state.used_screen_texture && !using_ssr && !using_sss; bool can_continue_depth = !scene_state.used_depth_texture && !using_ssr && !using_sss; @@ -1938,17 +1868,17 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf } RID framebuffer = using_separate_specular ? opaque_specular_framebuffer : opaque_framebuffer; - RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, depth_pre_pass ? (continue_depth ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CONTINUE) : RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); - _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(framebuffer), render_list.elements, render_list.element_count, false, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); - RD::get_singleton()->draw_list_end(); - + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), false, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + _render_list_with_threads(&render_list_params, framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, depth_pre_pass ? (continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP) : RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); if (will_continue_color && using_separate_specular) { // close the specular framebuffer, as it's no longer used - draw_list = RD::get_singleton()->draw_list_begin(render_buffer->specular_only_fb, RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_CONTINUE); + RD::get_singleton()->draw_list_begin(render_buffer->specular_only_fb, RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_CONTINUE); RD::get_singleton()->draw_list_end(); } } + RD::get_singleton()->draw_command_end_label(); + if (debug_giprobes) { //debug giprobes bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only); @@ -1958,9 +1888,11 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf dc.set_depth_correction(true); CameraMatrix cm = (dc * p_cam_projection) * CameraMatrix(p_cam_transform.affine_inverse()); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(opaque_framebuffer, RD::INITIAL_ACTION_CONTINUE, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); + RD::get_singleton()->draw_command_begin_label("Debug GIProbes"); for (int i = 0; i < (int)p_gi_probes.size(); i++) { _debug_giprobe(p_gi_probes[i], draw_list, opaque_framebuffer, cm, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION, 1.0); } + RD::get_singleton()->draw_command_end_label(); RD::get_singleton()->draw_list_end(); } @@ -1973,7 +1905,9 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf dc.set_depth_correction(true); CameraMatrix cm = (dc * p_cam_projection) * CameraMatrix(p_cam_transform.affine_inverse()); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(opaque_framebuffer, RD::INITIAL_ACTION_CONTINUE, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); + RD::get_singleton()->draw_command_begin_label("Debug SDFGI"); _debug_sdfgi_probes(p_render_buffer, draw_list, opaque_framebuffer, cm); + RD::get_singleton()->draw_command_end_label(); RD::get_singleton()->draw_list_end(); } @@ -1986,30 +1920,35 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf correction.set_depth_correction(true); projection = correction * p_cam_projection; } - + RD::get_singleton()->draw_command_begin_label("Draw Sky"); _draw_sky(can_continue_color, can_continue_depth, opaque_framebuffer, p_environment, projection, p_cam_transform); + RD::get_singleton()->draw_command_end_label(); } if (render_buffer && !can_continue_color && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); if (using_separate_specular) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->specular_msaa, render_buffer->specular, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->specular_msaa, render_buffer->specular); } } if (render_buffer && !can_continue_depth && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); } if (using_separate_specular) { if (using_sss) { RENDER_TIMESTAMP("Sub Surface Scattering"); + RD::get_singleton()->draw_command_begin_label("Process Sub Surface Scattering"); _process_sss(p_render_buffer, p_cam_projection); + RD::get_singleton()->draw_command_end_label(); } if (using_ssr) { RENDER_TIMESTAMP("Screen Space Reflection"); + RD::get_singleton()->draw_command_begin_label("Process Screen Space Reflections"); _process_ssr(p_render_buffer, render_buffer->color_fb, render_buffer->normal_roughness_buffer, render_buffer->specular, render_buffer->specular, Color(0, 0, 0, 1), p_environment, p_cam_projection, render_buffer->msaa == RS::VIEWPORT_MSAA_DISABLED); + RD::get_singleton()->draw_command_end_label(); } else { //just mix specular back RENDER_TIMESTAMP("Merge Specular"); @@ -2019,119 +1958,161 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf RENDER_TIMESTAMP("Render Transparent Pass"); - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_pixel_size, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); + RD::get_singleton()->draw_command_begin_label("Render Transparent Pass"); - render_list.sort_by_reverse_depth_and_priority(true); + rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps, true); - _fill_instances(&render_list.elements[render_list.max_elements - render_list.alpha_element_count], render_list.alpha_element_count, false, using_sdfgi); + _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); { - RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(alpha_framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); - _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(alpha_framebuffer), &render_list.elements[render_list.max_elements - render_list.alpha_element_count], render_list.alpha_element_count, false, PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); - RD::get_singleton()->draw_list_end(); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), false, PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + _render_list_with_threads(&render_list_params, alpha_framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); } + RD::get_singleton()->draw_command_end_label(); + + RD::get_singleton()->draw_command_begin_label("Resolve"); + if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); } -} -void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedArray<InstanceBase *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) { - RENDER_TIMESTAMP("Setup Rendering Shadow"); + RD::get_singleton()->draw_command_end_label(); +} +void RendererSceneRenderForward::_render_shadow_begin() { + scene_state.shadow_passes.clear(); + RD::get_singleton()->draw_command_begin_label("Shadow Setup"); _update_render_base_uniform_set(); - render_pass++; + render_list[RENDER_LIST_SECONDARY].clear(); + scene_state.instance_data[RENDER_LIST_SECONDARY].clear(); +} +void RendererSceneRenderForward::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end) { + uint32_t shadow_pass_index = scene_state.shadow_passes.size(); + + SceneState::ShadowPass shadow_pass; scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; - _setup_environment(RID(), RID(), p_projection, p_transform, RID(), true, Vector2(1, 1), RID(), true, Color(), 0, p_zfar, false, p_use_pancake); + _setup_environment(RID(), RID(), p_projection, p_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), !p_flip_y, Color(), 0, p_zfar, false, p_use_pancake, shadow_pass_index); if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { p_screen_lod_threshold = 0.0; } - render_list.clear(); - PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW; - _fill_render_list(p_instances, pass_mode, p_projection, p_transform); + uint32_t render_list_from = render_list[RENDER_LIST_SECONDARY].elements.size(); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_projection, p_transform, false, false, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, true); + uint32_t render_list_size = render_list[RENDER_LIST_SECONDARY].elements.size() - render_list_from; + render_list[RENDER_LIST_SECONDARY].sort_by_key_range(render_list_from, render_list_size); + _fill_instance_data(RENDER_LIST_SECONDARY, render_list_from, render_list_size, false); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>()); + { + //regular forward for now + bool flip_cull = p_use_dp_flip; + if (p_flip_y) { + flip_cull = !flip_cull; + } - RENDER_TIMESTAMP("Render Shadow"); + shadow_pass.element_from = render_list_from; + shadow_pass.element_count = render_list_size; + shadow_pass.flip_cull = flip_cull; + shadow_pass.pass_mode = pass_mode; - render_list.sort_by_key(false); + shadow_pass.rp_uniform_set = RID(); //will be filled later when instance buffer is complete + shadow_pass.camera_plane = p_camera_plane; + shadow_pass.screen_lod_threshold = p_screen_lod_threshold; + shadow_pass.lod_distance_multiplier = p_lod_distance_multiplier; - _fill_instances(render_list.elements, render_list.element_count, true); + shadow_pass.framebuffer = p_framebuffer; + shadow_pass.initial_depth_action = p_begin ? (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION : RD::INITIAL_ACTION_CLEAR) : (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION_CONTINUE : RD::INITIAL_ACTION_CONTINUE); + shadow_pass.final_depth_action = p_end ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE; + shadow_pass.rect = p_rect; - { - //regular forward for now - RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); - _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), render_list.elements, render_list.element_count, p_use_dp_flip, pass_mode, true, rp_uniform_set, false, Vector2(), p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold); - RD::get_singleton()->draw_list_end(); + scene_state.shadow_passes.push_back(shadow_pass); } } -void RendererSceneRenderForward::_render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<InstanceBase *> &p_instances) { - RENDER_TIMESTAMP("Setup Render Collider Heightfield"); +void RendererSceneRenderForward::_render_shadow_process() { + _update_instance_data_buffer(RENDER_LIST_SECONDARY); + //render shadows one after the other, so this can be done un-barriered and the driver can optimize (as well as allow us to run compute at the same time) - _update_render_base_uniform_set(); + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + //render passes need to be configured after instance buffer is done, since they need the latest version + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + shadow_pass.rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>(), false, i); + } - render_pass++; + RD::get_singleton()->draw_command_end_label(); +} +void RendererSceneRenderForward::_render_shadow_end(uint32_t p_barrier) { + RD::get_singleton()->draw_command_begin_label("Shadow Render"); - scene_state.ubo.dual_paraboloid_side = 0; + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, true, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); + _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); + } - _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), RID(), true, Color(), 0, p_cam_projection.get_z_far(), false, false); + if (p_barrier != RD::BARRIER_MASK_NO_BARRIER) { + RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, p_barrier); + } + RD::get_singleton()->draw_command_end_label(); +} - render_list.clear(); +void RendererSceneRenderForward::_render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) { + RENDER_TIMESTAMP("Setup Render Collider Heightfield"); - PassMode pass_mode = PASS_MODE_SHADOW; + RD::get_singleton()->draw_command_begin_label("Render Collider Heightfield"); - _fill_render_list(p_instances, pass_mode, p_cam_projection, p_cam_transform); + _update_render_base_uniform_set(); + scene_state.ubo.dual_paraboloid_side = 0; + + _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), true, Color(), 0, p_cam_projection.get_z_far(), false, false); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>()); + PassMode pass_mode = PASS_MODE_SHADOW; - RENDER_TIMESTAMP("Render Collider Heightield"); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_cam_projection, p_cam_transform); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); - render_list.sort_by_key(false); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); - _fill_instances(render_list.elements, render_list.element_count, true); + RENDER_TIMESTAMP("Render Collider Heightfield"); { //regular forward for now - RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); - _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_fb), render_list.elements, render_list.element_count, false, pass_mode, true, rp_uniform_set); - RD::get_singleton()->draw_list_end(); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, true, rp_uniform_set); + _render_list_with_threads(&render_list_params, p_fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); } + RD::get_singleton()->draw_command_end_label(); } -void RendererSceneRenderForward::_render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { +void RendererSceneRenderForward::_render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { RENDER_TIMESTAMP("Setup Rendering Material"); - _update_render_base_uniform_set(); + RD::get_singleton()->draw_command_begin_label("Render Material"); - render_pass++; + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; - scene_state.ubo.material_uv2_mode = true; - - _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), RID(), false, Color(), 0, 0); + scene_state.ubo.material_uv2_mode = false; - render_list.clear(); + _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; - _fill_render_list(p_instances, pass_mode, p_cam_projection, p_cam_transform); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_cam_projection, p_cam_transform); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); RENDER_TIMESTAMP("Render Material"); - render_list.sort_by_key(false); - - _fill_instances(render_list.elements, render_list.element_count, true); - { + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -2140,37 +2121,36 @@ void RendererSceneRenderForward::_render_material(const Transform &p_cam_transfo clear.push_back(Color(0, 0, 0, 0)); clear.push_back(Color(0, 0, 0, 0)); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, clear, 1.0, 0, p_region); - _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set); + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); RD::get_singleton()->draw_list_end(); } + + RD::get_singleton()->draw_command_end_label(); } -void RendererSceneRenderForward::_render_uv2(const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { +void RendererSceneRenderForward::_render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { RENDER_TIMESTAMP("Setup Rendering UV2"); - _update_render_base_uniform_set(); + RD::get_singleton()->draw_command_begin_label("Render UV2"); - render_pass++; + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; scene_state.ubo.material_uv2_mode = true; - _setup_environment(RID(), RID(), CameraMatrix(), Transform(), RID(), true, Vector2(1, 1), RID(), false, Color(), 0, 0); - - render_list.clear(); + _setup_environment(RID(), RID(), CameraMatrix(), Transform(), RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; - _fill_render_list(p_instances, pass_mode, CameraMatrix(), Transform()); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, CameraMatrix(), Transform()); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); RENDER_TIMESTAMP("Render Material"); - render_list.sort_by_key(false); - - _fill_instances(render_list.elements, render_list.element_count, true); - { + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set, true); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -2198,31 +2178,32 @@ void RendererSceneRenderForward::_render_uv2(const PagedArray<InstanceBase *> &p Vector2 ofs = uv_offsets[i]; ofs.x /= p_region.size.width; ofs.y /= p_region.size.height; - _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set, true, ofs); //first wireframe, for pseudo conservative + render_list_params.uv_offset = ofs; + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //first wireframe, for pseudo conservative } - _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set, false); //second regular triangles + render_list_params.uv_offset = Vector2(); + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //second regular triangles RD::get_singleton()->draw_list_end(); } + + RD::get_singleton()->draw_command_end_label(); } -void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<InstanceBase *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) { +void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) { RENDER_TIMESTAMP("Render SDFGI"); + RD::get_singleton()->draw_command_begin_label("Render SDFGI Voxel"); + _update_render_base_uniform_set(); RenderBufferDataForward *render_buffer = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); ERR_FAIL_COND(!render_buffer); - render_pass++; - render_list.clear(); - PassMode pass_mode = PASS_MODE_SDF; - _fill_render_list(p_instances, pass_mode, CameraMatrix(), Transform()); - render_list.sort_by_key(false); - _fill_instances(render_list.elements, render_list.element_count, true); - - RID rp_uniform_set = _setup_sdfgi_render_pass_uniform_set(p_albedo_texture, p_emission_texture, p_emission_aniso_texture, p_geom_facing_texture); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, CameraMatrix(), Transform()); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); Vector3 half_extents = p_bounds.size * 0.5; Vector3 center = p_bounds.position + half_extents; @@ -2273,7 +2254,9 @@ void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vecto RendererStorageRD::store_transform(to_bounds.affine_inverse() * cam_xform, scene_state.ubo.sdf_to_bounds); - _setup_environment(RID(), RID(), camera_proj, cam_xform, RID(), true, Vector2(1, 1), RID(), false, Color(), 0, 0); + _setup_environment(RID(), RID(), camera_proj, cam_xform, RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); + + RID rp_uniform_set = _setup_sdfgi_render_pass_uniform_set(p_albedo_texture, p_emission_texture, p_emission_aniso_texture, p_geom_facing_texture); Map<Size2i, RID>::Element *E = sdfgi_framebuffer_size_cache.find(fb_size); if (!E) { @@ -2281,10 +2264,11 @@ void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vecto E = sdfgi_framebuffer_size_cache.insert(fb_size, fb); } - RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(E->get(), RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, Vector<Color>(), 1.0, 0, Rect2(), sbs); - _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(E->get()), render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set, false); //second regular triangles - RD::get_singleton()->draw_list_end(); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set, false); + _render_list_with_threads(&render_list_params, E->get(), RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, Vector<Color>(), 1.0, 0, Rect2(), sbs); } + + RD::get_singleton()->draw_command_end_label(); } void RendererSceneRenderForward::_base_uniforms_changed() { @@ -2336,15 +2320,15 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { { RD::Uniform u; u.binding = 3; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.ids.push_back(scene_state.uniform_buffer); + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(get_omni_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; u.binding = 4; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.ids.push_back(scene_state.instance_buffer); + u.ids.push_back(get_spot_light_buffer()); uniforms.push_back(u); } @@ -2352,48 +2336,33 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { RD::Uniform u; u.binding = 5; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.ids.push_back(get_positional_light_buffer()); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.binding = 6; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_reflection_probe_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 7; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.ids.push_back(get_directional_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 10; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 11; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.ids = storage->lightmap_array_get_textures(); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 12; + u.binding = 8; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_capture_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 13; + u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture(); u.ids.push_back(decal_atlas); @@ -2401,7 +2370,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 14; + u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture_srgb(); u.ids.push_back(decal_atlas); @@ -2409,7 +2378,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 15; + u.binding = 11; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_decal_buffer()); uniforms.push_back(u); @@ -2417,35 +2386,8 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { { RD::Uniform u; - u.binding = 16; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.ids.push_back(get_cluster_builder_texture()); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 17; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.ids.push_back(get_cluster_builder_indices_buffer()); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.binding = 18; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - if (directional_shadow_get_texture().is_valid()) { - u.ids.push_back(directional_shadow_get_texture()); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); - } - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 19; + u.binding = 12; u.ids.push_back(storage->global_variables_get_storage_buffer()); uniforms.push_back(u); } @@ -2453,7 +2395,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { if (!low_end) { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 20; + u.binding = 13; u.ids.push_back(sdfgi_get_ubo()); uniforms.push_back(u); } @@ -2462,10 +2404,9 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { } } -RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, const PagedArray<RID> &p_gi_probes) { - if (render_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_set)) { - RD::get_singleton()->free(render_pass_uniform_set); - } +RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RenderListType p_render_list, RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, bool p_use_directional_shadow_atlas, int p_index) { + //there should always be enough uniform buffers for render passes, otherwise bugs + ERR_FAIL_INDEX_V(p_index, (int)scene_state.uniform_buffers.size(), RID()); RenderBufferDataForward *rb = nullptr; if (p_render_buffers.is_valid()) { @@ -2477,6 +2418,24 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff Vector<RD::Uniform> uniforms; { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(scene_state.uniform_buffers[p_index]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID instance_buffer = scene_state.instance_buffer[p_render_list]; + if (instance_buffer == RID()) { + instance_buffer = default_vec4_xform_buffer; // any buffer will do since its not used + } + u.ids.push_back(instance_buffer); + uniforms.push_back(u); + } + { RID radiance_texture; if (p_radiance_texture.is_valid()) { radiance_texture = p_radiance_texture; @@ -2484,7 +2443,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff radiance_texture = storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); } RD::Uniform u; - u.binding = 0; + u.binding = 2; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.push_back(radiance_texture); uniforms.push_back(u); @@ -2493,7 +2452,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RID ref_texture = p_reflection_atlas.is_valid() ? reflection_atlas_get_texture(p_reflection_atlas) : RID(); RD::Uniform u; - u.binding = 1; + u.binding = 3; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; if (ref_texture.is_valid()) { u.ids.push_back(ref_texture); @@ -2505,7 +2464,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 2; + u.binding = 4; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture; if (p_shadow_atlas.is_valid()) { @@ -2517,10 +2476,39 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff u.ids.push_back(texture); uniforms.push_back(u); } + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + if (p_use_directional_shadow_atlas && directional_shadow_get_texture().is_valid()) { + u.ids.push_back(directional_shadow_get_texture()); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.resize(scene_state.max_lightmaps); + RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + for (uint32_t i = 0; i < scene_state.max_lightmaps; i++) { + if (i < p_lightmaps.size()) { + RID base = lightmap_instance_get_lightmap(p_lightmaps[i]); + RID texture = storage->lightmap_get_texture(base); + RID rd_texture = storage->texture_get_rd_texture(texture); + u.ids.write[i] = rd_texture; + } else { + u.ids.write[i] = default_tex; + } + } + uniforms.push_back(u); + } { RD::Uniform u; - u.binding = 3; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(MAX_GI_PROBES); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); @@ -2541,7 +2529,16 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 4; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID cb = p_cluster_buffer.is_valid() ? p_cluster_buffer : default_vec4_xform_buffer; + u.ids.push_back(cb); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture = (false && rb && rb->depth.is_valid()) ? rb->depth : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); u.ids.push_back(texture); @@ -2549,17 +2546,18 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 5; + u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID bbt = rb ? render_buffers_get_back_buffer_texture(p_render_buffers) : RID(); RID texture = bbt.is_valid() ? bbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); } + if (!low_end) { { RD::Uniform u; - u.binding = 6; + u.binding = 11; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture = rb && rb->normal_roughness_buffer.is_valid() ? rb->normal_roughness_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_NORMAL); u.ids.push_back(texture); @@ -2568,7 +2566,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 7; + u.binding = 12; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID aot = rb ? render_buffers_get_ao_texture(p_render_buffers) : RID(); RID texture = aot.is_valid() ? aot : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); @@ -2578,24 +2576,26 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 8; + u.binding = 13; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID texture = rb && rb->ambient_buffer.is_valid() ? rb->ambient_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + RID ambient_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_ambient_texture(p_render_buffers) : RID(); + RID texture = ambient_buffer.is_valid() ? ambient_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 9; + u.binding = 14; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID texture = rb && rb->reflection_buffer.is_valid() ? rb->reflection_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + RID reflection_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_reflection_texture(p_render_buffers) : RID(); + RID texture = reflection_buffer.is_valid() ? reflection_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 10; + u.binding = 15; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID t; if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) { @@ -2608,7 +2608,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 11; + u.binding = 16; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) { u.ids.push_back(render_buffers_get_sdfgi_occlusion_texture(p_render_buffers)); @@ -2619,14 +2619,14 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 12; + u.binding = 17; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.ids.push_back(rb ? render_buffers_get_gi_probe_buffer(p_render_buffers) : render_buffers_get_default_gi_probe_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 13; + u.binding = 18; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID vfog = RID(); if (rb && render_buffers_has_volumetric_fog(p_render_buffers)) { @@ -2642,8 +2642,16 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } } - render_pass_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RENDER_PASS_UNIFORM_SET); - return render_pass_uniform_set; + if (p_index >= (int)render_pass_uniform_sets.size()) { + render_pass_uniform_sets.resize(p_index + 1); + } + + if (render_pass_uniform_sets[p_index].is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_sets[p_index])) { + RD::get_singleton()->free(render_pass_uniform_sets[p_index]); + } + + render_pass_uniform_sets[p_index] = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RENDER_PASS_UNIFORM_SET); + return render_pass_uniform_sets[p_index]; } RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture) { @@ -2654,10 +2662,24 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed Vector<RD::Uniform> uniforms; { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(scene_state.uniform_buffers[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(scene_state.instance_buffer[RENDER_LIST_SECONDARY]); + uniforms.push_back(u); + } + { // No radiance texture. RID radiance_texture = storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); RD::Uniform u; - u.binding = 0; + u.binding = 2; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.push_back(radiance_texture); uniforms.push_back(u); @@ -2667,7 +2689,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed // No reflection atlas. RID ref_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK); RD::Uniform u; - u.binding = 1; + u.binding = 3; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.push_back(ref_texture); uniforms.push_back(u); @@ -2676,7 +2698,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed { // No shadow atlas. RD::Uniform u; - u.binding = 2; + u.binding = 4; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); u.ids.push_back(texture); @@ -2684,9 +2706,33 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed } { + // No directional shadow atlas. + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + u.ids.push_back(texture); + uniforms.push_back(u); + } + + { + // No Lightmaps + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.resize(scene_state.max_lightmaps); + RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + for (uint32_t i = 0; i < scene_state.max_lightmaps; i++) { + u.ids.write[i] = default_tex; + } + + uniforms.push_back(u); + } + + { // No GIProbes RD::Uniform u; - u.binding = 3; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(MAX_GI_PROBES); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); @@ -2696,33 +2742,43 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed uniforms.push_back(u); } + + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID cb = default_vec4_xform_buffer; + u.ids.push_back(cb); + uniforms.push_back(u); + } + // actual sdfgi stuff { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 4; + u.binding = 9; u.ids.push_back(p_albedo_texture); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 5; + u.binding = 10; u.ids.push_back(p_emission_texture); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 6; + u.binding = 11; u.ids.push_back(p_emission_aniso_texture); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 7; + u.binding = 12; u.ids.push_back(p_geom_facing_texture); uniforms.push_back(u); } @@ -2746,23 +2802,530 @@ RID RendererSceneRenderForward::_render_buffers_get_normal_texture(RID p_render_ return rb->normal_roughness_buffer; } -RID RendererSceneRenderForward::_render_buffers_get_ambient_texture(RID p_render_buffers) { - RenderBufferDataForward *rb = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); +RendererSceneRenderForward *RendererSceneRenderForward::singleton = nullptr; - return rb->ambient_buffer; +void RendererSceneRenderForward::set_time(double p_time, double p_step) { + time = p_time; + RendererSceneRenderRD::set_time(p_time, p_step); } -RID RendererSceneRenderForward::_render_buffers_get_reflection_texture(RID p_render_buffers) { - RenderBufferDataForward *rb = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); +void RendererSceneRenderForward::_geometry_instance_mark_dirty(GeometryInstance *p_geometry_instance) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + if (ginstance->dirty_list_element.in_list()) { + return; + } + + //clear surface caches + GeometryInstanceSurfaceDataCache *surf = ginstance->surface_caches; + + while (surf) { + GeometryInstanceSurfaceDataCache *next = surf->next; + geometry_instance_surface_alloc.free(surf); + surf = next; + } + + ginstance->surface_caches = nullptr; - return rb->reflection_buffer; + geometry_instance_dirty_list.add(&ginstance->dirty_list_element); } -RendererSceneRenderForward *RendererSceneRenderForward::singleton = nullptr; +void RendererSceneRenderForward::_geometry_instance_add_surface_with_material(GeometryInstanceForward *ginstance, uint32_t p_surface, MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh) { + bool has_read_screen_alpha = p_material->shader_data->uses_screen_texture || p_material->shader_data->uses_depth_texture || p_material->shader_data->uses_normal_texture; + bool has_base_alpha = (p_material->shader_data->uses_alpha || has_read_screen_alpha); + bool has_blend_alpha = p_material->shader_data->uses_blend_alpha; + bool has_alpha = has_base_alpha || has_blend_alpha; -void RendererSceneRenderForward::set_time(double p_time, double p_step) { - time = p_time; - RendererSceneRenderRD::set_time(p_time, p_step); + uint32_t flags = 0; + + if (p_material->shader_data->uses_sss) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SUBSURFACE_SCATTERING; + } + + if (p_material->shader_data->uses_screen_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SCREEN_TEXTURE; + } + + if (p_material->shader_data->uses_depth_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_DEPTH_TEXTURE; + } + + if (p_material->shader_data->uses_normal_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_NORMAL_TEXTURE; + } + + if (ginstance->data->cast_double_sided_shaodows) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS; + } + + if (has_alpha || has_read_screen_alpha || p_material->shader_data->depth_draw == ShaderData::DEPTH_DRAW_DISABLED || p_material->shader_data->depth_test == ShaderData::DEPTH_TEST_DISABLED) { + //material is only meant for alpha pass + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA; + if (p_material->shader_data->uses_depth_pre_pass && !(p_material->shader_data->depth_draw == ShaderData::DEPTH_DRAW_DISABLED || p_material->shader_data->depth_test == ShaderData::DEPTH_TEST_DISABLED)) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW; + } + } else { + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW; + } + + MaterialData *material_shadow = nullptr; + void *surface_shadow = nullptr; + if (!p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_pre_pass) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SHARED_SHADOW_MATERIAL; + material_shadow = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); + + RID shadow_mesh = storage->mesh_get_shadow_mesh(p_mesh); + + if (shadow_mesh.is_valid()) { + surface_shadow = storage->mesh_get_surface(shadow_mesh, p_surface); + } + + } else { + material_shadow = p_material; + } + + GeometryInstanceSurfaceDataCache *sdcache = geometry_instance_surface_alloc.alloc(); + + sdcache->flags = flags; + + sdcache->shader = p_material->shader_data; + sdcache->material_uniform_set = p_material->uniform_set; + sdcache->surface = storage->mesh_get_surface(p_mesh, p_surface); + sdcache->primitive = storage->mesh_surface_get_primitive(sdcache->surface); + sdcache->surface_index = p_surface; + + if (ginstance->data->dirty_dependencies) { + storage->base_update_dependency(p_mesh, &ginstance->data->dependency_tracker); + } + + //shadow + sdcache->shader_shadow = material_shadow->shader_data; + sdcache->material_uniform_set_shadow = material_shadow->uniform_set; + + sdcache->surface_shadow = surface_shadow ? surface_shadow : sdcache->surface; + + sdcache->owner = ginstance; + + sdcache->next = ginstance->surface_caches; + ginstance->surface_caches = sdcache; + + //sortkey + + sdcache->sort.sort_key1 = 0; + sdcache->sort.sort_key2 = 0; + + sdcache->sort.surface_index = p_surface; + sdcache->sort.material_id_low = p_material_id & 0x3FFF; + sdcache->sort.material_id_hi = p_material_id >> 14; + sdcache->sort.shader_id = p_shader_id; + sdcache->sort.geometry_id = p_mesh.get_local_index(); //only meshes can repeat anyway + sdcache->sort.uses_forward_gi = ginstance->can_sdfgi; + sdcache->sort.priority = p_material->priority; +} + +void RendererSceneRenderForward::_geometry_instance_add_surface(GeometryInstanceForward *ginstance, uint32_t p_surface, RID p_material, RID p_mesh) { + RID m_src; + + m_src = ginstance->data->material_override.is_valid() ? ginstance->data->material_override : p_material; + + MaterialData *material = nullptr; + + if (m_src.is_valid()) { + material = (MaterialData *)storage->material_get_data(m_src, RendererStorageRD::SHADER_TYPE_3D); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (material) { + if (ginstance->data->dirty_dependencies) { + storage->material_update_dependency(m_src, &ginstance->data->dependency_tracker); + } + } else { + material = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); + m_src = default_material; + } + + ERR_FAIL_COND(!material); + + _geometry_instance_add_surface_with_material(ginstance, p_surface, material, m_src.get_local_index(), storage->material_get_shader_id(m_src), p_mesh); + + while (material->next_pass.is_valid()) { + RID next_pass = material->next_pass; + material = (MaterialData *)storage->material_get_data(next_pass, RendererStorageRD::SHADER_TYPE_3D); + if (!material || !material->shader_data->valid) { + break; + } + if (ginstance->data->dirty_dependencies) { + storage->material_update_dependency(next_pass, &ginstance->data->dependency_tracker); + } + _geometry_instance_add_surface_with_material(ginstance, p_surface, material, next_pass.get_local_index(), storage->material_get_shader_id(next_pass), p_mesh); + } +} + +void RendererSceneRenderForward::_geometry_instance_update(GeometryInstance *p_geometry_instance) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + + if (ginstance->data->dirty_dependencies) { + ginstance->data->dependency_tracker.update_begin(); + } + + //add geometry for drawing + switch (ginstance->data->base_type) { + case RS::INSTANCE_MESH: { + const RID *materials = nullptr; + uint32_t surface_count; + RID mesh = ginstance->data->base; + + materials = storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + //if no materials, no surfaces. + const RID *inst_materials = ginstance->data->surface_materials.ptr(); + uint32_t surf_mat_count = ginstance->data->surface_materials.size(); + + for (uint32_t j = 0; j < surface_count; j++) { + RID material = (j < surf_mat_count && inst_materials[j].is_valid()) ? inst_materials[j] : materials[j]; + _geometry_instance_add_surface(ginstance, j, material, mesh); + } + } + + ginstance->instance_count = 1; + + } break; + + case RS::INSTANCE_MULTIMESH: { + RID mesh = storage->multimesh_get_mesh(ginstance->data->base); + if (mesh.is_valid()) { + const RID *materials = nullptr; + uint32_t surface_count; + + materials = storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + for (uint32_t j = 0; j < surface_count; j++) { + _geometry_instance_add_surface(ginstance, j, materials[j], mesh); + } + } + + ginstance->instance_count = storage->multimesh_get_instances_to_draw(ginstance->data->base); + } + + } break; +#if 0 + case RS::INSTANCE_IMMEDIATE: { + RasterizerStorageGLES3::Immediate *immediate = storage->immediate_owner.getornull(inst->base); + ERR_CONTINUE(!immediate); + + _add_geometry(immediate, inst, nullptr, -1, p_depth_pass, p_shadow_pass); + + } break; +#endif + case RS::INSTANCE_PARTICLES: { + int draw_passes = storage->particles_get_draw_passes(ginstance->data->base); + + for (int j = 0; j < draw_passes; j++) { + RID mesh = storage->particles_get_draw_pass_mesh(ginstance->data->base, j); + if (!mesh.is_valid()) + continue; + + const RID *materials = nullptr; + uint32_t surface_count; + + materials = storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + for (uint32_t k = 0; k < surface_count; k++) { + _geometry_instance_add_surface(ginstance, k, materials[k], mesh); + } + } + } + + ginstance->instance_count = storage->particles_get_amount(ginstance->data->base); + + } break; + + default: { + } + } + + //Fill push constant + + bool store_transform = true; + + if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; + uint32_t stride; + if (storage->multimesh_get_transform_format(ginstance->data->base) == RS::MULTIMESH_TRANSFORM_2D) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; + stride = 2; + } else { + stride = 3; + } + if (storage->multimesh_uses_colors(ginstance->data->base)) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; + stride += 1; + } + if (storage->multimesh_uses_custom_data(ginstance->data->base)) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; + stride += 1; + } + + ginstance->base_flags |= (stride << INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_SHIFT); + ginstance->transforms_uniform_set = storage->multimesh_get_3d_uniform_set(ginstance->data->base, default_shader_rd, TRANSFORMS_UNIFORM_SET); + + } else if (ginstance->data->base_type == RS::INSTANCE_PARTICLES) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; + uint32_t stride; + if (false) { // 2D particles + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; + stride = 2; + } else { + stride = 3; + } + + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; + stride += 1; + + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; + stride += 1; + + ginstance->base_flags |= (stride << INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_SHIFT); + + if (!storage->particles_is_using_local_coords(ginstance->data->base)) { + store_transform = false; + } + ginstance->transforms_uniform_set = storage->particles_get_instance_buffer_uniform_set(ginstance->data->base, default_shader_rd, TRANSFORMS_UNIFORM_SET); + + } else if (ginstance->data->base_type == RS::INSTANCE_MESH) { + if (storage->skeleton_is_valid(ginstance->data->skeleton)) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_SKELETON; + ginstance->transforms_uniform_set = storage->skeleton_get_3d_uniform_set(ginstance->data->skeleton, default_shader_rd, TRANSFORMS_UNIFORM_SET); + if (ginstance->data->dirty_dependencies) { + storage->skeleton_update_dependency(ginstance->data->skeleton, &ginstance->data->dependency_tracker); + } + } + } + + ginstance->store_transform_cache = store_transform; + ginstance->can_sdfgi = false; + + if (!lightmap_instance_is_valid(ginstance->lightmap_instance) && !low_end) { + if (ginstance->gi_probes[0].is_null() && (ginstance->data->use_baked_light || ginstance->data->use_dynamic_gi)) { + ginstance->can_sdfgi = true; + } + } + + if (ginstance->data->dirty_dependencies) { + ginstance->data->dependency_tracker.update_end(); + ginstance->data->dirty_dependencies = false; + } + + ginstance->dirty_list_element.remove_from_list(); +} + +void RendererSceneRenderForward::_update_dirty_geometry_instances() { + while (geometry_instance_dirty_list.first()) { + _geometry_instance_update(geometry_instance_dirty_list.first()->self()); + } +} + +void RendererSceneRenderForward::_geometry_instance_dependency_changed(RendererStorage::DependencyChangedNotification p_notification, RendererStorage::DependencyTracker *p_tracker) { + switch (p_notification) { + case RendererStorage::DEPENDENCY_CHANGED_MATERIAL: + case RendererStorage::DEPENDENCY_CHANGED_MESH: + case RendererStorage::DEPENDENCY_CHANGED_MULTIMESH: + case RendererStorage::DEPENDENCY_CHANGED_SKELETON_DATA: { + static_cast<RendererSceneRenderForward *>(singleton)->_geometry_instance_mark_dirty(static_cast<GeometryInstance *>(p_tracker->userdata)); + } break; + case RendererStorage::DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES: { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_tracker->userdata); + if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { + ginstance->instance_count = static_cast<RendererSceneRenderForward *>(singleton)->storage->multimesh_get_instances_to_draw(ginstance->data->base); + } + } break; + default: { + //rest of notifications of no interest + } break; + } +} +void RendererSceneRenderForward::_geometry_instance_dependency_deleted(const RID &p_dependency, RendererStorage::DependencyTracker *p_tracker) { + static_cast<RendererSceneRenderForward *>(singleton)->_geometry_instance_mark_dirty(static_cast<GeometryInstance *>(p_tracker->userdata)); +} + +RendererSceneRender::GeometryInstance *RendererSceneRenderForward::geometry_instance_create(RID p_base) { + RS::InstanceType type = storage->get_base_type(p_base); + ERR_FAIL_COND_V(!((1 << type) & RS::INSTANCE_GEOMETRY_MASK), nullptr); + + GeometryInstanceForward *ginstance = geometry_instance_alloc.alloc(); + ginstance->data = memnew(GeometryInstanceForward::Data); + + ginstance->data->base = p_base; + ginstance->data->base_type = type; + ginstance->data->dependency_tracker.userdata = ginstance; + ginstance->data->dependency_tracker.changed_callback = _geometry_instance_dependency_changed; + ginstance->data->dependency_tracker.deleted_callback = _geometry_instance_dependency_deleted; + + _geometry_instance_mark_dirty(ginstance); + + return ginstance; +} +void RendererSceneRenderForward::geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->data->skeleton = p_skeleton; + _geometry_instance_mark_dirty(ginstance); + ginstance->data->dirty_dependencies = true; +} +void RendererSceneRenderForward::geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->data->material_override = p_override; + _geometry_instance_mark_dirty(ginstance); + ginstance->data->dirty_dependencies = true; +} +void RendererSceneRenderForward::geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->data->surface_materials = p_materials; + _geometry_instance_mark_dirty(ginstance); + ginstance->data->dirty_dependencies = true; +} +void RendererSceneRenderForward::geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->mesh_instance = p_mesh_instance; + _geometry_instance_mark_dirty(ginstance); +} +void RendererSceneRenderForward::geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->transform = p_transform; + ginstance->mirror = p_transform.basis.determinant() < 0; + ginstance->data->aabb = p_aabb; + ginstance->transformed_aabb = p_transformed_aabb; + + Vector3 model_scale_vec = p_transform.basis.get_scale_abs(); + // handle non uniform scale here + + float max_scale = MAX(model_scale_vec.x, MAX(model_scale_vec.y, model_scale_vec.z)); + float min_scale = MIN(model_scale_vec.x, MIN(model_scale_vec.y, model_scale_vec.z)); + ginstance->non_uniform_scale = max_scale >= 0.0 && (min_scale / max_scale) < 0.9; + + ginstance->lod_model_scale = max_scale; +} +void RendererSceneRenderForward::geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->lod_bias = p_lod_bias; +} +void RendererSceneRenderForward::geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->data->use_baked_light = p_enable; + _geometry_instance_mark_dirty(ginstance); +} +void RendererSceneRenderForward::geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->data->use_dynamic_gi = p_enable; + _geometry_instance_mark_dirty(ginstance); +} +void RendererSceneRenderForward::geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->lightmap_instance = p_lightmap_instance; + ginstance->lightmap_uv_scale = p_lightmap_uv_scale; + ginstance->lightmap_slice_index = p_lightmap_slice_index; + _geometry_instance_mark_dirty(ginstance); +} +void RendererSceneRenderForward::geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + if (p_sh9) { + if (ginstance->lightmap_sh == nullptr) { + ginstance->lightmap_sh = geometry_instance_lightmap_sh.alloc(); + } + + copymem(ginstance->lightmap_sh->sh, p_sh9, sizeof(Color) * 9); + } else { + if (ginstance->lightmap_sh != nullptr) { + geometry_instance_lightmap_sh.free(ginstance->lightmap_sh); + ginstance->lightmap_sh = nullptr; + } + } + _geometry_instance_mark_dirty(ginstance); +} +void RendererSceneRenderForward::geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->shader_parameters_offset = p_offset; + _geometry_instance_mark_dirty(ginstance); +} +void RendererSceneRenderForward::geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + + ginstance->data->cast_double_sided_shaodows = p_enable; + _geometry_instance_mark_dirty(ginstance); +} + +void RendererSceneRenderForward::geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->layer_mask = p_layer_mask; +} + +void RendererSceneRenderForward::geometry_instance_free(GeometryInstance *p_geometry_instance) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + if (ginstance->lightmap_sh != nullptr) { + geometry_instance_lightmap_sh.free(ginstance->lightmap_sh); + } + GeometryInstanceSurfaceDataCache *surf = ginstance->surface_caches; + while (surf) { + GeometryInstanceSurfaceDataCache *next = surf->next; + geometry_instance_surface_alloc.free(surf); + surf = next; + } + memdelete(ginstance->data); + geometry_instance_alloc.free(ginstance); +} + +uint32_t RendererSceneRenderForward::geometry_instance_get_pair_mask() { + return (1 << RS::INSTANCE_GI_PROBE); +} +void RendererSceneRenderForward::geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) { +} +void RendererSceneRenderForward::geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) { +} +void RendererSceneRenderForward::geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) { +} + +Transform RendererSceneRenderForward::geometry_instance_get_transform(GeometryInstance *p_instance) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_instance); + ERR_FAIL_COND_V(!ginstance, Transform()); + return ginstance->transform; +} +AABB RendererSceneRenderForward::geometry_instance_get_aabb(GeometryInstance *p_instance) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_instance); + ERR_FAIL_COND_V(!ginstance, AABB()); + return ginstance->data->aabb; +} + +void RendererSceneRenderForward::geometry_instance_pair_gi_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_gi_probe_instances, uint32_t p_gi_probe_instance_count) { + GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + if (p_gi_probe_instance_count > 0) { + ginstance->gi_probes[0] = p_gi_probe_instances[0]; + } else { + ginstance->gi_probes[0] = RID(); + } + + if (p_gi_probe_instance_count > 1) { + ginstance->gi_probes[1] = p_gi_probe_instances[1]; + } else { + ginstance->gi_probes[1] = RID(); + } } RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_storage) : @@ -2788,11 +3351,10 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor { //lightmaps - scene_state.max_lightmaps = storage->lightmap_array_get_size(); + scene_state.max_lightmaps = low_end ? 2 : MAX_LIGHTMAPS; defines += "\n#define MAX_LIGHTMAP_TEXTURES " + itos(scene_state.max_lightmaps) + "\n"; defines += "\n#define MAX_LIGHTMAPS " + itos(scene_state.max_lightmaps) + "\n"; - scene_state.lightmaps = memnew_arr(LightmapData, scene_state.max_lightmaps); scene_state.lightmap_buffer = RD::get_singleton()->storage_buffer_create(sizeof(LightmapData) * scene_state.max_lightmaps); } { @@ -3005,24 +3567,11 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; actions.default_repeat = ShaderLanguage::REPEAT_ENABLE; actions.global_buffer_array_variable = "global_variables.data"; - actions.instance_uniform_index_variable = "instances.data[instance_index].instance_uniforms_ofs"; + actions.instance_uniform_index_variable = "draw_call.instance_uniforms_ofs"; shader.compiler.initialize(actions); } - //render list - render_list.max_elements = GLOBAL_DEF_RST("rendering/limits/rendering/max_renderable_elements", (int)128000); - render_list.init(); - render_pass = 0; - - { - scene_state.max_instances = render_list.max_elements; - scene_state.instances = memnew_arr(InstanceData, scene_state.max_instances); - scene_state.instance_buffer = RD::get_singleton()->storage_buffer_create(sizeof(InstanceData) * scene_state.max_instances); - } - - scene_state.uniform_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SceneState::UBO)); - { //default material and shader default_shader = storage->shader_create(); @@ -3068,14 +3617,18 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor sampler.compare_op = RD::COMPARE_OP_LESS; shadow_sampler = RD::get_singleton()->sampler_create(sampler); } + + render_list_thread_threshold = GLOBAL_GET("rendering/forward_renderer/threaded_render_minimum_instances"); } RendererSceneRenderForward::~RendererSceneRenderForward() { directional_shadow_atlas_set_size(0); //clear base uniform set if still valid - if (render_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_set)) { - RD::get_singleton()->free(render_pass_uniform_set); + for (uint32_t i = 0; i < render_pass_uniform_sets.size(); i++) { + if (render_pass_uniform_sets[i].is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_sets[i])) { + RD::get_singleton()->free(render_pass_uniform_sets[i]); + } } if (sdfgi_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sdfgi_pass_uniform_set)) { @@ -3094,12 +3647,16 @@ RendererSceneRenderForward::~RendererSceneRenderForward() { storage->free(default_material); { - RD::get_singleton()->free(scene_state.uniform_buffer); - RD::get_singleton()->free(scene_state.instance_buffer); + for (uint32_t i = 0; i < scene_state.uniform_buffers.size(); i++) { + RD::get_singleton()->free(scene_state.uniform_buffers[i]); + } RD::get_singleton()->free(scene_state.lightmap_buffer); RD::get_singleton()->free(scene_state.lightmap_capture_buffer); - memdelete_arr(scene_state.instances); - memdelete_arr(scene_state.lightmaps); + for (uint32_t i = 0; i < RENDER_LIST_MAX; i++) { + if (scene_state.instance_buffer[i] != RID()) { + RD::get_singleton()->free(scene_state.instance_buffer[i]); + } + } memdelete_arr(scene_state.lightmap_captures); } diff --git a/servers/rendering/renderer_rd/renderer_scene_render_forward.h b/servers/rendering/renderer_rd/renderer_scene_render_forward.h index 4b37f4a391..af78c50fda 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_forward.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_forward.h @@ -31,6 +31,7 @@ #ifndef RENDERING_SERVER_SCENE_RENDER_FORWARD_H #define RENDERING_SERVER_SCENE_RENDER_FORWARD_H +#include "core/templates/paged_allocator.h" #include "servers/rendering/renderer_rd/pipeline_cache_rd.h" #include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" #include "servers/rendering/renderer_rd/renderer_storage_rd.h" @@ -46,7 +47,18 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { enum { SDFGI_MAX_CASCADES = 8, - MAX_GI_PROBES = 8 + MAX_GI_PROBES = 8, + MAX_LIGHTMAPS = 8, + MAX_GI_PROBES_PER_INSTANCE = 2, + INSTANCE_DATA_BUFFER_MIN_SIZE = 4096 + }; + + enum RenderListType { + RENDER_LIST_OPAQUE, //used for opaque objects + RENDER_LIST_ALPHA, //used for transparent objects + RENDER_LIST_SECONDARY, //used for shadows and other objects + RENDER_LIST_MAX + }; /* Scene Shader */ @@ -166,6 +178,8 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { virtual bool is_animated() const; virtual bool casts_shadows() const; virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + ShaderData(); virtual ~ShaderData(); }; @@ -197,14 +211,6 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { return static_cast<RendererSceneRenderForward *>(singleton)->_create_material_func(static_cast<ShaderData *>(p_shader)); } - /* Push Constant */ - - struct PushConstant { - uint32_t index; - uint32_t pad; - float bake_uv2_offset[2]; - }; - /* Framebuffer */ struct RenderBufferDataForward : public RenderBufferData { @@ -216,9 +222,6 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID normal_roughness_buffer; RID giprobe_buffer; - RID ambient_buffer; - RID reflection_buffer; - RS::ViewportMSAA msaa; RD::TextureSamples texture_samples; @@ -239,7 +242,6 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID render_sdfgi_uniform_set; void ensure_specular(); - void ensure_gi(); void ensure_giprobe(); void clear(); virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa); @@ -252,7 +254,7 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID shadow_sampler; RID render_base_uniform_set; - RID render_pass_uniform_set; + LocalVector<RID> render_pass_uniform_sets; RID sdfgi_pass_uniform_set; uint64_t lightmap_texture_array_version = 0xFFFFFFFF; @@ -261,12 +263,61 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { void _render_buffers_clear_uniform_set(RenderBufferDataForward *rb); virtual void _render_buffers_uniform_set_changed(RID p_render_buffers); virtual RID _render_buffers_get_normal_texture(RID p_render_buffers); - virtual RID _render_buffers_get_ambient_texture(RID p_render_buffers); - virtual RID _render_buffers_get_reflection_texture(RID p_render_buffers); void _update_render_base_uniform_set(); RID _setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture); - RID _setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, const PagedArray<RID> &p_gi_probes); + RID _setup_render_pass_uniform_set(RenderListType p_render_list, RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, bool p_use_directional_shadow_atlas = false, int p_index = 0); + + enum PassMode { + PASS_MODE_COLOR, + PASS_MODE_COLOR_SPECULAR, + PASS_MODE_COLOR_TRANSPARENT, + PASS_MODE_SHADOW, + PASS_MODE_SHADOW_DP, + PASS_MODE_DEPTH, + PASS_MODE_DEPTH_NORMAL_ROUGHNESS, + PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE, + PASS_MODE_DEPTH_MATERIAL, + PASS_MODE_SDF, + }; + + struct GeometryInstanceSurfaceDataCache; + struct RenderElementInfo; + + struct RenderListParameters { + GeometryInstanceSurfaceDataCache **elements = nullptr; + RenderElementInfo *element_info = nullptr; + int element_count = 0; + bool reverse_cull = false; + PassMode pass_mode = PASS_MODE_COLOR; + bool no_gi = false; + RID render_pass_uniform_set; + bool force_wireframe = false; + Vector2 uv_offset; + Plane lod_plane; + float lod_distance_multiplier = 0.0; + float screen_lod_threshold = 0.0; + RD::FramebufferFormatID framebuffer_format = 0; + uint32_t element_offset = 0; + uint32_t barrier = RD::BARRIER_MASK_ALL; + + RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { + elements = p_elements; + element_info = p_element_info; + element_count = p_element_count; + reverse_cull = p_reverse_cull; + pass_mode = p_pass_mode; + no_gi = p_no_gi; + render_pass_uniform_set = p_render_pass_uniform_set; + force_wireframe = p_force_wireframe; + uv_offset = p_uv_offset; + lod_plane = p_lod_plane; + lod_distance_multiplier = p_lod_distance_multiplier; + screen_lod_threshold = p_screen_lod_threshold; + element_offset = p_element_offset; + barrier = p_barrier; + } + }; struct LightmapData { float normal_xform[12]; @@ -292,16 +343,6 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { INSTANCE_DATA_FLAG_SKELETON = 1 << 19, }; - struct InstanceData { - float transform[16]; - float normal_transform[16]; - uint32_t flags; - uint32_t instance_uniforms_ofs; //instance_offset in instancing/skeleton buffer - uint32_t gi_offset; //GI information when using lightmapping (VCT or lightmap) - uint32_t mask; - float lightmap_uv_scale[4]; - }; - struct SceneState { struct UBO { float projection_matrix[16]; @@ -313,6 +354,11 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { float viewport_size[2]; float screen_pixel_size[2]; + uint32_t cluster_shift; + uint32_t cluster_width; + uint32_t cluster_type_size; + uint32_t max_cluster_element_count_div_32; + float directional_penumbra_shadow_kernel[128]; //32 vec4s float directional_soft_shadow_kernel[128]; float penumbra_shadow_kernel[128]; @@ -381,214 +427,340 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { uint32_t pancake_shadows; }; + struct PushConstant { + uint32_t base_index; // + uint32_t uv_offset; //packed + uint32_t pad[2]; + }; + + struct InstanceData { + float transform[16]; + uint32_t flags; + uint32_t instance_uniforms_ofs; //base offset in global buffer for instance variables + uint32_t gi_offset; //GI information when using lightmapping (VCT or lightmap index) + uint32_t layer_mask; + float lightmap_uv_scale[4]; + }; + UBO ubo; - RID uniform_buffer; + LocalVector<RID> uniform_buffers; - LightmapData *lightmaps; + LightmapData lightmaps[MAX_LIGHTMAPS]; + RID lightmap_ids[MAX_LIGHTMAPS]; + bool lightmap_has_sh[MAX_LIGHTMAPS]; + uint32_t lightmaps_used = 0; uint32_t max_lightmaps; RID lightmap_buffer; + RID instance_buffer[RENDER_LIST_MAX]; + uint32_t instance_buffer_size[RENDER_LIST_MAX] = { 0, 0, 0 }; + LocalVector<InstanceData> instance_data[RENDER_LIST_MAX]; + LightmapCaptureData *lightmap_captures; uint32_t max_lightmap_captures; RID lightmap_capture_buffer; - RID instance_buffer; - InstanceData *instances; - uint32_t max_instances; + RID giprobe_ids[MAX_GI_PROBES]; + uint32_t giprobes_used = 0; bool used_screen_texture = false; bool used_normal_texture = false; bool used_depth_texture = false; bool used_sss = false; - uint32_t current_shader_index = 0; - uint32_t current_material_index = 0; + + struct ShadowPass { + uint32_t element_from; + uint32_t element_count; + bool flip_cull; + PassMode pass_mode; + + RID rp_uniform_set; + Plane camera_plane; + float lod_distance_multiplier; + float screen_lod_threshold; + + RID framebuffer; + RD::InitialAction initial_depth_action; + RD::FinalAction final_depth_action; + Rect2i rect; + }; + + LocalVector<ShadowPass> shadow_passes; } scene_state; - /* Render List */ + static RendererSceneRenderForward *singleton; - struct RenderList { - int max_elements; - - struct Element { - RendererSceneRender::InstanceBase *instance; - MaterialData *material; - union { - struct { - //from least significant to most significant in sort, TODO: should be endian swapped on big endian - uint64_t geometry_index : 20; - uint64_t material_index : 15; - uint64_t shader_index : 12; - uint64_t uses_instancing : 1; - uint64_t uses_forward_gi : 1; - uint64_t uses_lightmap : 1; - uint64_t depth_layer : 4; - uint64_t priority : 8; - }; - - uint64_t sort_key; + double time; + RID default_shader; + RID default_material; + RID overdraw_material_shader; + RID overdraw_material; + RID wireframe_material_shader; + RID wireframe_material; + RID default_shader_rd; + RID default_shader_sdfgi_rd; + + RID default_vec4_xform_buffer; + RID default_vec4_xform_uniform_set; + + void _setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false, int p_index = 0); + void _setup_giprobes(const PagedArray<RID> &p_giprobes); + void _setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform &p_cam_transform); + + struct RenderElementInfo { + uint32_t repeat : 22; + uint32_t uses_lightmap : 1; + uint32_t uses_forward_gi : 1; + uint32_t lod_index : 8; + }; + + template <PassMode p_pass_mode> + _FORCE_INLINE_ void _render_list_template(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element); + + void _render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element); + + LocalVector<RD::DrawListID> thread_draw_lists; + void _render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params); + void _render_list_with_threads(RenderListParameters *p_params, RID p_framebuffer, RD::InitialAction p_initial_color_action, RD::FinalAction p_final_color_action, RD::InitialAction p_initial_depth_action, RD::FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values = Vector<Color>(), float p_clear_depth = 1.0, uint32_t p_clear_stencil = 0, const Rect2 &p_region = Rect2(), const Vector<RID> &p_storage_textures = Vector<RID>()); + + uint32_t render_list_thread_threshold = 500; + + void _update_instance_data_buffer(RenderListType p_render_list); + void _fill_instance_data(RenderListType p_render_list, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); + void _fill_render_list(RenderListType p_render_list, const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi = false, bool p_using_opaque_gi = false, const Plane &p_lod_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, bool p_append = false); + + Map<Size2i, RID> sdfgi_framebuffer_size_cache; + + struct GeometryInstanceData; + struct GeometryInstanceForward; + + struct GeometryInstanceLightmapSH { + Color sh[9]; + }; + + // Cached data for drawing surfaces + struct GeometryInstanceSurfaceDataCache { + enum { + FLAG_PASS_DEPTH = 1, + FLAG_PASS_OPAQUE = 2, + FLAG_PASS_ALPHA = 4, + FLAG_PASS_SHADOW = 8, + FLAG_USES_SHARED_SHADOW_MATERIAL = 128, + FLAG_USES_SUBSURFACE_SCATTERING = 2048, + FLAG_USES_SCREEN_TEXTURE = 4096, + FLAG_USES_DEPTH_TEXTURE = 8192, + FLAG_USES_NORMAL_TEXTURE = 16384, + FLAG_USES_DOUBLE_SIDED_SHADOWS = 32768, + }; + + union { + struct { + uint64_t lod_index : 8; + uint64_t surface_index : 10; + uint64_t geometry_id : 32; + uint64_t material_id_low : 14; + + uint64_t material_id_hi : 18; + uint64_t shader_id : 32; + uint64_t uses_forward_gi : 1; + uint64_t uses_lightmap : 1; + uint64_t depth_layer : 4; + uint64_t priority : 8; + }; + struct { + uint64_t sort_key1; + uint64_t sort_key2; }; - uint32_t surface_index; + } sort; + + RS::PrimitiveType primitive = RS::PRIMITIVE_MAX; + uint32_t flags = 0; + uint32_t surface_index = 0; + + void *surface = nullptr; + RID material_uniform_set; + ShaderData *shader = nullptr; + + void *surface_shadow = nullptr; + RID material_uniform_set_shadow; + ShaderData *shader_shadow = nullptr; + + GeometryInstanceSurfaceDataCache *next = nullptr; + GeometryInstanceForward *owner = nullptr; + }; + + struct GeometryInstanceForward : public GeometryInstance { + //used during rendering + bool mirror = false; + bool non_uniform_scale = false; + float lod_bias = 0.0; + float lod_model_scale = 1.0; + AABB transformed_aabb; //needed for LOD + float depth = 0; + uint32_t gi_offset_cache = 0; + uint32_t flags_cache = 0; + bool store_transform_cache = true; + int32_t shader_parameters_offset = -1; + uint32_t lightmap_slice_index; + Rect2 lightmap_uv_scale; + uint32_t layer_mask = 1; + RID transforms_uniform_set; + uint32_t instance_count = 0; + RID mesh_instance; + bool can_sdfgi = false; + //used during setup + uint32_t base_flags = 0; + Transform transform; + RID gi_probes[MAX_GI_PROBES_PER_INSTANCE]; + RID lightmap_instance; + GeometryInstanceLightmapSH *lightmap_sh = nullptr; + GeometryInstanceSurfaceDataCache *surface_caches = nullptr; + SelfList<GeometryInstanceForward> dirty_list_element; + + struct Data { + //data used less often goes into regular heap + RID base; + RS::InstanceType base_type; + + RID skeleton; + Vector<RID> surface_materials; + RID material_override; + AABB aabb; + + bool use_dynamic_gi = false; + bool use_baked_light = false; + bool cast_double_sided_shaodows = false; + bool mirror = false; + bool dirty_dependencies = false; + + RendererStorage::DependencyTracker dependency_tracker; }; - Element *base_elements; - Element **elements; + Data *data = nullptr; + + GeometryInstanceForward() : + dirty_list_element(this) {} + }; + + static void _geometry_instance_dependency_changed(RendererStorage::DependencyChangedNotification p_notification, RendererStorage::DependencyTracker *p_tracker); + static void _geometry_instance_dependency_deleted(const RID &p_dependency, RendererStorage::DependencyTracker *p_tracker); - int element_count; - int alpha_element_count; + SelfList<GeometryInstanceForward>::List geometry_instance_dirty_list; + + PagedAllocator<GeometryInstanceForward> geometry_instance_alloc; + PagedAllocator<GeometryInstanceSurfaceDataCache> geometry_instance_surface_alloc; + PagedAllocator<GeometryInstanceLightmapSH> geometry_instance_lightmap_sh; + + void _geometry_instance_add_surface_with_material(GeometryInstanceForward *ginstance, uint32_t p_surface, MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh); + void _geometry_instance_add_surface(GeometryInstanceForward *ginstance, uint32_t p_surface, RID p_material, RID p_mesh); + void _geometry_instance_mark_dirty(GeometryInstance *p_geometry_instance); + void _geometry_instance_update(GeometryInstance *p_geometry_instance); + void _update_dirty_geometry_instances(); + + bool low_end = false; + + /* Render List */ + + struct RenderList { + LocalVector<GeometryInstanceSurfaceDataCache *> elements; + LocalVector<RenderElementInfo> element_info; void clear() { - element_count = 0; - alpha_element_count = 0; + elements.clear(); + element_info.clear(); } //should eventually be replaced by radix struct SortByKey { - _FORCE_INLINE_ bool operator()(const Element *A, const Element *B) const { - return A->sort_key < B->sort_key; + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->sort.sort_key2 == B->sort.sort_key2) ? (A->sort.sort_key1 < B->sort.sort_key1) : (A->sort.sort_key2 < B->sort.sort_key2); } }; - void sort_by_key(bool p_alpha) { - SortArray<Element *, SortByKey> sorter; - if (p_alpha) { - sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count); - } else { - sorter.sort(elements, element_count); - } + void sort_by_key() { + SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + void sort_by_key_range(uint32_t p_from, uint32_t p_size) { + SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; + sorter.sort(elements.ptr() + p_from, p_size); } struct SortByDepth { - _FORCE_INLINE_ bool operator()(const Element *A, const Element *B) const { - return A->instance->depth < B->instance->depth; + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->owner->depth < B->owner->depth); } }; - void sort_by_depth(bool p_alpha) { //used for shadows + void sort_by_depth() { //used for shadows - SortArray<Element *, SortByDepth> sorter; - if (p_alpha) { - sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count); - } else { - sorter.sort(elements, element_count); - } + SortArray<GeometryInstanceSurfaceDataCache *, SortByDepth> sorter; + sorter.sort(elements.ptr(), elements.size()); } struct SortByReverseDepthAndPriority { - _FORCE_INLINE_ bool operator()(const Element *A, const Element *B) const { - uint32_t layer_A = uint32_t(A->priority); - uint32_t layer_B = uint32_t(B->priority); - if (layer_A == layer_B) { - return A->instance->depth > B->instance->depth; - } else { - return layer_A < layer_B; - } + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->sort.priority == B->sort.priority) ? (A->owner->depth > B->owner->depth) : (A->sort.priority < B->sort.priority); } }; void sort_by_reverse_depth_and_priority(bool p_alpha) { //used for alpha - SortArray<Element *, SortByReverseDepthAndPriority> sorter; - if (p_alpha) { - sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count); - } else { - sorter.sort(elements, element_count); - } + SortArray<GeometryInstanceSurfaceDataCache *, SortByReverseDepthAndPriority> sorter; + sorter.sort(elements.ptr(), elements.size()); } - _FORCE_INLINE_ Element *add_element() { - if (element_count + alpha_element_count >= max_elements) { - return nullptr; - } - elements[element_count] = &base_elements[element_count]; - return elements[element_count++]; - } - - _FORCE_INLINE_ Element *add_alpha_element() { - if (element_count + alpha_element_count >= max_elements) { - return nullptr; - } - int idx = max_elements - alpha_element_count - 1; - elements[idx] = &base_elements[idx]; - alpha_element_count++; - return elements[idx]; + _FORCE_INLINE_ void add_element(GeometryInstanceSurfaceDataCache *p_element) { + elements.push_back(p_element); } - - void init() { - element_count = 0; - alpha_element_count = 0; - elements = memnew_arr(Element *, max_elements); - base_elements = memnew_arr(Element, max_elements); - for (int i = 0; i < max_elements; i++) { - elements[i] = &base_elements[i]; // assign elements - } - } - - RenderList() { - max_elements = 0; - } - - ~RenderList() { - memdelete_arr(elements); - memdelete_arr(base_elements); - } - }; - - RenderList render_list; - - static RendererSceneRenderForward *singleton; - uint64_t render_pass; - double time; - RID default_shader; - RID default_material; - RID overdraw_material_shader; - RID overdraw_material; - RID wireframe_material_shader; - RID wireframe_material; - RID default_shader_rd; - RID default_shader_sdfgi_rd; - - RID default_vec4_xform_buffer; - RID default_vec4_xform_uniform_set; - - enum PassMode { - PASS_MODE_COLOR, - PASS_MODE_COLOR_SPECULAR, - PASS_MODE_COLOR_TRANSPARENT, - PASS_MODE_SHADOW, - PASS_MODE_SHADOW_DP, - PASS_MODE_DEPTH, - PASS_MODE_DEPTH_NORMAL_ROUGHNESS, - PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE, - PASS_MODE_DEPTH_MATERIAL, - PASS_MODE_SDF, }; - void _setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2 &p_screen_pixel_size, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false); - void _setup_lightmaps(const PagedArray<InstanceBase *> &p_lightmaps, const Transform &p_cam_transform); - - void _fill_instances(RenderList::Element **p_elements, int p_element_count, bool p_for_depth, bool p_has_sdfgi = false, bool p_has_opaque_gi = false); - void _render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderList::Element **p_elements, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0); - _FORCE_INLINE_ void _add_geometry(InstanceBase *p_instance, uint32_t p_surface, RID p_material, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi = false); - _FORCE_INLINE_ void _add_geometry_with_material(InstanceBase *p_instance, uint32_t p_surface, MaterialData *p_material, RID p_material_rid, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi = false); + RenderList render_list[RENDER_LIST_MAX]; - void _fill_render_list(const PagedArray<InstanceBase *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi = false); - - Map<Size2i, RID> sdfgi_framebuffer_size_cache; +protected: + virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_lod_threshold); - bool low_end = false; + virtual void _render_shadow_begin(); + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true); + virtual void _render_shadow_process(); + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL); -protected: - virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_lod_threshold); - virtual void _render_shadow(RID p_framebuffer, const PagedArray<InstanceBase *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0); - virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - virtual void _render_uv2(const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<InstanceBase *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture); - virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<InstanceBase *> &p_instances); + virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); + virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); + virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture); + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances); public: + virtual GeometryInstance *geometry_instance_create(RID p_base); + virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton); + virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override); + virtual void geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials); + virtual void geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance); + virtual void geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb); + virtual void geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask); + virtual void geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias); + virtual void geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable); + virtual void geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable); + virtual void geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index); + virtual void geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9); + virtual void geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset); + virtual void geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable); + + virtual Transform geometry_instance_get_transform(GeometryInstance *p_instance); + virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance); + + virtual void geometry_instance_free(GeometryInstance *p_geometry_instance); + + virtual uint32_t geometry_instance_get_pair_mask(); + virtual void geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count); + virtual void geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count); + virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count); + virtual void geometry_instance_pair_gi_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_gi_probe_instances, uint32_t p_gi_probe_instance_count); + virtual void set_time(double p_time, double p_step); virtual bool free(RID p_rid); diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index 8e55dea2b1..09d2c032a8 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -183,13 +183,11 @@ void RendererSceneRenderRD::_create_reflection_importance_sample(ReflectionData void RendererSceneRenderRD::_update_reflection_mipmaps(ReflectionData &rd, int p_start, int p_end) { for (int i = p_start; i < p_end; i++) { - for (int j = 0; j < rd.layers[i].mipmaps.size() - 1; j++) { - for (int k = 0; k < 6; k++) { - RID view = rd.layers[i].mipmaps[j].views[k]; - RID texture = rd.layers[i].mipmaps[j + 1].views[k]; - Size2i size = rd.layers[i].mipmaps[j + 1].size; - storage->get_effects()->make_mipmap(view, texture, size); - } + for (int j = 0; j < rd.layers[i].views.size() - 1; j++) { + RID view = rd.layers[i].views[j]; + RID texture = rd.layers[i].views[j + 1]; + Size2i size = rd.layers[i].mipmaps[j + 1].size; + storage->get_effects()->cubemap_downsample(view, texture, size); } } } @@ -1150,150 +1148,71 @@ void RendererSceneRenderRD::_sdfgi_update_cascades(RID p_render_buffers) { cascade_data[i].pad = 0; } - RD::get_singleton()->buffer_update(rb->sdfgi->cascades_ubo, 0, sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES, cascade_data, true); + RD::get_singleton()->buffer_update(rb->sdfgi->cascades_ubo, 0, sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES, cascade_data, RD::BARRIER_MASK_COMPUTE); } -void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count) { +void RendererSceneRenderRD::_sdfgi_update_light(RID p_render_buffers, RID p_environment) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(rb == nullptr); if (rb->sdfgi == nullptr) { return; } - Environment *env = environment_owner.getornull(p_environment); - - RENDER_TIMESTAMP(">SDFGI Update Probes"); - - /* Update Cascades UBO */ - _sdfgi_update_cascades(p_render_buffers); - /* Update Dynamic Lights Buffer */ - - RENDER_TIMESTAMP("Update Lights"); - - /* Update dynamic lights */ - - { - int32_t cascade_light_count[SDFGI::MAX_CASCADES]; - - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; - - SDGIShader::Light lights[SDFGI::MAX_DYNAMIC_LIGHTS]; - uint32_t idx = 0; - for (uint32_t j = 0; j < (uint32_t)p_directional_lights.size(); j++) { - if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { - break; - } - - LightInstance *li = light_instance_owner.getornull(p_directional_lights[j]); - ERR_CONTINUE(!li); - - if (storage->light_directional_is_sky_only(li->light)) { - continue; - } - - Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); - dir.y *= rb->sdfgi->y_mult; - dir.normalize(); - lights[idx].direction[0] = dir.x; - lights[idx].direction[1] = dir.y; - lights[idx].direction[2] = dir.z; - Color color = storage->light_get_color(li->light); - color = color.to_linear(); - lights[idx].color[0] = color.r; - lights[idx].color[1] = color.g; - lights[idx].color[2] = color.b; - lights[idx].type = RS::LIGHT_DIRECTIONAL; - lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); - lights[idx].has_shadow = storage->light_has_shadow(li->light); - idx++; - } + RD::get_singleton()->draw_command_begin_label("SDFGI Update dynamic Light"); - AABB cascade_aabb; - cascade_aabb.position = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + cascade.position)) * cascade.cell_size; - cascade_aabb.size = Vector3(1, 1, 1) * rb->sdfgi->cascade_size * cascade.cell_size; + /* Update dynamic light */ - for (uint32_t j = 0; j < p_positional_light_count; j++) { - if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { - break; - } - - LightInstance *li = light_instance_owner.getornull(p_positional_light_instances[j]); - ERR_CONTINUE(!li); + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.direct_light_pipeline[SDGIShader::DIRECT_LIGHT_MODE_DYNAMIC]); - uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light); - if (i > max_sdfgi_cascade) { - continue; - } + SDGIShader::DirectLightPushConstant push_constant; - if (!cascade_aabb.intersects(li->aabb)) { - continue; - } + push_constant.grid_size[0] = rb->sdfgi->cascade_size; + push_constant.grid_size[1] = rb->sdfgi->cascade_size; + push_constant.grid_size[2] = rb->sdfgi->cascade_size; + push_constant.max_cascades = rb->sdfgi->cascades.size(); + push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; + push_constant.multibounce = rb->sdfgi->uses_multibounce; + push_constant.y_mult = rb->sdfgi->y_mult; - Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); - //faster to not do this here - //dir.y *= rb->sdfgi->y_mult; - //dir.normalize(); - lights[idx].direction[0] = dir.x; - lights[idx].direction[1] = dir.y; - lights[idx].direction[2] = dir.z; - Vector3 pos = li->transform.origin; - pos.y *= rb->sdfgi->y_mult; - lights[idx].position[0] = pos.x; - lights[idx].position[1] = pos.y; - lights[idx].position[2] = pos.z; - Color color = storage->light_get_color(li->light); - color = color.to_linear(); - lights[idx].color[0] = color.r; - lights[idx].color[1] = color.g; - lights[idx].color[2] = color.b; - lights[idx].type = storage->light_get_type(li->light); - lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); - lights[idx].has_shadow = storage->light_has_shadow(li->light); - lights[idx].attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_ATTENUATION); - lights[idx].radius = storage->light_get_param(li->light, RS::LIGHT_PARAM_RANGE); - lights[idx].spot_angle = Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE)); - lights[idx].spot_attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { + SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; + push_constant.light_count = rb->sdfgi->cascade_dynamic_light_count[i]; + push_constant.cascade = i; - idx++; - } + if (rb->sdfgi->cascades[i].all_dynamic_lights_dirty || sdfgi_frames_to_update_light == RS::ENV_SDFGI_UPDATE_LIGHT_IN_1_FRAME) { + push_constant.process_offset = 0; + push_constant.process_increment = 1; + } else { + static uint32_t frames_to_update_table[RS::ENV_SDFGI_UPDATE_LIGHT_MAX] = { + 1, 2, 4, 8, 16 + }; - if (idx > 0) { - RD::get_singleton()->buffer_update(cascade.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights, true); - } + uint32_t frames_to_update = frames_to_update_table[sdfgi_frames_to_update_light]; - cascade_light_count[i] = idx; + push_constant.process_offset = RSG::rasterizer->get_frame_number() % frames_to_update; + push_constant.process_increment = frames_to_update; } + rb->sdfgi->cascades[i].all_dynamic_lights_dirty = false; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.direct_light_pipeline[SDGIShader::DIRECT_LIGHT_MODE_DYNAMIC]); - - SDGIShader::DirectLightPushConstant push_constant; - - push_constant.grid_size[0] = rb->sdfgi->cascade_size; - push_constant.grid_size[1] = rb->sdfgi->cascade_size; - push_constant.grid_size[2] = rb->sdfgi->cascade_size; - push_constant.max_cascades = rb->sdfgi->cascades.size(); - push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; - push_constant.multibounce = rb->sdfgi->uses_multibounce; - push_constant.y_mult = rb->sdfgi->y_mult; - - push_constant.process_offset = 0; - push_constant.process_increment = 1; - - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; - push_constant.light_count = cascade_light_count[i]; - push_constant.cascade = i; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascade.sdf_direct_light_uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::DirectLightPushConstant)); + RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cascade.solid_cell_dispatch_buffer, 0); + } + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_COMPUTE); + RD::get_singleton()->draw_command_end_label(); +} - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascade.sdf_direct_light_uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::DirectLightPushConstant)); - RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cascade.solid_cell_dispatch_buffer, 0); - } - RD::get_singleton()->compute_list_end(); +void RendererSceneRenderRD::_sdfgi_update_probes(RID p_render_buffers, RID p_environment) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(rb == nullptr); + if (rb->sdfgi == nullptr) { + return; } - RENDER_TIMESTAMP("Raytrace"); + RD::get_singleton()->draw_command_begin_label("SDFGI Update Probes"); + + Environment *env = environment_owner.getornull(p_environment); SDGIShader::IntegratePushConstant push_constant; push_constant.grid_size[1] = rb->sdfgi->cascade_size; @@ -1303,7 +1222,7 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; push_constant.history_index = rb->sdfgi->render_pass % rb->sdfgi->history_size; push_constant.history_size = rb->sdfgi->history_size; - static const uint32_t ray_count[RS::ENV_SDFGI_RAY_COUNT_MAX] = { 8, 16, 32, 64, 96, 128 }; + static const uint32_t ray_count[RS::ENV_SDFGI_RAY_COUNT_MAX] = { 4, 8, 16, 32, 64, 96, 128 }; push_constant.ray_count = ray_count[sdfgi_ray_count]; push_constant.ray_bias = rb->sdfgi->probe_bias; push_constant.image_size[0] = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count; @@ -1362,7 +1281,7 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi rb->sdfgi->render_pass++; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(true); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_PROCESS]); int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR; @@ -1376,14 +1295,47 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sky_uniform_set, 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::IntegratePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1); + } + + //end later after raster to avoid barriering on layout changes + //RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_sdfgi_store_probes(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(rb == nullptr); + if (rb->sdfgi == nullptr) { + return; } - RD::get_singleton()->compute_list_add_barrier(compute_list); //wait until done + RD::get_singleton()->barrier(RD::BARRIER_MASK_COMPUTE, RD::BARRIER_MASK_COMPUTE); + RD::get_singleton()->draw_command_begin_label("SDFGI Store Probes"); + + SDGIShader::IntegratePushConstant push_constant; + push_constant.grid_size[1] = rb->sdfgi->cascade_size; + push_constant.grid_size[2] = rb->sdfgi->cascade_size; + push_constant.grid_size[0] = rb->sdfgi->cascade_size; + push_constant.max_cascades = rb->sdfgi->cascades.size(); + push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; + push_constant.history_index = rb->sdfgi->render_pass % rb->sdfgi->history_size; + push_constant.history_size = rb->sdfgi->history_size; + static const uint32_t ray_count[RS::ENV_SDFGI_RAY_COUNT_MAX] = { 4, 8, 16, 32, 64, 96, 128 }; + push_constant.ray_count = ray_count[sdfgi_ray_count]; + push_constant.ray_bias = rb->sdfgi->probe_bias; + push_constant.image_size[0] = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count; + push_constant.image_size[1] = rb->sdfgi->probe_axis_count; + push_constant.store_ambient_texture = false; + + push_constant.sky_mode = 0; + push_constant.y_mult = rb->sdfgi->y_mult; // Then store values into the lightprobe texture. Separating these steps has a small performance hit, but it allows for multiple bounces RENDER_TIMESTAMP("Average Probes"); + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_STORE]); //convert to octahedral to store @@ -1393,20 +1345,22 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { push_constant.cascade = i; RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[i].integrate_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_sky_uniform_set, 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::IntegratePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1); } - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_COMPUTE); - RENDER_TIMESTAMP("<SDFGI Update Probes"); + RD::get_singleton()->draw_command_end_label(); } - void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used) { r_gi_probes_used = 0; RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(rb == nullptr); + RD::get_singleton()->draw_command_begin_label("GIProbes Setup"); + RID gi_probe_buffer = render_buffers_get_gi_probe_buffer(p_render_buffers); GI::GIProbeData gi_probe_data[RenderBuffers::MAX_GIPROBES]; @@ -1490,56 +1444,25 @@ void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transfor } if (p_gi_probes.size() > 0) { - RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GI::GIProbeData) * MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()), gi_probe_data, true); + RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GI::GIProbeData) * MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()), gi_probe_data, RD::BARRIER_MASK_COMPUTE); } -} -void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes) { - RENDER_TIMESTAMP("Render GI"); + RD::get_singleton()->draw_command_end_label(); +} +void RendererSceneRenderRD::_pre_process_gi(RID p_render_buffers, const Transform &p_transform) { + // Do the required buffer transfers and setup before the depth-pre pass, this way GI can + // run in parallel during depth-pre pass and shadow rendering. RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(rb == nullptr); - Environment *env = environment_owner.getornull(p_environment); - GI::PushConstant push_constant; - - push_constant.screen_size[0] = rb->width; - push_constant.screen_size[1] = rb->height; - push_constant.z_near = p_projection.get_z_near(); - push_constant.z_far = p_projection.get_z_far(); - push_constant.orthogonal = p_projection.is_orthogonal(); - push_constant.proj_info[0] = -2.0f / (rb->width * p_projection.matrix[0][0]); - push_constant.proj_info[1] = -2.0f / (rb->height * p_projection.matrix[1][1]); - push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0]; - push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1]; - push_constant.max_giprobes = MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()); - push_constant.high_quality_vct = gi_probe_quality == RS::GI_PROBE_QUALITY_HIGH; - push_constant.use_sdfgi = rb->sdfgi != nullptr; + /* Update Cascades UBO */ - if (env) { - push_constant.ao_color[0] = env->ao_color.r; - push_constant.ao_color[1] = env->ao_color.g; - push_constant.ao_color[2] = env->ao_color.b; - } else { - push_constant.ao_color[0] = 0; - push_constant.ao_color[1] = 0; - push_constant.ao_color[2] = 0; - } + if (rb->sdfgi) { + /* Update general SDFGI Buffer */ - push_constant.cam_rotation[0] = p_transform.basis[0][0]; - push_constant.cam_rotation[1] = p_transform.basis[1][0]; - push_constant.cam_rotation[2] = p_transform.basis[2][0]; - push_constant.cam_rotation[3] = 0; - push_constant.cam_rotation[4] = p_transform.basis[0][1]; - push_constant.cam_rotation[5] = p_transform.basis[1][1]; - push_constant.cam_rotation[6] = p_transform.basis[2][1]; - push_constant.cam_rotation[7] = 0; - push_constant.cam_rotation[8] = p_transform.basis[0][2]; - push_constant.cam_rotation[9] = p_transform.basis[1][2]; - push_constant.cam_rotation[10] = p_transform.basis[2][2]; - push_constant.cam_rotation[11] = 0; + _sdfgi_update_cascades(p_render_buffers); - if (rb->sdfgi) { GI::SDFGIData sdfgi_data; sdfgi_data.grid_size[0] = rb->sdfgi->cascade_size; @@ -1606,9 +1529,172 @@ void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_rough c.to_cell = 1.0 / rb->sdfgi->cascades[i].cell_size; } - RD::get_singleton()->buffer_update(gi.sdfgi_ubo, 0, sizeof(GI::SDFGIData), &sdfgi_data, true); + RD::get_singleton()->buffer_update(gi.sdfgi_ubo, 0, sizeof(GI::SDFGIData), &sdfgi_data, RD::BARRIER_MASK_COMPUTE); + + /* Update dynamic lights in SDFGI cascades */ + + for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { + SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; + + SDGIShader::Light lights[SDFGI::MAX_DYNAMIC_LIGHTS]; + uint32_t idx = 0; + for (uint32_t j = 0; j < (uint32_t)render_state.sdfgi_update_data->directional_lights->size(); j++) { + if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { + break; + } + + LightInstance *li = light_instance_owner.getornull(render_state.sdfgi_update_data->directional_lights->get(j)); + ERR_CONTINUE(!li); + + if (storage->light_directional_is_sky_only(li->light)) { + continue; + } + + Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); + dir.y *= rb->sdfgi->y_mult; + dir.normalize(); + lights[idx].direction[0] = dir.x; + lights[idx].direction[1] = dir.y; + lights[idx].direction[2] = dir.z; + Color color = storage->light_get_color(li->light); + color = color.to_linear(); + lights[idx].color[0] = color.r; + lights[idx].color[1] = color.g; + lights[idx].color[2] = color.b; + lights[idx].type = RS::LIGHT_DIRECTIONAL; + lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); + lights[idx].has_shadow = storage->light_has_shadow(li->light); + + idx++; + } + + AABB cascade_aabb; + cascade_aabb.position = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + cascade.position)) * cascade.cell_size; + cascade_aabb.size = Vector3(1, 1, 1) * rb->sdfgi->cascade_size * cascade.cell_size; + + for (uint32_t j = 0; j < render_state.sdfgi_update_data->positional_light_count; j++) { + if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { + break; + } + + LightInstance *li = light_instance_owner.getornull(render_state.sdfgi_update_data->positional_light_instances[j]); + ERR_CONTINUE(!li); + + uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light); + if (i > max_sdfgi_cascade) { + continue; + } + + if (!cascade_aabb.intersects(li->aabb)) { + continue; + } + + Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); + //faster to not do this here + //dir.y *= rb->sdfgi->y_mult; + //dir.normalize(); + lights[idx].direction[0] = dir.x; + lights[idx].direction[1] = dir.y; + lights[idx].direction[2] = dir.z; + Vector3 pos = li->transform.origin; + pos.y *= rb->sdfgi->y_mult; + lights[idx].position[0] = pos.x; + lights[idx].position[1] = pos.y; + lights[idx].position[2] = pos.z; + Color color = storage->light_get_color(li->light); + color = color.to_linear(); + lights[idx].color[0] = color.r; + lights[idx].color[1] = color.g; + lights[idx].color[2] = color.b; + lights[idx].type = storage->light_get_type(li->light); + lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); + lights[idx].has_shadow = storage->light_has_shadow(li->light); + lights[idx].attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_ATTENUATION); + lights[idx].radius = storage->light_get_param(li->light, RS::LIGHT_PARAM_RANGE); + lights[idx].cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE))); + lights[idx].inv_spot_attenuation = 1.0f / storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + + idx++; + } + + if (idx > 0) { + RD::get_singleton()->buffer_update(cascade.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights, RD::BARRIER_MASK_COMPUTE); + } + + rb->sdfgi->cascade_dynamic_light_count[i] = idx; + } + } +} + +void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes) { + RD::get_singleton()->draw_command_begin_label("GI Render"); + + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(rb == nullptr); + Environment *env = environment_owner.getornull(p_environment); + + if (rb->ambient_buffer.is_null() || rb->using_half_size_gi != gi.half_resolution) { + if (rb->ambient_buffer.is_valid()) { + RD::get_singleton()->free(rb->ambient_buffer); + RD::get_singleton()->free(rb->reflection_buffer); + } + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.width = rb->width; + tf.height = rb->height; + if (gi.half_resolution) { + tf.width >>= 1; + tf.height >>= 1; + } + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + rb->reflection_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); + rb->ambient_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); + rb->using_half_size_gi = gi.half_resolution; + + _render_buffers_uniform_set_changed(p_render_buffers); + } + + GI::PushConstant push_constant; + + push_constant.screen_size[0] = rb->width; + push_constant.screen_size[1] = rb->height; + push_constant.z_near = p_projection.get_z_near(); + push_constant.z_far = p_projection.get_z_far(); + push_constant.orthogonal = p_projection.is_orthogonal(); + push_constant.proj_info[0] = -2.0f / (rb->width * p_projection.matrix[0][0]); + push_constant.proj_info[1] = -2.0f / (rb->height * p_projection.matrix[1][1]); + push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0]; + push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1]; + push_constant.max_giprobes = MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()); + push_constant.high_quality_vct = gi_probe_quality == RS::GI_PROBE_QUALITY_HIGH; + + bool use_sdfgi = rb->sdfgi != nullptr; + bool use_giprobes = push_constant.max_giprobes > 0; + + if (env) { + push_constant.ao_color[0] = env->ao_color.r; + push_constant.ao_color[1] = env->ao_color.g; + push_constant.ao_color[2] = env->ao_color.b; + } else { + push_constant.ao_color[0] = 0; + push_constant.ao_color[1] = 0; + push_constant.ao_color[2] = 0; } + push_constant.cam_rotation[0] = p_transform.basis[0][0]; + push_constant.cam_rotation[1] = p_transform.basis[1][0]; + push_constant.cam_rotation[2] = p_transform.basis[2][0]; + push_constant.cam_rotation[3] = 0; + push_constant.cam_rotation[4] = p_transform.basis[0][1]; + push_constant.cam_rotation[5] = p_transform.basis[1][1]; + push_constant.cam_rotation[6] = p_transform.basis[2][1]; + push_constant.cam_rotation[7] = 0; + push_constant.cam_rotation[8] = p_transform.basis[0][2]; + push_constant.cam_rotation[9] = p_transform.basis[1][2]; + push_constant.cam_rotation[10] = p_transform.basis[2][2]; + push_constant.cam_rotation[11] = 0; + if (rb->gi_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(rb->gi_uniform_set)) { Vector<RD::Uniform> uniforms; { @@ -1693,7 +1779,7 @@ void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_rough RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; u.binding = 9; - u.ids.push_back(p_ambient_buffer); + u.ids.push_back(rb->ambient_buffer); uniforms.push_back(u); } @@ -1701,7 +1787,7 @@ void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_rough RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; u.binding = 10; - u.ids.push_back(p_reflection_buffer); + u.ids.push_back(rb->reflection_buffer); uniforms.push_back(u); } @@ -1765,12 +1851,26 @@ void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_rough rb->gi_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi.shader.version_get_shader(gi.shader_version, 0), 0); } - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi.pipelines[0]); + GI::Mode mode; + + if (rb->using_half_size_gi) { + mode = (use_sdfgi && use_giprobes) ? GI::MODE_HALF_RES_COMBINED : (use_sdfgi ? GI::MODE_HALF_RES_SDFGI : GI::MODE_HALF_RES_GIPROBE); + } else { + mode = (use_sdfgi && use_giprobes) ? GI::MODE_COMBINED : (use_sdfgi ? GI::MODE_SDFGI : GI::MODE_GIPROBE); + } + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(true); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi.pipelines[mode]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->gi_uniform_set, 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GI::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1, 8, 8, 1); - RD::get_singleton()->compute_list_end(); + + if (rb->using_half_size_gi) { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width >> 1, rb->height >> 1, 1); + } else { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1); + } + //do barrier later to allow oeverlap + //RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //no barriers, let other compute, raster and transfer happen at the same time + RD::get_singleton()->draw_command_end_label(); } RID RendererSceneRenderRD::sky_create() { @@ -2288,7 +2388,7 @@ void RendererSceneRenderRD::_setup_sky(RID p_environment, RID p_render_buffers, } if (light_data_dirty) { - RD::get_singleton()->buffer_update(sky_scene_state.directional_light_buffer, 0, sizeof(SkyDirectionalLightData) * sky_scene_state.max_directional_lights, sky_scene_state.directional_lights, true); + RD::get_singleton()->buffer_update(sky_scene_state.directional_light_buffer, 0, sizeof(SkyDirectionalLightData) * sky_scene_state.max_directional_lights, sky_scene_state.directional_lights); RendererSceneRenderRD::SkyDirectionalLightData *temp = sky_scene_state.last_frame_directional_lights; sky_scene_state.last_frame_directional_lights = sky_scene_state.directional_lights; @@ -2340,7 +2440,7 @@ void RendererSceneRenderRD::_setup_sky(RID p_environment, RID p_render_buffers, sky_scene_state.ubo.fog_light_color[2] = fog_color.b * fog_energy; sky_scene_state.ubo.fog_sun_scatter = environment_get_fog_sun_scatter(p_environment); - RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo, true); + RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo); } void RendererSceneRenderRD::_update_sky(RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform) { @@ -2674,6 +2774,12 @@ Variant RendererSceneRenderRD::SkyShaderData::get_default_parameter(const String return Variant(); } +RS::ShaderNativeSourceCode RendererSceneRenderRD::SkyShaderData::get_native_source_code() const { + RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; + + return scene_singleton->sky_shader.shader.version_get_native_source_code(version); +} + RendererSceneRenderRD::SkyShaderData::SkyShaderData() { valid = false; } @@ -3045,7 +3151,7 @@ float RendererSceneRenderRD::environment_get_fog_aerial_perspective(RID p_env) c return env->fog_aerial_perspective; } -void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RenderingServer::EnvVolumetricFogShadowFilter p_shadow_filter) { +void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) { Environment *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); @@ -3059,8 +3165,9 @@ void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_ena env->volumetric_fog_light_energy = p_light_energy; env->volumetric_fog_length = p_length; env->volumetric_fog_detail_spread = p_detail_spread; - env->volumetric_fog_shadow_filter = p_shadow_filter; env->volumetric_fog_gi_inject = p_gi_inject; + env->volumetric_fog_temporal_reprojection = p_temporal_reprojection; + env->volumetric_fog_temporal_reprojection_amount = p_temporal_reprojection_amount; } void RendererSceneRenderRD::environment_set_volumetric_fog_volume_size(int p_size, int p_depth) { @@ -3071,25 +3178,6 @@ void RendererSceneRenderRD::environment_set_volumetric_fog_volume_size(int p_siz void RendererSceneRenderRD::environment_set_volumetric_fog_filter_active(bool p_enable) { volumetric_fog_filter_active = p_enable; } -void RendererSceneRenderRD::environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size) { - p_shrink_size = nearest_power_of_2_templated(p_shrink_size); - if (volumetric_fog_directional_shadow_shrink == (uint32_t)p_shrink_size) { - return; - } - - _clear_shadow_shrink_stages(directional_shadow.shrink_stages); -} -void RendererSceneRenderRD::environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size) { - p_shrink_size = nearest_power_of_2_templated(p_shrink_size); - if (volumetric_fog_positional_shadow_shrink == (uint32_t)p_shrink_size) { - return; - } - - for (uint32_t i = 0; i < shadow_atlas_owner.get_rid_count(); i++) { - ShadowAtlas *sa = shadow_atlas_owner.get_ptr_by_index(i); - _clear_shadow_shrink_stages(sa->shrink_stages); - } -} void RendererSceneRenderRD::environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) { sdfgi_ray_count = p_ray_count; @@ -3098,6 +3186,9 @@ void RendererSceneRenderRD::environment_set_sdfgi_ray_count(RS::EnvironmentSDFGI void RendererSceneRenderRD::environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) { sdfgi_frames_to_converge = p_frames; } +void RendererSceneRenderRD::environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) { + sdfgi_frames_to_update_light = p_update; +} void RendererSceneRenderRD::environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance) { Environment *env = environment_owner.getornull(p_env); @@ -3227,6 +3318,10 @@ RID RendererSceneRenderRD::reflection_atlas_create() { ra.count = GLOBAL_GET("rendering/quality/reflection_atlas/reflection_count"); ra.size = GLOBAL_GET("rendering/quality/reflection_atlas/reflection_size"); + ra.cluster_builder = memnew(ClusterBuilderRD); + ra.cluster_builder->set_shared(&cluster_builder_shared); + ra.cluster_builder->setup(Size2i(ra.size, ra.size), max_cluster_elements, RID(), RID(), RID()); + return reflection_atlas_owner.make_rid(ra); } @@ -3238,6 +3333,8 @@ void RendererSceneRenderRD::reflection_atlas_set_size(RID p_ref_atlas, int p_ref return; //no changes } + ra->cluster_builder->setup(Size2i(ra->size, ra->size), max_cluster_elements, RID(), RID(), RID()); + ra->size = p_reflection_size; ra->count = p_reflection_count; @@ -3247,7 +3344,6 @@ void RendererSceneRenderRD::reflection_atlas_set_size(RID p_ref_atlas, int p_ref ra->reflection = RID(); RD::get_singleton()->free(ra->depth_buffer); ra->depth_buffer = RID(); - for (int i = 0; i < ra->reflections.size(); i++) { _clear_reflection_data(ra->reflections.write[i].data); if (ra->reflections[i].owner.is_null()) { @@ -3504,13 +3600,28 @@ RID RendererSceneRenderRD::shadow_atlas_create() { return shadow_atlas_owner.make_rid(ShadowAtlas()); } -void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size) { +void RendererSceneRenderRD::_update_shadow_atlas(ShadowAtlas *shadow_atlas) { + if (shadow_atlas->size > 0 && shadow_atlas->depth.is_null()) { + RD::TextureFormat tf; + tf.format = shadow_atlas->use_16_bits ? RD::DATA_FORMAT_D16_UNORM : RD::DATA_FORMAT_D32_SFLOAT; + tf.width = shadow_atlas->size; + tf.height = shadow_atlas->size; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + + shadow_atlas->depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + Vector<RID> fb_tex; + fb_tex.push_back(shadow_atlas->depth); + shadow_atlas->fb = RD::get_singleton()->framebuffer_create(fb_tex); + } +} + +void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits) { ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_atlas); ERR_FAIL_COND(!shadow_atlas); ERR_FAIL_COND(p_size < 0); p_size = next_power_of_2(p_size); - if (p_size == shadow_atlas->size) { + if (p_size == shadow_atlas->size && p_16_bits == shadow_atlas->use_16_bits) { return; } @@ -3518,7 +3629,6 @@ void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size) { if (shadow_atlas->depth.is_valid()) { RD::get_singleton()->free(shadow_atlas->depth); shadow_atlas->depth = RID(); - _clear_shadow_shrink_stages(shadow_atlas->shrink_stages); } for (int i = 0; i < 4; i++) { //clear subdivisions @@ -3537,16 +3647,7 @@ void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size) { shadow_atlas->shadow_owners.clear(); shadow_atlas->size = p_size; - - if (shadow_atlas->size) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = shadow_atlas->size; - tf.height = shadow_atlas->size; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - - shadow_atlas->depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } + shadow_atlas->use_16_bits = p_size; } void RendererSceneRenderRD::shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) { @@ -3801,10 +3902,24 @@ bool RendererSceneRenderRD::shadow_atlas_update_light(RID p_atlas, RID p_light_i return false; } -void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size) { +void RendererSceneRenderRD::_update_directional_shadow_atlas() { + if (directional_shadow.depth.is_null() && directional_shadow.size > 0) { + RD::TextureFormat tf; + tf.format = directional_shadow.use_16_bits ? RD::DATA_FORMAT_D16_UNORM : RD::DATA_FORMAT_D32_SFLOAT; + tf.width = directional_shadow.size; + tf.height = directional_shadow.size; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + + directional_shadow.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + Vector<RID> fb_tex; + fb_tex.push_back(directional_shadow.depth); + directional_shadow.fb = RD::get_singleton()->framebuffer_create(fb_tex); + } +} +void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size, bool p_16_bits) { p_size = nearest_power_of_2_templated(p_size); - if (directional_shadow.size == p_size) { + if (directional_shadow.size == p_size && directional_shadow.use_16_bits == p_16_bits) { return; } @@ -3812,21 +3927,9 @@ void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size) { if (directional_shadow.depth.is_valid()) { RD::get_singleton()->free(directional_shadow.depth); - _clear_shadow_shrink_stages(directional_shadow.shrink_stages); directional_shadow.depth = RID(); + _base_uniforms_changed(); } - - if (p_size > 0) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = p_size; - tf.height = p_size; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - - directional_shadow.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } - - _base_uniforms_changed(); } void RendererSceneRenderRD::set_directional_shadow_count(int p_count) { @@ -3946,11 +4049,7 @@ void RendererSceneRenderRD::light_instance_set_shadow_transform(RID p_light_inst LightInstance *light_instance = light_instance_owner.getornull(p_light_instance); ERR_FAIL_COND(!light_instance); - if (storage->light_get_type(light_instance->light) != RS::LIGHT_DIRECTIONAL) { - p_pass = 0; - } - - ERR_FAIL_INDEX(p_pass, 4); + ERR_FAIL_INDEX(p_pass, 6); light_instance->shadow_transform[p_pass].camera = p_projection; light_instance->shadow_transform[p_pass].transform = p_transform; @@ -3996,29 +4095,6 @@ RendererSceneRenderRD::ShadowCubemap *RendererSceneRenderRD::_get_shadow_cubemap return &shadow_cubemaps[p_size]; } -RendererSceneRenderRD::ShadowMap *RendererSceneRenderRD::_get_shadow_map(const Size2i &p_size) { - if (!shadow_maps.has(p_size)) { - ShadowMap sm; - { - RD::TextureFormat tf; - tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D32_SFLOAT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D32_SFLOAT : RD::DATA_FORMAT_X8_D24_UNORM_PACK32; - tf.width = p_size.width; - tf.height = p_size.height; - tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; - - sm.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } - - Vector<RID> fbtex; - fbtex.push_back(sm.depth); - sm.fb = RD::get_singleton()->framebuffer_create(fbtex); - - shadow_maps[p_size] = sm; - } - - return &shadow_maps[p_size]; -} - ////////////////////////// RID RendererSceneRenderRD::decal_instance_create(RID p_decal) { @@ -4035,6 +4111,19 @@ void RendererSceneRenderRD::decal_instance_set_transform(RID p_decal, const Tran ///////////////////////////////// +RID RendererSceneRenderRD::lightmap_instance_create(RID p_lightmap) { + LightmapInstance li; + li.lightmap = p_lightmap; + return lightmap_instance_owner.make_rid(li); +} +void RendererSceneRenderRD::lightmap_instance_set_transform(RID p_lightmap, const Transform &p_transform) { + LightmapInstance *li = lightmap_instance_owner.getornull(p_lightmap); + ERR_FAIL_COND(!li); + li->transform = p_transform; +} + +///////////////////////////////// + RID RendererSceneRenderRD::gi_probe_instance_create(RID p_base) { GIProbeInstance gi_probe; gi_probe.probe = p_base; @@ -4061,7 +4150,7 @@ bool RendererSceneRenderRD::gi_probe_needs_update(RID p_probe) const { return gi_probe->last_probe_version != storage->gi_probe_get_version(gi_probe->probe); } -void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<InstanceBase *> &p_dynamic_objects) { +void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<GeometryInstance *> &p_dynamic_objects) { GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe); ERR_FAIL_COND(!gi_probe); @@ -4106,7 +4195,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins gi_probe->texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, levels.size(), 0, 1, false); + RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, levels.size(), 0, 1); { int total_elements = 0; @@ -4397,7 +4486,10 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins } } - dmap.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_lighting_shader_version_shaders[(write && plot) ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT : write ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE : GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT], 0); + dmap.uniform_set = RD::get_singleton()->uniform_set_create( + uniforms, + giprobe_lighting_shader_version_shaders[(write && plot) ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT : (write ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE : GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT)], + 0); } gi_probe->dynamic_maps.push_back(dmap); @@ -4415,7 +4507,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins if (gi_probe->has_dynamic_object_data) { //if it has dynamic object data, it needs to be cleared - RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, gi_probe->mipmaps.size(), 0, 1, true); + RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, gi_probe->mipmaps.size(), 0, 1); } uint32_t light_count = 0; @@ -4447,8 +4539,8 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins l.color[1] = color.g; l.color[2] = color.b; - l.spot_angle_radians = Math::deg2rad(storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE)); - l.spot_attenuation = storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + l.cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE))); + l.inv_spot_attenuation = 1.0f / storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); Transform xform = light_instance_get_base_transform(light_instance); @@ -4466,7 +4558,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins l.has_shadow = storage->light_has_shadow(light); } - RD::get_singleton()->buffer_update(gi_probe_lights_uniform, 0, sizeof(GIProbeLight) * light_count, gi_probe_lights, true); + RD::get_singleton()->buffer_update(gi_probe_lights_uniform, 0, sizeof(GIProbeLight) * light_count, gi_probe_lights); } } @@ -4578,13 +4670,10 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins //this could probably be better parallelized in compute.. for (int i = 0; i < (int)p_dynamic_objects.size(); i++) { - InstanceBase *instance = p_dynamic_objects[i]; - //not used, so clear - instance->depth_layer = 0; - instance->depth = 0; + GeometryInstance *instance = p_dynamic_objects[i]; //transform aabb to giprobe - AABB aabb = (to_probe_xform * instance->transform).xform(instance->aabb); + AABB aabb = (to_probe_xform * geometry_instance_get_transform(instance)).xform(geometry_instance_get_aabb(instance)); //this needs to wrap to grid resolution to avoid jitter //also extend margin a bit just in case @@ -4834,7 +4923,16 @@ void RendererSceneRenderRD::_debug_giprobe(RID p_gi_probe, RD::DrawListID p_draw } giprobe_debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_debug_shader_version_shaders[0], 0); - RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, giprobe_debug_shader_version_pipelines[p_emission ? GI_PROBE_DEBUG_EMISSION : p_lighting ? (gi_probe->has_dynamic_object_data ? GI_PROBE_DEBUG_LIGHT_FULL : GI_PROBE_DEBUG_LIGHT) : GI_PROBE_DEBUG_COLOR].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer))); + + int giprobe_debug_pipeline = GI_PROBE_DEBUG_COLOR; + if (p_emission) { + giprobe_debug_pipeline = GI_PROBE_DEBUG_EMISSION; + } else if (p_lighting) { + giprobe_debug_pipeline = gi_probe->has_dynamic_object_data ? GI_PROBE_DEBUG_LIGHT_FULL : GI_PROBE_DEBUG_LIGHT; + } + RD::get_singleton()->draw_list_bind_render_pipeline( + p_draw_list, + giprobe_debug_shader_version_pipelines[giprobe_debug_pipeline].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer))); RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, giprobe_debug_uniform_set, 0); RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(GIProbeDebugPushConstant)); RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, 36); @@ -4861,7 +4959,7 @@ void RendererSceneRenderRD::_debug_sdfgi_probes(RID p_render_buffers, RD::DrawLi push_constant.band_power = 4; push_constant.sections_in_band = ((band_points / 2) - 1); push_constant.band_mask = band_points - 2; - push_constant.section_arc = (Math_PI * 2.0) / float(push_constant.sections_in_band); + push_constant.section_arc = Math_TAU / float(push_constant.sections_in_band); push_constant.y_mult = rb->sdfgi->y_mult; uint32_t total_points = push_constant.sections_in_band * band_points; @@ -5082,9 +5180,6 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { RD::get_singleton()->free(rb->luminance.reduce[i]); } - for (int i = 0; i < rb->luminance.reduce.size(); i++) { - RD::get_singleton()->free(rb->luminance.reduce[i]); - } rb->luminance.reduce.clear(); if (rb->luminance.current.is_valid()) { @@ -5125,6 +5220,13 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { RD::get_singleton()->free(rb->ssr.normal_scaled); rb->ssr.normal_scaled = RID(); } + + if (rb->ambient_buffer.is_valid()) { + RD::get_singleton()->free(rb->ambient_buffer); + RD::get_singleton()->free(rb->reflection_buffer); + rb->ambient_buffer = RID(); + rb->reflection_buffer = RID(); + } } void RendererSceneRenderRD::_process_sss(RID p_render_buffers, const CameraMatrix &p_camera) { @@ -5257,9 +5359,11 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.array_layers = 4; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.depth, "SSAO Depth"); for (uint32_t i = 0; i < tf.mipmaps; i++) { RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.depth, 0, i, RD::TEXTURE_SLICE_2D_ARRAY); rb->ssao.depth_slices.push_back(slice); + RD::get_singleton()->set_resource_name(rb->ssao.depth_slices[i], "SSAO Depth Mip " + itos(i) + " "); } } @@ -5272,9 +5376,11 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.array_layers = 4; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_deinterleaved = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved, "SSAO De-interleaved Array"); for (uint32_t i = 0; i < 4; i++) { RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.ao_deinterleaved, i, 0); rb->ssao.ao_deinterleaved_slices.push_back(slice); + RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved_slices[i], "SSAO De-interleaved Array Layer " + itos(i) + " "); } } @@ -5287,9 +5393,11 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.array_layers = 4; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_pong = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.ao_pong, "SSAO De-interleaved Array Pong"); for (uint32_t i = 0; i < 4; i++) { RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.ao_pong, i, 0); rb->ssao.ao_pong_slices.push_back(slice); + RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved_slices[i], "SSAO De-interleaved Array Layer " + itos(i) + " Pong"); } } @@ -5300,7 +5408,9 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.height = half_height; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.importance_map[0] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.importance_map[0], "SSAO Importance Map"); rb->ssao.importance_map[1] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.importance_map[1], "SSAO Importance Map Pong"); } { RD::TextureFormat tf; @@ -5309,6 +5419,7 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.height = rb->height; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_final = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.ao_final, "SSAO Final"); _render_buffers_uniform_set_changed(p_render_buffers); } ssao_using_half_size = ssao_half_size; @@ -5329,9 +5440,9 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen settings.blur_passes = ssao_blur_passes; settings.fadeout_from = ssao_fadeout_from; settings.fadeout_to = ssao_fadeout_to; - settings.screen_size = Size2i(rb->width, rb->height); + settings.full_screen_size = Size2i(rb->width, rb->height); settings.half_screen_size = Size2i(buffer_width, buffer_height); - settings.quarter_size = Size2i(half_width, half_height); + settings.quarter_screen_size = Size2i(half_width, half_height); storage->get_effects()->generate_ssao(rb->depth_texture, p_normal_buffer, rb->ssao.depth, rb->ssao.depth_slices, rb->ssao.ao_deinterleaved, rb->ssao.ao_deinterleaved_slices, rb->ssao.ao_pong, rb->ssao.ao_pong_slices, rb->ssao.ao_final, rb->ssao.importance_map[0], rb->ssao.importance_map[1], p_projection, settings, uniform_sets_are_invalid); } @@ -5524,10 +5635,10 @@ void RendererSceneRenderRD::_render_buffers_debug_draw(RID p_render_buffers, RID effects->copy_to_fb_rect(_render_buffers_get_normal_texture(p_render_buffers), storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, false); } - if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && _render_buffers_get_ambient_texture(p_render_buffers).is_valid()) { + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && rb->ambient_buffer.is_valid()) { Size2 rtsize = storage->render_target_get_size(rb->render_target); - RID ambient_texture = _render_buffers_get_ambient_texture(p_render_buffers); - RID reflection_texture = _render_buffers_get_reflection_texture(p_render_buffers); + RID ambient_texture = rb->ambient_buffer; + RID reflection_texture = rb->reflection_buffer; effects->copy_to_fb_rect(ambient_texture, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, false, false, true, reflection_texture); } } @@ -5683,7 +5794,7 @@ void RendererSceneRenderRD::_sdfgi_debug_draw(RID p_render_buffers, const Camera RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::DebugPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1); RD::get_singleton()->compute_list_end(); Size2 rtsize = storage->render_target_get_size(rb->render_target); @@ -5719,6 +5830,17 @@ RID RendererSceneRenderRD::render_buffers_get_default_gi_probe_buffer() { return default_giprobe_buffer; } +RID RendererSceneRenderRD::render_buffers_get_gi_ambient_texture(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND_V(!rb, RID()); + return rb->ambient_buffer; +} +RID RendererSceneRenderRD::render_buffers_get_gi_reflection_texture(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND_V(!rb, RID()); + return rb->reflection_buffer; +} + uint32_t RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_count(RID p_render_buffers) const { const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND_V(!rb, 0); @@ -5856,6 +5978,11 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p rb->msaa = p_msaa; rb->screen_space_aa = p_screen_space_aa; rb->use_debanding = p_use_debanding; + if (rb->cluster_builder == nullptr) { + rb->cluster_builder = memnew(ClusterBuilderRD); + } + rb->cluster_builder->set_shared(&cluster_builder_shared); + _free_render_buffer_data(rb); { @@ -5896,6 +6023,12 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p rb->data->configure(rb->texture, rb->depth_texture, p_width, p_height, p_msaa); _render_buffers_uniform_set_changed(p_render_buffers); + + rb->cluster_builder->setup(Size2i(p_width, p_height), max_cluster_elements, rb->depth_texture, storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED), rb->texture); +} + +void RendererSceneRenderRD::gi_set_use_half_resolution(bool p_enable) { + gi.half_resolution = p_enable; } void RendererSceneRenderRD::sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality) { @@ -6006,17 +6139,34 @@ RendererSceneRenderRD::RenderBufferData *RendererSceneRenderRD::render_buffers_g } void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment) { + cluster.reflection_count = 0; + for (uint32_t i = 0; i < (uint32_t)p_reflections.size(); i++) { - RID rpi = p_reflections[i]; + if (cluster.reflection_count == cluster.max_reflections) { + break; + } - if (i >= cluster.max_reflections) { - reflection_probe_instance_set_render_index(rpi, 0); //invalid, but something needs to be set + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_reflections[i]); + if (!rpi) { continue; } - reflection_probe_instance_set_render_index(rpi, i); + cluster.reflection_sort[cluster.reflection_count].instance = rpi; + cluster.reflection_sort[cluster.reflection_count].depth = -p_camera_inverse_transform.xform(rpi->transform.origin).z; + cluster.reflection_count++; + } + + if (cluster.reflection_count > 0) { + SortArray<Cluster::InstanceSort<ReflectionProbeInstance>> sort_array; + sort_array.sort(cluster.reflection_sort, cluster.reflection_count); + } + + for (uint32_t i = 0; i < cluster.reflection_count; i++) { + ReflectionProbeInstance *rpi = cluster.reflection_sort[i].instance; - RID base_probe = reflection_probe_instance_get_probe(rpi); + rpi->render_index = i; + + RID base_probe = rpi->probe; Cluster::ReflectionData &reflection_ubo = cluster.reflections[i]; @@ -6025,7 +6175,7 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti reflection_ubo.box_extents[0] = extents.x; reflection_ubo.box_extents[1] = extents.y; reflection_ubo.box_extents[2] = extents.z; - reflection_ubo.index = reflection_probe_instance_get_atlas_index(rpi); + reflection_ubo.index = rpi->atlas_index; Vector3 origin_offset = storage->reflection_probe_get_origin_offset(base_probe); @@ -6034,46 +6184,50 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti reflection_ubo.box_offset[2] = origin_offset.z; reflection_ubo.mask = storage->reflection_probe_get_cull_mask(base_probe); - float intensity = storage->reflection_probe_get_intensity(base_probe); - bool interior = storage->reflection_probe_is_interior(base_probe); - bool box_projection = storage->reflection_probe_is_box_projection(base_probe); + reflection_ubo.intensity = storage->reflection_probe_get_intensity(base_probe); + reflection_ubo.ambient_mode = storage->reflection_probe_get_ambient_mode(base_probe); - reflection_ubo.params[0] = intensity; - reflection_ubo.params[1] = 0; - reflection_ubo.params[2] = interior ? 1.0 : 0.0; - reflection_ubo.params[3] = box_projection ? 1.0 : 0.0; + reflection_ubo.exterior = !storage->reflection_probe_is_interior(base_probe); + reflection_ubo.box_project = storage->reflection_probe_is_box_projection(base_probe); Color ambient_linear = storage->reflection_probe_get_ambient_color(base_probe).to_linear(); float interior_ambient_energy = storage->reflection_probe_get_ambient_color_energy(base_probe); - uint32_t ambient_mode = storage->reflection_probe_get_ambient_mode(base_probe); reflection_ubo.ambient[0] = ambient_linear.r * interior_ambient_energy; reflection_ubo.ambient[1] = ambient_linear.g * interior_ambient_energy; reflection_ubo.ambient[2] = ambient_linear.b * interior_ambient_energy; - reflection_ubo.ambient_mode = ambient_mode; - Transform transform = reflection_probe_instance_get_transform(rpi); + Transform transform = rpi->transform; Transform proj = (p_camera_inverse_transform * transform).inverse(); RendererStorageRD::store_transform(proj, reflection_ubo.local_matrix); - cluster.builder.add_reflection_probe(transform, extents); + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_REFLECTION_PROBE, transform, extents); - reflection_probe_instance_set_render_pass(rpi, RSG::rasterizer->get_frame_number()); + rpi->last_pass = RSG::rasterizer->get_frame_number(); } - if (p_reflections.size()) { - RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, MIN(cluster.max_reflections, (unsigned int)p_reflections.size()) * sizeof(ReflectionData), cluster.reflections, true); + if (cluster.reflection_count) { + RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, cluster.reflection_count * sizeof(ReflectionData), cluster.reflections, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } -void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count) { - uint32_t light_count = 0; +void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count) { + Transform inverse_transform = p_camera_transform.affine_inverse(); + r_directional_light_count = 0; r_positional_light_count = 0; sky_scene_state.ubo.directional_light_count = 0; + Plane camera_plane(p_camera_transform.origin, -p_camera_transform.basis.get_axis(Vector3::AXIS_Z).normalized()); + + cluster.omni_light_count = 0; + cluster.spot_light_count = 0; + for (int i = 0; i < (int)p_lights.size(); i++) { - RID li = p_lights[i]; - RID base = light_instance_get_base_light(li); + LightInstance *li = light_instance_owner.getornull(p_lights[i]); + if (!li) { + continue; + } + RID base = li->light; ERR_CONTINUE(base.is_null()); @@ -6083,7 +6237,7 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const // Copy to SkyDirectionalLightData if (r_directional_light_count < sky_scene_state.max_directional_lights) { SkyDirectionalLightData &sky_light_data = sky_scene_state.directional_lights[r_directional_light_count]; - Transform light_transform = light_instance_get_base_transform(li); + Transform light_transform = li->transform; Vector3 world_direction = light_transform.basis.xform(Vector3(0, 0, 1)).normalized(); sky_light_data.direction[0] = world_direction.x; @@ -6119,9 +6273,9 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Cluster::DirectionalLightData &light_data = cluster.directional_lights[r_directional_light_count]; - Transform light_transform = light_instance_get_base_transform(li); + Transform light_transform = li->transform; - Vector3 direction = p_camera_inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, 1))).normalized(); + Vector3 direction = inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, 1))).normalized(); light_data.direction[0] = direction.x; light_data.direction[1] = direction.y; @@ -6200,28 +6354,28 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const int limit = smode == RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL ? 0 : (smode == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS ? 1 : 3); light_data.blend_splits = storage->light_directional_get_blend_splits(base); for (int j = 0; j < 4; j++) { - Rect2 atlas_rect = light_instance_get_directional_shadow_atlas_rect(li, j); - CameraMatrix matrix = light_instance_get_shadow_camera(li, j); - float split = light_instance_get_directional_shadow_split(li, MIN(limit, j)); + Rect2 atlas_rect = li->shadow_transform[j].atlas_rect; + CameraMatrix matrix = li->shadow_transform[j].camera; + float split = li->shadow_transform[MIN(limit, j)].split; CameraMatrix bias; bias.set_light_bias(); CameraMatrix rectm; rectm.set_light_atlas_rect(atlas_rect); - Transform modelview = (p_camera_inverse_transform * light_instance_get_shadow_transform(li, j)).inverse(); + Transform modelview = (inverse_transform * li->shadow_transform[j].transform).inverse(); CameraMatrix shadow_mtx = rectm * bias * matrix * modelview; light_data.shadow_split_offsets[j] = split; - float bias_scale = light_instance_get_shadow_bias_scale(li, j); + float bias_scale = li->shadow_transform[j].bias_scale; light_data.shadow_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * bias_scale; - light_data.shadow_normal_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * light_instance_get_directional_shadow_texel_size(li, j); + light_data.shadow_normal_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * li->shadow_transform[j].shadow_texel_size; light_data.shadow_transmittance_bias[j] = storage->light_get_transmittance_bias(base) * bias_scale; - light_data.shadow_z_range[j] = light_instance_get_shadow_range(li, j); - light_data.shadow_range_begin[j] = light_instance_get_shadow_range_begin(li, j); + light_data.shadow_z_range[j] = li->shadow_transform[j].farplane; + light_data.shadow_range_begin[j] = li->shadow_transform[j].range_begin; RendererStorageRD::store_camera(shadow_mtx, light_data.shadow_matrices[j]); - Vector2 uv_scale = light_instance_get_shadow_uv_scale(li, j); + Vector2 uv_scale = li->shadow_transform[j].uv_scale; uv_scale *= atlas_rect.size; //adapt to atlas size switch (j) { case 0: { @@ -6258,170 +6412,201 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const r_directional_light_count++; } break; - case RS::LIGHT_SPOT: case RS::LIGHT_OMNI: { - if (light_count >= cluster.max_lights) { + if (cluster.omni_light_count >= cluster.max_lights) { continue; } - Transform light_transform = light_instance_get_base_transform(li); + cluster.omni_light_sort[cluster.omni_light_count].instance = li; + cluster.omni_light_sort[cluster.omni_light_count].depth = camera_plane.distance_to(li->transform.origin); + cluster.omni_light_count++; + } break; + case RS::LIGHT_SPOT: { + if (cluster.spot_light_count >= cluster.max_lights) { + continue; + } - Cluster::LightData &light_data = cluster.lights[light_count]; - cluster.lights_instances[light_count] = li; + cluster.spot_light_sort[cluster.spot_light_count].instance = li; + cluster.spot_light_sort[cluster.spot_light_count].depth = camera_plane.distance_to(li->transform.origin); + cluster.spot_light_count++; + } break; + } - float sign = storage->light_is_negative(base) ? -1 : 1; - Color linear_col = storage->light_get_color(base).to_linear(); + li->last_pass = RSG::rasterizer->get_frame_number(); + } - light_data.attenuation_energy[0] = Math::make_half_float(storage->light_get_param(base, RS::LIGHT_PARAM_ATTENUATION)); - light_data.attenuation_energy[1] = Math::make_half_float(sign * storage->light_get_param(base, RS::LIGHT_PARAM_ENERGY) * Math_PI); + if (cluster.omni_light_count) { + SortArray<Cluster::InstanceSort<LightInstance>> sorter; + sorter.sort(cluster.omni_light_sort, cluster.omni_light_count); + } - light_data.color_specular[0] = MIN(uint32_t(linear_col.r * 255), 255); - light_data.color_specular[1] = MIN(uint32_t(linear_col.g * 255), 255); - light_data.color_specular[2] = MIN(uint32_t(linear_col.b * 255), 255); - light_data.color_specular[3] = MIN(uint32_t(storage->light_get_param(base, RS::LIGHT_PARAM_SPECULAR) * 255), 255); + if (cluster.spot_light_count) { + SortArray<Cluster::InstanceSort<LightInstance>> sorter; + sorter.sort(cluster.spot_light_sort, cluster.spot_light_count); + } - float radius = MAX(0.001, storage->light_get_param(base, RS::LIGHT_PARAM_RANGE)); - light_data.inv_radius = 1.0 / radius; + ShadowAtlas *shadow_atlas = nullptr; - Vector3 pos = p_camera_inverse_transform.xform(light_transform.origin); + if (p_shadow_atlas.is_valid() && p_using_shadows) { + shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); + } - light_data.position[0] = pos.x; - light_data.position[1] = pos.y; - light_data.position[2] = pos.z; + for (uint32_t i = 0; i < (cluster.omni_light_count + cluster.spot_light_count); i++) { + uint32_t index = (i < cluster.omni_light_count) ? i : i - (cluster.omni_light_count); + Cluster::LightData &light_data = (i < cluster.omni_light_count) ? cluster.omni_lights[index] : cluster.spot_lights[index]; + RS::LightType type = (i < cluster.omni_light_count) ? RS::LIGHT_OMNI : RS::LIGHT_SPOT; + LightInstance *li = (i < cluster.omni_light_count) ? cluster.omni_light_sort[index].instance : cluster.spot_light_sort[index].instance; + RID base = li->light; - Vector3 direction = p_camera_inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, -1))).normalized(); + Transform light_transform = li->transform; - light_data.direction[0] = direction.x; - light_data.direction[1] = direction.y; - light_data.direction[2] = direction.z; + float sign = storage->light_is_negative(base) ? -1 : 1; + Color linear_col = storage->light_get_color(base).to_linear(); - float size = storage->light_get_param(base, RS::LIGHT_PARAM_SIZE); + light_data.attenuation = storage->light_get_param(base, RS::LIGHT_PARAM_ATTENUATION); - light_data.size = size; + float energy = sign * storage->light_get_param(base, RS::LIGHT_PARAM_ENERGY) * Math_PI; - light_data.cone_attenuation_angle[0] = Math::make_half_float(storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ATTENUATION)); - float spot_angle = storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ANGLE); - light_data.cone_attenuation_angle[1] = Math::make_half_float(Math::cos(Math::deg2rad(spot_angle))); + light_data.color[0] = linear_col.r * energy; + light_data.color[1] = linear_col.g * energy; + light_data.color[2] = linear_col.b * energy; + light_data.specular_amount = storage->light_get_param(base, RS::LIGHT_PARAM_SPECULAR) * 2.0; - light_data.mask = storage->light_get_cull_mask(base); + float radius = MAX(0.001, storage->light_get_param(base, RS::LIGHT_PARAM_RANGE)); + light_data.inv_radius = 1.0 / radius; - light_data.atlas_rect[0] = 0; - light_data.atlas_rect[1] = 0; - light_data.atlas_rect[2] = 0; - light_data.atlas_rect[3] = 0; + Vector3 pos = inverse_transform.xform(light_transform.origin); - RID projector = storage->light_get_projector(base); + light_data.position[0] = pos.x; + light_data.position[1] = pos.y; + light_data.position[2] = pos.z; - if (projector.is_valid()) { - Rect2 rect = storage->decal_atlas_get_texture_rect(projector); + Vector3 direction = inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, -1))).normalized(); - if (type == RS::LIGHT_SPOT) { - light_data.projector_rect[0] = rect.position.x; - light_data.projector_rect[1] = rect.position.y + rect.size.height; //flip because shadow is flipped - light_data.projector_rect[2] = rect.size.width; - light_data.projector_rect[3] = -rect.size.height; - } else { - light_data.projector_rect[0] = rect.position.x; - light_data.projector_rect[1] = rect.position.y; - light_data.projector_rect[2] = rect.size.width; - light_data.projector_rect[3] = rect.size.height * 0.5; //used by dp, so needs to be half - } - } else { - light_data.projector_rect[0] = 0; - light_data.projector_rect[1] = 0; - light_data.projector_rect[2] = 0; - light_data.projector_rect[3] = 0; - } + light_data.direction[0] = direction.x; + light_data.direction[1] = direction.y; + light_data.direction[2] = direction.z; - if (p_using_shadows && p_shadow_atlas.is_valid() && shadow_atlas_owns_light_instance(p_shadow_atlas, li)) { - // fill in the shadow information + float size = storage->light_get_param(base, RS::LIGHT_PARAM_SIZE); - Color shadow_color = storage->light_get_shadow_color(base); + light_data.size = size; - light_data.shadow_color_enabled[0] = MIN(uint32_t(shadow_color.r * 255), 255); - light_data.shadow_color_enabled[1] = MIN(uint32_t(shadow_color.g * 255), 255); - light_data.shadow_color_enabled[2] = MIN(uint32_t(shadow_color.b * 255), 255); - light_data.shadow_color_enabled[3] = 255; + light_data.inv_spot_attenuation = 1.0f / storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ATTENUATION); + float spot_angle = storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ANGLE); + light_data.cos_spot_angle = Math::cos(Math::deg2rad(spot_angle)); - if (type == RS::LIGHT_SPOT) { - light_data.shadow_bias = (storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0); - float shadow_texel_size = Math::tan(Math::deg2rad(spot_angle)) * radius * 2.0; - shadow_texel_size *= light_instance_get_shadow_texel_size(li, p_shadow_atlas); + light_data.mask = storage->light_get_cull_mask(base); - light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size; + light_data.atlas_rect[0] = 0; + light_data.atlas_rect[1] = 0; + light_data.atlas_rect[2] = 0; + light_data.atlas_rect[3] = 0; - } else { //omni - light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0; - float shadow_texel_size = light_instance_get_shadow_texel_size(li, p_shadow_atlas); - light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size * 2.0; // applied in -1 .. 1 space - } + RID projector = storage->light_get_projector(base); - light_data.transmittance_bias = storage->light_get_transmittance_bias(base); + if (projector.is_valid()) { + Rect2 rect = storage->decal_atlas_get_texture_rect(projector); - Rect2 rect = light_instance_get_shadow_atlas_rect(li, p_shadow_atlas); + if (type == RS::LIGHT_SPOT) { + light_data.projector_rect[0] = rect.position.x; + light_data.projector_rect[1] = rect.position.y + rect.size.height; //flip because shadow is flipped + light_data.projector_rect[2] = rect.size.width; + light_data.projector_rect[3] = -rect.size.height; + } else { + light_data.projector_rect[0] = rect.position.x; + light_data.projector_rect[1] = rect.position.y; + light_data.projector_rect[2] = rect.size.width; + light_data.projector_rect[3] = rect.size.height * 0.5; //used by dp, so needs to be half + } + } else { + light_data.projector_rect[0] = 0; + light_data.projector_rect[1] = 0; + light_data.projector_rect[2] = 0; + light_data.projector_rect[3] = 0; + } - light_data.atlas_rect[0] = rect.position.x; - light_data.atlas_rect[1] = rect.position.y; - light_data.atlas_rect[2] = rect.size.width; - light_data.atlas_rect[3] = rect.size.height; + if (shadow_atlas && shadow_atlas->shadow_owners.has(li->self)) { + // fill in the shadow information - light_data.soft_shadow_scale = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BLUR); - light_data.shadow_volumetric_fog_fade = 1.0 / storage->light_get_shadow_volumetric_fog_fade(base); + light_data.shadow_enabled = true; - if (type == RS::LIGHT_OMNI) { - light_data.atlas_rect[3] *= 0.5; //one paraboloid on top of another - Transform proj = (p_camera_inverse_transform * light_transform).inverse(); + if (type == RS::LIGHT_SPOT) { + light_data.shadow_bias = (storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0); + float shadow_texel_size = Math::tan(Math::deg2rad(spot_angle)) * radius * 2.0; + shadow_texel_size *= light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); - RendererStorageRD::store_transform(proj, light_data.shadow_matrix); + light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size; - if (size > 0.0) { - light_data.soft_shadow_size = size; - } else { - light_data.soft_shadow_size = 0.0; - light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF - } + } else { //omni + light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0; + float shadow_texel_size = light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); + light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size * 2.0; // applied in -1 .. 1 space + } - } else if (type == RS::LIGHT_SPOT) { - Transform modelview = (p_camera_inverse_transform * light_transform).inverse(); - CameraMatrix bias; - bias.set_light_bias(); + light_data.transmittance_bias = storage->light_get_transmittance_bias(base); - CameraMatrix shadow_mtx = bias * light_instance_get_shadow_camera(li, 0) * modelview; - RendererStorageRD::store_camera(shadow_mtx, light_data.shadow_matrix); + Rect2 rect = light_instance_get_shadow_atlas_rect(li->self, p_shadow_atlas); - if (size > 0.0) { - CameraMatrix cm = light_instance_get_shadow_camera(li, 0); - float half_np = cm.get_z_near() * Math::tan(Math::deg2rad(spot_angle)); - light_data.soft_shadow_size = (size * 0.5 / radius) / (half_np / cm.get_z_near()) * rect.size.width; - } else { - light_data.soft_shadow_size = 0.0; - light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF - } - } + light_data.atlas_rect[0] = rect.position.x; + light_data.atlas_rect[1] = rect.position.y; + light_data.atlas_rect[2] = rect.size.width; + light_data.atlas_rect[3] = rect.size.height; + + light_data.soft_shadow_scale = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BLUR); + light_data.shadow_volumetric_fog_fade = 1.0 / storage->light_get_shadow_volumetric_fog_fade(base); + + if (type == RS::LIGHT_OMNI) { + light_data.atlas_rect[3] *= 0.5; //one paraboloid on top of another + Transform proj = (inverse_transform * light_transform).inverse(); + + RendererStorageRD::store_transform(proj, light_data.shadow_matrix); + + if (size > 0.0) { + light_data.soft_shadow_size = size; } else { - light_data.shadow_color_enabled[3] = 0; + light_data.soft_shadow_size = 0.0; + light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF } - light_instance_set_index(li, light_count); + } else if (type == RS::LIGHT_SPOT) { + Transform modelview = (inverse_transform * light_transform).inverse(); + CameraMatrix bias; + bias.set_light_bias(); - cluster.builder.add_light(type == RS::LIGHT_SPOT ? LightClusterBuilder::LIGHT_TYPE_SPOT : LightClusterBuilder::LIGHT_TYPE_OMNI, light_transform, radius, spot_angle); + CameraMatrix shadow_mtx = bias * li->shadow_transform[0].camera * modelview; + RendererStorageRD::store_camera(shadow_mtx, light_data.shadow_matrix); - light_count++; - r_positional_light_count++; - } break; + if (size > 0.0) { + CameraMatrix cm = li->shadow_transform[0].camera; + float half_np = cm.get_z_near() * Math::tan(Math::deg2rad(spot_angle)); + light_data.soft_shadow_size = (size * 0.5 / radius) / (half_np / cm.get_z_near()) * rect.size.width; + } else { + light_data.soft_shadow_size = 0.0; + light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF + } + } + } else { + light_data.shadow_enabled = false; } - light_instance_set_render_pass(li, RSG::rasterizer->get_frame_number()); + li->light_index = index; + + current_cluster_builder->add_light(type == RS::LIGHT_SPOT ? ClusterBuilderRD::LIGHT_TYPE_SPOT : ClusterBuilderRD::LIGHT_TYPE_OMNI, light_transform, radius, spot_angle); - //update UBO for forward rendering, blit to texture for clustered + r_positional_light_count++; } - if (light_count) { - RD::get_singleton()->buffer_update(cluster.light_buffer, 0, sizeof(Cluster::LightData) * light_count, cluster.lights, true); + //update without barriers + if (cluster.omni_light_count) { + RD::get_singleton()->buffer_update(cluster.omni_light_buffer, 0, sizeof(Cluster::LightData) * cluster.omni_light_count, cluster.omni_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + + if (cluster.spot_light_count) { + RD::get_singleton()->buffer_update(cluster.spot_light_buffer, 0, sizeof(Cluster::LightData) * cluster.spot_light_count, cluster.spot_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } if (r_directional_light_count) { - RD::get_singleton()->buffer_update(cluster.directional_light_buffer, 0, sizeof(Cluster::DirectionalLightData) * r_directional_light_count, cluster.directional_lights, true); + RD::get_singleton()->buffer_update(cluster.directional_light_buffer, 0, sizeof(Cluster::DirectionalLightData) * r_directional_light_count, cluster.directional_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } @@ -6430,18 +6615,26 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const uv_xform.basis.scale(Vector3(2.0, 1.0, 2.0)); uv_xform.origin = Vector3(-1.0, 0.0, -1.0); - uint32_t decal_count = MIN((uint32_t)p_decals.size(), cluster.max_decals); - int idx = 0; + uint32_t decal_count = p_decals.size(); + + cluster.decal_count = 0; + for (uint32_t i = 0; i < decal_count; i++) { - RID di = p_decals[i]; - RID decal = decal_instance_get_base(di); + if (cluster.decal_count == cluster.max_decals) { + break; + } - Transform xform = decal_instance_get_transform(di); + DecalInstance *di = decal_instance_owner.getornull(p_decals[i]); + if (!di) { + continue; + } + RID decal = di->decal; - float fade = 1.0; + Transform xform = di->transform; + + real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; if (storage->decal_is_distance_fade_enabled(decal)) { - real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; float fade_begin = storage->decal_get_distance_fade_begin(decal); float fade_length = storage->decal_get_distance_fade_length(decal); @@ -6449,18 +6642,43 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const if (distance > fade_begin + fade_length) { continue; // do not use this decal, its invisible } + } + } + cluster.decal_sort[cluster.decal_count].instance = di; + cluster.decal_sort[cluster.decal_count].depth = distance; + cluster.decal_count++; + } + + if (cluster.decal_count > 0) { + SortArray<Cluster::InstanceSort<DecalInstance>> sort_array; + sort_array.sort(cluster.decal_sort, cluster.decal_count); + } + + for (uint32_t i = 0; i < cluster.decal_count; i++) { + DecalInstance *di = cluster.decal_sort[i].instance; + RID decal = di->decal; + + Transform xform = di->transform; + float fade = 1.0; + + if (storage->decal_is_distance_fade_enabled(decal)) { + real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; + float fade_begin = storage->decal_get_distance_fade_begin(decal); + float fade_length = storage->decal_get_distance_fade_length(decal); + + if (distance > fade_begin) { fade = 1.0 - (distance - fade_begin) / fade_length; } } - Cluster::DecalData &dd = cluster.decals[idx]; + Cluster::DecalData &dd = cluster.decals[i]; Vector3 decal_extents = storage->decal_get_extents(decal); Transform scale_xform; scale_xform.basis.scale(Vector3(decal_extents.x, decal_extents.y, decal_extents.z)); - Transform to_decal_xform = (p_camera_inverse_xform * decal_instance_get_transform(di) * scale_xform * uv_xform).affine_inverse(); + Transform to_decal_xform = (p_camera_inverse_xform * di->transform * scale_xform * uv_xform).affine_inverse(); RendererStorageRD::store_transform(to_decal_xform, dd.xform); Vector3 normal = xform.basis.get_axis(Vector3::AXIS_Y).normalized(); @@ -6545,19 +6763,18 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const dd.upper_fade = storage->decal_get_upper_fade(decal); dd.lower_fade = storage->decal_get_lower_fade(decal); - cluster.builder.add_decal(xform, decal_extents); - - idx++; + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_DECAL, xform, decal_extents); } - if (idx > 0) { - RD::get_singleton()->buffer_update(cluster.decal_buffer, 0, sizeof(Cluster::DecalData) * idx, cluster.decals, true); + if (cluster.decal_count > 0) { + RD::get_singleton()->buffer_update(cluster.decal_buffer, 0, sizeof(Cluster::DecalData) * cluster.decal_count, cluster.decals, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) { ERR_FAIL_COND(!rb->volumetric_fog); + RD::get_singleton()->free(rb->volumetric_fog->prev_light_density_map); RD::get_singleton()->free(rb->volumetric_fog->light_density_map); RD::get_singleton()->free(rb->volumetric_fog->fog_map); @@ -6579,49 +6796,6 @@ void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) { rb->volumetric_fog = nullptr; } -void RendererSceneRenderRD::_allocate_shadow_shrink_stages(RID p_base, int p_base_size, Vector<ShadowShrinkStage> &shrink_stages, uint32_t p_target_size) { - //create fog mipmaps - uint32_t fog_texture_size = p_target_size; - uint32_t base_texture_size = p_base_size; - - ShadowShrinkStage first; - first.size = base_texture_size; - first.texture = p_base; - shrink_stages.push_back(first); //put depth first in case we dont find smaller ones - - while (fog_texture_size < base_texture_size) { - base_texture_size = MAX(base_texture_size / 8, fog_texture_size); - - ShadowShrinkStage s; - s.size = base_texture_size; - - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = base_texture_size; - tf.height = base_texture_size; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; - - if (base_texture_size == fog_texture_size) { - s.filter_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; - } - - s.texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - - shrink_stages.push_back(s); - } -} - -void RendererSceneRenderRD::_clear_shadow_shrink_stages(Vector<ShadowShrinkStage> &shrink_stages) { - for (int i = 1; i < shrink_stages.size(); i++) { - RD::get_singleton()->free(shrink_stages[i].texture); - if (shrink_stages[i].filter_texture.is_valid()) { - RD::get_singleton()->free(shrink_stages[i].filter_texture); - } - } - shrink_stages.clear(); -} - void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_gi_probe_count) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); @@ -6644,6 +6818,8 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e return; } + RENDER_TIMESTAMP(">Volumetric Fog"); + if (env && env->volumetric_fog_enabled && !rb->volumetric_fog) { //required volumetric fog but not existing, create rb->volumetric_fog = memnew(VolumetricFog); @@ -6657,11 +6833,16 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e tf.height = target_height; tf.depth = volumetric_fog_depth; tf.texture_type = RD::TEXTURE_TYPE_3D; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; rb->volumetric_fog->light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); - tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + + rb->volumetric_fog->prev_light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->texture_clear(rb->volumetric_fog->prev_light_density_map, Color(0, 0, 0, 0), 0, 1, 0, 1); + + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; rb->volumetric_fog->fog_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); _render_buffers_uniform_set_changed(p_render_buffers); @@ -6678,162 +6859,6 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e rb->volumetric_fog->sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_FOG); } - //update directional shadow - - if (p_use_directional_shadows) { - if (directional_shadow.shrink_stages.is_empty()) { - if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { - //invalidate uniform set, we will need a new one - RD::get_singleton()->free(rb->volumetric_fog->uniform_set); - rb->volumetric_fog->uniform_set = RID(); - } - _allocate_shadow_shrink_stages(directional_shadow.depth, directional_shadow.size, directional_shadow.shrink_stages, volumetric_fog_directional_shadow_shrink); - } - - if (directional_shadow.shrink_stages.size() > 1) { - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - for (int i = 1; i < directional_shadow.shrink_stages.size(); i++) { - int32_t src_size = directional_shadow.shrink_stages[i - 1].size; - int32_t dst_size = directional_shadow.shrink_stages[i].size; - Rect2i r(0, 0, src_size, src_size); - int32_t shrink_limit = 8 / (src_size / dst_size); - - storage->get_effects()->reduce_shadow(directional_shadow.shrink_stages[i - 1].texture, directional_shadow.shrink_stages[i].texture, Size2i(src_size, src_size), r, shrink_limit, compute_list); - RD::get_singleton()->compute_list_add_barrier(compute_list); - if (env->volumetric_fog_shadow_filter != RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED && directional_shadow.shrink_stages[i].filter_texture.is_valid()) { - Rect2i rf(0, 0, dst_size, dst_size); - storage->get_effects()->filter_shadow(directional_shadow.shrink_stages[i].texture, directional_shadow.shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), rf, env->volumetric_fog_shadow_filter, compute_list); - } - } - RD::get_singleton()->compute_list_end(); - } - } - - ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); - - if (shadow_atlas) { - //shrink shadows that need to be shrunk - - bool force_shrink_shadows = false; - - if (shadow_atlas->shrink_stages.is_empty()) { - if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { - //invalidate uniform set, we will need a new one - RD::get_singleton()->free(rb->volumetric_fog->uniform_set); - rb->volumetric_fog->uniform_set = RID(); - } - _allocate_shadow_shrink_stages(shadow_atlas->depth, shadow_atlas->size, shadow_atlas->shrink_stages, volumetric_fog_positional_shadow_shrink); - force_shrink_shadows = true; - } - - if (rb->volumetric_fog->last_shadow_filter != env->volumetric_fog_shadow_filter) { - //if shadow filter changed, invalidate caches - rb->volumetric_fog->last_shadow_filter = env->volumetric_fog_shadow_filter; - force_shrink_shadows = true; - } - - cluster.lights_shadow_rect_cache_count = 0; - - for (int i = 0; i < p_positional_light_count; i++) { - if (cluster.lights[i].shadow_color_enabled[3] > 127) { - RID li = cluster.lights_instances[i]; - - ERR_CONTINUE(!shadow_atlas->shadow_owners.has(li)); - - uint32_t key = shadow_atlas->shadow_owners[li]; - - uint32_t quadrant = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; - uint32_t shadow = key & ShadowAtlas::SHADOW_INDEX_MASK; - - ERR_CONTINUE((int)shadow >= shadow_atlas->quadrants[quadrant].shadows.size()); - - ShadowAtlas::Quadrant::Shadow &s = shadow_atlas->quadrants[quadrant].shadows.write[shadow]; - - if (!force_shrink_shadows && s.fog_version == s.version) { - continue; //do not update, no need - } - - s.fog_version = s.version; - - uint32_t quadrant_size = shadow_atlas->size >> 1; - - Rect2i atlas_rect; - - atlas_rect.position.x = (quadrant & 1) * quadrant_size; - atlas_rect.position.y = (quadrant >> 1) * quadrant_size; - - uint32_t shadow_size = (quadrant_size / shadow_atlas->quadrants[quadrant].subdivision); - atlas_rect.position.x += (shadow % shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; - atlas_rect.position.y += (shadow / shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; - - atlas_rect.size.x = shadow_size; - atlas_rect.size.y = shadow_size; - - cluster.lights_shadow_rect_cache[cluster.lights_shadow_rect_cache_count] = atlas_rect; - - cluster.lights_shadow_rect_cache_count++; - - if (cluster.lights_shadow_rect_cache_count == cluster.max_lights) { - break; //light limit reached - } - } - } - - if (cluster.lights_shadow_rect_cache_count > 0) { - //there are shadows to be shrunk, try to do them in parallel - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - - for (int i = 1; i < shadow_atlas->shrink_stages.size(); i++) { - int32_t base_size = shadow_atlas->shrink_stages[0].size; - int32_t src_size = shadow_atlas->shrink_stages[i - 1].size; - int32_t dst_size = shadow_atlas->shrink_stages[i].size; - - uint32_t rect_divisor = base_size / src_size; - - int32_t shrink_limit = 8 / (src_size / dst_size); - - //shrink in parallel for more performance - for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) { - Rect2i src_rect = cluster.lights_shadow_rect_cache[j]; - - src_rect.position /= rect_divisor; - src_rect.size /= rect_divisor; - - storage->get_effects()->reduce_shadow(shadow_atlas->shrink_stages[i - 1].texture, shadow_atlas->shrink_stages[i].texture, Size2i(src_size, src_size), src_rect, shrink_limit, compute_list); - } - - RD::get_singleton()->compute_list_add_barrier(compute_list); - - if (env->volumetric_fog_shadow_filter != RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED && shadow_atlas->shrink_stages[i].filter_texture.is_valid()) { - uint32_t filter_divisor = base_size / dst_size; - - //filter in parallel for more performance - for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) { - Rect2i dst_rect = cluster.lights_shadow_rect_cache[j]; - - dst_rect.position /= filter_divisor; - dst_rect.size /= filter_divisor; - - storage->get_effects()->filter_shadow(shadow_atlas->shrink_stages[i].texture, shadow_atlas->shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), dst_rect, env->volumetric_fog_shadow_filter, compute_list, true, false); - } - - RD::get_singleton()->compute_list_add_barrier(compute_list); - - for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) { - Rect2i dst_rect = cluster.lights_shadow_rect_cache[j]; - - dst_rect.position /= filter_divisor; - dst_rect.size /= filter_divisor; - - storage->get_effects()->filter_shadow(shadow_atlas->shrink_stages[i].texture, shadow_atlas->shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), dst_rect, env->volumetric_fog_shadow_filter, compute_list, false, true); - } - } - } - - RD::get_singleton()->compute_list_end(); - } - } - //update volumetric fog if (rb->volumetric_fog->uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { @@ -6845,10 +6870,11 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 1; - if (shadow_atlas == nullptr || shadow_atlas->shrink_stages.size() == 0) { + ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); + if (shadow_atlas == nullptr || shadow_atlas->depth.is_null()) { u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK)); } else { - u.ids.push_back(shadow_atlas->shrink_stages[shadow_atlas->shrink_stages.size() - 1].texture); + u.ids.push_back(shadow_atlas->depth); } uniforms.push_back(u); @@ -6858,10 +6884,10 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 2; - if (directional_shadow.shrink_stages.size() == 0) { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK)); + if (directional_shadow.depth.is_valid()) { + u.ids.push_back(directional_shadow.depth); } else { - u.ids.push_back(directional_shadow.shrink_stages[directional_shadow.shrink_stages.size() - 1].texture); + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK)); } uniforms.push_back(u); } @@ -6870,23 +6896,22 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 3; - u.ids.push_back(get_positional_light_buffer()); + u.ids.push_back(get_omni_light_buffer()); uniforms.push_back(u); } - { RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 4; - u.ids.push_back(get_directional_light_buffer()); + u.ids.push_back(get_spot_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.binding = 5; - u.ids.push_back(get_cluster_builder_texture()); + u.ids.push_back(get_directional_light_buffer()); uniforms.push_back(u); } @@ -6894,7 +6919,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 6; - u.ids.push_back(get_cluster_builder_indices_buffer()); + u.ids.push_back(rb->cluster_builder->get_cluster_buffer()); uniforms.push_back(u); } @@ -6954,6 +6979,20 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); uniforms.push_back(u); } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 14; + u.ids.push_back(volumetric_fog.params_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 15; + u.ids.push_back(rb->volumetric_fog->prev_light_density_map); + uniforms.push_back(u); + } rb->volumetric_fog->uniform_set = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.shader.version_get_shader(volumetric_fog.shader_version, 0), 0); @@ -6999,7 +7038,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e rb->volumetric_fog->length = env->volumetric_fog_length; rb->volumetric_fog->spread = env->volumetric_fog_detail_spread; - VolumetricFogShader::PushConstant push_constant; + VolumetricFogShader::ParamsUBO params; Vector2 frustum_near_size = p_cam_projection.get_viewport_half_extents(); Vector2 frustum_far_size = p_cam_projection.get_far_plane_half_extents(); @@ -7015,51 +7054,78 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e fog_near_size = Vector2(); } - push_constant.fog_frustum_size_begin[0] = fog_near_size.x; - push_constant.fog_frustum_size_begin[1] = fog_near_size.y; + params.fog_frustum_size_begin[0] = fog_near_size.x; + params.fog_frustum_size_begin[1] = fog_near_size.y; - push_constant.fog_frustum_size_end[0] = fog_far_size.x; - push_constant.fog_frustum_size_end[1] = fog_far_size.y; + params.fog_frustum_size_end[0] = fog_far_size.x; + params.fog_frustum_size_end[1] = fog_far_size.y; - push_constant.z_near = z_near; - push_constant.z_far = z_far; + params.z_near = z_near; + params.z_far = z_far; - push_constant.fog_frustum_end = fog_end; + params.fog_frustum_end = fog_end; - push_constant.fog_volume_size[0] = rb->volumetric_fog->width; - push_constant.fog_volume_size[1] = rb->volumetric_fog->height; - push_constant.fog_volume_size[2] = rb->volumetric_fog->depth; + params.fog_volume_size[0] = rb->volumetric_fog->width; + params.fog_volume_size[1] = rb->volumetric_fog->height; + params.fog_volume_size[2] = rb->volumetric_fog->depth; - push_constant.directional_light_count = p_directional_light_count; + params.directional_light_count = p_directional_light_count; Color light = env->volumetric_fog_light.to_linear(); - push_constant.light_energy[0] = light.r * env->volumetric_fog_light_energy; - push_constant.light_energy[1] = light.g * env->volumetric_fog_light_energy; - push_constant.light_energy[2] = light.b * env->volumetric_fog_light_energy; - push_constant.base_density = env->volumetric_fog_density; + params.light_energy[0] = light.r * env->volumetric_fog_light_energy; + params.light_energy[1] = light.g * env->volumetric_fog_light_energy; + params.light_energy[2] = light.b * env->volumetric_fog_light_energy; + params.base_density = env->volumetric_fog_density; + + params.detail_spread = env->volumetric_fog_detail_spread; + params.gi_inject = env->volumetric_fog_gi_inject; + + params.cam_rotation[0] = p_cam_transform.basis[0][0]; + params.cam_rotation[1] = p_cam_transform.basis[1][0]; + params.cam_rotation[2] = p_cam_transform.basis[2][0]; + params.cam_rotation[3] = 0; + params.cam_rotation[4] = p_cam_transform.basis[0][1]; + params.cam_rotation[5] = p_cam_transform.basis[1][1]; + params.cam_rotation[6] = p_cam_transform.basis[2][1]; + params.cam_rotation[7] = 0; + params.cam_rotation[8] = p_cam_transform.basis[0][2]; + params.cam_rotation[9] = p_cam_transform.basis[1][2]; + params.cam_rotation[10] = p_cam_transform.basis[2][2]; + params.cam_rotation[11] = 0; + params.filter_axis = 0; + params.max_gi_probes = env->volumetric_fog_gi_inject > 0.001 ? p_gi_probe_count : 0; + params.temporal_frame = RSG::rasterizer->get_frame_number() % VolumetricFog::MAX_TEMPORAL_FRAMES; + + Transform to_prev_cam_view = rb->volumetric_fog->prev_cam_transform.affine_inverse() * p_cam_transform; + storage->store_transform(to_prev_cam_view, params.to_prev_view); + + params.use_temporal_reprojection = env->volumetric_fog_temporal_reprojection; + params.temporal_blend = env->volumetric_fog_temporal_reprojection_amount; - push_constant.detail_spread = env->volumetric_fog_detail_spread; - push_constant.gi_inject = env->volumetric_fog_gi_inject; + { + uint32_t cluster_size = rb->cluster_builder->get_cluster_size(); + params.cluster_shift = get_shift_from_power_of_2(cluster_size); - push_constant.cam_rotation[0] = p_cam_transform.basis[0][0]; - push_constant.cam_rotation[1] = p_cam_transform.basis[1][0]; - push_constant.cam_rotation[2] = p_cam_transform.basis[2][0]; - push_constant.cam_rotation[3] = 0; - push_constant.cam_rotation[4] = p_cam_transform.basis[0][1]; - push_constant.cam_rotation[5] = p_cam_transform.basis[1][1]; - push_constant.cam_rotation[6] = p_cam_transform.basis[2][1]; - push_constant.cam_rotation[7] = 0; - push_constant.cam_rotation[8] = p_cam_transform.basis[0][2]; - push_constant.cam_rotation[9] = p_cam_transform.basis[1][2]; - push_constant.cam_rotation[10] = p_cam_transform.basis[2][2]; - push_constant.cam_rotation[11] = 0; - push_constant.filter_axis = 0; - push_constant.max_gi_probes = env->volumetric_fog_gi_inject > 0.001 ? p_gi_probe_count : 0; + uint32_t cluster_screen_width = (rb->width - 1) / cluster_size + 1; + uint32_t cluster_screen_height = (rb->height - 1) / cluster_size + 1; + params.cluster_type_size = cluster_screen_width * cluster_screen_height * (32 + 32); + params.cluster_width = cluster_screen_width; + params.max_cluster_element_count_div_32 = max_cluster_elements / 32; + + params.screen_size[0] = rb->width; + params.screen_size[1] = rb->height; + } /* Vector2 dssize = directional_shadow_get_size(); push_constant.directional_shadow_pixel_size[0] = 1.0 / dssize.x; push_constant.directional_shadow_pixel_size[1] = 1.0 / dssize.y; */ + + RD::get_singleton()->draw_command_begin_label("Render Volumetric Fog"); + + RENDER_TIMESTAMP("Render Fog"); + RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), ¶ms, RD::BARRIER_MASK_COMPUTE); + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); bool use_filter = volumetric_fog_filter_active; @@ -7067,93 +7133,212 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[using_sdfgi ? VOLUMETRIC_FOG_SHADER_DENSITY_WITH_SDFGI : VOLUMETRIC_FOG_SHADER_DENSITY]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0); + if (using_sdfgi) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->sdfgi_uniform_set, 1); } - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VolumetricFogShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth); - RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); + + RD::get_singleton()->compute_list_end(); + + RD::get_singleton()->texture_copy(rb->volumetric_fog->light_density_map, rb->volumetric_fog->prev_light_density_map, Vector3(0, 0, 0), Vector3(0, 0, 0), Vector3(rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth), 0, 0, 0, 0); + + compute_list = RD::get_singleton()->compute_list_begin(); if (use_filter) { + RD::get_singleton()->draw_command_begin_label("Filter Fog"); + + RENDER_TIMESTAMP("Filter Fog"); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[VOLUMETRIC_FOG_SHADER_FILTER]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VolumetricFogShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth); - RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->compute_list_end(); + //need restart for buffer update - push_constant.filter_axis = 1; + params.filter_axis = 1; + RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), ¶ms); + compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[VOLUMETRIC_FOG_SHADER_FILTER]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set2, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VolumetricFogShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 8, 8, 1); + if (using_sdfgi) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->sdfgi_uniform_set, 1); + } + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth); RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); } + RENDER_TIMESTAMP("Integrate Fog"); + RD::get_singleton()->draw_command_begin_label("Integrate Fog"); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[VOLUMETRIC_FOG_SHADER_FOG]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VolumetricFogShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, 1); - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_RASTER); + + RENDER_TIMESTAMP("<Volumetric Fog"); + RD::get_singleton()->draw_command_end_label(); + + rb->volumetric_fog->prev_cam_transform = p_cam_transform; } -void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) { - Color clear_color; - if (p_render_buffers.is_valid()) { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(!rb); - clear_color = storage->render_target_get_clear_request_color(rb->render_target); - } else { - clear_color = storage->get_default_clear_color(); +uint32_t RendererSceneRenderRD::_get_render_state_directional_light_count() const { + return render_state.directional_light_count; +} + +bool RendererSceneRenderRD::_needs_post_prepass_render(bool p_use_gi) { + if (render_state.render_buffers.is_valid()) { + RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); + if (rb->sdfgi != nullptr) { + return true; + } } + return false; +} - //assign render indices to giprobes - for (uint32_t i = 0; i < (uint32_t)p_gi_probes.size(); i++) { - GIProbeInstance *giprobe_inst = gi_probe_instance_owner.getornull(p_gi_probes[i]); - if (giprobe_inst) { - giprobe_inst->render_index = i; +void RendererSceneRenderRD::_post_prepass_render(bool p_use_gi) { + if (render_state.render_buffers.is_valid()) { + if (p_use_gi) { + _sdfgi_update_probes(render_state.render_buffers, render_state.environment); } } +} - const PagedArray<RID> *lights = &p_lights; - const PagedArray<RID> *reflections = &p_reflection_probes; - const PagedArray<RID> *gi_probes = &p_gi_probes; +void RendererSceneRenderRD::_pre_resolve_render(bool p_use_gi) { + if (render_state.render_buffers.is_valid()) { + if (p_use_gi) { + RD::get_singleton()->compute_list_end(); + } + } +} - PagedArray<RID> empty; +void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_gi_probe_buffer) { + // Render shadows while GI is rendering, due to how barriers are handled, this should happen at the same time - if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { - lights = ∅ - reflections = ∅ - gi_probes = ∅ + if (render_state.render_buffers.is_valid() && p_use_gi) { + _sdfgi_store_probes(render_state.render_buffers); + } + + render_state.cube_shadows.clear(); + render_state.shadows.clear(); + render_state.directional_shadows.clear(); + + Plane camera_plane(render_state.cam_transform.origin, -render_state.cam_transform.basis.get_axis(Vector3::AXIS_Z)); + float lod_distance_multiplier = render_state.cam_projection.get_lod_multiplier(); + + { + for (int i = 0; i < render_state.render_shadow_count; i++) { + LightInstance *li = light_instance_owner.getornull(render_state.render_shadows[i].light); + + if (storage->light_get_type(li->light) == RS::LIGHT_DIRECTIONAL) { + render_state.directional_shadows.push_back(i); + } else if (storage->light_get_type(li->light) == RS::LIGHT_OMNI && storage->light_omni_get_shadow_mode(li->light) == RS::LIGHT_OMNI_SHADOW_CUBE) { + render_state.cube_shadows.push_back(i); + } else { + render_state.shadows.push_back(i); + } + } + + //cube shadows are rendered in their own way + for (uint32_t i = 0; i < render_state.cube_shadows.size(); i++) { + _render_shadow_pass(render_state.render_shadows[render_state.cube_shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.cube_shadows[i]].pass, render_state.render_shadows[render_state.cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, true, true, true); + } + + if (render_state.directional_shadows.size()) { + //open the pass for directional shadows + _update_directional_shadow_atlas(); + RD::get_singleton()->draw_list_begin(directional_shadow.fb, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE); + RD::get_singleton()->draw_list_end(); + } + } + + // Render GI + + bool render_shadows = render_state.directional_shadows.size() || render_state.shadows.size(); + bool render_gi = render_state.render_buffers.is_valid() && p_use_gi; + + if (render_shadows && render_gi) { + RENDER_TIMESTAMP("Render GI + Render Shadows (parallel)"); + } else if (render_shadows) { + RENDER_TIMESTAMP("Render Shadows"); + } else if (render_gi) { + RENDER_TIMESTAMP("Render GI"); + } + + //prepare shadow rendering + if (render_shadows) { + _render_shadow_begin(); + + //render directional shadows + for (uint32_t i = 0; i < render_state.directional_shadows.size(); i++) { + _render_shadow_pass(render_state.render_shadows[render_state.directional_shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.directional_shadows[i]].pass, render_state.render_shadows[render_state.directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, false, i == render_state.directional_shadows.size() - 1, false); + } + //render positional shadows + for (uint32_t i = 0; i < render_state.shadows.size(); i++) { + _render_shadow_pass(render_state.render_shadows[render_state.shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.shadows[i]].pass, render_state.render_shadows[render_state.shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, i == 0, i == render_state.shadows.size() - 1, true); + } + + _render_shadow_process(); + } + + //start GI + if (render_gi) { + _process_gi(render_state.render_buffers, p_normal_roughness_buffer, p_gi_probe_buffer, render_state.environment, render_state.cam_projection, render_state.cam_transform, *render_state.gi_probes); + } + + //Do shadow rendering (in parallel with GI) + if (render_shadows) { + _render_shadow_end(RD::BARRIER_MASK_NO_BARRIER); + } + + if (render_gi) { + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //use a later barrier + } + + if (render_state.render_buffers.is_valid()) { + if (p_use_ssao) { + _process_ssao(render_state.render_buffers, render_state.environment, p_normal_roughness_buffer, render_state.cam_projection); + } } - cluster.builder.begin(p_cam_transform.affine_inverse(), p_cam_projection); //prepare cluster + //full barrier here, we need raster, transfer and compute and it depends from the previous work + RD::get_singleton()->barrier(RD::BARRIER_MASK_ALL, RD::BARRIER_MASK_ALL); + + if (current_cluster_builder) { + current_cluster_builder->begin(render_state.cam_transform, render_state.cam_projection, !render_state.reflection_probe.is_valid()); + } bool using_shadows = true; - if (p_reflection_probe.is_valid()) { - if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(p_reflection_probe))) { + if (render_state.reflection_probe.is_valid()) { + if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(render_state.reflection_probe))) { using_shadows = false; } } else { //do not render reflections when rendering a reflection probe - _setup_reflections(*reflections, p_cam_transform.affine_inverse(), p_environment); + _setup_reflections(*render_state.reflection_probes, render_state.cam_transform.affine_inverse(), render_state.environment); } uint32_t directional_light_count = 0; uint32_t positional_light_count = 0; - _setup_lights(*lights, p_cam_transform.affine_inverse(), p_shadow_atlas, using_shadows, directional_light_count, positional_light_count); - _setup_decals(p_decals, p_cam_transform.affine_inverse()); - cluster.builder.bake_cluster(); //bake to cluster + _setup_lights(*render_state.lights, render_state.cam_transform, render_state.shadow_atlas, using_shadows, directional_light_count, positional_light_count); + _setup_decals(*render_state.decals, render_state.cam_transform.affine_inverse()); - uint32_t gi_probe_count = 0; - _setup_giprobes(p_render_buffers, p_cam_transform, *gi_probes, gi_probe_count); + render_state.directional_light_count = directional_light_count; - if (p_render_buffers.is_valid()) { + if (current_cluster_builder) { + current_cluster_builder->bake_cluster(); + } + + if (render_state.render_buffers.is_valid()) { bool directional_shadows = false; for (uint32_t i = 0; i < directional_light_count; i++) { if (cluster.directional_lights[i].shadow_enabled) { @@ -7161,12 +7346,126 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform & break; } } - _update_volumetric_fog(p_render_buffers, p_environment, p_cam_projection, p_cam_transform, p_shadow_atlas, directional_light_count, directional_shadows, positional_light_count, gi_probe_count); + _update_volumetric_fog(render_state.render_buffers, render_state.environment, render_state.cam_projection, render_state.cam_transform, render_state.shadow_atlas, directional_light_count, directional_shadows, positional_light_count, render_state.gi_probe_count); + } +} + +void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data) { + //assign render data + { + render_state.render_buffers = p_render_buffers; + render_state.cam_transform = p_cam_transform; + render_state.cam_projection = p_cam_projection; + render_state.cam_ortogonal = p_cam_projection.is_orthogonal(); + render_state.instances = &p_instances; + render_state.lights = &p_lights; + render_state.reflection_probes = &p_reflection_probes; + render_state.gi_probes = &p_gi_probes; + render_state.decals = &p_decals; + render_state.lightmaps = &p_lightmaps; + render_state.environment = p_environment; + render_state.camera_effects = p_camera_effects; + render_state.shadow_atlas = p_shadow_atlas; + render_state.reflection_atlas = p_reflection_atlas; + render_state.reflection_probe = p_reflection_probe; + render_state.reflection_probe_pass = p_reflection_probe_pass; + render_state.screen_lod_threshold = p_screen_lod_threshold; + + render_state.render_shadows = p_render_shadows; + render_state.render_shadow_count = p_render_shadow_count; + render_state.render_sdfgi_regions = p_render_sdfgi_regions; + render_state.render_sdfgi_region_count = p_render_sdfgi_region_count; + render_state.sdfgi_update_data = p_sdfgi_update_data; + } + + PagedArray<RID> empty; + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { + render_state.lights = ∅ + render_state.reflection_probes = ∅ + render_state.gi_probes = ∅ + } + + //sdfgi first + if (p_render_buffers.is_valid()) { + for (int i = 0; i < render_state.render_sdfgi_region_count; i++) { + _render_sdfgi_region(p_render_buffers, render_state.render_sdfgi_regions[i].region, render_state.render_sdfgi_regions[i].instances); + } + if (render_state.sdfgi_update_data->update_static) { + _render_sdfgi_static_lights(p_render_buffers, render_state.sdfgi_update_data->static_cascade_count, p_sdfgi_update_data->static_cascade_indices, render_state.sdfgi_update_data->static_positional_lights); + } + } + + Color clear_color; + if (p_render_buffers.is_valid()) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(!rb); + clear_color = storage->render_target_get_clear_request_color(rb->render_target); + } else { + clear_color = storage->get_default_clear_color(); + } + + //assign render indices to giprobes + for (uint32_t i = 0; i < (uint32_t)p_gi_probes.size(); i++) { + GIProbeInstance *giprobe_inst = gi_probe_instance_owner.getornull(p_gi_probes[i]); + if (giprobe_inst) { + giprobe_inst->render_index = i; + } } - _render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, directional_light_count, *gi_probes, p_lightmaps, p_environment, p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold); + if (render_buffers_owner.owns(render_state.render_buffers)) { + RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); + current_cluster_builder = rb->cluster_builder; + } else if (reflection_probe_instance_owner.owns(render_state.reflection_probe)) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(render_state.reflection_probe); + ReflectionAtlas *ra = reflection_atlas_owner.getornull(rpi->atlas); + if (!ra) { + ERR_PRINT("reflection probe has no reflection atlas! Bug?"); + current_cluster_builder = nullptr; + } else { + current_cluster_builder = ra->cluster_builder; + } + } else { + ERR_PRINT("No cluster builder, bug"); //should never happen, will crash + current_cluster_builder = nullptr; + } if (p_render_buffers.is_valid()) { + _pre_process_gi(p_render_buffers, p_cam_transform); + } + + render_state.gi_probe_count = 0; + if (render_state.render_buffers.is_valid()) { + _setup_giprobes(render_state.render_buffers, render_state.cam_transform, *render_state.gi_probes, render_state.gi_probe_count); + _sdfgi_update_light(render_state.render_buffers, render_state.environment); + } + + render_state.depth_prepass_used = false; + //calls _pre_opaque_render between depth pre-pass and opaque pass + _render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, *render_state.gi_probes, p_lightmaps, p_environment, current_cluster_builder->get_cluster_buffer(), current_cluster_builder->get_cluster_size(), current_cluster_builder->get_max_cluster_elements(), p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold); + + if (p_render_buffers.is_valid()) { + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) { + ClusterBuilderRD::ElementType elem_type = ClusterBuilderRD::ELEMENT_TYPE_MAX; + switch (debug_draw) { + case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS: + elem_type = ClusterBuilderRD::ELEMENT_TYPE_OMNI_LIGHT; + break; + case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS: + elem_type = ClusterBuilderRD::ELEMENT_TYPE_SPOT_LIGHT; + break; + case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS: + elem_type = ClusterBuilderRD::ELEMENT_TYPE_DECAL; + break; + case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES: + elem_type = ClusterBuilderRD::ELEMENT_TYPE_REFLECTION_PROBE; + break; + default: { + } + } + current_cluster_builder->debug(elem_type); + } + RENDER_TIMESTAMP("Tonemap"); _render_buffers_post_process_and_tonemap(p_render_buffers, p_environment, p_camera_effects, p_cam_projection); @@ -7177,27 +7476,26 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform & } } -void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<InstanceBase *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) { +void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_open_pass, bool p_close_pass, bool p_clear_region) { LightInstance *light_instance = light_instance_owner.getornull(p_light); ERR_FAIL_COND(!light_instance); Rect2i atlas_rect; - RID atlas_texture; + uint32_t atlas_size; + RID atlas_fb; bool using_dual_paraboloid = false; bool using_dual_paraboloid_flip = false; - float znear = 0; - float zfar = 0; RID render_fb; RID render_texture; - float bias = 0; - float normal_bias = 0; + float zfar; bool use_pancake = false; - bool use_linear_depth = false; bool render_cubemap = false; bool finalize_cubemap = false; + bool flip_y = false; + CameraMatrix light_projection; Transform light_transform; @@ -7230,7 +7528,6 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p atlas_rect.position.x += atlas_rect.size.width; atlas_rect.position.y += atlas_rect.size.height; } - } else if (storage->light_directional_get_shadow_mode(light_instance->light) == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS) { atlas_rect.size.height /= 2; @@ -7245,15 +7542,11 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p light_instance->shadow_transform[p_pass].atlas_rect.position /= directional_shadow.size; light_instance->shadow_transform[p_pass].atlas_rect.size /= directional_shadow.size; - float bias_mult = light_instance->shadow_transform[p_pass].bias_scale; zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE); - bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_BIAS) * bias_mult; - normal_bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * bias_mult; - ShadowMap *shadow_map = _get_shadow_map(atlas_rect.size); - render_fb = shadow_map->fb; - render_texture = shadow_map->depth; - atlas_texture = directional_shadow.depth; + render_fb = directional_shadow.fb; + render_texture = RID(); + flip_y = true; } else { //set from shadow atlas @@ -7262,6 +7555,8 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p ERR_FAIL_COND(!shadow_atlas); ERR_FAIL_COND(!shadow_atlas->shadow_owners.has(p_light)); + _update_shadow_atlas(shadow_atlas); + uint32_t key = shadow_atlas->shadow_owners[p_light]; uint32_t quadrant = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; @@ -7280,11 +7575,8 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p atlas_rect.size.width = shadow_size; atlas_rect.size.height = shadow_size; - atlas_texture = shadow_atlas->depth; zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE); - bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_BIAS); - normal_bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS); if (storage->light_get_type(light_instance->light) == RS::LIGHT_OMNI) { if (storage->light_omni_get_shadow_mode(light_instance->light) == RS::LIGHT_OMNI_SHADOW_CUBE) { @@ -7293,10 +7585,17 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p render_fb = cubemap->side_fb[p_pass]; render_texture = cubemap->cubemap; - light_projection = light_instance->shadow_transform[0].camera; - light_transform = light_instance->shadow_transform[0].transform; + light_projection = light_instance->shadow_transform[p_pass].camera; + light_transform = light_instance->shadow_transform[p_pass].transform; render_cubemap = true; finalize_cubemap = p_pass == 5; + atlas_fb = shadow_atlas->fb; + + atlas_size = shadow_atlas->size; + + if (p_pass == 0) { + _render_shadow_begin(); + } } else { light_projection = light_instance->shadow_transform[0].camera; @@ -7307,57 +7606,52 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p using_dual_paraboloid = true; using_dual_paraboloid_flip = p_pass == 1; - - ShadowMap *shadow_map = _get_shadow_map(atlas_rect.size); - render_fb = shadow_map->fb; - render_texture = shadow_map->depth; + render_fb = shadow_atlas->fb; + flip_y = true; } } else if (storage->light_get_type(light_instance->light) == RS::LIGHT_SPOT) { light_projection = light_instance->shadow_transform[0].camera; light_transform = light_instance->shadow_transform[0].transform; - ShadowMap *shadow_map = _get_shadow_map(atlas_rect.size); - render_fb = shadow_map->fb; - render_texture = shadow_map->depth; + render_fb = shadow_atlas->fb; - znear = light_instance->shadow_transform[0].camera.get_z_near(); - use_linear_depth = true; + flip_y = true; } } if (render_cubemap) { //rendering to cubemap - _render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold); + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, Rect2(), false, true, true, true); if (finalize_cubemap) { + _render_shadow_process(); + _render_shadow_end(); //reblit - atlas_rect.size.height /= 2; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_texture, atlas_rect, light_projection.get_z_near(), light_projection.get_z_far(), 0.0, false); - atlas_rect.position.y += atlas_rect.size.height; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_texture, atlas_rect, light_projection.get_z_near(), light_projection.get_z_far(), 0.0, true); - } - } else { - //render shadow - - _render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, bias, normal_bias, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold); + Rect2 atlas_rect_norm = atlas_rect; + atlas_rect_norm.position.x /= float(atlas_size); + atlas_rect_norm.position.y /= float(atlas_size); + atlas_rect_norm.size.x /= float(atlas_size); + atlas_rect_norm.size.y /= float(atlas_size); + atlas_rect_norm.size.height /= 2; + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), false); + atlas_rect_norm.position.y += atlas_rect_norm.size.height; + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), true); - //copy to atlas - if (use_linear_depth) { - storage->get_effects()->copy_depth_to_rect_and_linearize(render_texture, atlas_texture, atlas_rect, true, znear, zfar); - } else { - storage->get_effects()->copy_depth_to_rect(render_texture, atlas_texture, atlas_rect, true); + //restore transform so it can be properly used + light_instance_set_shadow_transform(p_light, CameraMatrix(), light_instance->transform, zfar, 0, 0, 0); } - //does not work from depth to color - //RD::get_singleton()->texture_copy(render_texture, atlas_texture, Vector3(0, 0, 0), Vector3(atlas_rect.position.x, atlas_rect.position.y, 0), Vector3(atlas_rect.size.x, atlas_rect.size.y, 1), 0, 0, 0, 0, true); + } else { + //render shadow + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, atlas_rect, flip_y, p_clear_region, p_open_pass, p_close_pass); } } -void RendererSceneRenderRD::render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { +void RendererSceneRenderRD::render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { _render_material(p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, p_framebuffer, p_region); } -void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<InstanceBase *> &p_instances) { +void RendererSceneRenderRD::_render_sdfgi_region(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances) { //print_line("rendering region " + itos(p_region)); RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); @@ -7373,16 +7667,18 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con if (cascade_prev != cascade) { //initialize render - RD::get_singleton()->texture_clear(rb->sdfgi->render_albedo, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->render_emission, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->render_emission_aniso, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->render_geom_facing, Color(0, 0, 0, 0), 0, 1, 0, 1, true); + RD::get_singleton()->texture_clear(rb->sdfgi->render_albedo, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->render_emission, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->render_emission_aniso, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->render_geom_facing, Color(0, 0, 0, 0), 0, 1, 0, 1); } //print_line("rendering cascade " + itos(p_region) + " objects: " + itos(p_cull_count) + " bounds: " + bounds + " from: " + from + " size: " + size + " cell size: " + rtos(rb->sdfgi->cascades[cascade].cell_size)); _render_sdfgi(p_render_buffers, from, size, bounds, p_instances, rb->sdfgi->render_albedo, rb->sdfgi->render_emission, rb->sdfgi->render_emission_aniso, rb->sdfgi->render_geom_facing); if (cascade_next != cascade) { + RD::get_singleton()->draw_command_begin_label("SDFGI Pre-Process Cascade"); + RENDER_TIMESTAMP(">SDFGI Update SDF"); //done rendering! must update SDF //clear dispatch indirect data @@ -7405,6 +7701,9 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con push_constant.scroll[1] = 0; push_constant.scroll[2] = 0; } + + rb->sdfgi->cascades[cascade].all_dynamic_lights_dirty = true; + push_constant.grid_size = rb->sdfgi->cascade_size; push_constant.cascade = cascade; @@ -7429,7 +7728,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con groups.z = rb->sdfgi->cascade_size - ABS(dirty.z); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, groups.x, groups.y, groups.z, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, groups.x, groups.y, groups.z); //no barrier, continue together @@ -7471,7 +7770,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].integrate_uniform_set, 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_sky_uniform_set, 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDGIShader::IntegratePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -7479,7 +7778,24 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].integrate_uniform_set, 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_sky_uniform_set, 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDGIShader::IntegratePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + if (rb->sdfgi->uses_multibounce) { + //multibounce requires this to be stored so direct light can read from it + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_STORE]); + + //convert to octahedral to store + ipush_constant.image_size[0] *= SDFGI::LIGHTPROBE_OCT_SIZE; + ipush_constant.image_size[1] *= SDFGI::LIGHTPROBE_OCT_SIZE; + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].integrate_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_sky_uniform_set, 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDGIShader::IntegratePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1); + } } //ok finally barrier @@ -7488,7 +7804,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con //clear dispatch indirect data uint32_t dispatch_indirct_data[4] = { 0, 0, 0, 0 }; - RD::get_singleton()->buffer_update(rb->sdfgi->cascades[cascade].solid_cell_dispatch_buffer, 0, sizeof(uint32_t) * 4, dispatch_indirct_data, true); + RD::get_singleton()->buffer_update(rb->sdfgi->cascades[cascade].solid_cell_dispatch_buffer, 0, sizeof(uint32_t) * 4, dispatch_indirct_data); RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); @@ -7502,7 +7818,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE_HALF]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->sdf_initialize_half_uniform_set, 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size); RD::get_singleton()->compute_list_add_barrier(compute_list); //must start with regular jumpflood @@ -7522,7 +7838,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con push_constant.step_size = s; RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_half_uniform_set[jf_us], 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size); RD::get_singleton()->compute_list_add_barrier(compute_list); jf_us = jf_us == 0 ? 1 : 0; @@ -7540,7 +7856,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con push_constant.step_size = s; RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_half_uniform_set[jf_us], 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size, optimized_jf_group_size, optimized_jf_group_size, optimized_jf_group_size); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size); RD::get_singleton()->compute_list_add_barrier(compute_list); jf_us = jf_us == 0 ? 1 : 0; } @@ -7552,7 +7868,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_UPSCALE]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->sdf_upscale_uniform_set, 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); //run one pass of fullsize jumpflood to fix up half size arctifacts @@ -7562,7 +7878,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_OPTIMIZED]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_uniform_set[rb->sdfgi->upscale_jfa_uniform_set_index], 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, optimized_jf_group_size, optimized_jf_group_size, optimized_jf_group_size); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); } else { @@ -7572,7 +7888,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->sdf_initialize_uniform_set, 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -7589,7 +7905,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con push_constant.step_size = s; RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_uniform_set[jf_us], 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); jf_us = jf_us == 0 ? 1 : 0; @@ -7607,7 +7923,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con push_constant.step_size = s; RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_uniform_set[jf_us], 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, optimized_jf_group_size, optimized_jf_group_size, optimized_jf_group_size); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); jf_us = jf_us == 0 ? 1 : 0; } @@ -7654,14 +7970,14 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_STORE]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].sdf_store_uniform_set, 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_end(); //clear these textures, as they will have previous garbage on next draw - RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_tex, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_0_tex, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_1_tex, Color(0, 0, 0, 0), 0, 1, 0, 1, true); + RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_0_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_1_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); #if 0 Vector<uint8_t> data = RD::get_singleton()->texture_get_data(rb->sdfgi->cascades[cascade].sdf, 0); @@ -7691,10 +8007,11 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con #endif RENDER_TIMESTAMP("<SDFGI Update SDF"); + RD::get_singleton()->draw_command_end_label(); } } -void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<InstanceBase *> &p_instances) { +void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances) { ERR_FAIL_COND(!storage->particles_collision_is_heightfield(p_collider)); Vector3 extents = storage->particles_collision_get_extents(p_collider) * p_transform.basis.get_scale(); CameraMatrix cm; @@ -7711,32 +8028,17 @@ void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, _render_particle_collider_heightfield(fb, cam_xform, cm, p_instances); } -void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result) { +void RendererSceneRenderRD::_render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); ERR_FAIL_COND(!rb->sdfgi); - _sdfgi_update_cascades(p_render_buffers); //need cascades updated for this - - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.direct_light_pipeline[SDGIShader::DIRECT_LIGHT_MODE_STATIC]); - - SDGIShader::DirectLightPushConstant dl_push_constant; + RD::get_singleton()->draw_command_begin_label("SDFGI Render Static Lighs"); - dl_push_constant.grid_size[0] = rb->sdfgi->cascade_size; - dl_push_constant.grid_size[1] = rb->sdfgi->cascade_size; - dl_push_constant.grid_size[2] = rb->sdfgi->cascade_size; - dl_push_constant.max_cascades = rb->sdfgi->cascades.size(); - dl_push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; - dl_push_constant.multibounce = false; // this is static light, do not multibounce yet - dl_push_constant.y_mult = rb->sdfgi->y_mult; - - //all must be processed - dl_push_constant.process_offset = 0; - dl_push_constant.process_increment = 1; + _sdfgi_update_cascades(p_render_buffers); //need cascades updated for this SDGIShader::Light lights[SDFGI::MAX_STATIC_LIGHTS]; + uint32_t light_count[SDFGI::MAX_STATIC_LIGHTS]; for (uint32_t i = 0; i < p_cascade_count; i++) { ERR_CONTINUE(p_cascade_indices[i] >= rb->sdfgi->cascades.size()); @@ -7792,18 +8094,45 @@ void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uin lights[idx].has_shadow = storage->light_has_shadow(li->light); lights[idx].attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_ATTENUATION); lights[idx].radius = storage->light_get_param(li->light, RS::LIGHT_PARAM_RANGE); - lights[idx].spot_angle = Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE)); - lights[idx].spot_attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + lights[idx].cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE))); + lights[idx].inv_spot_attenuation = 1.0f / storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION); idx++; } if (idx > 0) { - RD::get_singleton()->buffer_update(cc.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights, true); + RD::get_singleton()->buffer_update(cc.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights); } - dl_push_constant.light_count = idx; + + light_count[i] = idx; } + } + + /* Static Lights */ + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.direct_light_pipeline[SDGIShader::DIRECT_LIGHT_MODE_STATIC]); + + SDGIShader::DirectLightPushConstant dl_push_constant; + dl_push_constant.grid_size[0] = rb->sdfgi->cascade_size; + dl_push_constant.grid_size[1] = rb->sdfgi->cascade_size; + dl_push_constant.grid_size[2] = rb->sdfgi->cascade_size; + dl_push_constant.max_cascades = rb->sdfgi->cascades.size(); + dl_push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; + dl_push_constant.multibounce = false; // this is static light, do not multibounce yet + dl_push_constant.y_mult = rb->sdfgi->y_mult; + + //all must be processed + dl_push_constant.process_offset = 0; + dl_push_constant.process_increment = 1; + + for (uint32_t i = 0; i < p_cascade_count; i++) { + ERR_CONTINUE(p_cascade_indices[i] >= rb->sdfgi->cascades.size()); + + SDFGI::Cascade &cc = rb->sdfgi->cascades[p_cascade_indices[i]]; + + dl_push_constant.light_count = light_count[i]; dl_push_constant.cascade = p_cascade_indices[i]; if (dl_push_constant.light_count > 0) { @@ -7814,6 +8143,8 @@ void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uin } RD::get_singleton()->compute_list_end(); + + RD::get_singleton()->draw_command_end_label(); } bool RendererSceneRenderRD::free(RID p_rid) { @@ -7827,6 +8158,9 @@ bool RendererSceneRenderRD::free(RID p_rid) { if (rb->volumetric_fog) { _volumetric_fog_erase(rb); } + if (rb->cluster_builder) { + memdelete(rb->cluster_builder); + } render_buffers_owner.free(p_rid); } else if (environment_owner.owns(p_rid)) { //not much to delete, just free it @@ -7836,6 +8170,10 @@ bool RendererSceneRenderRD::free(RID p_rid) { camera_effects_owner.free(p_rid); } else if (reflection_atlas_owner.owns(p_rid)) { reflection_atlas_set_size(p_rid, 0, 0); + ReflectionAtlas *ra = reflection_atlas_owner.getornull(p_rid); + if (ra->cluster_builder) { + memdelete(ra->cluster_builder); + } reflection_atlas_owner.free(p_rid); } else if (reflection_probe_instance_owner.owns(p_rid)) { //not much to delete, just free it @@ -7844,6 +8182,8 @@ bool RendererSceneRenderRD::free(RID p_rid) { reflection_probe_instance_owner.free(p_rid); } else if (decal_instance_owner.owns(p_rid)) { decal_instance_owner.free(p_rid); + } else if (lightmap_instance_owner.owns(p_rid)) { + lightmap_instance_owner.free(p_rid); } else if (gi_probe_instance_owner.owns(p_rid)) { GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_rid); if (gi_probe->texture.is_valid()) { @@ -7979,23 +8319,28 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto //RID sampled_light; - InstanceBase ins; + GeometryInstance *gi = geometry_instance_create(p_base); + + uint32_t sc = RSG::storage->mesh_get_surface_count(p_base); + Vector<RID> materials; + materials.resize(sc); - ins.base_type = RSG::storage->get_base_type(p_base); - ins.base = p_base; - ins.materials.resize(RSG::storage->mesh_get_surface_count(p_base)); - for (int i = 0; i < ins.materials.size(); i++) { - if (i < p_material_overrides.size()) { - ins.materials.write[i] = p_material_overrides[i]; + for (uint32_t i = 0; i < sc; i++) { + if (i < (uint32_t)p_material_overrides.size()) { + materials.write[i] = p_material_overrides[i]; } } + geometry_instance_set_surface_materials(gi, materials); + if (cull_argument.size() == 0) { cull_argument.push_back(nullptr); } - cull_argument[0] = &ins; + cull_argument[0] = gi; _render_uv2(cull_argument, fb, Rect2i(0, 0, p_image_size.width, p_image_size.height)); + geometry_instance_free(gi); + TypedArray<Image> ret; { @@ -8047,20 +8392,17 @@ void RendererSceneRenderRD::sdfgi_set_debug_probe_select(const Vector3 &p_positi RendererSceneRenderRD *RendererSceneRenderRD::singleton = nullptr; -RID RendererSceneRenderRD::get_cluster_builder_texture() { - return cluster.builder.get_cluster_texture(); -} - -RID RendererSceneRenderRD::get_cluster_builder_indices_buffer() { - return cluster.builder.get_cluster_indices_buffer(); -} - RID RendererSceneRenderRD::get_reflection_probe_buffer() { return cluster.reflection_buffer; } -RID RendererSceneRenderRD::get_positional_light_buffer() { - return cluster.light_buffer; +RID RendererSceneRenderRD::get_omni_light_buffer() { + return cluster.omni_light_buffer; } + +RID RendererSceneRenderRD::get_spot_light_buffer() { + return cluster.spot_light_buffer; +} + RID RendererSceneRenderRD::get_directional_light_buffer() { return cluster.directional_light_buffer; } @@ -8076,13 +8418,21 @@ bool RendererSceneRenderRD::is_low_end() const { } RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { + max_cluster_elements = GLOBAL_GET("rendering/cluster_builder/max_clustered_elements"); + storage = p_storage; singleton = this; roughness_layers = GLOBAL_GET("rendering/quality/reflections/roughness_layers"); sky_ggx_samples_quality = GLOBAL_GET("rendering/quality/reflections/ggx_samples"); sky_use_cubemap_array = GLOBAL_GET("rendering/quality/reflections/texture_array_reflections"); - // sky_use_cubemap_array = false; + + sdfgi_ray_count = RS::EnvironmentSDFGIRayCount(CLAMP(int32_t(GLOBAL_GET("rendering/sdfgi/probe_ray_count")), 0, int32_t(RS::ENV_SDFGI_RAY_COUNT_MAX - 1))); + sdfgi_frames_to_converge = RS::EnvironmentSDFGIFramesToConverge(CLAMP(int32_t(GLOBAL_GET("rendering/sdfgi/frames_to_converge")), 0, int32_t(RS::ENV_SDFGI_CONVERGE_MAX - 1))); + sdfgi_frames_to_update_light = RS::EnvironmentSDFGIFramesToUpdateLight(CLAMP(int32_t(GLOBAL_GET("rendering/sdfgi/frames_to_update_lights")), 0, int32_t(RS::ENV_SDFGI_UPDATE_LIGHT_MAX - 1))); + + directional_shadow.size = GLOBAL_GET("rendering/quality/directional_shadow/size"); + directional_shadow.use_16_bits = GLOBAL_GET("rendering/quality/directional_shadow/16_bits"); uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); @@ -8376,6 +8726,9 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { //calculate tables String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; defines += "\n#define SH_SIZE " + itos(SDFGI::SH_SIZE) + "\n"; + if (sky_use_cubemap_array) { + defines += "\n#define USE_CUBEMAP_ARRAY\n"; + } Vector<String> integrate_modes; integrate_modes.push_back("\n#define MODE_PROCESS\n"); @@ -8410,11 +8763,18 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { sdfgi_shader.integrate_default_sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.integrate.version_get_shader(sdfgi_shader.integrate_shader, 0), 1); } } + //GK { //calculate tables String defines = "\n#define SDFGI_OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; Vector<String> gi_modes; - gi_modes.push_back(""); + gi_modes.push_back("\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define USE_SDFGI\n"); + gi_modes.push_back("\n#define USE_SDFGI\n\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n\n#define USE_GIPROBES\n"); + gi.shader.initialize(gi_modes, defines); gi.shader_version = gi.shader.version_create(); for (int i = 0; i < GI::MODE_MAX; i++) { @@ -8458,31 +8818,33 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { default_giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GI::GIProbeData) * RenderBuffers::MAX_GIPROBES); } - //cluster setup - uint32_t uniform_max_size = RD::get_singleton()->limit_get(RD::LIMIT_MAX_UNIFORM_BUFFER_SIZE); + { //decals + cluster.max_decals = max_cluster_elements; + uint32_t decal_buffer_size = cluster.max_decals * sizeof(Cluster::DecalData); + cluster.decals = memnew_arr(Cluster::DecalData, cluster.max_decals); + cluster.decal_sort = memnew_arr(Cluster::InstanceSort<DecalInstance>, cluster.max_decals); + cluster.decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size); + } { //reflections - uint32_t reflection_buffer_size; - if (uniform_max_size < 65536) { - //Yes, you guessed right, ARM again - reflection_buffer_size = uniform_max_size; - } else { - reflection_buffer_size = 65536; - } - cluster.max_reflections = reflection_buffer_size / sizeof(Cluster::ReflectionData); + cluster.max_reflections = max_cluster_elements; cluster.reflections = memnew_arr(Cluster::ReflectionData, cluster.max_reflections); - cluster.reflection_buffer = RD::get_singleton()->storage_buffer_create(reflection_buffer_size); + cluster.reflection_sort = memnew_arr(Cluster::InstanceSort<ReflectionProbeInstance>, cluster.max_reflections); + cluster.reflection_buffer = RD::get_singleton()->storage_buffer_create(sizeof(Cluster::ReflectionData) * cluster.max_reflections); } { //lights - cluster.max_lights = MIN(1024 * 1024, uniform_max_size) / sizeof(Cluster::LightData); //1mb of lights + cluster.max_lights = max_cluster_elements; + uint32_t light_buffer_size = cluster.max_lights * sizeof(Cluster::LightData); - cluster.lights = memnew_arr(Cluster::LightData, cluster.max_lights); - cluster.light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size); + cluster.omni_lights = memnew_arr(Cluster::LightData, cluster.max_lights); + cluster.omni_light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size); + cluster.omni_light_sort = memnew_arr(Cluster::InstanceSort<LightInstance>, cluster.max_lights); + cluster.spot_lights = memnew_arr(Cluster::LightData, cluster.max_lights); + cluster.spot_light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size); + cluster.spot_light_sort = memnew_arr(Cluster::InstanceSort<LightInstance>, cluster.max_lights); //defines += "\n#define MAX_LIGHT_DATA_STRUCTS " + itos(cluster.max_lights) + "\n"; - cluster.lights_instances = memnew_arr(RID, cluster.max_lights); - cluster.lights_shadow_rect_cache = memnew_arr(Rect2i, cluster.max_lights); cluster.max_directional_lights = MAX_DIRECTIONAL_LIGHTS; uint32_t directional_light_buffer_size = cluster.max_directional_lights * sizeof(Cluster::DirectionalLightData); @@ -8490,15 +8852,6 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { cluster.directional_light_buffer = RD::get_singleton()->uniform_buffer_create(directional_light_buffer_size); } - { //decals - cluster.max_decals = MIN(1024 * 1024, uniform_max_size) / sizeof(Cluster::DecalData); //1mb of decals - uint32_t decal_buffer_size = cluster.max_decals * sizeof(Cluster::DecalData); - cluster.decals = memnew_arr(Cluster::DecalData, cluster.max_decals); - cluster.decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size); - } - - cluster.builder.setup(16, 8, 24); - if (!low_end) { String defines = "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(cluster.max_directional_lights) + "\n"; Vector<String> volumetric_fog_modes; @@ -8511,6 +8864,7 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { for (int i = 0; i < VOLUMETRIC_FOG_SHADER_MAX; i++) { volumetric_fog.pipelines[i] = RD::get_singleton()->compute_pipeline_create(volumetric_fog.shader.version_get_shader(volumetric_fog.shader_version, i)); } + volumetric_fog.params_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(VolumetricFogShader::ParamsUBO)); } { @@ -8543,16 +8897,13 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { environment_set_volumetric_fog_volume_size(GLOBAL_GET("rendering/volumetric_fog/volume_size"), GLOBAL_GET("rendering/volumetric_fog/volume_depth")); environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/volumetric_fog/use_filter")); - environment_set_volumetric_fog_directional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/directional_shadow_shrink")); - environment_set_volumetric_fog_positional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/positional_shadow_shrink")); cull_argument.set_page_pool(&cull_argument_pool); + + gi.half_resolution = GLOBAL_GET("rendering/quality/gi/use_half_resolution"); } RendererSceneRenderRD::~RendererSceneRenderRD() { - for (Map<Vector2i, ShadowMap>::Element *E = shadow_maps.front(); E; E = E->next()) { - RD::get_singleton()->free(E->get().depth); - } for (Map<int, ShadowCubemap>::Element *E = shadow_cubemaps.front(); E; E = E->next()) { RD::get_singleton()->free(E->get().cubemap); } @@ -8576,6 +8927,7 @@ RendererSceneRenderRD::~RendererSceneRenderRD() { sdfgi_shader.preprocess.version_free(sdfgi_shader.preprocess_shader); volumetric_fog.shader.version_free(volumetric_fog.shader_version); + RD::get_singleton()->free(volumetric_fog.params_ubo); memdelete_arr(gi_probe_lights); } @@ -8597,15 +8949,19 @@ RendererSceneRenderRD::~RendererSceneRenderRD() { { RD::get_singleton()->free(cluster.directional_light_buffer); - RD::get_singleton()->free(cluster.light_buffer); + RD::get_singleton()->free(cluster.omni_light_buffer); + RD::get_singleton()->free(cluster.spot_light_buffer); RD::get_singleton()->free(cluster.reflection_buffer); RD::get_singleton()->free(cluster.decal_buffer); memdelete_arr(cluster.directional_lights); - memdelete_arr(cluster.lights); - memdelete_arr(cluster.lights_shadow_rect_cache); - memdelete_arr(cluster.lights_instances); + memdelete_arr(cluster.omni_lights); + memdelete_arr(cluster.spot_lights); + memdelete_arr(cluster.omni_light_sort); + memdelete_arr(cluster.spot_light_sort); memdelete_arr(cluster.reflections); + memdelete_arr(cluster.reflection_sort); memdelete_arr(cluster.decals); + memdelete_arr(cluster.decal_sort); } RD::get_singleton()->free(shadow_sampler); diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h index af35e1b3b4..cdcdb73132 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -34,7 +34,7 @@ #include "core/templates/local_vector.h" #include "core/templates/rid_owner.h" #include "servers/rendering/renderer_compositor.h" -#include "servers/rendering/renderer_rd/light_cluster_builder.h" +#include "servers/rendering/renderer_rd/cluster_builder_rd.h" #include "servers/rendering/renderer_rd/renderer_storage_rd.h" #include "servers/rendering/renderer_rd/shaders/gi.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/giprobe.glsl.gen.h" @@ -104,17 +104,22 @@ protected: }; virtual RenderBufferData *_create_render_buffer_data() = 0; - void _setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count); + void _setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count); void _setup_decals(const PagedArray<RID> &p_decals, const Transform &p_camera_inverse_xform); void _setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment); void _setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used); - virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_color, float p_screen_lod_threshold) = 0; - virtual void _render_shadow(RID p_framebuffer, const PagedArray<InstanceBase *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0) = 0; - virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; - virtual void _render_uv2(const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; - virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<InstanceBase *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) = 0; - virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<InstanceBase *> &p_instances) = 0; + virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_cluster_max_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_color, float p_screen_lod_threshold) = 0; + + virtual void _render_shadow_begin() = 0; + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true) = 0; + virtual void _render_shadow_process() = 0; + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL) = 0; + + virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; + virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; + virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) = 0; + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) = 0; virtual void _debug_giprobe(RID p_gi_probe, RenderingDevice::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha); void _debug_sdfgi_probes(RID p_render_buffers, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform); @@ -124,8 +129,6 @@ protected: virtual void _base_uniforms_changed() = 0; virtual void _render_buffers_uniform_set_changed(RID p_render_buffers) = 0; virtual RID _render_buffers_get_normal_texture(RID p_render_buffers) = 0; - virtual RID _render_buffers_get_ambient_texture(RID p_render_buffers) = 0; - virtual RID _render_buffers_get_reflection_texture(RID p_render_buffers) = 0; void _process_ssao(RID p_render_buffers, RID p_environment, RID p_normal_buffer, const CameraMatrix &p_projection); void _process_ssr(RID p_render_buffers, RID p_dest_framebuffer, RID p_normal_buffer, RID p_specular_buffer, RID p_metallic, const Color &p_metallic_mask, RID p_environment, const CameraMatrix &p_projection, bool p_use_additive); @@ -134,11 +137,19 @@ protected: void _setup_sky(RID p_environment, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size); void _update_sky(RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform); void _draw_sky(bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform); - void _process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes); + void _pre_process_gi(RID p_render_buffers, const Transform &p_transform); + void _process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes); + + bool _needs_post_prepass_render(bool p_use_gi); + void _post_prepass_render(bool p_use_gi); + void _pre_resolve_render(bool p_use_gi); + + void _pre_opaque_render(bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_gi_probe_buffer); + uint32_t _get_render_state_directional_light_count() const; // needed for a single argument calls (material and uv2) - PagedArrayPool<InstanceBase *> cull_argument_pool; - PagedArray<InstanceBase *> cull_argument; //need this to exist + PagedArrayPool<GeometryInstance *> cull_argument_pool; + PagedArray<GeometryInstance *> cull_argument; //need this to exist private: RS::ViewportDebugDraw debug_draw = RS::VIEWPORT_DEBUG_DRAW_DISABLED; double time_step = 0; @@ -233,6 +244,7 @@ private: virtual bool is_animated() const; virtual bool casts_shadows() const; virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; SkyShaderData(); virtual ~SkyShaderData(); }; @@ -340,6 +352,8 @@ private: }; Vector<Reflection> reflections; + + ClusterBuilderRD *cluster_builder = nullptr; }; mutable RID_Owner<ReflectionAtlas> reflection_atlas_owner; @@ -374,6 +388,15 @@ private: mutable RID_Owner<DecalInstance> decal_instance_owner; + /* LIGHTMAP INSTANCE */ + + struct LightmapInstance { + RID lightmap; + Transform transform; + }; + + mutable RID_Owner<LightmapInstance> lightmap_instance_owner; + /* GIPROBE INSTANCE */ struct GIProbeLight { @@ -383,10 +406,10 @@ private: float attenuation; float color[3]; - float spot_angle_radians; + float cos_spot_angle; float position[3]; - float spot_attenuation; + float inv_spot_attenuation; float direction[3]; uint32_t has_shadow; @@ -562,17 +585,18 @@ private: uint32_t smallest_subdiv = 0; int size = 0; + bool use_16_bits = false; RID depth; RID fb; //for copying Map<RID, uint32_t> shadow_owners; - - Vector<ShadowShrinkStage> shrink_stages; }; RID_Owner<ShadowAtlas> shadow_atlas_owner; + void _update_shadow_atlas(ShadowAtlas *shadow_atlas); + bool _shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow); RS::ShadowQuality shadows_quality = RS::SHADOW_QUALITY_MAX; //So it always updates when first set @@ -593,17 +617,16 @@ private: struct DirectionalShadow { RID depth; + RID fb; //when renderign direct int light_count = 0; int size = 0; + bool use_16_bits = false; int current_light = 0; - Vector<ShadowShrinkStage> shrink_stages; - } directional_shadow; - void _allocate_shadow_shrink_stages(RID p_base, int p_base_size, Vector<ShadowShrinkStage> &shrink_stages, uint32_t p_target_size); - void _clear_shadow_shrink_stages(Vector<ShadowShrinkStage> &shrink_stages); + void _update_directional_shadow_atlas(); /* SHADOW CUBEMAPS */ @@ -615,14 +638,6 @@ private: Map<int, ShadowCubemap> shadow_cubemaps; ShadowCubemap *_get_shadow_cubemap(int p_size); - struct ShadowMap { - RID depth; - RID fb; - }; - - Map<Vector2i, ShadowMap> shadow_maps; - ShadowMap *_get_shadow_map(const Size2i &p_size); - void _create_shadow_cubemaps(); /* LIGHT INSTANCE */ @@ -642,7 +657,7 @@ private: RS::LightType light_type = RS::LIGHT_DIRECTIONAL; - ShadowTransform shadow_transform[4]; + ShadowTransform shadow_transform[6]; AABB aabb; RID self; @@ -721,8 +736,9 @@ private: float volumetric_fog_light_energy = 0.0; float volumetric_fog_length = 64.0; float volumetric_fog_detail_spread = 2.0; - RS::EnvVolumetricFogShadowFilter volumetric_fog_shadow_filter = RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_LOW; float volumetric_fog_gi_inject = 0.0; + bool volumetric_fog_temporal_reprojection = true; + float volumetric_fog_temporal_reprojection_amount = 0.9; /// Glow @@ -823,6 +839,9 @@ private: /* RENDER BUFFERS */ + ClusterBuilderSharedDataRD cluster_builder_shared; + ClusterBuilderRD *current_cluster_builder = nullptr; + struct SDFGI; struct VolumetricFog; @@ -848,6 +867,8 @@ private: SDFGI *sdfgi = nullptr; VolumetricFog *volumetric_fog = nullptr; + ClusterBuilderRD *cluster_builder = nullptr; + //built-in textures used for ping pong image processing and blurring struct Blur { RID texture; @@ -887,6 +908,16 @@ private: RID giprobe_textures[MAX_GIPROBES]; RID giprobe_buffer; + + RID ambient_buffer; + RID reflection_buffer; + bool using_half_size_gi = false; + + struct GI { + RID full_buffer; + RID full_dispatch; + RID full_mask; + } gi; }; RID default_giprobe_buffer; @@ -948,6 +979,8 @@ private: RID scroll_occlusion_uniform_set; RID integrate_uniform_set; RID lights_buffer; + + bool all_dynamic_lights_dirty = true; }; //used for rendering (voxelization) @@ -1005,10 +1038,18 @@ private: float y_mult = 1.0; uint32_t render_pass = 0; + + int32_t cascade_dynamic_light_count[SDFGI::MAX_CASCADES]; //used dynamically }; + void _sdfgi_update_light(RID p_render_buffers, RID p_environment); + void _sdfgi_update_probes(RID p_render_buffers, RID p_environment); + void _sdfgi_store_probes(RID p_render_buffers); + RS::EnvironmentSDFGIRayCount sdfgi_ray_count = RS::ENV_SDFGI_RAY_COUNT_16; RS::EnvironmentSDFGIFramesToConverge sdfgi_frames_to_converge = RS::ENV_SDFGI_CONVERGE_IN_10_FRAMES; + RS::EnvironmentSDFGIFramesToUpdateLight sdfgi_frames_to_update_light = RS::ENV_SDFGI_UPDATE_LIGHT_IN_4_FRAMES; + float sdfgi_solid_cell_ratio = 0.25; Vector3 sdfgi_debug_probe_pos; Vector3 sdfgi_debug_probe_dir; @@ -1105,8 +1146,8 @@ private: float attenuation; uint32_t type; - float spot_angle; - float spot_attenuation; + float cos_spot_angle; + float inv_spot_attenuation; float radius; float shadow_color[4]; @@ -1246,23 +1287,28 @@ private: float z_far; float proj_info[4]; - + float ao_color[3]; uint32_t max_giprobes; + uint32_t high_quality_vct; - uint32_t use_sdfgi; uint32_t orthogonal; - - float ao_color[3]; - uint32_t pad; + uint32_t pad[2]; float cam_rotation[12]; }; RID sdfgi_ubo; - enum { - MODE_MAX = 1 + enum Mode { + MODE_GIPROBE, + MODE_SDFGI, + MODE_COMBINED, + MODE_HALF_RES_GIPROBE, + MODE_HALF_RES_SDFGI, + MODE_HALF_RES_COMBINED, + MODE_MAX }; + bool half_resolution = false; GiShaderRD shader; RID shader_version; RID pipelines[MODE_MAX]; @@ -1287,14 +1333,23 @@ private: struct Cluster { /* Scene State UBO */ - struct ReflectionData { //should always be 128 bytes + enum { + REFLECTION_AMBIENT_DISABLED = 0, + REFLECTION_AMBIENT_ENVIRONMENT = 1, + REFLECTION_AMBIENT_COLOR = 2, + }; + + struct ReflectionData { float box_extents[3]; float index; float box_offset[3]; uint32_t mask; - float params[4]; // intensity, 0, interior , boxproject float ambient[3]; // ambient color, + float intensity; + bool exterior; + bool box_project; uint32_t ambient_mode; + uint32_t pad; float local_matrix[16]; // up to here for spot and omni, rest is for directional }; @@ -1303,10 +1358,15 @@ private: float inv_radius; float direction[3]; float size; - uint16_t attenuation_energy[2]; //16 bits attenuation, then energy - uint8_t color_specular[4]; //rgb color, a specular (8 bit unorm) - uint16_t cone_attenuation_angle[2]; // attenuation and angle, (16bit float) - uint8_t shadow_color_enabled[4]; //shadow rgb color, a>0.5 enabled (8bit unorm) + + float color[3]; + float attenuation; + + float inv_spot_attenuation; + float cos_spot_angle; + float specular_amount; + uint32_t shadow_enabled; + float atlas_rect[4]; // in omni, used for atlas uv, in spot, used for projector uv float shadow_matrix[16]; float shadow_bias; @@ -1370,31 +1430,85 @@ private: float normal_fade; }; + template <class T> + struct InstanceSort { + float depth; + T *instance; + bool operator<(const InstanceSort &p_sort) const { + return depth < p_sort.depth; + } + }; + ReflectionData *reflections; + InstanceSort<ReflectionProbeInstance> *reflection_sort; uint32_t max_reflections; RID reflection_buffer; uint32_t max_reflection_probes_per_instance; + uint32_t reflection_count = 0; DecalData *decals; + InstanceSort<DecalInstance> *decal_sort; uint32_t max_decals; RID decal_buffer; + uint32_t decal_count; + + LightData *omni_lights; + LightData *spot_lights; - LightData *lights; + InstanceSort<LightInstance> *omni_light_sort; + InstanceSort<LightInstance> *spot_light_sort; uint32_t max_lights; - RID light_buffer; - RID *lights_instances; - Rect2i *lights_shadow_rect_cache; - uint32_t lights_shadow_rect_cache_count = 0; + RID omni_light_buffer; + RID spot_light_buffer; + uint32_t omni_light_count = 0; + uint32_t spot_light_count = 0; DirectionalLightData *directional_lights; uint32_t max_directional_lights; RID directional_light_buffer; - LightClusterBuilder builder; - } cluster; + struct RenderState { + RID render_buffers; + Transform cam_transform; + CameraMatrix cam_projection; + bool cam_ortogonal = false; + const PagedArray<GeometryInstance *> *instances = nullptr; + const PagedArray<RID> *lights = nullptr; + const PagedArray<RID> *reflection_probes = nullptr; + const PagedArray<RID> *gi_probes = nullptr; + const PagedArray<RID> *decals = nullptr; + const PagedArray<RID> *lightmaps = nullptr; + RID environment; + RID camera_effects; + RID shadow_atlas; + RID reflection_atlas; + RID reflection_probe; + int reflection_probe_pass = 0; + float screen_lod_threshold = 0.0; + + const RenderShadowData *render_shadows = nullptr; + int render_shadow_count = 0; + const RenderSDFGIData *render_sdfgi_regions = nullptr; + int render_sdfgi_region_count = 0; + const RenderSDFGIUpdateData *sdfgi_update_data = nullptr; + + uint32_t directional_light_count = 0; + uint32_t gi_probe_count = 0; + + LocalVector<int> cube_shadows; + LocalVector<int> shadows; + LocalVector<int> directional_shadows; + + bool depth_prepass_used; + } render_state; + struct VolumetricFog { + enum { + MAX_TEMPORAL_FRAMES = 16 + }; + uint32_t width = 0; uint32_t height = 0; uint32_t depth = 0; @@ -1403,6 +1517,8 @@ private: float spread; RID light_density_map; + RID prev_light_density_map; + RID fog_map; RID uniform_set; RID uniform_set2; @@ -1410,6 +1526,8 @@ private: RID sky_uniform_set; int last_shadow_filter = -1; + + Transform prev_cam_transform; }; enum { @@ -1421,7 +1539,7 @@ private: }; struct VolumetricFogShader { - struct PushConstant { + struct ParamsUBO { float fog_frustum_size_begin[2]; float fog_frustum_size_end[2]; @@ -1439,13 +1557,24 @@ private: float detail_spread; float gi_inject; uint32_t max_gi_probes; - uint32_t pad; + uint32_t cluster_type_size; + + float screen_size[2]; + uint32_t cluster_shift; + uint32_t cluster_width; + + uint32_t max_cluster_element_count_div_32; + uint32_t use_temporal_reprojection; + uint32_t temporal_frame; + float temporal_blend; float cam_rotation[12]; + float to_prev_view[16]; }; VolumetricFogShaderRD shader; + RID params_ubo; RID shader_version; RID pipelines[VOLUMETRIC_FOG_SHADER_MAX]; @@ -1453,9 +1582,7 @@ private: uint32_t volumetric_fog_depth = 128; uint32_t volumetric_fog_size = 128; - bool volumetric_fog_filter_active = false; - uint32_t volumetric_fog_directional_shadow_shrink = 512; - uint32_t volumetric_fog_positional_shadow_shrink = 512; + bool volumetric_fog_filter_active = true; void _volumetric_fog_erase(RenderBuffers *rb); void _update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_gi_probe_count); @@ -1470,13 +1597,21 @@ private: float weight; }; + uint32_t max_cluster_elements = 512; bool low_end = false; + void _render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0, bool p_open_pass = true, bool p_close_pass = true, bool p_clear_region = true); + void _render_sdfgi_region(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances); + void _render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result); + public: + virtual Transform geometry_instance_get_transform(GeometryInstance *p_instance) = 0; + virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance) = 0; + /* SHADOW ATLAS API */ RID shadow_atlas_create(); - void shadow_atlas_set_size(RID p_atlas, int p_size); + void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits = false); void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision); bool shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version); _FORCE_INLINE_ bool shadow_atlas_owns_light_instance(RID p_atlas, RID p_light_intance) { @@ -1497,7 +1632,7 @@ public: return Size2(atlas->size, atlas->size); } - void directional_shadow_atlas_set_size(int p_size); + void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false); int get_directional_light_shadow_size(RID p_light_intance); void set_directional_shadow_count(int p_count); @@ -1516,7 +1651,6 @@ public: virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const; virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const; virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const; - virtual void sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count); RID sdfgi_get_ubo() const { return gi.sdfgi_ubo; } /* SKY API */ @@ -1573,12 +1707,10 @@ public: float environment_get_fog_height_density(RID p_env) const; float environment_get_fog_aerial_perspective(RID p_env) const; - void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RS::EnvVolumetricFogShadowFilter p_shadow_filter); + void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount); virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth); virtual void environment_set_volumetric_fog_filter_active(bool p_enable); - virtual void environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size); - virtual void environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size); void environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance); void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect); @@ -1592,6 +1724,7 @@ public: virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias); virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count); virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames); + virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update); void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality); RS::EnvironmentSSRRoughnessQuality environment_get_ssr_roughness_quality() const; @@ -1822,10 +1955,25 @@ public: return decal->transform; } + virtual RID lightmap_instance_create(RID p_lightmap); + virtual void lightmap_instance_set_transform(RID p_lightmap, const Transform &p_transform); + _FORCE_INLINE_ bool lightmap_instance_is_valid(RID p_lightmap_instance) { + return lightmap_instance_owner.getornull(p_lightmap_instance) != nullptr; + } + + _FORCE_INLINE_ RID lightmap_instance_get_lightmap(RID p_lightmap_instance) { + LightmapInstance *li = lightmap_instance_owner.getornull(p_lightmap_instance); + return li->lightmap; + } + _FORCE_INLINE_ Transform lightmap_instance_get_transform(RID p_lightmap_instance) { + LightmapInstance *li = lightmap_instance_owner.getornull(p_lightmap_instance); + return li->transform; + } + RID gi_probe_instance_create(RID p_base); void gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform); bool gi_probe_needs_update(RID p_probe) const; - void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::InstanceBase *> &p_dynamic_objects); + void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects); void gi_probe_set_quality(RS::GIProbeQuality p_quality) { gi_probe_quality = p_quality; } @@ -1875,11 +2023,14 @@ public: */ RID render_buffers_create(); void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding); + void gi_set_use_half_resolution(bool p_enable); RID render_buffers_get_ao_texture(RID p_render_buffers); RID render_buffers_get_back_buffer_texture(RID p_render_buffers); RID render_buffers_get_gi_probe_buffer(RID p_render_buffers); RID render_buffers_get_default_gi_probe_buffer(); + RID render_buffers_get_gi_ambient_texture(RID p_render_buffers); + RID render_buffers_get_gi_reflection_texture(RID p_render_buffers); uint32_t render_buffers_get_sdfgi_cascade_count(RID p_render_buffers) const; bool render_buffers_is_sdfgi_enabled(RID p_render_buffers) const; @@ -1900,16 +2051,11 @@ public: float render_buffers_get_volumetric_fog_end(RID p_render_buffers); float render_buffers_get_volumetric_fog_detail_spread(RID p_render_buffers); - void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold); - - void render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<InstanceBase *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0); - - void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region); + void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data = nullptr); - void render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<InstanceBase *> &p_instances); - void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result); + void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<InstanceBase *> &p_instances); + void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances); virtual void set_scene_pass(uint64_t p_pass) { scene_pass = p_pass; @@ -1960,10 +2106,9 @@ public: virtual void set_time(double p_time, double p_step); - RID get_cluster_builder_texture(); - RID get_cluster_builder_indices_buffer(); RID get_reflection_probe_buffer(); - RID get_positional_light_buffer(); + RID get_omni_light_buffer(); + RID get_spot_light_buffer(); RID get_directional_light_buffer(); RID get_decal_buffer(); int get_max_directional_lights() const; diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.cpp b/servers/rendering/renderer_rd/renderer_storage_rd.cpp index 68983da408..a1358f94fa 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_storage_rd.cpp @@ -873,7 +873,7 @@ void RendererStorageRD::_texture_2d_update(RID p_texture, const Ref<Image> &p_im TextureToRDFormat f; Ref<Image> validated = _validate_texture_format(p_image, f); - RD::get_singleton()->texture_update(tex->rd_texture, p_layer, validated->get_data(), !p_immediate); + RD::get_singleton()->texture_update(tex->rd_texture, p_layer, validated->get_data()); } void RendererStorageRD::texture_2d_update_immediate(RID p_texture, const Ref<Image> &p_image, int p_layer) { @@ -918,7 +918,7 @@ void RendererStorageRD::texture_3d_update(RID p_texture, const Vector<Ref<Image> } } - RD::get_singleton()->texture_update(tex->rd_texture, 0, all_data, true); + RD::get_singleton()->texture_update(tex->rd_texture, 0, all_data); } void RendererStorageRD::texture_proxy_update(RID p_texture, RID p_proxy_to) { @@ -1438,7 +1438,7 @@ void RendererStorageRD::shader_set_code(RID p_shader, const String &p_code) { for (Set<Material *>::Element *E = shader->owners.front(); E; E = E->next()) { Material *material = E->get(); - material->instance_dependency.instance_notify_changed(false, true); + material->dependency.changed_notify(DEPENDENCY_CHANGED_MATERIAL); _material_queue_update(material, true, true); } } @@ -1499,6 +1499,15 @@ void RendererStorageRD::shader_set_data_request_function(ShaderType p_shader_typ shader_data_request_func[p_shader_type] = p_function; } +RS::ShaderNativeSourceCode RendererStorageRD::shader_get_native_source_code(RID p_shader) const { + Shader *shader = shader_owner.getornull(p_shader); + ERR_FAIL_COND_V(!shader, RS::ShaderNativeSourceCode()); + if (shader->data) { + return shader->data->get_native_source_code(); + } + return RS::ShaderNativeSourceCode(); +} + /* COMMON MATERIAL API */ RID RendererStorageRD::material_create() { @@ -1547,7 +1556,8 @@ void RendererStorageRD::material_set_shader(RID p_material, RID p_shader) { } if (p_shader.is_null()) { - material->instance_dependency.instance_notify_changed(false, true); + material->dependency.changed_notify(DEPENDENCY_CHANGED_MATERIAL); + material->shader_id = 0; return; } @@ -1555,6 +1565,7 @@ void RendererStorageRD::material_set_shader(RID p_material, RID p_shader) { ERR_FAIL_COND(!shader); material->shader = shader; material->shader_type = shader->type; + material->shader_id = p_shader.get_local_index(); shader->owners.insert(material); if (shader->type == SHADER_TYPE_MAX) { @@ -1568,7 +1579,7 @@ void RendererStorageRD::material_set_shader(RID p_material, RID p_shader) { material->data->set_next_pass(material->next_pass); material->data->set_render_priority(material->priority); //updating happens later - material->instance_dependency.instance_notify_changed(false, true); + material->dependency.changed_notify(DEPENDENCY_CHANGED_MATERIAL); _material_queue_update(material, true, true); } @@ -1613,7 +1624,7 @@ void RendererStorageRD::material_set_next_pass(RID p_material, RID p_next_materi material->data->set_next_pass(p_next_material); } - material->instance_dependency.instance_notify_changed(false, true); + material->dependency.changed_notify(DEPENDENCY_CHANGED_MATERIAL); } void RendererStorageRD::material_set_render_priority(RID p_material, int priority) { @@ -1663,10 +1674,10 @@ void RendererStorageRD::material_get_instance_shader_parameters(RID p_material, } } -void RendererStorageRD::material_update_dependency(RID p_material, InstanceBaseDependency *p_instance) { +void RendererStorageRD::material_update_dependency(RID p_material, DependencyTracker *p_instance) { Material *material = material_owner.getornull(p_material); ERR_FAIL_COND(!material); - p_instance->update_dependency(&material->instance_dependency); + p_instance->update_dependency(&material->dependency); if (material->next_pass.is_valid()) { material_update_dependency(material->next_pass, p_instance); } @@ -2596,7 +2607,13 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su _mesh_instance_add_surface(mi, mesh, mesh->surface_count - 1); } - mesh->instance_dependency.instance_notify_changed(true, true); + mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + + for (Set<Mesh *>::Element *E = mesh->shadow_owners.front(); E; E = E->next()) { + Mesh *shadow_owner = E->get(); + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + } mesh->material_cache.clear(); } @@ -2638,7 +2655,7 @@ void RendererStorageRD::mesh_surface_set_material(RID p_mesh, int p_surface, RID ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); mesh->surfaces[p_surface]->material = p_material; - mesh->instance_dependency.instance_notify_changed(false, true); + mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MATERIAL); mesh->material_cache.clear(); } @@ -2813,6 +2830,25 @@ AABB RendererStorageRD::mesh_get_aabb(RID p_mesh, RID p_skeleton) { return aabb; } +void RendererStorageRD::mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh) { + Mesh *mesh = mesh_owner.getornull(p_mesh); + ERR_FAIL_COND(!mesh); + + Mesh *shadow_mesh = mesh_owner.getornull(mesh->shadow_mesh); + if (shadow_mesh) { + shadow_mesh->shadow_owners.erase(mesh); + } + mesh->shadow_mesh = p_shadow_mesh; + + shadow_mesh = mesh_owner.getornull(mesh->shadow_mesh); + + if (shadow_mesh) { + shadow_mesh->shadow_owners.insert(mesh); + } + + mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); +} + void RendererStorageRD::mesh_clear(RID p_mesh) { Mesh *mesh = mesh_owner.getornull(p_mesh); ERR_FAIL_COND(!mesh); @@ -2858,8 +2894,14 @@ void RendererStorageRD::mesh_clear(RID p_mesh) { MeshInstance *mi = E->get(); _mesh_instance_clear(mi); } - mesh->instance_dependency.instance_notify_changed(true, true); mesh->has_bone_weights = false; + mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + + for (Set<Mesh *>::Element *E = mesh->shadow_owners.front(); E; E = E->next()) { + Mesh *shadow_owner = E->get(); + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + } } bool RendererStorageRD::mesh_needs_instance(RID p_mesh, bool p_has_skeleton) { @@ -3002,7 +3044,7 @@ void RendererStorageRD::update_mesh_instances() { MeshInstance *mi = dirty_mesh_instance_weights.first()->self(); if (mi->blend_weights_buffer.is_valid()) { - RD::get_singleton()->buffer_update(mi->blend_weights_buffer, 0, mi->blend_weights.size() * sizeof(float), mi->blend_weights.ptr(), true); + RD::get_singleton()->buffer_update(mi->blend_weights_buffer, 0, mi->blend_weights.size() * sizeof(float), mi->blend_weights.ptr()); } dirty_mesh_instance_weights.remove(&mi->weight_update_list); mi->weights_dirty = false; @@ -3056,7 +3098,7 @@ void RendererStorageRD::update_mesh_instances() { RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SkeletonShader::PushConstant)); //dispatch without barrier, so all is done at the same time - RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.vertex_count, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.vertex_count, 1, 1); } mi->dirty = false; @@ -3298,6 +3340,8 @@ void RendererStorageRD::multimesh_allocate(RID p_multimesh, int p_instances, RS: if (multimesh->instances) { multimesh->buffer = RD::get_singleton()->storage_buffer_create(multimesh->instances * multimesh->stride_cache * 4); } + + multimesh->dependency.changed_notify(DEPENDENCY_CHANGED_MULTIMESH); } int RendererStorageRD::multimesh_get_instance_count(RID p_multimesh) const { @@ -3331,7 +3375,7 @@ void RendererStorageRD::multimesh_set_mesh(RID p_multimesh, RID p_mesh) { } } - multimesh->instance_dependency.instance_notify_changed(true, true); + multimesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); } #define MULTIMESH_DIRTY_REGION_SIZE 512 @@ -3668,7 +3712,7 @@ void RendererStorageRD::multimesh_set_buffer(RID p_multimesh, const Vector<float { const float *r = p_buffer.ptr(); - RD::get_singleton()->buffer_update(multimesh->buffer, 0, p_buffer.size() * sizeof(float), r, false); + RD::get_singleton()->buffer_update(multimesh->buffer, 0, p_buffer.size() * sizeof(float), r); multimesh->buffer_set = true; } @@ -3690,7 +3734,7 @@ void RendererStorageRD::multimesh_set_buffer(RID p_multimesh, const Vector<float const float *data = p_buffer.ptr(); _multimesh_re_create_aabb(multimesh, data, multimesh->instances); - multimesh->instance_dependency.instance_notify_changed(true, false); + multimesh->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } } @@ -3731,6 +3775,8 @@ void RendererStorageRD::multimesh_set_visible_instances(RID p_multimesh, int p_v } multimesh->visible_instances = p_visible; + + multimesh->dependency.changed_notify(DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES); } int RendererStorageRD::multimesh_get_visible_instances(RID p_multimesh) const { @@ -3765,14 +3811,14 @@ void RendererStorageRD::_update_dirty_multimeshes() { if (multimesh->data_cache_used_dirty_regions > 32 || multimesh->data_cache_used_dirty_regions > visible_region_count / 2) { //if there too many dirty regions, or represent the majority of regions, just copy all, else transfer cost piles up too much - RD::get_singleton()->buffer_update(multimesh->buffer, 0, MIN(visible_region_count * region_size, multimesh->instances * multimesh->stride_cache * sizeof(float)), data, false); + RD::get_singleton()->buffer_update(multimesh->buffer, 0, MIN(visible_region_count * region_size, multimesh->instances * multimesh->stride_cache * sizeof(float)), data); } else { //not that many regions? update them all for (uint32_t i = 0; i < visible_region_count; i++) { if (multimesh->data_cache_dirty_regions[i]) { uint64_t offset = i * region_size; uint64_t size = multimesh->stride_cache * multimesh->instances * sizeof(float); - RD::get_singleton()->buffer_update(multimesh->buffer, offset, MIN(region_size, size - offset), &data[i * region_size], false); + RD::get_singleton()->buffer_update(multimesh->buffer, offset, MIN(region_size, size - offset), &data[i * region_size]); } } } @@ -3788,7 +3834,7 @@ void RendererStorageRD::_update_dirty_multimeshes() { //aabb is dirty.. _multimesh_re_create_aabb(multimesh, data, visible_instances); multimesh->aabb_dirty = false; - multimesh->instance_dependency.instance_notify_changed(true, false); + multimesh->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } } @@ -3926,7 +3972,7 @@ void RendererStorageRD::particles_set_custom_aabb(RID p_particles, const AABB &p Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->custom_aabb = p_aabb; - particles->instance_dependency.instance_notify_changed(true, false); + particles->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } void RendererStorageRD::particles_set_speed_scale(RID p_particles, float p_scale) { @@ -4155,24 +4201,18 @@ RID RendererStorageRD::particles_get_draw_pass_mesh(RID p_particles, int p_pass) return particles->draw_passes[p_pass]; } -void RendererStorageRD::particles_add_collision(RID p_particles, InstanceBaseDependency *p_instance) { - RendererSceneRender::InstanceBase *instance = static_cast<RendererSceneRender::InstanceBase *>(p_instance); - +void RendererStorageRD::particles_add_collision(RID p_particles, RID p_particles_collision_instance) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); - ERR_FAIL_COND(instance->base_type != RS::INSTANCE_PARTICLES_COLLISION); - - particles->collisions.insert(instance); + particles->collisions.insert(p_particles_collision_instance); } -void RendererStorageRD::particles_remove_collision(RID p_particles, InstanceBaseDependency *p_instance) { - RendererSceneRender::InstanceBase *instance = static_cast<RendererSceneRender::InstanceBase *>(p_instance); - +void RendererStorageRD::particles_remove_collision(RID p_particles, RID p_particles_collision_instance) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); - particles->collisions.erase(instance); + particles->collisions.erase(p_particles_collision_instance); } void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta) { @@ -4272,9 +4312,15 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta to_particles = p_particles->emission_transform.affine_inverse(); } uint32_t collision_3d_textures_used = 0; - for (const Set<RendererSceneRender::InstanceBase *>::Element *E = p_particles->collisions.front(); E; E = E->next()) { - ParticlesCollision *pc = particles_collision_owner.getornull(E->get()->base); - Transform to_collider = E->get()->transform; + for (const Set<RID>::Element *E = p_particles->collisions.front(); E; E = E->next()) { + ParticlesCollisionInstance *pci = particles_collision_instance_owner.getornull(E->get()); + if (!pci || !pci->active) { + continue; + } + ParticlesCollision *pc = particles_collision_owner.getornull(pci->collision); + ERR_CONTINUE(!pc); + + Transform to_collider = pci->transform; if (p_particles->use_local_coords) { to_collider = to_particles * to_collider; } @@ -4463,7 +4509,7 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta if (sub_emitter && sub_emitter->emission_storage_buffer.is_valid()) { // print_line("updating subemitter buffer"); int32_t zero[4] = { 0, sub_emitter->amount, 0, 0 }; - RD::get_singleton()->buffer_update(sub_emitter->emission_storage_buffer, 0, sizeof(uint32_t) * 4, zero, true); + RD::get_singleton()->buffer_update(sub_emitter->emission_storage_buffer, 0, sizeof(uint32_t) * 4, zero); push_constant.can_emit = true; if (sub_emitter->emitting) { @@ -4481,13 +4527,13 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta } if (p_particles->emission_buffer && p_particles->emission_buffer->particle_count) { - RD::get_singleton()->buffer_update(p_particles->emission_storage_buffer, 0, sizeof(uint32_t) * 4 + sizeof(ParticleEmissionBuffer::Data) * p_particles->emission_buffer->particle_count, p_particles->emission_buffer, true); + RD::get_singleton()->buffer_update(p_particles->emission_storage_buffer, 0, sizeof(uint32_t) * 4 + sizeof(ParticleEmissionBuffer::Data) * p_particles->emission_buffer->particle_count, p_particles->emission_buffer); p_particles->emission_buffer->particle_count = 0; } p_particles->clear = false; - RD::get_singleton()->buffer_update(p_particles->frame_params_buffer, 0, sizeof(ParticlesFrameParams), &frame_params, true); + RD::get_singleton()->buffer_update(p_particles->frame_params_buffer, 0, sizeof(ParticlesFrameParams), &frame_params); ParticlesMaterialData *m = (ParticlesMaterialData *)material_get_data(p_particles->process_material, SHADER_TYPE_PARTICLES); if (!m) { @@ -4509,7 +4555,7 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ParticlesShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_particles->amount, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); } @@ -4563,7 +4609,7 @@ void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 & RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); @@ -4575,7 +4621,7 @@ void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 & RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); } @@ -4682,12 +4728,12 @@ void RendererStorageRD::update_particles() { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); } - particles->instance_dependency.instance_notify_changed(true, false); //make sure shadows are updated + particles->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } } @@ -4817,6 +4863,10 @@ Variant RendererStorageRD::ParticlesShaderData::get_default_parameter(const Stri return Variant(); } +RS::ShaderNativeSourceCode RendererStorageRD::ParticlesShaderData::get_native_source_code() const { + return base_singleton->particles_shader.shader.version_get_native_source_code(version); +} + RendererStorageRD::ParticlesShaderData::ParticlesShaderData() { valid = false; } @@ -4986,7 +5036,7 @@ void RendererStorageRD::particles_collision_set_collision_type(RID p_particles_c particles_collision->heightfield_texture = RID(); } particles_collision->type = p_type; - particles_collision->instance_dependency.instance_notify_changed(true, false); + particles_collision->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } void RendererStorageRD::particles_collision_set_cull_mask(RID p_particles_collision, uint32_t p_cull_mask) { @@ -5000,7 +5050,7 @@ void RendererStorageRD::particles_collision_set_sphere_radius(RID p_particles_co ERR_FAIL_COND(!particles_collision); particles_collision->radius = p_radius; - particles_collision->instance_dependency.instance_notify_changed(true, false); + particles_collision->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } void RendererStorageRD::particles_collision_set_box_extents(RID p_particles_collision, const Vector3 &p_extents) { @@ -5008,7 +5058,7 @@ void RendererStorageRD::particles_collision_set_box_extents(RID p_particles_coll ERR_FAIL_COND(!particles_collision); particles_collision->extents = p_extents; - particles_collision->instance_dependency.instance_notify_changed(true, false); + particles_collision->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } void RendererStorageRD::particles_collision_set_attractor_strength(RID p_particles_collision, float p_strength) { @@ -5042,7 +5092,7 @@ void RendererStorageRD::particles_collision_set_field_texture(RID p_particles_co void RendererStorageRD::particles_collision_height_field_update(RID p_particles_collision) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_particles_collision); ERR_FAIL_COND(!particles_collision); - particles_collision->instance_dependency.instance_notify_changed(true, false); + particles_collision->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } void RendererStorageRD::particles_collision_set_height_field_resolution(RID p_particles_collision, RS::ParticlesCollisionHeightfieldResolution p_resolution) { @@ -5096,6 +5146,22 @@ bool RendererStorageRD::particles_collision_is_heightfield(RID p_particles_colli return particles_collision->type == RS::PARTICLES_COLLISION_TYPE_HEIGHTFIELD_COLLIDE; } +RID RendererStorageRD::particles_collision_instance_create(RID p_collision) { + ParticlesCollisionInstance pci; + pci.collision = p_collision; + return particles_collision_instance_owner.make_rid(pci); +} +void RendererStorageRD::particles_collision_instance_set_transform(RID p_collision_instance, const Transform &p_transform) { + ParticlesCollisionInstance *pci = particles_collision_instance_owner.getornull(p_collision_instance); + ERR_FAIL_COND(!pci); + pci->transform = p_transform; +} +void RendererStorageRD::particles_collision_instance_set_active(RID p_collision_instance, bool p_active) { + ParticlesCollisionInstance *pci = particles_collision_instance_owner.getornull(p_collision_instance); + ERR_FAIL_COND(!pci); + pci->active = p_active; +} + /* SKELETON API */ RID RendererStorageRD::skeleton_create() { @@ -5149,6 +5215,8 @@ void RendererStorageRD::skeleton_allocate(RID p_skeleton, int p_bones, bool p_2d skeleton->uniform_set_mi = RD::get_singleton()->uniform_set_create(uniforms, skeleton_shader.version_shader[0], SkeletonShader::UNIFORM_SET_SKELETON); } } + + skeleton->dependency.changed_notify(DEPENDENCY_CHANGED_SKELETON_DATA); } int RendererStorageRD::skeleton_get_bone_count(RID p_skeleton) const { @@ -5264,12 +5332,13 @@ void RendererStorageRD::_update_dirty_skeletons() { Skeleton *skeleton = skeleton_dirty_list; if (skeleton->size) { - RD::get_singleton()->buffer_update(skeleton->buffer, 0, skeleton->data.size() * sizeof(float), skeleton->data.ptr(), false); + RD::get_singleton()->buffer_update(skeleton->buffer, 0, skeleton->data.size() * sizeof(float), skeleton->data.ptr()); } skeleton_dirty_list = skeleton->dirty_list; - skeleton->instance_dependency.instance_notify_changed(true, false); + skeleton->dependency.changed_notify(DEPENDENCY_CHANGED_SKELETON_BONES); + skeleton->version++; skeleton->dirty = false; @@ -5290,17 +5359,20 @@ RID RendererStorageRD::light_create(RS::LightType p_type) { light.param[RS::LIGHT_PARAM_SPECULAR] = 0.5; light.param[RS::LIGHT_PARAM_RANGE] = 1.0; light.param[RS::LIGHT_PARAM_SIZE] = 0.0; + light.param[RS::LIGHT_PARAM_ATTENUATION] = 1.0; light.param[RS::LIGHT_PARAM_SPOT_ANGLE] = 45; + light.param[RS::LIGHT_PARAM_SPOT_ATTENUATION] = 1.0; light.param[RS::LIGHT_PARAM_SHADOW_MAX_DISTANCE] = 0; light.param[RS::LIGHT_PARAM_SHADOW_SPLIT_1_OFFSET] = 0.1; light.param[RS::LIGHT_PARAM_SHADOW_SPLIT_2_OFFSET] = 0.3; light.param[RS::LIGHT_PARAM_SHADOW_SPLIT_3_OFFSET] = 0.6; light.param[RS::LIGHT_PARAM_SHADOW_FADE_START] = 0.8; - light.param[RS::LIGHT_PARAM_SHADOW_BIAS] = 0.02; light.param[RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS] = 1.0; + light.param[RS::LIGHT_PARAM_SHADOW_BIAS] = 0.02; + light.param[RS::LIGHT_PARAM_SHADOW_BLUR] = 0; light.param[RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE] = 20.0; + light.param[RS::LIGHT_PARAM_SHADOW_VOLUMETRIC_FOG_FADE] = 0.1; light.param[RS::LIGHT_PARAM_TRANSMITTANCE_BIAS] = 0.05; - light.param[RS::LIGHT_PARAM_SHADOW_VOLUMETRIC_FOG_FADE] = 1.0; return light_owner.make_rid(light); } @@ -5328,7 +5400,7 @@ void RendererStorageRD::light_set_param(RID p_light, RS::LightParam p_param, flo case RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE: case RS::LIGHT_PARAM_SHADOW_BIAS: { light->version++; - light->instance_dependency.instance_notify_changed(true, false); + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } break; default: { } @@ -5343,7 +5415,7 @@ void RendererStorageRD::light_set_shadow(RID p_light, bool p_enabled) { light->shadow = p_enabled; light->version++; - light->instance_dependency.instance_notify_changed(true, false); + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } void RendererStorageRD::light_set_shadow_color(RID p_light, const Color &p_color) { @@ -5385,7 +5457,7 @@ void RendererStorageRD::light_set_cull_mask(RID p_light, uint32_t p_mask) { light->cull_mask = p_mask; light->version++; - light->instance_dependency.instance_notify_changed(true, false); + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } void RendererStorageRD::light_set_reverse_cull_face_mode(RID p_light, bool p_enabled) { @@ -5395,7 +5467,7 @@ void RendererStorageRD::light_set_reverse_cull_face_mode(RID p_light, bool p_ena light->reverse_cull = p_enabled; light->version++; - light->instance_dependency.instance_notify_changed(true, false); + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } void RendererStorageRD::light_set_bake_mode(RID p_light, RS::LightBakeMode p_bake_mode) { @@ -5405,7 +5477,7 @@ void RendererStorageRD::light_set_bake_mode(RID p_light, RS::LightBakeMode p_bak light->bake_mode = p_bake_mode; light->version++; - light->instance_dependency.instance_notify_changed(true, false); + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } void RendererStorageRD::light_set_max_sdfgi_cascade(RID p_light, uint32_t p_cascade) { @@ -5415,7 +5487,7 @@ void RendererStorageRD::light_set_max_sdfgi_cascade(RID p_light, uint32_t p_casc light->max_sdfgi_cascade = p_cascade; light->version++; - light->instance_dependency.instance_notify_changed(true, false); + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } void RendererStorageRD::light_omni_set_shadow_mode(RID p_light, RS::LightOmniShadowMode p_mode) { @@ -5425,7 +5497,7 @@ void RendererStorageRD::light_omni_set_shadow_mode(RID p_light, RS::LightOmniSha light->omni_shadow_mode = p_mode; light->version++; - light->instance_dependency.instance_notify_changed(true, false); + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } RS::LightOmniShadowMode RendererStorageRD::light_omni_get_shadow_mode(RID p_light) { @@ -5441,7 +5513,7 @@ void RendererStorageRD::light_directional_set_shadow_mode(RID p_light, RS::Light light->directional_shadow_mode = p_mode; light->version++; - light->instance_dependency.instance_notify_changed(true, false); + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } void RendererStorageRD::light_directional_set_blend_splits(RID p_light, bool p_enable) { @@ -5450,7 +5522,7 @@ void RendererStorageRD::light_directional_set_blend_splits(RID p_light, bool p_e light->directional_blend_splits = p_enable; light->version++; - light->instance_dependency.instance_notify_changed(true, false); + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } bool RendererStorageRD::light_directional_get_blend_splits(RID p_light) const { @@ -5549,7 +5621,7 @@ void RendererStorageRD::reflection_probe_set_update_mode(RID p_probe, RS::Reflec ERR_FAIL_COND(!reflection_probe); reflection_probe->update_mode = p_mode; - reflection_probe->instance_dependency.instance_notify_changed(true, false); + reflection_probe->dependency.changed_notify(DEPENDENCY_CHANGED_REFLECTION_PROBE); } void RendererStorageRD::reflection_probe_set_intensity(RID p_probe, float p_intensity) { @@ -5586,7 +5658,7 @@ void RendererStorageRD::reflection_probe_set_max_distance(RID p_probe, float p_d reflection_probe->max_distance = p_distance; - reflection_probe->instance_dependency.instance_notify_changed(true, false); + reflection_probe->dependency.changed_notify(DEPENDENCY_CHANGED_REFLECTION_PROBE); } void RendererStorageRD::reflection_probe_set_extents(RID p_probe, const Vector3 &p_extents) { @@ -5597,7 +5669,7 @@ void RendererStorageRD::reflection_probe_set_extents(RID p_probe, const Vector3 return; } reflection_probe->extents = p_extents; - reflection_probe->instance_dependency.instance_notify_changed(true, false); + reflection_probe->dependency.changed_notify(DEPENDENCY_CHANGED_REFLECTION_PROBE); } void RendererStorageRD::reflection_probe_set_origin_offset(RID p_probe, const Vector3 &p_offset) { @@ -5605,7 +5677,7 @@ void RendererStorageRD::reflection_probe_set_origin_offset(RID p_probe, const Ve ERR_FAIL_COND(!reflection_probe); reflection_probe->origin_offset = p_offset; - reflection_probe->instance_dependency.instance_notify_changed(true, false); + reflection_probe->dependency.changed_notify(DEPENDENCY_CHANGED_REFLECTION_PROBE); } void RendererStorageRD::reflection_probe_set_as_interior(RID p_probe, bool p_enable) { @@ -5613,7 +5685,7 @@ void RendererStorageRD::reflection_probe_set_as_interior(RID p_probe, bool p_ena ERR_FAIL_COND(!reflection_probe); reflection_probe->interior = p_enable; - reflection_probe->instance_dependency.instance_notify_changed(true, false); + reflection_probe->dependency.changed_notify(DEPENDENCY_CHANGED_REFLECTION_PROBE); } void RendererStorageRD::reflection_probe_set_enable_box_projection(RID p_probe, bool p_enable) { @@ -5628,7 +5700,7 @@ void RendererStorageRD::reflection_probe_set_enable_shadows(RID p_probe, bool p_ ERR_FAIL_COND(!reflection_probe); reflection_probe->enable_shadows = p_enable; - reflection_probe->instance_dependency.instance_notify_changed(true, false); + reflection_probe->dependency.changed_notify(DEPENDENCY_CHANGED_REFLECTION_PROBE); } void RendererStorageRD::reflection_probe_set_cull_mask(RID p_probe, uint32_t p_layers) { @@ -5636,7 +5708,7 @@ void RendererStorageRD::reflection_probe_set_cull_mask(RID p_probe, uint32_t p_l ERR_FAIL_COND(!reflection_probe); reflection_probe->cull_mask = p_layers; - reflection_probe->instance_dependency.instance_notify_changed(true, false); + reflection_probe->dependency.changed_notify(DEPENDENCY_CHANGED_REFLECTION_PROBE); } void RendererStorageRD::reflection_probe_set_resolution(RID p_probe, int p_resolution) { @@ -5653,7 +5725,7 @@ void RendererStorageRD::reflection_probe_set_lod_threshold(RID p_probe, float p_ reflection_probe->lod_threshold = p_ratio; - reflection_probe->instance_dependency.instance_notify_changed(true, false); + reflection_probe->dependency.changed_notify(DEPENDENCY_CHANGED_REFLECTION_PROBE); } AABB RendererStorageRD::reflection_probe_get_aabb(RID p_probe) const { @@ -5771,7 +5843,7 @@ void RendererStorageRD::decal_set_extents(RID p_decal, const Vector3 &p_extents) Decal *decal = decal_owner.getornull(p_decal); ERR_FAIL_COND(!decal); decal->extents = p_extents; - decal->instance_dependency.instance_notify_changed(true, false); + decal->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } void RendererStorageRD::decal_set_texture(RID p_decal, RS::DecalTexture p_type, RID p_texture) { @@ -5795,7 +5867,7 @@ void RendererStorageRD::decal_set_texture(RID p_decal, RS::DecalTexture p_type, texture_add_to_decal_atlas(decal->textures[p_type]); } - decal->instance_dependency.instance_notify_changed(false, true); + decal->dependency.changed_notify(DEPENDENCY_CHANGED_DECAL); } void RendererStorageRD::decal_set_emission_energy(RID p_decal, float p_energy) { @@ -5820,7 +5892,7 @@ void RendererStorageRD::decal_set_cull_mask(RID p_decal, uint32_t p_layers) { Decal *decal = decal_owner.getornull(p_decal); ERR_FAIL_COND(!decal); decal->cull_mask = p_layers; - decal->instance_dependency.instance_notify_changed(true, false); + decal->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } void RendererStorageRD::decal_set_distance_fade(RID p_decal, bool p_enabled, float p_begin, float p_length) { @@ -5977,7 +6049,7 @@ void RendererStorageRD::gi_probe_allocate(RID p_gi_probe, const Transform &p_to_ gi_probe->version++; gi_probe->data_version++; - gi_probe->instance_dependency.instance_notify_changed(true, false); + gi_probe->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } AABB RendererStorageRD::gi_probe_get_bounds(RID p_gi_probe) const { @@ -6908,7 +6980,7 @@ void RendererStorageRD::render_target_sdf_process(RID p_render_target) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rt->sdf_buffer_process_uniform_sets[1], 0); //fill [0] RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(RenderTargetSDF::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); /* Process */ @@ -6924,7 +6996,7 @@ void RendererStorageRD::render_target_sdf_process(RID p_render_target) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rt->sdf_buffer_process_uniform_sets[swap ? 1 : 0], 0); push_constant.stride = stride; RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(RenderTargetSDF::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); stride /= 2; swap = !swap; RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -6935,7 +7007,7 @@ void RendererStorageRD::render_target_sdf_process(RID p_render_target) { RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, rt_sdf.pipelines[shrink ? RenderTargetSDF::SHADER_STORE_SHRINK : RenderTargetSDF::SHADER_STORE]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rt->sdf_buffer_process_uniform_sets[swap ? 1 : 0], 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(RenderTargetSDF::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); RD::get_singleton()->compute_list_end(); } @@ -7055,45 +7127,45 @@ void RendererStorageRD::render_target_set_backbuffer_uniform_set(RID p_render_ta rt->backbuffer_uniform_set = p_uniform_set; } -void RendererStorageRD::base_update_dependency(RID p_base, InstanceBaseDependency *p_instance) { +void RendererStorageRD::base_update_dependency(RID p_base, DependencyTracker *p_instance) { if (mesh_owner.owns(p_base)) { Mesh *mesh = mesh_owner.getornull(p_base); - p_instance->update_dependency(&mesh->instance_dependency); + p_instance->update_dependency(&mesh->dependency); } else if (multimesh_owner.owns(p_base)) { MultiMesh *multimesh = multimesh_owner.getornull(p_base); - p_instance->update_dependency(&multimesh->instance_dependency); + p_instance->update_dependency(&multimesh->dependency); if (multimesh->mesh.is_valid()) { base_update_dependency(multimesh->mesh, p_instance); } } else if (reflection_probe_owner.owns(p_base)) { ReflectionProbe *rp = reflection_probe_owner.getornull(p_base); - p_instance->update_dependency(&rp->instance_dependency); + p_instance->update_dependency(&rp->dependency); } else if (decal_owner.owns(p_base)) { Decal *decal = decal_owner.getornull(p_base); - p_instance->update_dependency(&decal->instance_dependency); + p_instance->update_dependency(&decal->dependency); } else if (gi_probe_owner.owns(p_base)) { GIProbe *gip = gi_probe_owner.getornull(p_base); - p_instance->update_dependency(&gip->instance_dependency); + p_instance->update_dependency(&gip->dependency); } else if (lightmap_owner.owns(p_base)) { Lightmap *lm = lightmap_owner.getornull(p_base); - p_instance->update_dependency(&lm->instance_dependency); + p_instance->update_dependency(&lm->dependency); } else if (light_owner.owns(p_base)) { Light *l = light_owner.getornull(p_base); - p_instance->update_dependency(&l->instance_dependency); + p_instance->update_dependency(&l->dependency); } else if (particles_owner.owns(p_base)) { Particles *p = particles_owner.getornull(p_base); - p_instance->update_dependency(&p->instance_dependency); + p_instance->update_dependency(&p->dependency); } else if (particles_collision_owner.owns(p_base)) { ParticlesCollision *pc = particles_collision_owner.getornull(p_base); - p_instance->update_dependency(&pc->instance_dependency); + p_instance->update_dependency(&pc->dependency); } } -void RendererStorageRD::skeleton_update_dependency(RID p_skeleton, InstanceBaseDependency *p_instance) { +void RendererStorageRD::skeleton_update_dependency(RID p_skeleton, DependencyTracker *p_instance) { Skeleton *skeleton = skeleton_owner.getornull(p_skeleton); ERR_FAIL_COND(!skeleton); - p_instance->update_dependency(&skeleton->instance_dependency); + p_instance->update_dependency(&skeleton->dependency); } RS::InstanceType RendererStorageRD::get_base_type(RID p_rid) const { @@ -7299,6 +7371,7 @@ void RendererStorageRD::_update_decal_atlas() { tformat.shareable_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_SRGB); decal_atlas.texture = RD::get_singleton()->texture_create(tformat, RD::TextureView()); + RD::get_singleton()->texture_clear(decal_atlas.texture, Color(0, 0, 0, 0), 0, decal_atlas.mipmaps, 0, 1); { //create the framebuffer @@ -7353,7 +7426,7 @@ void RendererStorageRD::_update_decal_atlas() { prev_texture = mm.texture; } } else { - RD::get_singleton()->texture_clear(mm.texture, clear_color, 0, 1, 0, 1, false); + RD::get_singleton()->texture_clear(mm.texture, clear_color, 0, 1, 0, 1); } } } @@ -8114,21 +8187,31 @@ bool RendererStorageRD::free(RID p_rid) { _update_queued_materials(); } material_set_shader(p_rid, RID()); //clean up shader - material->instance_dependency.instance_notify_deleted(p_rid); + material->dependency.deleted_notify(p_rid); + material_owner.free(p_rid); } else if (mesh_owner.owns(p_rid)) { mesh_clear(p_rid); + mesh_set_shadow_mesh(p_rid, RID()); Mesh *mesh = mesh_owner.getornull(p_rid); - mesh->instance_dependency.instance_notify_deleted(p_rid); + mesh->dependency.deleted_notify(p_rid); if (mesh->instances.size()) { ERR_PRINT("deleting mesh with active instances"); } + if (mesh->shadow_owners.size()) { + for (Set<Mesh *>::Element *E = mesh->shadow_owners.front(); E; E = E->next()) { + Mesh *shadow_owner = E->get(); + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + } + } mesh_owner.free(p_rid); } else if (mesh_instance_owner.owns(p_rid)) { MeshInstance *mi = mesh_instance_owner.getornull(p_rid); _mesh_instance_clear(mi); mi->mesh->instances.erase(mi->I); mi->I = nullptr; + mesh_instance_owner.free(p_rid); memdelete(mi); @@ -8136,17 +8219,17 @@ bool RendererStorageRD::free(RID p_rid) { _update_dirty_multimeshes(); multimesh_allocate(p_rid, 0, RS::MULTIMESH_TRANSFORM_2D); MultiMesh *multimesh = multimesh_owner.getornull(p_rid); - multimesh->instance_dependency.instance_notify_deleted(p_rid); + multimesh->dependency.deleted_notify(p_rid); multimesh_owner.free(p_rid); } else if (skeleton_owner.owns(p_rid)) { _update_dirty_skeletons(); skeleton_allocate(p_rid, 0); Skeleton *skeleton = skeleton_owner.getornull(p_rid); - skeleton->instance_dependency.instance_notify_deleted(p_rid); + skeleton->dependency.deleted_notify(p_rid); skeleton_owner.free(p_rid); } else if (reflection_probe_owner.owns(p_rid)) { ReflectionProbe *reflection_probe = reflection_probe_owner.getornull(p_rid); - reflection_probe->instance_dependency.instance_notify_deleted(p_rid); + reflection_probe->dependency.deleted_notify(p_rid); reflection_probe_owner.free(p_rid); } else if (decal_owner.owns(p_rid)) { Decal *decal = decal_owner.getornull(p_rid); @@ -8155,30 +8238,30 @@ bool RendererStorageRD::free(RID p_rid) { texture_remove_from_decal_atlas(decal->textures[i]); } } - decal->instance_dependency.instance_notify_deleted(p_rid); + decal->dependency.deleted_notify(p_rid); decal_owner.free(p_rid); } else if (gi_probe_owner.owns(p_rid)) { gi_probe_allocate(p_rid, Transform(), AABB(), Vector3i(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<int>()); //deallocate GIProbe *gi_probe = gi_probe_owner.getornull(p_rid); - gi_probe->instance_dependency.instance_notify_deleted(p_rid); + gi_probe->dependency.deleted_notify(p_rid); gi_probe_owner.free(p_rid); } else if (lightmap_owner.owns(p_rid)) { lightmap_set_textures(p_rid, RID(), false); Lightmap *lightmap = lightmap_owner.getornull(p_rid); - lightmap->instance_dependency.instance_notify_deleted(p_rid); + lightmap->dependency.deleted_notify(p_rid); lightmap_owner.free(p_rid); } else if (light_owner.owns(p_rid)) { light_set_projector(p_rid, RID()); //clear projector // delete the texture Light *light = light_owner.getornull(p_rid); - light->instance_dependency.instance_notify_deleted(p_rid); + light->dependency.deleted_notify(p_rid); light_owner.free(p_rid); } else if (particles_owner.owns(p_rid)) { Particles *particles = particles_owner.getornull(p_rid); _particles_free_data(particles); - particles->instance_dependency.instance_notify_deleted(p_rid); + particles->dependency.deleted_notify(p_rid); particles_owner.free(p_rid); } else if (particles_collision_owner.owns(p_rid)) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_rid); @@ -8186,8 +8269,10 @@ bool RendererStorageRD::free(RID p_rid) { if (particles_collision->heightfield_texture.is_valid()) { RD::get_singleton()->free(particles_collision->heightfield_texture); } - particles_collision->instance_dependency.instance_notify_deleted(p_rid); + particles_collision->dependency.deleted_notify(p_rid); particles_collision_owner.free(p_rid); + } else if (particles_collision_instance_owner.owns(p_rid)) { + particles_collision_instance_owner.free(p_rid); } else if (render_target_owner.owns(p_rid)) { RenderTarget *rt = render_target_owner.getornull(p_rid); @@ -8212,11 +8297,11 @@ EffectsRD *RendererStorageRD::get_effects() { } void RendererStorageRD::capture_timestamps_begin() { - RD::get_singleton()->capture_timestamp("Frame Begin", false); + RD::get_singleton()->capture_timestamp("Frame Begin"); } void RendererStorageRD::capture_timestamp(const String &p_name) { - RD::get_singleton()->capture_timestamp(p_name, true); + RD::get_singleton()->capture_timestamp(p_name); } uint32_t RendererStorageRD::get_captured_timestamps_count() const { @@ -8736,7 +8821,7 @@ RendererStorageRD::RendererStorageRD() { actions.renames["RESTART_VELOCITY"] = "restart_velocity"; actions.renames["RESTART_COLOR"] = "restart_color"; actions.renames["RESTART_CUSTOM"] = "restart_custom"; - actions.renames["emit_particle"] = "emit_particle"; + actions.renames["emit_subparticle"] = "emit_subparticle"; actions.renames["COLLIDED"] = "collided"; actions.renames["COLLISION_NORMAL"] = "collision_normal"; actions.renames["COLLISION_DEPTH"] = "collision_depth"; diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.h b/servers/rendering/renderer_rd/renderer_storage_rd.h index 6d1587185e..48d43568c4 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.h +++ b/servers/rendering/renderer_rd/renderer_storage_rd.h @@ -95,6 +95,21 @@ public: p_array[11] = 0; } + static _FORCE_INLINE_ void store_transform_transposed_3x4(const Transform &p_mtx, float *p_array) { + p_array[0] = p_mtx.basis.elements[0][0]; + p_array[1] = p_mtx.basis.elements[0][1]; + p_array[2] = p_mtx.basis.elements[0][2]; + p_array[3] = p_mtx.origin.x; + p_array[4] = p_mtx.basis.elements[1][0]; + p_array[5] = p_mtx.basis.elements[1][1]; + p_array[6] = p_mtx.basis.elements[1][2]; + p_array[7] = p_mtx.origin.y; + p_array[8] = p_mtx.basis.elements[2][0]; + p_array[9] = p_mtx.basis.elements[2][1]; + p_array[10] = p_mtx.basis.elements[2][2]; + p_array[11] = p_mtx.origin.z; + } + static _FORCE_INLINE_ void store_camera(const CameraMatrix &p_mtx, float *p_array) { for (int i = 0; i < 4; i++) { for (int j = 0; j < 4; j++) { @@ -127,6 +142,8 @@ public: virtual bool is_animated() const = 0; virtual bool casts_shadows() const = 0; virtual Variant get_default_parameter(const StringName &p_parameter) const = 0; + virtual RS::ShaderNativeSourceCode get_native_source_code() const { return RS::ShaderNativeSourceCode(); } + virtual ~ShaderData() {} }; @@ -360,6 +377,7 @@ private: Shader *shader; //shortcut to shader data and type ShaderType shader_type; + uint32_t shader_id = 0; bool update_requested; bool uniform_dirty; bool texture_dirty; @@ -367,7 +385,7 @@ private: Map<StringName, Variant> params; int32_t priority; RID next_pass; - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; }; MaterialDataRequestFunction material_data_request_func[SHADER_TYPE_MAX]; @@ -460,7 +478,10 @@ private: List<MeshInstance *> instances; - RendererStorage::InstanceDependency instance_dependency; + RID shadow_mesh; + Set<Mesh *> shadow_owners; + + Dependency dependency; }; mutable RID_Owner<Mesh> mesh_owner; @@ -563,7 +584,7 @@ private: bool dirty = false; MultiMesh *dirty_list = nullptr; - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; }; mutable RID_Owner<MultiMesh> multimesh_owner; @@ -734,7 +755,7 @@ private: ParticleEmissionBuffer *emission_buffer = nullptr; RID emission_storage_buffer; - Set<RendererSceneRender::InstanceBase *> collisions; + Set<RID> collisions; Particles() : inactive(true), @@ -761,7 +782,7 @@ private: clear(true) { } - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; ParticlesFrameParams frame_params; }; @@ -839,6 +860,8 @@ private: virtual bool is_animated() const; virtual bool casts_shadows() const; virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + ParticlesShaderData(); virtual ~ParticlesShaderData(); }; @@ -889,11 +912,19 @@ private: RS::ParticlesCollisionHeightfieldResolution heightfield_resolution = RS::PARTICLES_COLLISION_HEIGHTFIELD_RESOLUTION_1024; - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; }; mutable RID_Owner<ParticlesCollision> particles_collision_owner; + struct ParticlesCollisionInstance { + RID collision; + Transform transform; + bool active = false; + }; + + mutable RID_Owner<ParticlesCollisionInstance> particles_collision_instance_owner; + /* Skeleton */ struct Skeleton { @@ -911,7 +942,7 @@ private: uint64_t version = 1; - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; }; mutable RID_Owner<Skeleton> skeleton_owner; @@ -943,7 +974,7 @@ private: bool directional_sky_only = false; uint64_t version = 0; - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; }; mutable RID_Owner<Light> light_owner; @@ -966,7 +997,7 @@ private: uint32_t cull_mask = (1 << 20) - 1; float lod_threshold = 0.01; - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; }; mutable RID_Owner<ReflectionProbe> reflection_probe_owner; @@ -987,7 +1018,7 @@ private: float distance_fade_length = 1; float normal_fade = 0.0; - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; }; mutable RID_Owner<Decal> decal_owner; @@ -1025,7 +1056,7 @@ private: uint32_t version = 1; uint32_t data_version = 1; - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; }; GiprobeSdfShaderRD giprobe_sdf_shader; @@ -1054,7 +1085,7 @@ private: int32_t over = EMPTY_LEAF, under = EMPTY_LEAF; }; - RendererStorage::InstanceDependency instance_dependency; + Dependency dependency; }; bool using_lightmap_array; //high end uses this @@ -1330,6 +1361,8 @@ public: Variant shader_get_param_default(RID p_shader, const StringName &p_param) const; void shader_set_data_request_function(ShaderType p_shader_type, ShaderDataRequestFunction p_function); + virtual RS::ShaderNativeSourceCode shader_get_native_source_code(RID p_shader) const; + /* COMMON MATERIAL API */ RID material_create(); @@ -1347,11 +1380,16 @@ public: void material_get_instance_shader_parameters(RID p_material, List<InstanceShaderParam> *r_parameters); - void material_update_dependency(RID p_material, InstanceBaseDependency *p_instance); + void material_update_dependency(RID p_material, DependencyTracker *p_instance); void material_force_update_textures(RID p_material, ShaderType p_shader_type); void material_set_data_request_function(ShaderType p_shader_type, MaterialDataRequestFunction p_function); + _FORCE_INLINE_ uint32_t material_get_shader_id(RID p_material) { + Material *material = material_owner.getornull(p_material); + return material->shader_id; + } + _FORCE_INLINE_ MaterialData *material_get_data(RID p_material, ShaderType p_shader_type) { Material *material = material_owner.getornull(p_material); if (!material || material->shader_type != p_shader_type) { @@ -1388,6 +1426,7 @@ public: virtual AABB mesh_get_custom_aabb(RID p_mesh) const; virtual AABB mesh_get_aabb(RID p_mesh, RID p_skeleton = RID()); + virtual void mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh); virtual void mesh_clear(RID p_mesh); @@ -1426,6 +1465,13 @@ public: return mesh->surfaces[p_surface_index]; } + _FORCE_INLINE_ RID mesh_get_shadow_mesh(RID p_mesh) { + Mesh *mesh = mesh_owner.getornull(p_mesh); + ERR_FAIL_COND_V(!mesh, RID()); + + return mesh->shadow_mesh; + } + _FORCE_INLINE_ RS::PrimitiveType mesh_surface_get_primitive(void *p_surface) { Mesh::Surface *surface = reinterpret_cast<Mesh::Surface *>(p_surface); return surface->primitive; @@ -1436,13 +1482,7 @@ public: return s->lod_count > 0; } - _FORCE_INLINE_ RID mesh_surface_get_index_array(void *p_surface) const { - Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); - - return s->index_array; - } - - _FORCE_INLINE_ RID mesh_surface_get_index_array_with_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold) const { + _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold) const { Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); int32_t current_lod = -1; @@ -1454,9 +1494,19 @@ public: current_lod = i; } if (current_lod == -1) { + return 0; + } else { + return current_lod + 1; + } + } + + _FORCE_INLINE_ RID mesh_surface_get_index_array(void *p_surface, uint32_t p_lod) const { + Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + + if (p_lod == 0) { return s->index_array; } else { - return s->lods[current_lod].index_array; + return s->lods[p_lod - 1].index_array; } } @@ -1664,6 +1714,10 @@ public: void skeleton_bone_set_transform_2d(RID p_skeleton, int p_bone, const Transform2D &p_transform); Transform2D skeleton_bone_get_transform_2d(RID p_skeleton, int p_bone) const; + _FORCE_INLINE_ bool skeleton_is_valid(RID p_skeleton) { + return skeleton_owner.getornull(p_skeleton) != nullptr; + } + _FORCE_INLINE_ RID skeleton_get_3d_uniform_set(RID p_skeleton, RID p_shader, uint32_t p_set) const { Skeleton *skeleton = skeleton_owner.getornull(p_skeleton); ERR_FAIL_COND_V(!skeleton, RID()); @@ -1827,8 +1881,8 @@ public: Color reflection_probe_get_ambient_color(RID p_probe) const; float reflection_probe_get_ambient_color_energy(RID p_probe) const; - void base_update_dependency(RID p_base, InstanceBaseDependency *p_instance); - void skeleton_update_dependency(RID p_skeleton, InstanceBaseDependency *p_instance); + void base_update_dependency(RID p_base, DependencyTracker *p_instance); + void skeleton_update_dependency(RID p_skeleton, DependencyTracker *p_instance); /* DECAL API */ @@ -1977,7 +2031,11 @@ public: _FORCE_INLINE_ float lightmap_get_probe_capture_update_speed() const { return lightmap_probe_capture_update_speed; } - + _FORCE_INLINE_ RID lightmap_get_texture(RID p_lightmap) const { + const Lightmap *lm = lightmap_owner.getornull(p_lightmap); + ERR_FAIL_COND_V(!lm, RID()); + return lm->light_texture; + } _FORCE_INLINE_ int32_t lightmap_get_array_index(RID p_lightmap) const { ERR_FAIL_COND_V(!using_lightmap_array, -1); //only for arrays const Lightmap *lm = lightmap_owner.getornull(p_lightmap); @@ -2078,8 +2136,8 @@ public: return particles->particles_transforms_buffer_uniform_set; } - virtual void particles_add_collision(RID p_particles, InstanceBaseDependency *p_instance); - virtual void particles_remove_collision(RID p_particles, InstanceBaseDependency *p_instance); + virtual void particles_add_collision(RID p_particles, RID p_particles_collision_instance); + virtual void particles_remove_collision(RID p_particles, RID p_particles_collision_instance); /* PARTICLES COLLISION */ @@ -2099,6 +2157,11 @@ public: virtual bool particles_collision_is_heightfield(RID p_particles_collision) const; RID particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const; + //used from 2D and 3D + virtual RID particles_collision_instance_create(RID p_collision); + virtual void particles_collision_instance_set_transform(RID p_collision_instance, const Transform &p_transform); + virtual void particles_collision_instance_set_active(RID p_collision_instance, bool p_active); + /* GLOBAL VARIABLES API */ virtual void global_variable_add(const StringName &p_name, RS::GlobalVariableType p_type, const Variant &p_value); diff --git a/servers/rendering/renderer_rd/shader_rd.cpp b/servers/rendering/renderer_rd/shader_rd.cpp index d1f07a354f..e4a39ff813 100644 --- a/servers/rendering/renderer_rd/shader_rd.cpp +++ b/servers/rendering/renderer_rd/shader_rd.cpp @@ -301,6 +301,7 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { builder.append(compute_codev.get_data()); // version info (if exists) builder.append("\n"); //make sure defines begin at newline + builder.append(base_compute_defines.get_data()); builder.append(general_defines.get_data()); builder.append(variant_defines[p_variant].get_data()); @@ -351,6 +352,127 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { } } +RS::ShaderNativeSourceCode ShaderRD::version_get_native_source_code(RID p_version) { + Version *version = version_owner.getornull(p_version); + RS::ShaderNativeSourceCode source_code; + ERR_FAIL_COND_V(!version, source_code); + + source_code.versions.resize(variant_defines.size()); + + for (int i = 0; i < source_code.versions.size(); i++) { + if (!is_compute) { + //vertex stage + + StringBuilder builder; + + builder.append(vertex_codev.get_data()); // version info (if exists) + builder.append("\n"); //make sure defines begin at newline + builder.append(general_defines.get_data()); + builder.append(variant_defines[i].get_data()); + + for (int j = 0; j < version->custom_defines.size(); j++) { + builder.append(version->custom_defines[j].get_data()); + } + + builder.append(vertex_code0.get_data()); //first part of vertex + + builder.append(version->uniforms.get_data()); //uniforms (same for vertex and fragment) + + builder.append(vertex_code1.get_data()); //second part of vertex + + builder.append(version->vertex_globals.get_data()); // vertex globals + + builder.append(vertex_code2.get_data()); //third part of vertex + + builder.append(version->vertex_code.get_data()); // code + + builder.append(vertex_code3.get_data()); //fourth of vertex + + RS::ShaderNativeSourceCode::Version::Stage stage; + stage.name = "vertex"; + stage.code = builder.as_string(); + + source_code.versions.write[i].stages.push_back(stage); + } + + if (!is_compute) { + //fragment stage + + StringBuilder builder; + + builder.append(fragment_codev.get_data()); // version info (if exists) + builder.append("\n"); //make sure defines begin at newline + builder.append(general_defines.get_data()); + builder.append(variant_defines[i].get_data()); + for (int j = 0; j < version->custom_defines.size(); j++) { + builder.append(version->custom_defines[j].get_data()); + } + + builder.append(fragment_code0.get_data()); //first part of fragment + + builder.append(version->uniforms.get_data()); //uniforms (same for fragment and fragment) + + builder.append(fragment_code1.get_data()); //first part of fragment + + builder.append(version->fragment_globals.get_data()); // fragment globals + + builder.append(fragment_code2.get_data()); //third part of fragment + + builder.append(version->fragment_light.get_data()); // fragment light + + builder.append(fragment_code3.get_data()); //fourth part of fragment + + builder.append(version->fragment_code.get_data()); // fragment code + + builder.append(fragment_code4.get_data()); //fourth part of fragment + + RS::ShaderNativeSourceCode::Version::Stage stage; + stage.name = "fragment"; + stage.code = builder.as_string(); + + source_code.versions.write[i].stages.push_back(stage); + } + + if (is_compute) { + //compute stage + + StringBuilder builder; + + builder.append(compute_codev.get_data()); // version info (if exists) + builder.append("\n"); //make sure defines begin at newline + builder.append(base_compute_defines.get_data()); + builder.append(general_defines.get_data()); + builder.append(variant_defines[i].get_data()); + + for (int j = 0; j < version->custom_defines.size(); j++) { + builder.append(version->custom_defines[j].get_data()); + } + + builder.append(compute_code0.get_data()); //first part of compute + + builder.append(version->uniforms.get_data()); //uniforms (same for compute and fragment) + + builder.append(compute_code1.get_data()); //second part of compute + + builder.append(version->compute_globals.get_data()); // compute globals + + builder.append(compute_code2.get_data()); //third part of compute + + builder.append(version->compute_code.get_data()); // code + + builder.append(compute_code3.get_data()); //fourth of compute + + RS::ShaderNativeSourceCode::Version::Stage stage; + stage.name = "compute"; + stage.code = builder.as_string(); + + source_code.versions.write[i].stages.push_back(stage); + } + } + + return source_code; +} + void ShaderRD::_compile_version(Version *p_version) { _clear_version(p_version); @@ -360,7 +482,7 @@ void ShaderRD::_compile_version(Version *p_version) { p_version->variants = memnew_arr(RID, variant_defines.size()); #if 1 - RendererCompositorRD::thread_work_pool.do_work(variant_defines.size(), this, &ShaderRD::_compile_variant, p_version); + RendererThreadPool::singleton->thread_work_pool.do_work(variant_defines.size(), this, &ShaderRD::_compile_variant, p_version); #else for (int i = 0; i < variant_defines.size(); i++) { _compile_variant(i, p_version); @@ -475,6 +597,22 @@ bool ShaderRD::is_variant_enabled(int p_variant) const { return variants_enabled[p_variant]; } +ShaderRD::ShaderRD() { + // Do not feel forced to use this, in most cases it makes little to no difference. + bool use_32_threads = false; + if (RD::get_singleton()->get_device_vendor_name() == "NVIDIA") { + use_32_threads = true; + } + String base_compute_define_text; + if (use_32_threads) { + base_compute_define_text = "\n#define NATIVE_LOCAL_GROUP_SIZE 32\n#define NATIVE_LOCAL_SIZE_2D_X 8\n#define NATIVE_LOCAL_SIZE_2D_Y 4\n"; + } else { + base_compute_define_text = "\n#define NATIVE_LOCAL_GROUP_SIZE 64\n#define NATIVE_LOCAL_SIZE_2D_X 8\n#define NATIVE_LOCAL_SIZE_2D_Y 8\n"; + } + + base_compute_defines = base_compute_define_text.ascii(); +} + void ShaderRD::initialize(const Vector<String> &p_variant_defines, const String &p_general_defines) { ERR_FAIL_COND(variant_defines.size()); ERR_FAIL_COND(p_variant_defines.size() == 0); diff --git a/servers/rendering/renderer_rd/shader_rd.h b/servers/rendering/renderer_rd/shader_rd.h index a80d08050a..e0f4dcf2d0 100644 --- a/servers/rendering/renderer_rd/shader_rd.h +++ b/servers/rendering/renderer_rd/shader_rd.h @@ -36,6 +36,7 @@ #include "core/templates/map.h" #include "core/templates/rid_owner.h" #include "core/variant/variant.h" +#include "servers/rendering_server.h" #include <stdio.h> /** @@ -98,8 +99,10 @@ class ShaderRD { const char *name; + CharString base_compute_defines; + protected: - ShaderRD() {} + ShaderRD(); void setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name); public: @@ -133,6 +136,8 @@ public: void set_variant_enabled(int p_variant, bool p_enabled); bool is_variant_enabled(int p_variant) const; + RS::ShaderNativeSourceCode version_get_native_source_code(RID p_version); + void initialize(const Vector<String> &p_variant_defines, const String &p_general_defines = ""); virtual ~ShaderRD(); }; diff --git a/servers/rendering/renderer_rd/shaders/SCsub b/servers/rendering/renderer_rd/shaders/SCsub index deaa9668df..c192574ff2 100644 --- a/servers/rendering/renderer_rd/shaders/SCsub +++ b/servers/rendering/renderer_rd/shaders/SCsub @@ -39,8 +39,10 @@ if "RD_GLSL" in env["BUILDERS"]: env.RD_GLSL("sdfgi_debug.glsl") env.RD_GLSL("sdfgi_debug_probes.glsl") env.RD_GLSL("volumetric_fog.glsl") - env.RD_GLSL("shadow_reduce.glsl") env.RD_GLSL("particles.glsl") env.RD_GLSL("particles_copy.glsl") env.RD_GLSL("sort.glsl") env.RD_GLSL("skeleton.glsl") + env.RD_GLSL("cluster_render.glsl") + env.RD_GLSL("cluster_store.glsl") + env.RD_GLSL("cluster_debug.glsl") diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl index 2a7cae3b4c..3b39edc70e 100644 --- a/servers/rendering/renderer_rd/shaders/canvas.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -396,7 +396,7 @@ vec4 light_shadow_compute(uint light_base, vec4 light_color, vec4 shadow_uv vec4 shadow_color = unpackUnorm4x8(light_array.data[light_base].shadow_color); #ifdef LIGHT_SHADER_CODE_USED - shadow_color *= shadow_modulate; + shadow_color.rgb *= shadow_modulate; #endif shadow_color.a *= light_color.a; //respect light alpha @@ -497,7 +497,7 @@ void main() { vec2 shadow_vertex = vertex; { - float normal_depth = 1.0; + float normal_map_depth = 1.0; #if defined(NORMAL_MAP_USED) vec3 normal_map = vec3(0.0, 0.0, 1.0); @@ -511,7 +511,7 @@ FRAGMENT_SHADER_CODE /* clang-format on */ #if defined(NORMAL_MAP_USED) - normal = mix(vec3(0.0, 0.0, 1.0), normal_map * vec3(2.0, -2.0, 1.0) - vec3(1.0, -1.0, 0.0), normal_depth); + normal = mix(vec3(0.0, 0.0, 1.0), normal_map * vec3(2.0, -2.0, 1.0) - vec3(1.0, -1.0, 0.0), normal_map_depth); #endif } @@ -546,7 +546,7 @@ FRAGMENT_SHADER_CODE #ifdef LIGHT_SHADER_CODE_USED vec4 shadow_modulate = vec4(1.0); - light_color = light_compute(light_vertex, direction, normal, light_color, light_color.a, specular_shininess, shadow_modulate, screen_uv, color, uv, true); + light_color = light_compute(light_vertex, vec3(direction, light_array.data[light_base].height), normal, light_color, light_color.a, specular_shininess, shadow_modulate, screen_uv, uv, color, true); #else if (normal_used) { @@ -563,7 +563,7 @@ FRAGMENT_SHADER_CODE light_color = light_shadow_compute(light_base, light_color, shadow_uv #ifdef LIGHT_SHADER_CODE_USED , - shadow_modulate + shadow_modulate.rgb #endif ); } @@ -605,7 +605,7 @@ FRAGMENT_SHADER_CODE vec3 light_position = vec3(light_array.data[light_base].position, light_array.data[light_base].height); light_color.rgb *= light_base_color.rgb; - light_color = light_compute(light_vertex, light_position, normal, light_color, light_base_color.a, specular_shininess, shadow_modulate, screen_uv, color, uv, false); + light_color = light_compute(light_vertex, light_position, normal, light_color, light_base_color.a, specular_shininess, shadow_modulate, screen_uv, uv, color, false); #else light_color.rgb *= light_base_color.rgb * light_base_color.a; @@ -659,7 +659,7 @@ FRAGMENT_SHADER_CODE light_color = light_shadow_compute(light_base, light_color, shadow_uv #ifdef LIGHT_SHADER_CODE_USED , - shadow_modulate + shadow_modulate.rgb #endif ); } diff --git a/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl index e723468dd8..3a4bf4da07 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl @@ -6,12 +6,18 @@ struct LightData { //this structure needs to be as packed as possible vec3 position; float inv_radius; + vec3 direction; float size; - uint attenuation_energy; //attenuation - uint color_specular; //rgb color, a specular (8 bit unorm) - uint cone_attenuation_angle; // attenuation and angle, (16bit float) - uint shadow_color_enabled; //shadow rgb color, a>0.5 enabled (8bit unorm) + + vec3 color; + float attenuation; + + float cone_attenuation; + float cone_angle; + float specular_amount; + bool shadow_enabled; + vec4 atlas_rect; // rect in the shadow atlas mat4 shadow_matrix; float shadow_bias; @@ -34,9 +40,13 @@ struct ReflectionData { float index; vec3 box_offset; uint mask; - vec4 params; // intensity, 0, interior , boxproject vec3 ambient; // ambient color + float intensity; + bool exterior; + bool box_project; uint ambient_mode; + uint pad; + //0-8 is intensity,8-9 is ambient, mode mat4 local_matrix; // up to here for spot and omni, rest is for directional // notes: for ambientblend, use distance to edge to blend between already existing global environment }; diff --git a/servers/rendering/renderer_rd/shaders/cluster_debug.glsl b/servers/rendering/renderer_rd/shaders/cluster_debug.glsl new file mode 100644 index 0000000000..70a875192c --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cluster_debug.glsl @@ -0,0 +1,115 @@ +#[compute] + +#version 450 + +VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +const vec3 usage_gradient[33] = vec3[]( // 1 (none) + 32 + vec3(0.14, 0.17, 0.23), + vec3(0.24, 0.44, 0.83), + vec3(0.23, 0.57, 0.84), + vec3(0.22, 0.71, 0.84), + vec3(0.22, 0.85, 0.83), + vec3(0.21, 0.85, 0.72), + vec3(0.21, 0.85, 0.57), + vec3(0.20, 0.85, 0.42), + vec3(0.20, 0.85, 0.27), + vec3(0.27, 0.86, 0.19), + vec3(0.51, 0.85, 0.19), + vec3(0.57, 0.86, 0.19), + vec3(0.62, 0.85, 0.19), + vec3(0.67, 0.86, 0.20), + vec3(0.73, 0.85, 0.20), + vec3(0.78, 0.85, 0.20), + vec3(0.83, 0.85, 0.20), + vec3(0.85, 0.82, 0.20), + vec3(0.85, 0.76, 0.20), + vec3(0.85, 0.81, 0.20), + vec3(0.85, 0.65, 0.20), + vec3(0.84, 0.60, 0.21), + vec3(0.84, 0.56, 0.21), + vec3(0.84, 0.51, 0.21), + vec3(0.84, 0.46, 0.21), + vec3(0.84, 0.41, 0.21), + vec3(0.84, 0.36, 0.21), + vec3(0.84, 0.31, 0.21), + vec3(0.84, 0.27, 0.21), + vec3(0.83, 0.22, 0.22), + vec3(0.83, 0.22, 0.27), + vec3(0.83, 0.22, 0.32), + vec3(1.00, 0.63, 0.70)); +layout(push_constant, binding = 0, std430) uniform Params { + uvec2 screen_size; + uvec2 cluster_screen_size; + + uint cluster_shift; + uint cluster_type; + float z_near; + float z_far; + + bool orthogonal; + uint max_cluster_element_count_div_32; + uint pad1; + uint pad2; +} +params; + +layout(set = 0, binding = 1, std430) buffer restrict readonly ClusterData { + uint data[]; +} +cluster_data; + +layout(rgba16f, set = 0, binding = 2) uniform restrict writeonly image2D screen_buffer; +layout(set = 0, binding = 3) uniform texture2D depth_buffer; +layout(set = 0, binding = 4) uniform sampler depth_buffer_sampler; + +void main() { + uvec2 screen_pos = gl_GlobalInvocationID.xy; + if (any(greaterThanEqual(screen_pos, params.screen_size))) { + return; + } + + uvec2 cluster_pos = screen_pos >> params.cluster_shift; + + uint offset = cluster_pos.y * params.cluster_screen_size.x + cluster_pos.x; + offset += params.cluster_screen_size.x * params.cluster_screen_size.y * params.cluster_type; + offset *= (params.max_cluster_element_count_div_32 + 32); + + //depth buffers generally can't be accessed via image API + float depth = texelFetch(sampler2D(depth_buffer, depth_buffer_sampler), ivec2(screen_pos), 0).r * 2.0 - 1.0; + + if (params.orthogonal) { + depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + } + depth /= params.z_far; + + uint slice = uint(clamp(floor(depth * 32.0), 0.0, 31.0)); + uint slice_minmax = cluster_data.data[offset + params.max_cluster_element_count_div_32 + slice]; + uint item_min = slice_minmax & 0xFFFF; + uint item_max = slice_minmax >> 16; + + uint item_count = 0; + for (uint i = 0; i < params.max_cluster_element_count_div_32; i++) { + uint slice_bits = cluster_data.data[offset + i]; + while (slice_bits != 0) { + uint bit = findLSB(slice_bits); + uint item = i * 32 + bit; + if ((item >= item_min && item < item_max)) { + item_count++; + } + slice_bits &= ~(1 << bit); + } + } + + item_count = min(item_count, 32); + + vec3 color = usage_gradient[item_count]; + + color = mix(color * 1.2, color * 0.3, float(slice) / 31.0); + + imageStore(screen_buffer, ivec2(screen_pos), vec4(color, 1.0)); +} diff --git a/servers/rendering/renderer_rd/shaders/cluster_render.glsl b/servers/rendering/renderer_rd/shaders/cluster_render.glsl new file mode 100644 index 0000000000..8723ea78e4 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cluster_render.glsl @@ -0,0 +1,168 @@ +#[vertex] + +#version 450 + +VERSION_DEFINES + +layout(location = 0) in vec3 vertex_attrib; + +layout(location = 0) out float depth_interp; +layout(location = 1) out flat uint element_index; + +layout(push_constant, binding = 0, std430) uniform Params { + uint base_index; + uint pad0; + uint pad1; + uint pad2; +} +params; + +layout(set = 0, binding = 1, std140) uniform State { + mat4 projection; + + float inv_z_far; + uint screen_to_clusters_shift; // shift to obtain coordinates in block indices + uint cluster_screen_width; // + uint cluster_data_size; // how much data for a single cluster takes + + uint cluster_depth_offset; + uint pad0; + uint pad1; + uint pad2; +} +state; + +struct RenderElement { + uint type; //0-4 + bool touches_near; + bool touches_far; + uint original_index; + mat3x4 transform_inv; + vec3 scale; + uint pad; +}; + +layout(set = 0, binding = 2, std430) buffer restrict readonly RenderElements { + RenderElement data[]; +} +render_elements; + +void main() { + element_index = params.base_index + gl_InstanceIndex; + + vec3 vertex = vertex_attrib; + vertex *= render_elements.data[element_index].scale; + + vertex = vec4(vertex, 1.0) * render_elements.data[element_index].transform_inv; + depth_interp = -vertex.z; + + gl_Position = state.projection * vec4(vertex, 1.0); +} + +#[fragment] + +#version 450 + +VERSION_DEFINES + +#if defined(GL_KHR_shader_subgroup_ballot) && defined(GL_KHR_shader_subgroup_arithmetic) && defined(GL_KHR_shader_subgroup_vote) + +#extension GL_KHR_shader_subgroup_ballot : enable +#extension GL_KHR_shader_subgroup_arithmetic : enable +#extension GL_KHR_shader_subgroup_vote : enable + +#define USE_SUBGROUPS +#endif + +layout(location = 0) in float depth_interp; +layout(location = 1) in flat uint element_index; + +layout(set = 0, binding = 1, std140) uniform State { + mat4 projection; + float inv_z_far; + uint screen_to_clusters_shift; // shift to obtain coordinates in block indices + uint cluster_screen_width; // + uint cluster_data_size; // how much data for a single cluster takes + uint cluster_depth_offset; + uint pad0; + uint pad1; + uint pad2; +} +state; + +//cluster data is layout linearly, each cell contains the follow information: +// - list of bits for every element to mark as used, so (max_elem_count/32)*4 uints +// - a uint for each element to mark the depth bits used when rendering (0-31) + +layout(set = 0, binding = 3, std430) buffer restrict ClusterRender { + uint data[]; +} +cluster_render; + +void main() { + //convert from screen to cluster + uvec2 cluster = uvec2(gl_FragCoord.xy) >> state.screen_to_clusters_shift; + + //get linear cluster offset from screen poss + uint cluster_offset = cluster.x + state.cluster_screen_width * cluster.y; + //multiply by data size to position at the beginning of the element list for this cluster + cluster_offset *= state.cluster_data_size; + + //find the current element in the list and plot the bit to mark it as used + uint usage_write_offset = cluster_offset + (element_index >> 5); + uint usage_write_bit = 1 << (element_index & 0x1F); + +#ifdef USE_SUBGROUPS + + uint cluster_thread_group_index; + + if (!gl_HelperInvocation) { + //http://advances.realtimerendering.com/s2017/2017_Sig_Improved_Culling_final.pdf + + uvec4 mask; + + while (true) { + // find the cluster offset of the first active thread + // threads that did break; go inactive and no longer count + uint first = subgroupBroadcastFirst(cluster_offset); + // update the mask for thread that match this cluster + mask = subgroupBallot(first == cluster_offset); + if (first == cluster_offset) { + // This thread belongs to the group of threads that match this offset, + // so exit the loop. + break; + } + } + + cluster_thread_group_index = subgroupBallotExclusiveBitCount(mask); + + if (cluster_thread_group_index == 0) { + atomicOr(cluster_render.data[usage_write_offset], usage_write_bit); + } + } +#else + if (!gl_HelperInvocation) { + atomicOr(cluster_render.data[usage_write_offset], usage_write_bit); + } +#endif + //find the current element in the depth usage list and mark the current depth as used + float unit_depth = depth_interp * state.inv_z_far; + + uint z_bit = clamp(uint(floor(unit_depth * 32.0)), 0, 31); + + uint z_write_offset = cluster_offset + state.cluster_depth_offset + element_index; + uint z_write_bit = 1 << z_bit; + +#ifdef USE_SUBGROUPS + if (!gl_HelperInvocation) { + z_write_bit = subgroupOr(z_write_bit); //merge all Zs + if (cluster_thread_group_index == 0) { + atomicOr(cluster_render.data[z_write_offset], z_write_bit); + } + } +#else + if (!gl_HelperInvocation) { + atomicOr(cluster_render.data[z_write_offset], z_write_bit); + } +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/cluster_store.glsl b/servers/rendering/renderer_rd/shaders/cluster_store.glsl new file mode 100644 index 0000000000..5be0893c4f --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cluster_store.glsl @@ -0,0 +1,119 @@ +#[compute] + +#version 450 + +VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(push_constant, binding = 0, std430) uniform Params { + uint cluster_render_data_size; // how much data for a single cluster takes + uint max_render_element_count_div_32; //divided by 32 + uvec2 cluster_screen_size; + uint render_element_count_div_32; //divided by 32 + + uint max_cluster_element_count_div_32; //divided by 32 + uint pad1; + uint pad2; +} +params; + +layout(set = 0, binding = 1, std430) buffer restrict readonly ClusterRender { + uint data[]; +} +cluster_render; + +layout(set = 0, binding = 2, std430) buffer restrict ClusterStore { + uint data[]; +} +cluster_store; + +struct RenderElement { + uint type; //0-4 + bool touches_near; + bool touches_far; + uint original_index; + mat3x4 transform_inv; + vec3 scale; + uint pad; +}; + +layout(set = 0, binding = 3, std430) buffer restrict readonly RenderElements { + RenderElement data[]; +} +render_elements; + +void main() { + uvec2 pos = gl_GlobalInvocationID.xy; + if (any(greaterThanEqual(pos, params.cluster_screen_size))) { + return; + } + + //counter for each type of render_element + + //base offset for this cluster + uint base_offset = (pos.x + params.cluster_screen_size.x * pos.y); + uint src_offset = base_offset * params.cluster_render_data_size; + + uint render_element_offset = 0; + + //check all render_elements and see which one was written to + while (render_element_offset < params.render_element_count_div_32) { + uint bits = cluster_render.data[src_offset + render_element_offset]; + while (bits != 0) { + //if bits exist, check the render_element + uint index_bit = findLSB(bits); + uint index = render_element_offset * 32 + index_bit; + uint type = render_elements.data[index].type; + + uint z_range_offset = src_offset + params.max_render_element_count_div_32 + index; + uint z_range = cluster_render.data[z_range_offset]; + + //if object was written, z was written, but check just in case + if (z_range != 0) { //should always be > 0 + + uint from_z = findLSB(z_range); + uint to_z = findMSB(z_range) + 1; + + if (render_elements.data[index].touches_near) { + from_z = 0; + } + + if (render_elements.data[index].touches_far) { + to_z = 32; + } + + // find cluster offset in the buffer used for indexing in the renderer + uint dst_offset = (base_offset + type * (params.cluster_screen_size.x * params.cluster_screen_size.y)) * (params.max_cluster_element_count_div_32 + 32); + + uint orig_index = render_elements.data[index].original_index; + //store this index in the Z slices by setting the relevant bit + for (uint i = from_z; i < to_z; i++) { + uint slice_ofs = dst_offset + params.max_cluster_element_count_div_32 + i; + + uint minmax = cluster_store.data[slice_ofs]; + + if (minmax == 0) { + minmax = 0xFFFF; //min 0, max 0xFFFF + } + + uint elem_min = min(orig_index, minmax & 0xFFFF); + uint elem_max = max(orig_index + 1, minmax >> 16); //always store plus one, so zero means range is empty when not written to + + minmax = elem_min | (elem_max << 16); + cluster_store.data[slice_ofs] = minmax; + } + + uint store_word = orig_index >> 5; + uint store_bit = orig_index & 0x1F; + + //store the actual render_element index at the end, so the rendering code can reference it + cluster_store.data[dst_offset + store_word] |= 1 << store_bit; + } + + bits &= ~(1 << index_bit); //clear the bit to continue iterating + } + + render_element_offset++; + } +} diff --git a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl index 54d67db6c6..c3ac0bee57 100644 --- a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl +++ b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl @@ -1,33 +1,48 @@ -#[compute] +#[vertex] #version 450 VERSION_DEFINES -layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +layout(push_constant, binding = 1, std430) uniform Params { + float z_far; + float z_near; + bool z_flip; + uint pad; + vec4 screen_rect; +} +params; + +layout(location = 0) out vec2 uv_interp; + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + vec2 screen_pos = uv_interp * params.screen_rect.zw + params.screen_rect.xy; + gl_Position = vec4(screen_pos * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +VERSION_DEFINES + +layout(location = 0) in vec2 uv_interp; layout(set = 0, binding = 0) uniform samplerCube source_cube; layout(push_constant, binding = 1, std430) uniform Params { - ivec2 screen_size; - ivec2 offset; - float bias; float z_far; float z_near; bool z_flip; + uint pad; + vec4 screen_rect; } params; -layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D depth_buffer; - void main() { - ivec2 pos = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThan(pos, params.screen_size))) { //too large, do nothing - return; - } - - vec2 pixel_size = 1.0 / vec2(params.screen_size); - vec2 uv = (vec2(pos) + 0.5) * pixel_size; + vec2 uv = uv_interp; vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); @@ -65,5 +80,5 @@ void main() { float linear_depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); depth = (linear_depth * depth_fix) / params.z_far; - imageStore(depth_buffer, pos + params.offset, vec4(depth)); + gl_FragDepth = depth; } diff --git a/servers/rendering/renderer_rd/shaders/gi.glsl b/servers/rendering/renderer_rd/shaders/gi.glsl index 8011dadc72..92a5682572 100644 --- a/servers/rendering/renderer_rd/shaders/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/gi.glsl @@ -97,13 +97,12 @@ layout(push_constant, binding = 0, std430) uniform Params { vec4 proj_info; + vec3 ao_color; uint max_giprobes; + bool high_quality_vct; - bool use_sdfgi; bool orthogonal; - - vec3 ao_color; - uint pad; + uint pad[2]; mat3x4 cam_rotation; } @@ -331,7 +330,7 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o } ambient_light.rgb = diffuse; -#if 1 + if (roughness < 0.2) { vec3 pos_to_uvw = 1.0 / sdfgi.grid_size; vec4 light_accum = vec4(0.0); @@ -363,59 +362,63 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o //ray_pos += ray_dir * (bias / sdfgi.cascades[cascade].to_cell); //bias to avoid self occlusion ray_pos += (ray_dir * 1.0 / max(abs_ray_dir.x, max(abs_ray_dir.y, abs_ray_dir.z)) + cam_normal * 1.4) * bias / sdfgi.cascades[cascade].to_cell; } - float softness = 0.2 + min(1.0, roughness * 5.0) * 4.0; //approximation to roughness so it does not seem like a hard fade - while (length(ray_pos) < max_distance) { - for (uint i = 0; i < sdfgi.max_cascades; i++) { - if (i >= cascade && length(ray_pos) < radius_sizes[i]) { - cascade = max(i, cascade); //never go down - - vec3 pos = ray_pos - sdfgi.cascades[i].position; - pos *= sdfgi.cascades[i].to_cell * pos_to_uvw; - - float distance = texture(sampler3D(sdf_cascades[i], linear_sampler), pos).r * 255.0 - 1.1; - - vec4 hit_light = vec4(0.0); - if (distance < softness) { - hit_light.rgb = texture(sampler3D(light_cascades[i], linear_sampler), pos).rgb; - hit_light.rgb *= 0.5; //approximation given value read is actually meant for anisotropy - hit_light.a = clamp(1.0 - (distance / softness), 0.0, 1.0); - hit_light.rgb *= hit_light.a; - } + uint i = 0; + bool found = false; + while (true) { + if (length(ray_pos) >= max_distance || light_accum.a > 0.99) { + break; + } + if (!found && i >= cascade && length(ray_pos) < radius_sizes[i]) { + uint next_i = min(i + 1, sdfgi.max_cascades - 1); + cascade = max(i, cascade); //never go down - distance /= sdfgi.cascades[i].to_cell; + vec3 pos = ray_pos - sdfgi.cascades[i].position; + pos *= sdfgi.cascades[i].to_cell * pos_to_uvw; - if (i < (sdfgi.max_cascades - 1)) { - pos = ray_pos - sdfgi.cascades[i + 1].position; - pos *= sdfgi.cascades[i + 1].to_cell * pos_to_uvw; + float fdistance = textureLod(sampler3D(sdf_cascades[i], linear_sampler), pos, 0.0).r * 255.0 - 1.1; - float distance2 = texture(sampler3D(sdf_cascades[i + 1], linear_sampler), pos).r * 255.0 - 1.1; + vec4 hit_light = vec4(0.0); + if (fdistance < softness) { + hit_light.rgb = textureLod(sampler3D(light_cascades[i], linear_sampler), pos, 0.0).rgb; + hit_light.rgb *= 0.5; //approximation given value read is actually meant for anisotropy + hit_light.a = clamp(1.0 - (fdistance / softness), 0.0, 1.0); + hit_light.rgb *= hit_light.a; + } - vec4 hit_light2 = vec4(0.0); - if (distance2 < softness) { - hit_light2.rgb = texture(sampler3D(light_cascades[i + 1], linear_sampler), pos).rgb; - hit_light2.rgb *= 0.5; //approximation given value read is actually meant for anisotropy - hit_light2.a = clamp(1.0 - (distance2 / softness), 0.0, 1.0); - hit_light2.rgb *= hit_light2.a; - } + fdistance /= sdfgi.cascades[i].to_cell; - float prev_radius = i == 0 ? 0.0 : radius_sizes[i - 1]; - float blend = clamp((length(ray_pos) - prev_radius) / (radius_sizes[i] - prev_radius), 0.0, 1.0); + if (i < (sdfgi.max_cascades - 1)) { + pos = ray_pos - sdfgi.cascades[next_i].position; + pos *= sdfgi.cascades[next_i].to_cell * pos_to_uvw; - distance2 /= sdfgi.cascades[i + 1].to_cell; + float fdistance2 = textureLod(sampler3D(sdf_cascades[next_i], linear_sampler), pos, 0.0).r * 255.0 - 1.1; - hit_light = mix(hit_light, hit_light2, blend); - distance = mix(distance, distance2, blend); + vec4 hit_light2 = vec4(0.0); + if (fdistance2 < softness) { + hit_light2.rgb = textureLod(sampler3D(light_cascades[next_i], linear_sampler), pos, 0.0).rgb; + hit_light2.rgb *= 0.5; //approximation given value read is actually meant for anisotropy + hit_light2.a = clamp(1.0 - (fdistance2 / softness), 0.0, 1.0); + hit_light2.rgb *= hit_light2.a; } - light_accum += hit_light; - ray_pos += ray_dir * distance; - break; + float prev_radius = i == 0 ? 0.0 : radius_sizes[max(0, i - 1)]; + float blend = clamp((length(ray_pos) - prev_radius) / (radius_sizes[i] - prev_radius), 0.0, 1.0); + + fdistance2 /= sdfgi.cascades[next_i].to_cell; + + hit_light = mix(hit_light, hit_light2, blend); + fdistance = mix(fdistance, fdistance2, blend); } - } - if (light_accum.a > 0.99) { - break; + light_accum += hit_light; + ray_pos += ray_dir * fdistance; + found = true; + } + i++; + if (i == sdfgi.max_cascades) { + i = 0; + found = false; } } @@ -434,8 +437,6 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o } } -#endif - reflection_light.rgb = specular; ambient_light.rgb *= sdfgi.energy; @@ -597,35 +598,24 @@ vec4 fetch_normal_and_roughness(ivec2 pos) { return normal_roughness; } -void main() { - // Pixel being shaded - ivec2 pos = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing - return; - } - - vec3 vertex = reconstruct_position(pos); - vertex.y = -vertex.y; - +void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 reflection_light) { vec4 normal_roughness = fetch_normal_and_roughness(pos); - vec3 normal = normal_roughness.xyz; - vec4 ambient_light = vec4(0.0), reflection_light = vec4(0.0); + vec3 normal = normal_roughness.xyz; if (normal.length() > 0.5) { //valid normal, can do GI float roughness = normal_roughness.w; - vertex = mat3(params.cam_rotation) * vertex; normal = normalize(mat3(params.cam_rotation) * normal); - vec3 reflection = normalize(reflect(normalize(vertex), normal)); - if (params.use_sdfgi) { - sdfgi_process(vertex, normal, reflection, roughness, ambient_light, reflection_light); - } +#ifdef USE_SDFGI + sdfgi_process(vertex, normal, reflection, roughness, ambient_light, reflection_light); +#endif - if (params.max_giprobes > 0) { +#ifdef USE_GIPROBES + { uvec2 giprobe_tex = texelFetch(usampler2D(giprobe_buffer, linear_sampler), pos, 0).rg; roughness *= roughness; //find arbitrary tangent and bitangent, then build a matrix @@ -648,16 +638,40 @@ void main() { spec_accum /= blend_accum; } - if (params.use_sdfgi) { - reflection_light = blend_color(spec_accum, reflection_light); - ambient_light = blend_color(amb_accum, ambient_light); - } else { - reflection_light = spec_accum; - ambient_light = amb_accum; - } +#ifdef USE_SDFGI + reflection_light = blend_color(spec_accum, reflection_light); + ambient_light = blend_color(amb_accum, ambient_light); +#else + reflection_light = spec_accum; + ambient_light = amb_accum; +#endif } +#endif + } +} + +void main() { + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + +#ifdef MODE_HALF_RES + pos <<= 1; +#endif + if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing + return; } + vec4 ambient_light = vec4(0.0); + vec4 reflection_light = vec4(0.0); + + vec3 vertex = reconstruct_position(pos); + vertex.y = -vertex.y; + + process_gi(pos, vertex, ambient_light, reflection_light); + +#ifdef MODE_HALF_RES + pos >>= 1; +#endif + imageStore(ambient_buffer, pos, ambient_light); imageStore(reflection_buffer, pos, reflection_light); } diff --git a/servers/rendering/renderer_rd/shaders/giprobe.glsl b/servers/rendering/renderer_rd/shaders/giprobe.glsl index ea4237a45e..b931461b31 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe.glsl @@ -51,10 +51,10 @@ struct Light { float attenuation; vec3 color; - float spot_angle_radians; + float cos_spot_angle; vec3 position; - float spot_attenuation; + float inv_spot_attenuation; vec3 direction; bool has_shadow; @@ -208,6 +208,15 @@ float raymarch(float distance, float distance_adv, vec3 from, vec3 direction) { return occlusion; //max(0.0,distance); } +float get_omni_attenuation(float distance, float inv_range, float decay) { + float nd = distance * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(distance, 0.0001), -decay); +} + bool compute_light_vector(uint light, vec3 pos, out float attenuation, out vec3 light_pos) { if (lights.data[light].type == LIGHT_TYPE_DIRECTIONAL) { light_pos = pos - lights.data[light].direction * length(vec3(params.limits)); @@ -220,17 +229,19 @@ bool compute_light_vector(uint light, vec3 pos, out float attenuation, out vec3 return false; } - attenuation = pow(clamp(1.0 - distance / lights.data[light].radius, 0.0001, 1.0), lights.data[light].attenuation); + attenuation = get_omni_attenuation(distance, 1.0 / lights.data[light].radius, lights.data[light].attenuation); if (lights.data[light].type == LIGHT_TYPE_SPOT) { vec3 rel = normalize(pos - light_pos); - float angle = acos(dot(rel, lights.data[light].direction)); - if (angle > lights.data[light].spot_angle_radians) { + float cos_spot_angle = lights.data[light].cos_spot_angle; + float cos_angle = dot(rel, lights.data[light].direction); + if (cos_angle < cos_spot_angle) { return false; } - float d = clamp(angle / lights.data[light].spot_angle_radians, 0, 1); - attenuation *= pow(1.0 - d, lights.data[light].spot_attenuation); + float scos = max(cos_angle, cos_spot_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cos_spot_angle)); + attenuation *= 1.0 - pow(spot_rim, lights.data[light].inv_spot_attenuation); } } diff --git a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl index 9c794f1bcc..56b3b7ccb4 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl @@ -43,10 +43,10 @@ struct Light { float attenuation; vec3 color; - float spot_angle_radians; + float cos_spot_angle; vec3 position; - float spot_attenuation; + float inv_spot_attenuation; vec3 direction; bool has_shadow; @@ -146,13 +146,15 @@ bool compute_light_vector(uint light, uint cell, vec3 pos, out float attenuation if (lights.data[light].type == LIGHT_TYPE_SPOT) { vec3 rel = normalize(pos - light_pos); - float angle = acos(dot(rel, lights.data[light].direction)); - if (angle > lights.data[light].spot_angle_radians) { + float cos_spot_angle = lights.data[light].cos_spot_angle; + float cos_angle = dot(rel, lights.data[light].direction); + if (cos_angle < cos_spot_angle) { return false; } - float d = clamp(angle / lights.data[light].spot_angle_radians, 0, 1); - attenuation *= pow(1.0 - d, lights.data[light].spot_attenuation); + float scos = max(cos_angle, cos_spot_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cos_spot_angle)); + attenuation *= 1.0 - pow(spot_rim, lights.data[light].inv_spot_attenuation); } } diff --git a/servers/rendering/renderer_rd/shaders/particles.glsl b/servers/rendering/renderer_rd/shaders/particles.glsl index 926c7ef9fc..cb6d8dc7f6 100644 --- a/servers/rendering/renderer_rd/shaders/particles.glsl +++ b/servers/rendering/renderer_rd/shaders/particles.glsl @@ -173,7 +173,7 @@ uint hash(uint x) { return x; } -bool emit_particle(mat4 p_xform, vec3 p_velocity, vec4 p_color, vec4 p_custom, uint p_flags) { +bool emit_subparticle(mat4 p_xform, vec3 p_velocity, vec4 p_color, vec4 p_custom, uint p_flags) { if (!params.can_emit) { return false; } diff --git a/servers/rendering/renderer_rd/shaders/resolve.glsl b/servers/rendering/renderer_rd/shaders/resolve.glsl index 9429a66dc9..e83c4ca93b 100644 --- a/servers/rendering/renderer_rd/shaders/resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/resolve.glsl @@ -58,6 +58,116 @@ void main() { #else +#if 1 + + vec4 group1; + vec4 group2; + vec4 group3; + vec4 group4; + int best_index = 0; + + //2X + group1.x = texelFetch(source_depth, pos, 0).r; + group1.y = texelFetch(source_depth, pos, 1).r; + + //4X + if (params.sample_count >= 4) { + group1.z = texelFetch(source_depth, pos, 2).r; + group1.w = texelFetch(source_depth, pos, 3).r; + } + //8X + if (params.sample_count >= 8) { + group2.x = texelFetch(source_depth, pos, 4).r; + group2.y = texelFetch(source_depth, pos, 5).r; + group2.z = texelFetch(source_depth, pos, 6).r; + group2.w = texelFetch(source_depth, pos, 7).r; + } + //16X + if (params.sample_count >= 16) { + group3.x = texelFetch(source_depth, pos, 8).r; + group3.y = texelFetch(source_depth, pos, 9).r; + group3.z = texelFetch(source_depth, pos, 10).r; + group3.w = texelFetch(source_depth, pos, 11).r; + + group4.x = texelFetch(source_depth, pos, 12).r; + group4.y = texelFetch(source_depth, pos, 13).r; + group4.z = texelFetch(source_depth, pos, 14).r; + group4.w = texelFetch(source_depth, pos, 15).r; + } + + if (params.sample_count == 2) { + best_index = (pos.x & 1) ^ ((pos.y >> 1) & 1); //not much can be done here + } else if (params.sample_count == 4) { + vec4 freq = vec4(equal(group1, vec4(group1.x))); + freq += vec4(equal(group1, vec4(group1.y))); + freq += vec4(equal(group1, vec4(group1.z))); + freq += vec4(equal(group1, vec4(group1.w))); + + float min_f = freq.x; + best_index = 0; + if (freq.y < min_f) { + best_index = 1; + min_f = freq.y; + } + if (freq.z < min_f) { + best_index = 2; + min_f = freq.z; + } + if (freq.w < min_f) { + best_index = 3; + } + } else if (params.sample_count == 8) { + vec4 freq0 = vec4(equal(group1, vec4(group1.x))); + vec4 freq1 = vec4(equal(group2, vec4(group1.x))); + freq0 += vec4(equal(group1, vec4(group1.y))); + freq1 += vec4(equal(group2, vec4(group1.y))); + freq0 += vec4(equal(group1, vec4(group1.z))); + freq1 += vec4(equal(group2, vec4(group1.z))); + freq0 += vec4(equal(group1, vec4(group1.w))); + freq1 += vec4(equal(group2, vec4(group1.w))); + freq0 += vec4(equal(group1, vec4(group2.x))); + freq1 += vec4(equal(group2, vec4(group2.x))); + freq0 += vec4(equal(group1, vec4(group2.y))); + freq1 += vec4(equal(group2, vec4(group2.y))); + freq0 += vec4(equal(group1, vec4(group2.z))); + freq1 += vec4(equal(group2, vec4(group2.z))); + freq0 += vec4(equal(group1, vec4(group2.w))); + freq1 += vec4(equal(group2, vec4(group2.w))); + + float min_f0 = freq0.x; + int best_index0 = 0; + if (freq0.y < min_f0) { + best_index0 = 1; + min_f0 = freq0.y; + } + if (freq0.z < min_f0) { + best_index0 = 2; + min_f0 = freq0.z; + } + if (freq0.w < min_f0) { + best_index0 = 3; + min_f0 = freq0.w; + } + + float min_f1 = freq1.x; + int best_index1 = 4; + if (freq1.y < min_f1) { + best_index1 = 5; + min_f1 = freq1.y; + } + if (freq1.z < min_f1) { + best_index1 = 6; + min_f1 = freq1.z; + } + if (freq1.w < min_f1) { + best_index1 = 7; + min_f1 = freq1.w; + } + + best_index = mix(best_index0, best_index1, min_f0 < min_f1); + } + +#else float depths[16]; int depth_indices[16]; int depth_amount[16]; @@ -91,7 +201,7 @@ void main() { depth_least = depth_amount[j]; } } - +#endif best_depth = texelFetch(source_depth, pos, best_index).r; best_normal_roughness = texelFetch(source_normal_roughness, pos, best_index); #ifdef GIPROBE_RESOLVE diff --git a/servers/rendering/renderer_rd/shaders/scene_forward.glsl b/servers/rendering/renderer_rd/shaders/scene_forward.glsl index 05f7637478..adccf1e712 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward.glsl @@ -89,33 +89,45 @@ MATERIAL_UNIFORMS } material; #endif -/* clang-format off */ - -VERTEX_SHADER_GLOBALS - -/* clang-format on */ - invariant gl_Position; -layout(location = 7) flat out uint instance_index; - #ifdef MODE_DUAL_PARABOLOID layout(location = 8) out float dp_clip; #endif +layout(location = 9) out flat uint instance_index; + +/* clang-format off */ + +VERTEX_SHADER_GLOBALS + +/* clang-format on */ + void main() { - instance_index = draw_call.instance_index; vec4 instance_custom = vec4(0.0); #if defined(COLOR_USED) color_interp = color_attrib; #endif + instance_index = draw_call.instance_index; + + bool is_multimesh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH); + if (!is_multimesh) { + instance_index += gl_InstanceIndex; + } + mat4 world_matrix = instances.data[instance_index].transform; - mat3 world_normal_matrix = mat3(instances.data[instance_index].normal_transform); - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH)) { + mat3 world_normal_matrix; + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { + world_normal_matrix = inverse(mat3(world_matrix)); + } else { + world_normal_matrix = mat3(world_matrix); + } + + if (is_multimesh) { //multimesh, instances are for it uint offset = (instances.data[instance_index].flags >> INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT) & INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK; offset *= gl_InstanceIndex; @@ -144,10 +156,6 @@ void main() { matrix = transpose(matrix); world_matrix = world_matrix * matrix; world_normal_matrix = world_normal_matrix * mat3(matrix); - - } else { - //not a multimesh, instances are for multiple draw calls - instance_index += gl_InstanceIndex; } vec3 vertex = vertex_attrib; @@ -195,7 +203,7 @@ void main() { uv2_interp = uv2_attrib; #endif -#ifdef USE_OVERRIDE_POSITION +#ifdef OVERRIDE_POSITION vec4 position; #endif @@ -290,7 +298,7 @@ VERTEX_SHADER_CODE #endif //MODE_RENDER_DEPTH -#ifdef USE_OVERRIDE_POSITION +#ifdef OVERRIDE_POSITION gl_Position = position; #else gl_Position = projection_matrix * vec4(vertex_interp, 1.0); @@ -305,7 +313,8 @@ VERTEX_SHADER_CODE #endif #ifdef MODE_RENDER_MATERIAL if (scene_data.material_uv2_mode) { - gl_Position.xy = (uv2_attrib.xy + draw_call.bake_uv2_offset) * 2.0 - 1.0; + vec2 uv_offset = unpackHalf2x16(draw_call.uv_offset); + gl_Position.xy = (uv2_attrib.xy + uv_offset) * 2.0 - 1.0; gl_Position.z = 0.00001; gl_Position.w = 1.0; } @@ -345,18 +354,17 @@ layout(location = 5) in vec3 tangent_interp; layout(location = 6) in vec3 binormal_interp; #endif -layout(location = 7) flat in uint instance_index; - #ifdef MODE_DUAL_PARABOLOID layout(location = 8) in float dp_clip; #endif +layout(location = 9) in flat uint instance_index; + //defines to keep compatibility with vertex #define world_matrix instances.data[instance_index].transform -#define world_normal_matrix instances.data[instance_index].normal_transform #define projection_matrix scene_data.projection_matrix #if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) @@ -545,7 +553,7 @@ vec3 F0(float metallic, float specular, vec3 albedo) { return mix(vec3(dielectric), albedo, vec3(metallic)); } -void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float attenuation, vec3 shadow_attenuation, vec3 diffuse_color, float roughness, float metallic, float specular, float specular_blob_intensity, +void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, vec3 f0, uint orms, float specular_amount, #ifdef LIGHT_BACKLIGHT_USED vec3 backlight, #endif @@ -557,7 +565,7 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte float transmittance_z, #endif #ifdef LIGHT_RIM_USED - float rim, float rim_tint, + float rim, float rim_tint, vec3 rim_color, #endif #ifdef LIGHT_CLEARCOAT_USED float clearcoat, float clearcoat_gloss, @@ -565,6 +573,9 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte #ifdef LIGHT_ANISOTROPY_USED vec3 B, vec3 T, float anisotropy, #endif +#ifdef USE_SOFT_SHADOWS + float A, +#endif #ifdef USE_SHADOW_TO_OPACITY inout float alpha, #endif @@ -574,7 +585,6 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte // light is written by the light shader vec3 normal = N; - vec3 albedo = diffuse_color; vec3 light = L; vec3 view = V; @@ -585,7 +595,12 @@ LIGHT_SHADER_CODE /* clang-format on */ #else + +#ifdef USE_SOFT_SHADOWS float NdotL = min(A + dot(N, L), 1.0); +#else + float NdotL = dot(N, L); +#endif float cNdotL = max(NdotL, 0.0); // clamped NdotL float NdotV = dot(N, V); float cNdotV = max(NdotV, 0.0); @@ -595,14 +610,25 @@ LIGHT_SHADER_CODE #endif #if defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) +#ifdef USE_SOFT_SHADOWS float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); +#else + float cNdotH = clamp(dot(N, H), 0.0, 1.0); +#endif #endif #if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) +#ifdef USE_SOFT_SHADOWS float cLdotH = clamp(A + dot(L, H), 0.0, 1.0); +#else + float cLdotH = clamp(dot(L, H), 0.0, 1.0); +#endif #endif + float metallic = unpackUnorm4x8(orms).z; if (metallic < 1.0) { + float roughness = unpackUnorm4x8(orms).y; + #if defined(DIFFUSE_OREN_NAYAR) vec3 diffuse_brdf_NL; #else @@ -612,23 +638,6 @@ LIGHT_SHADER_CODE #if defined(DIFFUSE_LAMBERT_WRAP) // energy conserving lambert wrap shader diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))); - -#elif defined(DIFFUSE_OREN_NAYAR) - - { - // see http://mimosa-pudica.net/improved-oren-nayar.html - float LdotV = dot(L, V); - - float s = LdotV - NdotL * NdotV; - float t = mix(1.0, max(NdotL, NdotV), step(0.0, s)); - - float sigma2 = roughness * roughness; // TODO: this needs checking - vec3 A = 1.0 + sigma2 * (-0.5 / (sigma2 + 0.33) + 0.17 * diffuse_color / (sigma2 + 0.13)); - float B = 0.45 * sigma2 / (sigma2 + 0.09); - - diffuse_brdf_NL = cNdotL * (A + vec3(B) * s / t) * (1.0 / M_PI); - } - #elif defined(DIFFUSE_TOON) diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL); @@ -656,15 +665,15 @@ LIGHT_SHADER_CODE diffuse_brdf_NL = cNdotL * (1.0 / M_PI); #endif - diffuse_light += light_color * diffuse_color * shadow_attenuation * diffuse_brdf_NL * attenuation; + diffuse_light += light_color * diffuse_brdf_NL * attenuation; #if defined(LIGHT_BACKLIGHT_USED) - diffuse_light += light_color * diffuse_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * backlight * attenuation; + diffuse_light += light_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * backlight * attenuation; #endif #if defined(LIGHT_RIM_USED) float rim_light = pow(max(0.0, 1.0 - cNdotV), max(0.0, (1.0 - roughness) * 16.0)); - diffuse_light += rim_light * rim * mix(vec3(1.0), diffuse_color, rim_tint) * light_color; + diffuse_light += rim_light * rim * mix(vec3(1.0), rim_color, rim_tint) * light_color; #endif #ifdef LIGHT_TRANSMITTANCE_USED @@ -682,7 +691,7 @@ LIGHT_SHADER_CODE vec3(0.358, 0.004, 0.0) * exp(dd / 1.99) + vec3(0.078, 0.0, 0.0) * exp(dd / 7.41); - diffuse_light += profile * transmittance_color.a * diffuse_color * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI) * attenuation; + diffuse_light += profile * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); } #else @@ -692,7 +701,7 @@ LIGHT_SHADER_CODE fade = pow(max(0.0, 1.0 - fade), transmittance_curve); fade *= clamp(transmittance_boost - NdotL, 0.0, 1.0); - diffuse_light += diffuse_color * transmittance_color.rgb * light_color * (1.0 / M_PI) * transmittance_color.a * fade * attenuation; + diffuse_light += transmittance_color.rgb * light_color * (1.0 / M_PI) * transmittance_color.a * fade; } #endif //SSS_MODE_SKIN @@ -700,6 +709,7 @@ LIGHT_SHADER_CODE #endif //LIGHT_TRANSMITTANCE_USED } + float roughness = unpackUnorm4x8(orms).y; if (roughness > 0.0) { // FIXME: roughness == 0 should not disable specular light entirely // D @@ -712,7 +722,7 @@ LIGHT_SHADER_CODE blinn *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); float intensity = blinn; - specular_light += light_color * shadow_attenuation * intensity * specular_blob_intensity * attenuation; + specular_light += light_color * intensity * attenuation * specular_amount; #elif defined(SPECULAR_PHONG) @@ -723,7 +733,7 @@ LIGHT_SHADER_CODE phong *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); float intensity = (phong) / max(4.0 * cNdotV * cNdotL, 0.75); - specular_light += light_color * shadow_attenuation * intensity * specular_blob_intensity * attenuation; + specular_light += light_color * intensity * attenuation * specular_amount; #elif defined(SPECULAR_TOON) @@ -732,7 +742,7 @@ LIGHT_SHADER_CODE float mid = 1.0 - roughness; mid *= mid; float intensity = smoothstep(mid - roughness * 0.5, mid + roughness * 0.5, RdotV) * mid; - diffuse_light += light_color * shadow_attenuation * intensity * specular_blob_intensity * attenuation; // write to diffuse_light, as in toon shading you generally want no reflection + diffuse_light += light_color * intensity * attenuation * specular_amount; // write to diffuse_light, as in toon shading you generally want no reflection #elif defined(SPECULAR_DISABLED) // none.. @@ -757,13 +767,12 @@ LIGHT_SHADER_CODE float G = G_GGX_2cos(cNdotL, alpha_ggx) * G_GGX_2cos(cNdotV, alpha_ggx); #endif // F - vec3 f0 = F0(metallic, specular, diffuse_color); float cLdotH5 = SchlickFresnel(cLdotH); vec3 F = mix(vec3(cLdotH5), vec3(1.0), f0); vec3 specular_brdf_NL = cNdotL * D * F * G; - specular_light += specular_brdf_NL * light_color * shadow_attenuation * specular_blob_intensity * attenuation; + specular_light += specular_brdf_NL * light_color * attenuation * specular_amount; #endif #if defined(LIGHT_CLEARCOAT_USED) @@ -777,12 +786,12 @@ LIGHT_SHADER_CODE float clearcoat_specular_brdf_NL = 0.25 * clearcoat * Gr * Fr * Dr * cNdotL; - specular_light += clearcoat_specular_brdf_NL * light_color * shadow_attenuation * specular_blob_intensity * attenuation; + specular_light += clearcoat_specular_brdf_NL * light_color * attenuation * specular_amount; #endif } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(1.0 - length(shadow_attenuation * attenuation), 0.0, 1.0)); + alpha = min(alpha, clamp(1.0 - attenuation), 0.0, 1.0)); #endif #endif //defined(USE_LIGHT_SHADER_CODE) @@ -895,69 +904,39 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex #endif //USE_NO_SHADOWS -void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 albedo, float roughness, float metallic, float specular, float p_blob_intensity, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 binormal, vec3 tangent, float anisotropy, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, inout vec3 specular_light) { - vec3 light_rel_vec = lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - float normalized_distance = light_length * lights.data[idx].inv_radius; - vec2 attenuation_energy = unpackHalf2x16(lights.data[idx].attenuation_energy); - float omni_attenuation = pow(max(1.0 - normalized_distance, 0.0), attenuation_energy.x); - float light_attenuation = omni_attenuation; - vec3 shadow_attenuation = vec3(1.0); - vec4 color_specular = unpackUnorm4x8(lights.data[idx].color_specular); - color_specular.rgb *= attenuation_energy.y; - float size_A = 0.0; - - if (lights.data[idx].size > 0.0) { - float t = lights.data[idx].size / max(0.001, light_length); - size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); - } - -#ifdef LIGHT_TRANSMITTANCE_USED - float transmittance_z = transmittance_depth; //no transmittance by default -#endif +float get_omni_attenuation(float distance, float inv_range, float decay) { + float nd = distance * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(distance, 0.0001), -decay); +} +float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { #ifndef USE_NO_SHADOWS - vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[idx].shadow_color_enabled); - if (shadow_color_enabled.w > 0.5) { + if (omni_lights.data[idx].shadow_enabled) { // there is a shadowmap + vec3 light_rel_vec = omni_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + vec4 v = vec4(vertex, 1.0); - vec4 splane = (lights.data[idx].shadow_matrix * v); + vec4 splane = (omni_lights.data[idx].shadow_matrix * v); float shadow_len = length(splane.xyz); //need to remember shadow len from here { - vec3 nofs = normal_interp * lights.data[idx].shadow_normal_bias / lights.data[idx].inv_radius; + vec3 nofs = normal_interp * omni_lights.data[idx].shadow_normal_bias / omni_lights.data[idx].inv_radius; nofs *= (1.0 - max(0.0, dot(normalize(light_rel_vec), normalize(normal_interp)))); v.xyz += nofs; - splane = (lights.data[idx].shadow_matrix * v); + splane = (omni_lights.data[idx].shadow_matrix * v); } float shadow; - if (lights.data[idx].soft_shadow_size > 0.0) { +#ifdef USE_SOFT_SHADOWS + if (omni_lights.data[idx].soft_shadow_size > 0.0) { //soft shadow //find blocker @@ -977,10 +956,10 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); vec3 tangent = normalize(cross(v0, normal)); vec3 bitangent = normalize(cross(tangent, normal)); - float z_norm = shadow_len * lights.data[idx].inv_radius; + float z_norm = shadow_len * omni_lights.data[idx].inv_radius; - tangent *= lights.data[idx].soft_shadow_size * lights.data[idx].soft_shadow_scale; - bitangent *= lights.data[idx].soft_shadow_size * lights.data[idx].soft_shadow_scale; + tangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; + bitangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; @@ -988,7 +967,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; pos = normalize(pos); - vec4 uv_rect = lights.data[idx].atlas_rect; + vec4 uv_rect = omni_lights.data[idx].atlas_rect; if (pos.z >= 0.0) { pos.z += 1.0; @@ -1016,7 +995,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v tangent *= penumbra; bitangent *= penumbra; - z_norm -= lights.data[idx].inv_radius * lights.data[idx].shadow_bias; + z_norm -= omni_lights.data[idx].inv_radius * omni_lights.data[idx].shadow_bias; shadow = 0.0; for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { @@ -1024,7 +1003,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; pos = normalize(pos); - vec4 uv_rect = lights.data[idx].atlas_rect; + vec4 uv_rect = omni_lights.data[idx].atlas_rect; if (pos.z >= 0.0) { pos.z += 1.0; @@ -1047,8 +1026,9 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v shadow = 1.0; } } else { +#endif splane.xyz = normalize(splane.xyz); - vec4 clamp_rect = lights.data[idx].atlas_rect; + vec4 clamp_rect = omni_lights.data[idx].atlas_rect; if (splane.z >= 0.0) { splane.z += 1.0; @@ -1062,101 +1042,149 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v splane.xy /= splane.z; splane.xy = splane.xy * 0.5 + 0.5; - splane.z = (shadow_len - lights.data[idx].shadow_bias) * lights.data[idx].inv_radius; + splane.z = (shadow_len - omni_lights.data[idx].shadow_bias) * omni_lights.data[idx].inv_radius; splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; splane.w = 1.0; //needed? i think it should be 1 already - shadow = sample_pcf_shadow(shadow_atlas, lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, splane); + shadow = sample_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, splane); +#ifdef USE_SOFT_SHADOWS } +#endif + + return shadow; + } +#endif + return 1.0; +} + +void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif #ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 clamp_rect = lights.data[idx].atlas_rect; + vec4 transmittance_color, + float transmittance_depth, + float transmittance_curve, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, vec3 rim_color, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + inout float alpha, +#endif + inout vec3 diffuse_light, inout vec3 specular_light) { + vec3 light_rel_vec = omni_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + float omni_attenuation = get_omni_attenuation(light_length, omni_lights.data[idx].inv_radius, omni_lights.data[idx].attenuation); + float light_attenuation = omni_attenuation; + vec3 color = omni_lights.data[idx].color; - //redo shadowmapping, but shrink the model a bit to avoid arctifacts - splane = (lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * lights.data[idx].transmittance_bias, 1.0)); +#ifdef USE_SOFT_SHADOWS + float size_A = 0.0; - shadow_len = length(splane.xyz); - splane = normalize(splane.xyz); + if (omni_lights.data[idx].size > 0.0) { + float t = omni_lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } +#endif - if (splane.z >= 0.0) { - splane.z += 1.0; +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; //no transmittance by default + transmittance_color.a *= light_attenuation; + { + vec4 clamp_rect = omni_lights.data[idx].atlas_rect; - } else { - splane.z = 1.0 - splane.z; - } + //redo shadowmapping, but shrink the model a bit to avoid arctifacts + vec4 splane = (omni_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * omni_lights.data[idx].transmittance_bias, 1.0)); - splane.xy /= splane.z; - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = shadow_len * lights.data[idx].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already + shadow_len = length(splane.xyz); + splane = normalize(splane.xyz); + + if (splane.z >= 0.0) { + splane.z += 1.0; - float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; - transmittance_z = (splane.z - shadow_z) / lights.data[idx].inv_radius; + } else { + splane.z = 1.0 - splane.z; } -#endif - vec3 no_shadow = vec3(1.0); + splane.xy /= splane.z; + splane.xy = splane.xy * 0.5 + 0.5; + splane.z = shadow_len * omni_lights.data[idx].inv_radius; + splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; + splane.w = 1.0; //needed? i think it should be 1 already - if (lights.data[idx].projector_rect != vec4(0.0)) { - vec3 local_v = (lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; - local_v = normalize(local_v); + float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; + transmittance_z = (splane.z - shadow_z) / omni_lights.data[idx].inv_radius; + } +#endif - vec4 atlas_rect = lights.data[idx].projector_rect; +#if 0 - if (local_v.z >= 0.0) { - local_v.z += 1.0; - atlas_rect.y += atlas_rect.w; + if (omni_lights.data[idx].projector_rect != vec4(0.0)) { + vec3 local_v = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; + local_v = normalize(local_v); - } else { - local_v.z = 1.0 - local_v.z; - } + vec4 atlas_rect = omni_lights.data[idx].projector_rect; - local_v.xy /= local_v.z; - local_v.xy = local_v.xy * 0.5 + 0.5; - vec2 proj_uv = local_v.xy * atlas_rect.zw; + if (local_v.z >= 0.0) { + local_v.z += 1.0; + atlas_rect.y += atlas_rect.w; - vec2 proj_uv_ddx; - vec2 proj_uv_ddy; - { - vec3 local_v_ddx = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; - local_v_ddx = normalize(local_v_ddx); + } else { + local_v.z = 1.0 - local_v.z; + } - if (local_v_ddx.z >= 0.0) { - local_v_ddx.z += 1.0; - } else { - local_v_ddx.z = 1.0 - local_v_ddx.z; - } + local_v.xy /= local_v.z; + local_v.xy = local_v.xy * 0.5 + 0.5; + vec2 proj_uv = local_v.xy * atlas_rect.zw; - local_v_ddx.xy /= local_v_ddx.z; - local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; + vec2 proj_uv_ddx; + vec2 proj_uv_ddy; + { + vec3 local_v_ddx = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; + local_v_ddx = normalize(local_v_ddx); - proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; + if (local_v_ddx.z >= 0.0) { + local_v_ddx.z += 1.0; + } else { + local_v_ddx.z = 1.0 - local_v_ddx.z; + } - vec3 local_v_ddy = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; - local_v_ddy = normalize(local_v_ddy); + local_v_ddx.xy /= local_v_ddx.z; + local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; - if (local_v_ddy.z >= 0.0) { - local_v_ddy.z += 1.0; - } else { - local_v_ddy.z = 1.0 - local_v_ddy.z; - } + proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; - local_v_ddy.xy /= local_v_ddy.z; - local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; + vec3 local_v_ddy = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; + local_v_ddy = normalize(local_v_ddy); - proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; + if (local_v_ddy.z >= 0.0) { + local_v_ddy.z += 1.0; + } else { + local_v_ddy.z = 1.0 - local_v_ddy.z; } - vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); - no_shadow = mix(no_shadow, proj.rgb, proj.a); + local_v_ddy.xy /= local_v_ddy.z; + local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; + + proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; } - shadow_attenuation = mix(shadow_color_enabled.rgb, no_shadow, shadow); + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); + no_shadow = mix(no_shadow, proj.rgb, proj.a); } -#endif //USE_NO_SHADOWS +#endif + + light_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color_specular.rgb, light_attenuation, shadow_attenuation, albedo, roughness, metallic, specular, color_specular.a * p_blob_intensity, + light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -1168,7 +1196,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim * omni_attenuation, rim_tint, + rim * omni_attenuation, rim_tint, rim_color, #endif #ifdef LIGHT_CLEARCOAT_USED clearcoat, clearcoat_gloss, @@ -1176,6 +1204,9 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, #endif +#ifdef USE_SOFT_SHADOWS + size_A, +#endif #ifdef USE_SHADOW_TO_OPACITY alpha, #endif @@ -1183,89 +1214,39 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v specular_light); } -void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 albedo, float roughness, float metallic, float specular, float p_blob_intensity, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 binormal, vec3 tangent, float anisotropy, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, - inout vec3 specular_light) { - vec3 light_rel_vec = lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - float normalized_distance = light_length * lights.data[idx].inv_radius; - vec2 attenuation_energy = unpackHalf2x16(lights.data[idx].attenuation_energy); - float spot_attenuation = pow(max(1.0 - normalized_distance, 0.001), attenuation_energy.x); - vec3 spot_dir = lights.data[idx].direction; - vec2 spot_att_angle = unpackHalf2x16(lights.data[idx].cone_attenuation_angle); - float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_att_angle.y); - float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_att_angle.y)); - spot_attenuation *= 1.0 - pow(spot_rim, spot_att_angle.x); - float light_attenuation = spot_attenuation; - vec3 shadow_attenuation = vec3(1.0); - vec4 color_specular = unpackUnorm4x8(lights.data[idx].color_specular); - color_specular.rgb *= attenuation_energy.y; - - float size_A = 0.0; - - if (lights.data[idx].size > 0.0) { - float t = lights.data[idx].size / max(0.001, light_length); - size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); - } -/* - if (lights.data[idx].atlas_rect!=vec4(0.0)) { - //use projector texture - } - */ -#ifdef LIGHT_TRANSMITTANCE_USED - float transmittance_z = transmittance_depth; -#endif - +float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { #ifndef USE_NO_SHADOWS - vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[idx].shadow_color_enabled); - if (shadow_color_enabled.w > 0.5) { + if (spot_lights.data[idx].shadow_enabled) { + vec3 light_rel_vec = spot_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + vec3 spot_dir = spot_lights.data[idx].direction; //there is a shadowmap vec4 v = vec4(vertex, 1.0); - v.xyz -= spot_dir * lights.data[idx].shadow_bias; + v.xyz -= spot_dir * spot_lights.data[idx].shadow_bias; - float z_norm = dot(spot_dir, -light_rel_vec) * lights.data[idx].inv_radius; + float z_norm = dot(spot_dir, -light_rel_vec) * spot_lights.data[idx].inv_radius; float depth_bias_scale = 1.0 / (max(0.0001, z_norm)); //the closer to the light origin, the more you have to offset to reach 1px in the map - vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(spot_dir, -normalize(normal_interp)))) * lights.data[idx].shadow_normal_bias * depth_bias_scale; + vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(spot_dir, -normalize(normal_interp)))) * spot_lights.data[idx].shadow_normal_bias * depth_bias_scale; normal_bias -= spot_dir * dot(spot_dir, normal_bias); //only XY, no Z v.xyz += normal_bias; //adjust with bias - z_norm = dot(spot_dir, v.xyz - lights.data[idx].position) * lights.data[idx].inv_radius; + z_norm = dot(spot_dir, v.xyz - spot_lights.data[idx].position) * spot_lights.data[idx].inv_radius; float shadow; - vec4 splane = (lights.data[idx].shadow_matrix * v); + vec4 splane = (spot_lights.data[idx].shadow_matrix * v); splane /= splane.w; - if (lights.data[idx].soft_shadow_size > 0.0) { +#ifdef USE_SOFT_SHADOWS + if (spot_lights.data[idx].soft_shadow_size > 0.0) { //soft shadow //find blocker - vec2 shadow_uv = splane.xy * lights.data[idx].atlas_rect.zw + lights.data[idx].atlas_rect.xy; + vec2 shadow_uv = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; float blocker_count = 0.0; float blocker_average = 0.0; @@ -1278,11 +1259,11 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); } - float uv_size = lights.data[idx].soft_shadow_size * z_norm * lights.data[idx].soft_shadow_scale; - vec2 clamp_max = lights.data[idx].atlas_rect.xy + lights.data[idx].atlas_rect.zw; + float uv_size = spot_lights.data[idx].soft_shadow_size * z_norm * spot_lights.data[idx].soft_shadow_scale; + vec2 clamp_max = spot_lights.data[idx].atlas_rect.xy + spot_lights.data[idx].atlas_rect.zw; for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; - suv = clamp(suv, lights.data[idx].atlas_rect.xy, clamp_max); + suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; if (d < z_norm) { blocker_average += d; @@ -1299,7 +1280,7 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v shadow = 0.0; for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; - suv = clamp(suv, lights.data[idx].atlas_rect.xy, clamp_max); + suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, z_norm, 1.0)); } @@ -1311,54 +1292,93 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v } } else { +#endif //hard shadow - vec4 shadow_uv = vec4(splane.xy * lights.data[idx].atlas_rect.zw + lights.data[idx].atlas_rect.xy, z_norm, 1.0); + vec4 shadow_uv = vec4(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z, 1.0); - shadow = sample_pcf_shadow(shadow_atlas, lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); + shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); +#ifdef USE_SOFT_SHADOWS } +#endif - vec3 no_shadow = vec3(1.0); + return shadow; + } - if (lights.data[idx].projector_rect != vec4(0.0)) { - splane = (lights.data[idx].shadow_matrix * vec4(vertex, 1.0)); - splane /= splane.w; +#endif //USE_NO_SHADOWS - vec2 proj_uv = splane.xy * lights.data[idx].projector_rect.zw; + return 1.0; +} - //ensure we have proper mipmaps - vec4 splane_ddx = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)); - splane_ddx /= splane_ddx.w; - vec2 proj_uv_ddx = splane_ddx.xy * lights.data[idx].projector_rect.zw - proj_uv; +void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_curve, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, vec3 rim_color, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + inout float alpha, +#endif + inout vec3 diffuse_light, + inout vec3 specular_light) { + vec3 light_rel_vec = spot_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + float spot_attenuation = get_omni_attenuation(light_length, spot_lights.data[idx].inv_radius, spot_lights.data[idx].attenuation); + vec3 spot_dir = spot_lights.data[idx].direction; + float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[idx].cone_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[idx].cone_angle)); + spot_attenuation *= 1.0 - pow(spot_rim, spot_lights.data[idx].cone_attenuation); + float light_attenuation = spot_attenuation; + vec3 color = spot_lights.data[idx].color; + float specular_amount = spot_lights.data[idx].specular_amount; - vec4 splane_ddy = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)); - splane_ddy /= splane_ddy.w; - vec2 proj_uv_ddy = splane_ddy.xy * lights.data[idx].projector_rect.zw - proj_uv; +#ifdef USE_SOFT_SHADOWS + float size_A = 0.0; - vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + lights.data[idx].projector_rect.xy, proj_uv_ddx, proj_uv_ddy); - no_shadow = mix(no_shadow, proj.rgb, proj.a); - } + if (spot_lights.data[idx].size > 0.0) { + float t = spot_lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } +#endif - shadow_attenuation = mix(shadow_color_enabled.rgb, no_shadow, shadow); + /* + if (spot_lights.data[idx].atlas_rect!=vec4(0.0)) { + //use projector texture + } + */ #ifdef LIGHT_TRANSMITTANCE_USED - { - splane = (lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * lights.data[idx].transmittance_bias, 1.0)); - splane /= splane.w; - splane.xy = splane.xy * lights.data[idx].atlas_rect.zw + lights.data[idx].atlas_rect.xy; - - float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; - //reconstruct depth - shadow_z /= lights.data[idx].inv_radius; - //distance to light plane - float z = dot(spot_dir, -light_rel_vec); - transmittance_z = z - shadow_z; - } -#endif //LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; + transmittance_color.a *= light_attenuation; + { + splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * spot_lights.data[idx].transmittance_bias, 1.0)); + splane /= splane.w; + splane.xy = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; + + float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; + //reconstruct depth + shadow_z /= spot_lights.data[idx].inv_radius; + //distance to light plane + float z = dot(spot_dir, -light_rel_vec); + transmittance_z = z - shadow_z; } +#endif //LIGHT_TRANSMITTANCE_USED -#endif //USE_NO_SHADOWS + light_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color_specular.rgb, light_attenuation, shadow_attenuation, albedo, roughness, metallic, specular, color_specular.a * p_blob_intensity, + light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -1370,7 +1390,7 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim * spot_attenuation, rim_tint, + rim * spot_attenuation, rim_tint, rim_color, #endif #ifdef LIGHT_CLEARCOAT_USED clearcoat, clearcoat_gloss, @@ -1378,6 +1398,9 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, #endif +#ifdef USE_SOFT_SHADOW + size_A, +#endif #ifdef USE_SHADOW_TO_OPACITY alpha, #endif @@ -1401,11 +1424,11 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes blend *= blend; blend = max(0.0, 1.0 - blend); - if (reflections.data[ref_index].params.x > 0.0) { // compute reflection + if (reflections.data[ref_index].intensity > 0.0) { // compute reflection vec3 local_ref_vec = (reflections.data[ref_index].local_matrix * vec4(ref_vec, 0.0)).xyz; - if (reflections.data[ref_index].params.w > 0.5) { //box project + if (reflections.data[ref_index].box_project) { //box project vec3 nrdir = normalize(local_ref_vec); vec3 rbmax = (box_extents - local_pos) / nrdir; @@ -1422,11 +1445,11 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes reflection.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_ref_vec, reflections.data[ref_index].index), roughness * MAX_ROUGHNESS_LOD).rgb; - if (reflections.data[ref_index].params.z < 0.5) { + if (reflections.data[ref_index].exterior) { reflection.rgb = mix(specular_light, reflection.rgb, blend); } - reflection.rgb *= reflections.data[ref_index].params.x; + reflection.rgb *= reflections.data[ref_index].intensity; //intensity reflection.a = blend; reflection.rgb *= reflection.a; @@ -1445,7 +1468,7 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes ambient_out.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_amb_vec, reflections.data[ref_index].index), MAX_ROUGHNESS_LOD).rgb; ambient_out.a = blend; - if (reflections.data[ref_index].params.z < 0.5) { //interior + if (reflections.data[ref_index].exterior) { ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); } @@ -1456,7 +1479,7 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes vec4 ambient_out; ambient_out.a = blend; ambient_out.rgb = reflections.data[ref_index].ambient; - if (reflections.data[ref_index].params.z < 0.5) { + if (reflections.data[ref_index].exterior) { ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); } ambient_out.rgb *= ambient_out.a; @@ -1759,7 +1782,7 @@ vec4 fog_process(vec3 vertex) { } } - float fog_amount = 1.0 - exp(vertex.z * scene_data.fog_density); + float fog_amount = 1.0 - exp(min(0.0, vertex.z * scene_data.fog_density)); if (abs(scene_data.fog_height_density) > 0.001) { float y = (scene_data.camera_matrix * vec4(vertex, 1.0)).y; @@ -1774,7 +1797,43 @@ vec4 fog_process(vec3 vertex) { return vec4(fog_color, fog_amount); } +void cluster_get_item_range(uint p_offset, out uint item_min, out uint item_max, out uint item_from, out uint item_to) { + uint item_min_max = cluster_buffer.data[p_offset]; + item_min = item_min_max & 0xFFFF; + item_max = item_min_max >> 16; + ; + + item_from = item_min >> 5; + item_to = (item_max == 0) ? 0 : ((item_max - 1) >> 5) + 1; //side effect of how it is stored, as item_max 0 means no elements +} + +uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { + int local_min = clamp(int(z_min) - int(i) * 32, 0, 31); + int mask_width = min(int(z_max) - int(z_min), 32 - local_min); + return bitfieldInsert(uint(0), uint(0xFFFFFFFF), local_min, mask_width); +} + +float blur_shadow(float shadow) { + return shadow; +#if 0 + //disabling for now, will investigate later + float interp_shadow = shadow; + if (gl_HelperInvocation) { + interp_shadow = -4.0; // technically anything below -4 will do but just to make sure + } + + uvec2 fc2 = uvec2(gl_FragCoord.xy); + interp_shadow -= dFdx(interp_shadow) * (float(fc2.x & 1) - 0.5); + interp_shadow -= dFdy(interp_shadow) * (float(fc2.y & 1) - 0.5); + + if (interp_shadow >= 0.0) { + shadow = interp_shadow; + } + return shadow; #endif +} + +#endif //!MODE_RENDER DEPTH void main() { #ifdef MODE_DUAL_PARABOLOID @@ -1802,9 +1861,7 @@ void main() { float clearcoat_gloss = 0.0; float anisotropy = 0.0; vec2 anisotropy_flow = vec2(1.0, 0.0); -#if defined(CUSTOM_FOG_USED) - vec4 custom_fog = vec4(0.0); -#endif + vec4 fog = vec4(0.0); #if defined(CUSTOM_RADIANCE_USED) vec4 custom_radiance = vec4(0.0); #endif @@ -1812,10 +1869,8 @@ void main() { vec4 custom_irradiance = vec4(0.0); #endif -#if defined(AO_USED) float ao = 1.0; float ao_light_affect = 0.0; -#endif float alpha = 1.0; @@ -1855,7 +1910,7 @@ void main() { vec3 normal_map = vec3(0.5); #endif - float normal_depth = 1.0; + float normal_map_depth = 1.0; vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size + scene_data.screen_pixel_size * 0.5; //account for center @@ -1931,7 +1986,7 @@ FRAGMENT_SHADER_CODE normal_map.xy = normal_map.xy * 2.0 - 1.0; normal_map.z = sqrt(max(0.0, 1.0 - dot(normal_map.xy, normal_map.xy))); //always ignore Z, as it can be RG packed, Z may be pos/neg, etc. - normal = normalize(mix(normal, tangent * normal_map.x + binormal * normal_map.y + normal * normal_map.z, normal_depth)); + normal = normalize(mix(normal, tangent * normal_map.x + binormal * normal_map.y + normal * normal_map.z, normal_map_depth)); #endif @@ -1953,77 +2008,147 @@ FRAGMENT_SHADER_CODE discard; } #endif + + /////////////////////// FOG ////////////////////// +#ifndef MODE_RENDER_DEPTH + +#ifndef CUSTOM_FOG_USED + // fog must be processed as early as possible and then packed. + // to maximize VGPR usage + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + + if (scene_data.fog_enabled) { + fog = fog_process(vertex); + } + +#ifndef LOW_END_MODE + if (scene_data.volumetric_fog_enabled) { + vec4 volumetric_fog = volumetric_fog_process(screen_uv, -vertex.z); + if (scene_data.fog_enabled) { + //must use the full blending equation here to blend fogs + vec4 res; + float sa = 1.0 - volumetric_fog.a; + res.a = fog.a * sa + volumetric_fog.a; + if (res.a == 0.0) { + res.rgb = vec3(0.0); + } else { + res.rgb = (fog.rgb * fog.a * sa + volumetric_fog.rgb * volumetric_fog.a) / res.a; + } + fog = res; + } else { + fog = volumetric_fog; + } + } +#endif //!LOW_END_MODE +#endif //!CUSTOM_FOG_USED + + uint fog_rg = packHalf2x16(fog.rg); + uint fog_ba = packHalf2x16(fog.ba); + +#endif //!MODE_RENDER_DEPTH + /////////////////////// DECALS //////////////////////////////// #ifndef MODE_RENDER_DEPTH - uvec4 cluster_cell = texture(usampler3D(cluster_texture, material_samplers[SAMPLER_NEAREST_CLAMP]), vec3(screen_uv, (abs(vertex.z) - scene_data.z_near) / (scene_data.z_far - scene_data.z_near))); + uvec2 cluster_pos = uvec2(gl_FragCoord.xy) >> scene_data.cluster_shift; + uint cluster_offset = (scene_data.cluster_width * cluster_pos.y + cluster_pos.x) * (scene_data.max_cluster_element_count_div_32 + 32); + + uint cluster_z = uint(clamp((-vertex.z / scene_data.z_far) * 32.0, 0.0, 31.0)); + //used for interpolating anything cluster related vec3 vertex_ddx = dFdx(vertex); vec3 vertex_ddy = dFdy(vertex); { // process decals - uint decal_count = cluster_cell.w >> CLUSTER_COUNTER_SHIFT; - uint decal_pointer = cluster_cell.w & CLUSTER_POINTER_MASK; + uint cluster_decal_offset = cluster_offset + scene_data.cluster_type_size * 2; - //do outside for performance and avoiding arctifacts + uint item_min; + uint item_max; + uint item_from; + uint item_to; - for (uint i = 0; i < decal_count; i++) { - uint decal_index = cluster_data.indices[decal_pointer + i]; - if (!bool(decals.data[decal_index].mask & instances.data[instance_index].layer_mask)) { - continue; //not masked - } + cluster_get_item_range(cluster_decal_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); - vec3 uv_local = (decals.data[decal_index].xform * vec4(vertex, 1.0)).xyz; - if (any(lessThan(uv_local, vec3(0.0, -1.0, 0.0))) || any(greaterThan(uv_local, vec3(1.0)))) { - continue; //out of decal - } +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif - //we need ddx/ddy for mipmaps, so simulate them - vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; - vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_decal_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif - float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + uint decal_index = 32 * i + bit; - if (decals.data[decal_index].normal_fade > 0.0) { - fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); - } + if (!bool(decals.data[decal_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } - if (decals.data[decal_index].albedo_rect != vec4(0.0)) { - //has albedo - vec4 decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); - decal_albedo *= decals.data[decal_index].modulate; - decal_albedo.a *= fade; - albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); - - if (decals.data[decal_index].normal_rect != vec4(0.0)) { - vec3 decal_normal = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; - decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software - decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); - //convert to view space, use xzy because y is up - decal_normal = (decals.data[decal_index].normal_xform * decal_normal.xzy).xyz; - - normal = normalize(mix(normal, decal_normal, decal_albedo.a)); + vec3 uv_local = (decals.data[decal_index].xform * vec4(vertex, 1.0)).xyz; + if (any(lessThan(uv_local, vec3(0.0, -1.0, 0.0))) || any(greaterThan(uv_local, vec3(1.0)))) { + continue; //out of decal } - if (decals.data[decal_index].orm_rect != vec4(0.0)) { - vec3 decal_orm = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; -#if defined(AO_USED) - ao = mix(ao, decal_orm.r, decal_albedo.a); -#endif - roughness = mix(roughness, decal_orm.g, decal_albedo.a); - metallic = mix(metallic, decal_orm.b, decal_albedo.a); + //we need ddx/ddy for mipmaps, so simulate them + vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; + vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + + float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); + + if (decals.data[decal_index].normal_fade > 0.0) { + fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); } - } - if (decals.data[decal_index].emission_rect != vec4(0.0)) { - //emission is additive, so its independent from albedo - emission += textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { + //has albedo + vec4 decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + decal_albedo *= decals.data[decal_index].modulate; + decal_albedo.a *= fade; + albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); + + if (decals.data[decal_index].normal_rect != vec4(0.0)) { + vec3 decal_normal = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software + decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); + //convert to view space, use xzy because y is up + decal_normal = (decals.data[decal_index].normal_xform * decal_normal.xzy).xyz; + + normal = normalize(mix(normal, decal_normal, decal_albedo.a)); + } + + if (decals.data[decal_index].orm_rect != vec4(0.0)) { + vec3 decal_orm = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + ao = mix(ao, decal_orm.r, decal_albedo.a); + roughness = mix(roughness, decal_orm.g, decal_albedo.a); + metallic = mix(metallic, decal_orm.b, decal_albedo.a); + } + } + + if (decals.data[decal_index].emission_rect != vec4(0.0)) { + //emission is additive, so its independent from albedo + emission += textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + } } } } + //pack albedo until needed again, saves 2 VGPRs in the meantime + #endif //not render depth /////////////////////// LIGHTING ////////////////////////////// @@ -2091,12 +2216,7 @@ FRAGMENT_SHADER_CODE //radiance - float specular_blob_intensity = 1.0; - -#if defined(SPECULAR_TOON) - specular_blob_intensity *= specular * 2.0; -#endif - +/// GI /// #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) #ifdef USE_LIGHTMAP @@ -2124,10 +2244,10 @@ FRAGMENT_SHADER_CODE } else if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { // has actual lightmap bool uses_sh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SH_LIGHTMAP); - uint ofs = instances.data[instance_index].gi_offset & 0xFFF; + uint ofs = instances.data[instance_index].gi_offset & 0xFFFF; vec3 uvw; uvw.xy = uv2 * instances.data[instance_index].lightmap_uv_scale.zw + instances.data[instance_index].lightmap_uv_scale.xy; - uvw.z = float((instances.data[instance_index].gi_offset >> 12) & 0xFF); + uvw.z = float((instances.data[instance_index].gi_offset >> 16) & 0xFFFF); if (uses_sh) { uvw.z *= 4.0; //SH textures use 4 times more data @@ -2263,17 +2383,17 @@ FRAGMENT_SHADER_CODE if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers - ivec2 coord; + vec2 coord; if (scene_data.gi_upscale_for_msaa) { - ivec2 base_coord = ivec2(gl_FragCoord.xy); - ivec2 closest_coord = base_coord; - float closest_ang = dot(normal, texelFetch(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0).xyz * 2.0 - 1.0); + vec2 base_coord = screen_uv; + vec2 closest_coord = base_coord; + float closest_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0.0).xyz * 2.0 - 1.0); for (int i = 0; i < 4; i++) { - const ivec2 neighbours[4] = ivec2[](ivec2(-1, 0), ivec2(1, 0), ivec2(0, -1), ivec2(0, 1)); - ivec2 neighbour_coord = base_coord + neighbours[i]; - float neighbour_ang = dot(normal, texelFetch(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0).xyz * 2.0 - 1.0); + const vec2 neighbours[4] = vec2[](vec2(-1, 0), vec2(1, 0), vec2(0, -1), vec2(0, 1)); + vec2 neighbour_coord = base_coord + neighbours[i] * scene_data.screen_pixel_size; + float neighbour_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0.0).xyz * 2.0 - 1.0); if (neighbour_ang > closest_ang) { closest_ang = neighbour_ang; closest_coord = neighbour_coord; @@ -2283,28 +2403,69 @@ FRAGMENT_SHADER_CODE coord = closest_coord; } else { - coord = ivec2(gl_FragCoord.xy); + coord = screen_uv; } - vec4 buffer_ambient = texelFetch(sampler2D(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0); - vec4 buffer_reflection = texelFetch(sampler2D(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0); + vec4 buffer_ambient = textureLod(sampler2D(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); + vec4 buffer_reflection = textureLod(sampler2D(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a); specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a); } #endif +#ifndef LOW_END_MODE + if (scene_data.ssao_enabled) { + float ssao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; + ao = min(ao, ssao); + ao_light_affect = mix(ao_light_affect, max(ao_light_affect, scene_data.ssao_light_affect), scene_data.ssao_ao_affect); + } +#endif //LOW_END_MODE + { // process reflections vec4 reflection_accum = vec4(0.0, 0.0, 0.0, 0.0); vec4 ambient_accum = vec4(0.0, 0.0, 0.0, 0.0); - uint reflection_probe_count = cluster_cell.z >> CLUSTER_COUNTER_SHIFT; - uint reflection_probe_pointer = cluster_cell.z & CLUSTER_POINTER_MASK; + uint cluster_reflection_offset = cluster_offset + scene_data.cluster_type_size * 3; + + uint item_min; + uint item_max; + uint item_from; + uint item_to; - for (uint i = 0; i < reflection_probe_count; i++) { - uint ref_index = cluster_data.indices[reflection_probe_pointer + i]; - reflection_process(ref_index, vertex, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); + cluster_get_item_range(cluster_reflection_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_reflection_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + uint reflection_index = 32 * i + bit; + + if (!bool(reflections.data[reflection_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + reflection_process(reflection_index, vertex, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); + } } if (reflection_accum.a > 0.0) { @@ -2318,6 +2479,16 @@ FRAGMENT_SHADER_CODE #endif } + //finalize ambient light here + ambient_light *= albedo.rgb; + ambient_light *= ao; + + // convert ao to direct light ao + ao = mix(1.0, ao, ao_light_affect); + + //this saves some VGPRs + vec3 f0 = F0(metallic, specular, albedo); + { #if defined(DIFFUSE_TOON) //simplify for toon, as @@ -2335,24 +2506,39 @@ FRAGMENT_SHADER_CODE float a004 = min(r.x * r.x, exp2(-9.28 * ndotv)) * r.x + r.y; vec2 env = vec2(-1.04, 1.04) * a004 + r.zw; - vec3 f0 = F0(metallic, specular, albedo); specular_light *= env.x * f0 + env.y; #endif } +#endif //GI !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#if !defined(MODE_RENDER_DEPTH) + //this saves some VGPRs + uint orms = packUnorm4x8(vec4(ao, roughness, metallic, specular)); +#endif + +// LIGHTING +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + { //directional light - for (uint i = 0; i < scene_data.directional_light_count; i++) { + // Do shadow and lighting in two passes to reduce register pressure + uint shadow0 = 0; + uint shadow1 = 0; + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + if (!bool(directional_lights.data[i].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } - vec3 shadow_attenuation = vec3(1.0); - -#ifdef LIGHT_TRANSMITTANCE_USED - float transmittance_z = transmittance_depth; -#endif + float shadow = 1.0; +#ifdef USE_SOFT_SHADOWS + //version with soft shadows, more expensive if (directional_lights.data[i].shadow_enabled) { float depth_z = -vertex.z; @@ -2366,8 +2552,6 @@ FRAGMENT_SHADER_CODE normal_bias -= light_dir * dot(light_dir, normal_bias); \ m_var.xyz += normal_bias; - float shadow = 0.0; - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { vec4 v = vec4(vertex, 1.0); @@ -2388,19 +2572,6 @@ FRAGMENT_SHADER_CODE shadow_color = directional_lights.data[i].shadow_color1.rgb; -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.x, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix1 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_z_range.x; - float z = trans_coord.z * directional_lights.data[i].shadow_z_range.x; - - transmittance_z = z - shadow_z; - } -#endif } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); @@ -2420,19 +2591,6 @@ FRAGMENT_SHADER_CODE } shadow_color = directional_lights.data[i].shadow_color2.rgb; -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.y, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix2 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_z_range.y; - float z = trans_coord.z * directional_lights.data[i].shadow_z_range.y; - - transmittance_z = z - shadow_z; - } -#endif } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); @@ -2452,19 +2610,6 @@ FRAGMENT_SHADER_CODE } shadow_color = directional_lights.data[i].shadow_color3.rgb; -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.z, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix3 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_z_range.z; - float z = trans_coord.z * directional_lights.data[i].shadow_z_range.z; - - transmittance_z = z - shadow_z; - } -#endif } else { vec4 v = vec4(vertex, 1.0); @@ -2485,20 +2630,6 @@ FRAGMENT_SHADER_CODE } shadow_color = directional_lights.data[i].shadow_color4.rgb; - -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.w, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix4 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_z_range.w; - float z = trans_coord.z * directional_lights.data[i].shadow_z_range.w; - - transmittance_z = z - shadow_z; - } -#endif } if (directional_lights.data[i].blend_splits) { @@ -2572,130 +2703,407 @@ FRAGMENT_SHADER_CODE shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance - shadow_attenuation = mix(shadow_color, vec3(1.0), shadow); +#undef BIAS_FUNC + } +#else + // Soft shadow disabled version + + if (directional_lights.data[i].shadow_enabled) { + float depth_z = -vertex.z; + + vec4 pssm_coord; + vec3 light_dir = directional_lights.data[i].direction; + vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); + +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = base_normal_bias * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ + m_var.xyz += normal_bias; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); +#ifdef LIGHT_TRANSMITTANCE_USED + { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.x, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix1 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.x; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.x; + + transmittance_z = z - shadow_z; + } +#endif + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); +#ifdef LIGHT_TRANSMITTANCE_USED + { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.y, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix2 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.y; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.y; + + transmittance_z = z - shadow_z; + } +#endif + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); +#ifdef LIGHT_TRANSMITTANCE_USED + { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.z, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix3 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.z; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.z; + + transmittance_z = z - shadow_z; + } +#endif + + } else { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); +#ifdef LIGHT_TRANSMITTANCE_USED + { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.w, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix4 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.w; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; + + transmittance_z = z - shadow_z; + } +#endif + } + + pssm_coord /= pssm_coord.w; + + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + } + + pssm_coord /= pssm_coord.w; + + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + shadow = mix(shadow, shadow2, pssm_blend); + } + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance #undef BIAS_FUNC } +#endif + + if (i < 4) { + shadow0 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << (i * 8); + } else { + shadow1 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << ((i - 4) * 8); + } + } + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + + if (!bool(directional_lights.data[i].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; + + if (directional_lights.data[i].shadow_enabled) { + float depth_z = -vertex.z; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.x, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix1 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.x; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.x; + + transmittance_z = z - shadow_z; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.y, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix2 * trans_vertex; + trans_coord /= trans_coord.w; - light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].size, directional_lights.data[i].color * directional_lights.data[i].energy, 1.0, shadow_attenuation, albedo, roughness, metallic, specular, directional_lights.data[i].specular * specular_blob_intensity, + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.y; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.y; + + transmittance_z = z - shadow_z; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.z, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix3 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.z; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.z; + + transmittance_z = z - shadow_z; + + } else { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.w, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix4 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.w; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; + + transmittance_z = z - shadow_z; + } +#endif + + float shadow = 1.0; + + if (i < 4) { + shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; + } else { + shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; + } + + blur_shadow(shadow); + + light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, - transmittance_z, + transmittance_color, + transmittance_depth, + transmittance_curve, + transmittance_boost, + transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim, rim_tint, + rim, rim_tint, albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, + binormal, tangent, anisotropy, +#endif +#ifdef USE_SOFT_SHADOW + directional_lights.data[i].size, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, - specular_light); + diffuse_light, + specular_light); + } } - } - { //omni lights + { //omni lights - uint omni_light_count = cluster_cell.x >> CLUSTER_COUNTER_SHIFT; - uint omni_light_pointer = cluster_cell.x & CLUSTER_POINTER_MASK; + uint cluster_omni_offset = cluster_offset; - for (uint i = 0; i < omni_light_count; i++) { - uint light_index = cluster_data.indices[omni_light_pointer + i]; + uint item_min; + uint item_max; + uint item_from; + uint item_to; - if (!bool(lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { - continue; //not masked - } + cluster_get_item_range(cluster_omni_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_omni_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif - light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, albedo, roughness, metallic, specular, specular_blob_intensity, + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + uint light_index = 32 * i + bit; + + if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + float shadow = light_process_omni_shadow(light_index, vertex, view); + + shadow = blur_shadow(shadow); + + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_curve, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, + rim, + rim_tint, + albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, binormal, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, specular_light); + diffuse_light, specular_light); + } + } } - } - { //spot lights - uint spot_light_count = cluster_cell.y >> CLUSTER_COUNTER_SHIFT; - uint spot_light_pointer = cluster_cell.y & CLUSTER_POINTER_MASK; + { //spot lights - for (uint i = 0; i < spot_light_count; i++) { - uint light_index = cluster_data.indices[spot_light_pointer + i]; + uint cluster_spot_offset = cluster_offset + scene_data.cluster_type_size; - if (!bool(lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { - continue; //not masked - } + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_spot_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); - light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, albedo, roughness, metallic, specular, specular_blob_intensity, +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_spot_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + + uint light_index = 32 * i + bit; + + if (!bool(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + float shadow = light_process_spot_shadow(light_index, vertex, view); + + shadow = blur_shadow(shadow); + + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_curve, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, + rim, + rim_tint, + albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, binormal, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, specular_light); + diffuse_light, specular_light); + } + } } - } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); + alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); #if defined(ALPHA_SCISSOR_USED) - if (alpha < alpha_scissor) { - discard; - } + if (alpha < alpha_scissor) { + discard; + } #endif // ALPHA_SCISSOR_USED #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { - discard; - } + if (alpha < opaque_prepass_threshold) { + discard; + } #endif // USE_OPAQUE_PREPASS @@ -2707,173 +3115,149 @@ FRAGMENT_SHADER_CODE #ifdef MODE_RENDER_SDF - { - vec3 local_pos = (scene_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; - ivec3 grid_pos = scene_data.sdf_offset + ivec3(local_pos * vec3(scene_data.sdf_size)); - - uint albedo16 = 0x1; //solid flag - albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; - albedo16 |= clamp(uint(albedo.g * 31.0), 0, 31) << 6; - albedo16 |= clamp(uint(albedo.b * 31.0), 0, 31) << 1; - - imageStore(albedo_volume_grid, grid_pos, uvec4(albedo16)); - - uint facing_bits = 0; - const vec3 aniso_dir[6] = vec3[]( - vec3(1, 0, 0), - vec3(0, 1, 0), - vec3(0, 0, 1), - vec3(-1, 0, 0), - vec3(0, -1, 0), - vec3(0, 0, -1)); - - vec3 cam_normal = mat3(scene_data.camera_matrix) * normalize(normal_interp); - - float closest_dist = -1e20; - - for (uint i = 0; i < 6; i++) { - float d = dot(cam_normal, aniso_dir[i]); - if (d > closest_dist) { - closest_dist = d; - facing_bits = (1 << i); + { + vec3 local_pos = (scene_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; + ivec3 grid_pos = scene_data.sdf_offset + ivec3(local_pos * vec3(scene_data.sdf_size)); + + uint albedo16 = 0x1; //solid flag + albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; + albedo16 |= clamp(uint(albedo.g * 31.0), 0, 31) << 6; + albedo16 |= clamp(uint(albedo.b * 31.0), 0, 31) << 1; + + imageStore(albedo_volume_grid, grid_pos, uvec4(albedo16)); + + uint facing_bits = 0; + const vec3 aniso_dir[6] = vec3[]( + vec3(1, 0, 0), + vec3(0, 1, 0), + vec3(0, 0, 1), + vec3(-1, 0, 0), + vec3(0, -1, 0), + vec3(0, 0, -1)); + + vec3 cam_normal = mat3(scene_data.camera_matrix) * normalize(normal_interp); + + float closest_dist = -1e20; + + for (uint i = 0; i < 6; i++) { + float d = dot(cam_normal, aniso_dir[i]); + if (d > closest_dist) { + closest_dist = d; + facing_bits = (1 << i); + } } - } - imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits + imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits - if (length(emission) > 0.001) { - float lumas[6]; - vec3 light_total = vec3(0); + if (length(emission) > 0.001) { + float lumas[6]; + vec3 light_total = vec3(0); - for (int i = 0; i < 6; i++) { - float strength = max(0.0, dot(cam_normal, aniso_dir[i])); - vec3 light = emission * strength; - light_total += light; - lumas[i] = max(light.r, max(light.g, light.b)); - } + for (int i = 0; i < 6; i++) { + float strength = max(0.0, dot(cam_normal, aniso_dir[i])); + vec3 light = emission * strength; + light_total += light; + lumas[i] = max(light.r, max(light.g, light.b)); + } - float luma_total = max(light_total.r, max(light_total.g, light_total.b)); + float luma_total = max(light_total.r, max(light_total.g, light_total.b)); - uint light_aniso = 0; + uint light_aniso = 0; - for (int i = 0; i < 6; i++) { - light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); - } + for (int i = 0; i < 6; i++) { + light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); + } - //compress to RGBE9995 to save space + //compress to RGBE9995 to save space - const float pow2to9 = 512.0f; - const float B = 15.0f; - const float N = 9.0f; - const float LN2 = 0.6931471805599453094172321215; + const float pow2to9 = 512.0f; + const float B = 15.0f; + const float N = 9.0f; + const float LN2 = 0.6931471805599453094172321215; - float cRed = clamp(light_total.r, 0.0, 65408.0); - float cGreen = clamp(light_total.g, 0.0, 65408.0); - float cBlue = clamp(light_total.b, 0.0, 65408.0); + float cRed = clamp(light_total.r, 0.0, 65408.0); + float cGreen = clamp(light_total.g, 0.0, 65408.0); + float cBlue = clamp(light_total.b, 0.0, 65408.0); - float cMax = max(cRed, max(cGreen, cBlue)); + float cMax = max(cRed, max(cGreen, cBlue)); - float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; + float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; - float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); + float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); - float exps = expp + 1.0f; + float exps = expp + 1.0f; - if (0.0 <= sMax && sMax < pow2to9) { - exps = expp; - } + if (0.0 <= sMax && sMax < pow2to9) { + exps = expp; + } - float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); - float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); - float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); - //store as 8985 to have 2 extra neighbour bits - uint light_rgbe = ((uint(sRed) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); + float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); + float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); + float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); + //store as 8985 to have 2 extra neighbour bits + uint light_rgbe = ((uint(sRed) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); - imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); - imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); + imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); + imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); + } } - } #endif #ifdef MODE_RENDER_MATERIAL - albedo_output_buffer.rgb = albedo; - albedo_output_buffer.a = alpha; + albedo_output_buffer.rgb = albedo; + albedo_output_buffer.a = alpha; - normal_output_buffer.rgb = normal * 0.5 + 0.5; - normal_output_buffer.a = 0.0; - depth_output_buffer.r = -vertex.z; + normal_output_buffer.rgb = normal * 0.5 + 0.5; + normal_output_buffer.a = 0.0; + depth_output_buffer.r = -vertex.z; -#if defined(AO_USED) - orm_output_buffer.r = ao; -#else - orm_output_buffer.r = 0.0; -#endif - orm_output_buffer.g = roughness; - orm_output_buffer.b = metallic; - orm_output_buffer.a = sss_strength; + orm_output_buffer.r = ao; + orm_output_buffer.g = roughness; + orm_output_buffer.b = metallic; + orm_output_buffer.a = sss_strength; - emission_output_buffer.rgb = emission; - emission_output_buffer.a = 0.0; + emission_output_buffer.rgb = emission; + emission_output_buffer.a = 0.0; #endif #ifdef MODE_RENDER_NORMAL_ROUGHNESS - normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); + normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); #ifdef MODE_RENDER_GIPROBE - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes - uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; - uint index2 = instances.data[instance_index].gi_offset >> 16; - giprobe_buffer.x = index1 & 0xFF; - giprobe_buffer.y = index2 & 0xFF; - } else { - giprobe_buffer.x = 0xFF; - giprobe_buffer.y = 0xFF; - } + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes + uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; + uint index2 = instances.data[instance_index].gi_offset >> 16; + giprobe_buffer.x = index1 & 0xFF; + giprobe_buffer.y = index2 & 0xFF; + } else { + giprobe_buffer.x = 0xFF; + giprobe_buffer.y = 0xFF; + } #endif -#endif //MODE_RENDER_NORMAL +#endif //MODE_RENDER_NORMAL_ROUGHNESS //nothing happens, so a tree-ssa optimizer will result in no fragment shader :) #else - specular_light *= scene_data.reflection_multiplier; - ambient_light *= albedo; //ambient must be multiplied by albedo at the end - -//ambient occlusion -#if defined(AO_USED) - -#ifndef LOW_END_MODE - if (scene_data.ssao_enabled && scene_data.ssao_ao_affect > 0.0) { - float ssao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; - ao = mix(ao, min(ao, ssao), scene_data.ssao_ao_affect); - ao_light_affect = mix(ao_light_affect, max(ao_light_affect, scene_data.ssao_light_affect), scene_data.ssao_ao_affect); - } -#endif //LOW_END_MODE - - ambient_light = mix(scene_data.ao_color.rgb, ambient_light, ao); - ao_light_affect = mix(1.0, ao, ao_light_affect); - specular_light = mix(scene_data.ao_color.rgb, specular_light, ao_light_affect); - diffuse_light = mix(scene_data.ao_color.rgb, diffuse_light, ao_light_affect); -#else - -#ifndef LOW_END_MODE - if (scene_data.ssao_enabled) { - float ao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; - ambient_light = mix(scene_data.ao_color.rgb, ambient_light, ao); - float ao_light_affect = mix(1.0, ao, scene_data.ssao_light_affect); - specular_light = mix(scene_data.ao_color.rgb, specular_light, ao_light_affect); - diffuse_light = mix(scene_data.ao_color.rgb, diffuse_light, ao_light_affect); - } -#endif //LOW_END_MODE + // multiply by albedo + diffuse_light *= albedo; // ambient must be multiplied by albedo at the end -#endif // AO_USED + // apply direct light AO + ao = unpackUnorm4x8(orms).x; + specular_light *= ao; + diffuse_light *= ao; - // base color remapping - diffuse_light *= 1.0 - metallic; // TODO: avoid all diffuse and ambient light calculations when metallic == 1 up to this point + // apply metallic + metallic = unpackUnorm4x8(orms).z; + diffuse_light *= 1.0 - metallic; ambient_light *= 1.0 - metallic; + //restore fog + fog = vec4(unpackHalf2x16(fog_rg), unpackHalf2x16(fog_ba)); + #ifdef MODE_MULTIPLE_RENDER_TARGETS #ifdef MODE_UNSHADED @@ -2889,25 +3273,8 @@ FRAGMENT_SHADER_CODE specular_buffer = vec4(specular_light, metallic); #endif - // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. - if (scene_data.fog_enabled) { - vec4 fog = fog_process(vertex); - diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); - specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); - } - -#ifndef LOW_END_MODE - if (scene_data.volumetric_fog_enabled) { - vec4 fog = volumetric_fog_process(screen_uv, -vertex.z); - diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); - specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); - } -#endif // LOW_END_MODE - -#if defined(CUSTOM_FOG_USED) - diffuse_buffer.rgb = mix(diffuse_buffer.rgb, custom_fog.rgb, custom_fog.a); - specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), custom_fog.a); -#endif //CUSTOM_FOG_USED + diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); + specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); #else //MODE_MULTIPLE_RENDER_TARGETS @@ -2919,22 +3286,10 @@ FRAGMENT_SHADER_CODE #endif //USE_NO_SHADING // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. - if (scene_data.fog_enabled) { - vec4 fog = fog_process(vertex); - frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); - } -#ifndef LOW_END_MODE - if (scene_data.volumetric_fog_enabled) { - vec4 fog = volumetric_fog_process(screen_uv, -vertex.z); - frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); - } -#endif - -#if defined(CUSTOM_FOG_USED) - frag_color.rgb = mix(frag_color.rgb, custom_fog.rgb, custom_fog.a); -#endif //CUSTOM_FOG_USED + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + ; #endif //MODE_MULTIPLE_RENDER_TARGETS #endif //MODE_RENDER_DEPTH -} + } diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl index 17ed22f58a..d78890fa9e 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl @@ -3,6 +3,15 @@ #define MAX_GI_PROBES 8 +#if defined(GL_KHR_shader_subgroup_ballot) && defined(GL_KHR_shader_subgroup_arithmetic) + +#extension GL_KHR_shader_subgroup_ballot : enable +#extension GL_KHR_shader_subgroup_arithmetic : enable + +#define USE_SUBGROUPS + +#endif + #include "cluster_data_inc.glsl" #if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(MODE_RENDER_SDF) || defined(MODE_RENDER_NORMAL_ROUGHNESS) || defined(MODE_RENDER_GIPROBE) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) @@ -13,8 +22,9 @@ layout(push_constant, binding = 0, std430) uniform DrawCall { uint instance_index; - uint pad; //16 bits minimum size - vec2 bake_uv2_offset; //used for bake to uv2, ignored otherwise + uint uv_offset; + uint pad0; + uint pad1; } draw_call; @@ -33,91 +43,13 @@ draw_call; #define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 #define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 -layout(set = 0, binding = 1) uniform sampler material_samplers[12]; - -layout(set = 0, binding = 2) uniform sampler shadow_sampler; - #define SDFGI_MAX_CASCADES 8 -layout(set = 0, binding = 3, std140) uniform SceneData { - mat4 projection_matrix; - mat4 inv_projection_matrix; - - mat4 camera_matrix; - mat4 inv_camera_matrix; - - vec2 viewport_size; - vec2 screen_pixel_size; - - //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted - vec4 directional_penumbra_shadow_kernel[32]; - vec4 directional_soft_shadow_kernel[32]; - vec4 penumbra_shadow_kernel[32]; - vec4 soft_shadow_kernel[32]; - - uint directional_penumbra_shadow_samples; - uint directional_soft_shadow_samples; - uint penumbra_shadow_samples; - uint soft_shadow_samples; - - vec4 ambient_light_color_energy; - - float ambient_color_sky_mix; - bool use_ambient_light; - bool use_ambient_cubemap; - bool use_reflection_cubemap; - - mat3 radiance_inverse_xform; - - vec2 shadow_atlas_pixel_size; - vec2 directional_shadow_pixel_size; - - uint directional_light_count; - float dual_paraboloid_side; - float z_far; - float z_near; - - bool ssao_enabled; - float ssao_light_affect; - float ssao_ao_affect; - bool roughness_limiter_enabled; - - float roughness_limiter_amount; - float roughness_limiter_limit; - uvec2 roughness_limiter_pad; - - vec4 ao_color; - - mat4 sdf_to_bounds; - - ivec3 sdf_offset; - bool material_uv2_mode; - - ivec3 sdf_size; - bool gi_upscale_for_msaa; - - bool volumetric_fog_enabled; - float volumetric_fog_inv_length; - float volumetric_fog_detail_spread; - uint volumetric_fog_pad; - - bool fog_enabled; - float fog_density; - float fog_height; - float fog_height_density; - - vec3 fog_light_color; - float fog_sun_scatter; - - float fog_aerial_perspective; - - float time; - float reflection_multiplier; // one normally, zero when rendering reflections +/* Set 1: Base Pass (never changes) */ - bool pancake_shadows; -} +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; -scene_data; +layout(set = 0, binding = 2) uniform sampler shadow_sampler; #define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6) #define INSTANCE_FLAGS_USE_SDFGI (1 << 7) @@ -134,33 +66,24 @@ scene_data; #define INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK 0x7 #define INSTANCE_FLAGS_SKELETON (1 << 19) +#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 20) -struct InstanceData { - mat4 transform; - mat4 normal_transform; - uint flags; - uint instance_uniforms_ofs; //base offset in global buffer for instance variables - uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) - uint layer_mask; - vec4 lightmap_uv_scale; -}; - -layout(set = 0, binding = 4, std430) restrict readonly buffer Instances { - InstanceData data[]; +layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { + LightData data[]; } -instances; +omni_lights; -layout(set = 0, binding = 5, std430) restrict readonly buffer Lights { +layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { LightData data[]; } -lights; +spot_lights; -layout(set = 0, binding = 6) buffer restrict readonly ReflectionProbeData { +layout(set = 0, binding = 5) buffer restrict readonly ReflectionProbeData { ReflectionData data[]; } reflections; -layout(set = 0, binding = 7, std140) uniform DirectionalLights { +layout(set = 0, binding = 6, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } directional_lights; @@ -172,40 +95,29 @@ struct Lightmap { mat3 normal_xform; }; -layout(set = 0, binding = 10, std140) restrict readonly buffer Lightmaps { +layout(set = 0, binding = 7, std140) restrict readonly buffer Lightmaps { Lightmap data[]; } lightmaps; -layout(set = 0, binding = 11) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; - struct LightmapCapture { vec4 sh[9]; }; -layout(set = 0, binding = 12, std140) restrict readonly buffer LightmapCaptures { +layout(set = 0, binding = 8, std140) restrict readonly buffer LightmapCaptures { LightmapCapture data[]; } lightmap_captures; -layout(set = 0, binding = 13) uniform texture2D decal_atlas; -layout(set = 0, binding = 14) uniform texture2D decal_atlas_srgb; +layout(set = 0, binding = 9) uniform texture2D decal_atlas; +layout(set = 0, binding = 10) uniform texture2D decal_atlas_srgb; -layout(set = 0, binding = 15, std430) restrict readonly buffer Decals { +layout(set = 0, binding = 11, std430) restrict readonly buffer Decals { DecalData data[]; } decals; -layout(set = 0, binding = 16) uniform utexture3D cluster_texture; - -layout(set = 0, binding = 17, std430) restrict readonly buffer ClusterData { - uint indices[]; -} -cluster_data; - -layout(set = 0, binding = 18) uniform texture2D directional_shadow_atlas; - -layout(set = 0, binding = 19, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 12, std430) restrict readonly buffer GlobalVariableData { vec4 data[]; } global_variables; @@ -219,7 +131,7 @@ struct SDFGIProbeCascadeData { float to_cell; // 1/bounds * grid_size }; -layout(set = 0, binding = 20, std140) uniform SDFGI { +layout(set = 0, binding = 13, std140) uniform SDFGI { vec3 grid_size; uint max_cascades; @@ -249,38 +161,140 @@ sdfgi; #endif //LOW_END_MODE -// decal atlas +/* Set 2: Render Pass (changes per render pass) */ + +layout(set = 1, binding = 0, std140) uniform SceneData { + mat4 projection_matrix; + mat4 inv_projection_matrix; + + mat4 camera_matrix; + mat4 inv_camera_matrix; + + vec2 viewport_size; + vec2 screen_pixel_size; + + uint cluster_shift; + uint cluster_width; + uint cluster_type_size; + uint max_cluster_element_count_div_32; + + //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted + vec4 directional_penumbra_shadow_kernel[32]; + vec4 directional_soft_shadow_kernel[32]; + vec4 penumbra_shadow_kernel[32]; + vec4 soft_shadow_kernel[32]; + + uint directional_penumbra_shadow_samples; + uint directional_soft_shadow_samples; + uint penumbra_shadow_samples; + uint soft_shadow_samples; + + vec4 ambient_light_color_energy; + + float ambient_color_sky_mix; + bool use_ambient_light; + bool use_ambient_cubemap; + bool use_reflection_cubemap; -/* Set 1, Radiance */ + mat3 radiance_inverse_xform; + + vec2 shadow_atlas_pixel_size; + vec2 directional_shadow_pixel_size; + + uint directional_light_count; + float dual_paraboloid_side; + float z_far; + float z_near; + + bool ssao_enabled; + float ssao_light_affect; + float ssao_ao_affect; + bool roughness_limiter_enabled; + + float roughness_limiter_amount; + float roughness_limiter_limit; + uvec2 roughness_limiter_pad; + + vec4 ao_color; + + mat4 sdf_to_bounds; + + ivec3 sdf_offset; + bool material_uv2_mode; + + ivec3 sdf_size; + bool gi_upscale_for_msaa; + + bool volumetric_fog_enabled; + float volumetric_fog_inv_length; + float volumetric_fog_detail_spread; + uint volumetric_fog_pad; + + bool fog_enabled; + float fog_density; + float fog_height; + float fog_height_density; + + vec3 fog_light_color; + float fog_sun_scatter; + + float fog_aerial_perspective; + + float time; + float reflection_multiplier; // one normally, zero when rendering reflections + + bool pancake_shadows; +} + +scene_data; + +struct InstanceData { + mat4 transform; + uint flags; + uint instance_uniforms_ofs; //base offset in global buffer for instance variables + uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) + uint layer_mask; + vec4 lightmap_uv_scale; +}; + +layout(set = 1, binding = 1, std430) buffer restrict readonly InstanceDataBuffer { + InstanceData data[]; +} +instances; #ifdef USE_RADIANCE_CUBEMAP_ARRAY -layout(set = 1, binding = 0) uniform textureCubeArray radiance_cubemap; +layout(set = 1, binding = 2) uniform textureCubeArray radiance_cubemap; #else -layout(set = 1, binding = 0) uniform textureCube radiance_cubemap; +layout(set = 1, binding = 2) uniform textureCube radiance_cubemap; #endif -/* Set 2, Reflection and Shadow Atlases (view dependent) */ +layout(set = 1, binding = 3) uniform textureCubeArray reflection_atlas; -layout(set = 1, binding = 1) uniform textureCubeArray reflection_atlas; +layout(set = 1, binding = 4) uniform texture2D shadow_atlas; -layout(set = 1, binding = 2) uniform texture2D shadow_atlas; +layout(set = 1, binding = 5) uniform texture2D directional_shadow_atlas; -#ifndef LOW_END_MODE -layout(set = 1, binding = 3) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; +layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; + +#ifndef LOW_END_MOD +layout(set = 1, binding = 7) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; #endif -/* Set 3, Render Buffers */ +layout(set = 1, binding = 8, std430) buffer restrict readonly ClusterBuffer { + uint data[]; +} +cluster_buffer; #ifdef MODE_RENDER_SDF -layout(r16ui, set = 1, binding = 4) uniform restrict writeonly uimage3D albedo_volume_grid; -layout(r32ui, set = 1, binding = 5) uniform restrict writeonly uimage3D emission_grid; -layout(r32ui, set = 1, binding = 6) uniform restrict writeonly uimage3D emission_aniso_grid; -layout(r32ui, set = 1, binding = 7) uniform restrict uimage3D geom_facing_grid; +layout(r16ui, set = 1, binding = 9) uniform restrict writeonly uimage3D albedo_volume_grid; +layout(r32ui, set = 1, binding = 10) uniform restrict writeonly uimage3D emission_grid; +layout(r32ui, set = 1, binding = 11) uniform restrict writeonly uimage3D emission_aniso_grid; +layout(r32ui, set = 1, binding = 12) uniform restrict uimage3D geom_facing_grid; //still need to be present for shaders that use it, so remap them to something #define depth_buffer shadow_atlas @@ -289,17 +303,17 @@ layout(r32ui, set = 1, binding = 7) uniform restrict uimage3D geom_facing_grid; #else -layout(set = 1, binding = 4) uniform texture2D depth_buffer; -layout(set = 1, binding = 5) uniform texture2D color_buffer; +layout(set = 1, binding = 9) uniform texture2D depth_buffer; +layout(set = 1, binding = 10) uniform texture2D color_buffer; #ifndef LOW_END_MODE -layout(set = 1, binding = 6) uniform texture2D normal_roughness_buffer; -layout(set = 1, binding = 7) uniform texture2D ao_buffer; -layout(set = 1, binding = 8) uniform texture2D ambient_buffer; -layout(set = 1, binding = 9) uniform texture2D reflection_buffer; -layout(set = 1, binding = 10) uniform texture2DArray sdfgi_lightprobe_texture; -layout(set = 1, binding = 11) uniform texture3D sdfgi_occlusion_cascades; +layout(set = 1, binding = 11) uniform texture2D normal_roughness_buffer; +layout(set = 1, binding = 12) uniform texture2D ao_buffer; +layout(set = 1, binding = 13) uniform texture2D ambient_buffer; +layout(set = 1, binding = 14) uniform texture2D reflection_buffer; +layout(set = 1, binding = 15) uniform texture2DArray sdfgi_lightprobe_texture; +layout(set = 1, binding = 16) uniform texture3D sdfgi_occlusion_cascades; struct GIProbeData { mat4 xform; @@ -317,22 +331,22 @@ struct GIProbeData { uint mipmaps; }; -layout(set = 1, binding = 12, std140) uniform GIProbes { +layout(set = 1, binding = 17, std140) uniform GIProbes { GIProbeData data[MAX_GI_PROBES]; } gi_probes; -layout(set = 1, binding = 13) uniform texture3D volumetric_fog_texture; +layout(set = 1, binding = 18) uniform texture3D volumetric_fog_texture; #endif // LOW_END_MODE #endif -/* Set 4 Skeleton & Instancing (Multimesh) */ +/* Set 2 Skeleton & Instancing (can change per item) */ layout(set = 2, binding = 0, std430) restrict readonly buffer Transforms { vec4 data[]; } transforms; -/* Set 5 User Material */ +/* Set 3 User Material */ diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl index 813ea29fa1..e4c3f3a84b 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl @@ -97,6 +97,8 @@ void main() { float blend = 0.0; #if 1 + // No interpolation + vec3 inv_dir = 1.0 / ray_dir; float rough = 0.5; @@ -161,114 +163,11 @@ void main() { hit_light *= (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); - if (blend > 0.0) { - light = mix(light, hit_light, blend); - blend = 0.0; - } else { - light = hit_light; - - //process blend - float blend_from = (float(params.probe_axis_size - 1) / 2.0) - 2.5; - float blend_to = blend_from + 2.0; - - vec3 cam_pos = params.cam_transform[3].xyz - cascades.data[i].offset; - cam_pos *= cascades.data[i].to_cell; - - pos += ray_dir * min(advance, max_advance); - vec3 inner_pos = pos - cam_pos; - - inner_pos = inner_pos * float(params.probe_axis_size - 1) / params.grid_size.x; - - float len = length(inner_pos); - - inner_pos = abs(normalize(inner_pos)); - len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z)); - - if (len >= blend_from) { - blend = smoothstep(blend_from, blend_to, len); - - pos /= cascades.data[i].to_cell; - pos += cascades.data[i].offset; - ray_pos = pos; - hit = false; //continue trace for blend - - continue; - } - } + light = hit_light; break; } - light = mix(light, vec3(0.0), blend); - -#else - - vec3 inv_dir = 1.0 / ray_dir; - - bool hit = false; - vec4 light_accum = vec4(0.0); - - float blend_size = (params.grid_size.x / float(params.probe_axis_size - 1)) * 0.5; - - float radius_sizes[MAX_CASCADES]; - for (uint i = 0; i < params.max_cascades; i++) { - radius_sizes[i] = (1.0 / cascades.data[i].to_cell) * (params.grid_size.x * 0.5 - blend_size); - } - - float max_distance = radius_sizes[params.max_cascades - 1]; - float advance = 0; - while (advance < max_distance) { - for (uint i = 0; i < params.max_cascades; i++) { - if (advance < radius_sizes[i]) { - vec3 pos = (ray_pos + ray_dir * advance) - cascades.data[i].offset; - pos *= cascades.data[i].to_cell * pos_to_uvw; - - float distance = texture(sampler3D(sdf_cascades[i], linear_sampler), pos).r * 255.0 - 1.0; - - vec4 hit_light = vec4(0.0); - if (distance < 1.0) { - hit_light.a = max(0.0, 1.0 - distance); - hit_light.rgb = texture(sampler3D(light_cascades[i], linear_sampler), pos).rgb; - hit_light.rgb *= hit_light.a; - } - - distance /= cascades.data[i].to_cell; - - if (i < (params.max_cascades - 1)) { - pos = (ray_pos + ray_dir * advance) - cascades.data[i + 1].offset; - pos *= cascades.data[i + 1].to_cell * pos_to_uvw; - - float distance2 = texture(sampler3D(sdf_cascades[i + 1], linear_sampler), pos).r * 255.0 - 1.0; - - vec4 hit_light2 = vec4(0.0); - if (distance2 < 1.0) { - hit_light2.a = max(0.0, 1.0 - distance2); - hit_light2.rgb = texture(sampler3D(light_cascades[i + 1], linear_sampler), pos).rgb; - hit_light2.rgb *= hit_light2.a; - } - - float prev_radius = i == 0 ? 0.0 : radius_sizes[i - 1]; - float blend = (advance - prev_radius) / (radius_sizes[i] - prev_radius); - - distance2 /= cascades.data[i + 1].to_cell; - - hit_light = mix(hit_light, hit_light2, blend); - distance = mix(distance, distance2, blend); - } - - light_accum += hit_light; - advance += distance; - break; - } - } - - if (light_accum.a > 0.98) { - break; - } - } - - light = light_accum.rgb / light_accum.a; - #endif imageStore(screen_buffer, screen_pos, vec4(linear_to_srgb(light), 1.0)); diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl index 61e4bf5e18..5e8934adb4 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl @@ -67,8 +67,8 @@ struct Light { float attenuation; uint type; - float spot_angle; - float spot_attenuation; + float cos_spot_angle; + float inv_spot_attenuation; float radius; vec4 shadow_color; @@ -112,11 +112,23 @@ vec2 octahedron_encode(vec3 n) { return n.xy; } +float get_omni_attenuation(float distance, float inv_range, float decay) { + float nd = distance * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(distance, 0.0001), -decay); +} + void main() { uint voxel_index = uint(gl_GlobalInvocationID.x); //used for skipping voxels every N frames - voxel_index = params.process_offset + voxel_index * params.process_increment; + if (params.process_increment > 1) { + voxel_index *= params.process_increment; + voxel_index += params.process_offset; + } if (voxel_index >= dispatch_data.total_count) { return; @@ -134,10 +146,78 @@ void main() { uint voxel_albedo = process_voxels.data[voxel_index].albedo; vec3 albedo = vec3(uvec3(voxel_albedo >> 10, voxel_albedo >> 5, voxel_albedo) & uvec3(0x1F)) / float(0x1F); - vec3 light_accum[6]; - + vec3 light_accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); uint valid_aniso = (voxel_albedo >> 15) & 0x3F; + const vec3 aniso_dir[6] = vec3[]( + vec3(1, 0, 0), + vec3(0, 1, 0), + vec3(0, 0, 1), + vec3(-1, 0, 0), + vec3(0, -1, 0), + vec3(0, 0, -1)); + + // Add indirect light first, in order to save computation resources +#ifdef MODE_PROCESS_DYNAMIC + if (params.multibounce) { + vec3 pos = (vec3(positioni) + vec3(0.5)) * float(params.probe_axis_size - 1) / params.grid_size; + ivec3 probe_base_pos = ivec3(pos); + + float weight_accum[6] = float[](0, 0, 0, 0, 0, 0); + + ivec3 tex_pos = ivec3(probe_base_pos.xy, int(params.cascade)); + tex_pos.x += probe_base_pos.z * int(params.probe_axis_size); + + tex_pos.xy = tex_pos.xy * (OCT_SIZE + 2) + ivec2(1); + + vec3 base_tex_posf = vec3(tex_pos); + vec2 tex_pixel_size = 1.0 / vec2(ivec2((OCT_SIZE + 2) * params.probe_axis_size * params.probe_axis_size, (OCT_SIZE + 2) * params.probe_axis_size)); + vec3 probe_uv_offset = (ivec3(OCT_SIZE + 2, OCT_SIZE + 2, (OCT_SIZE + 2) * params.probe_axis_size)) * tex_pixel_size.xyx; + + for (uint j = 0; j < 8; j++) { + ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); + ivec3 probe_posi = probe_base_pos; + probe_posi += offset; + + // Compute weight + + vec3 probe_pos = vec3(probe_posi); + vec3 probe_to_pos = pos - probe_pos; + vec3 probe_dir = normalize(-probe_to_pos); + + // Compute lightprobe texture position + + vec3 trilinear = vec3(1.0) - abs(probe_to_pos); + + for (uint k = 0; k < 6; k++) { + if (bool(valid_aniso & (1 << k))) { + vec3 n = aniso_dir[k]; + float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(n, probe_dir)); + + vec3 tex_posf = base_tex_posf + vec3(octahedron_encode(n) * float(OCT_SIZE), 0.0); + tex_posf.xy *= tex_pixel_size; + + vec3 pos_uvw = tex_posf; + pos_uvw.xy += vec2(offset.xy) * probe_uv_offset.xy; + pos_uvw.x += float(offset.z) * probe_uv_offset.z; + vec3 indirect_light = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb; + + light_accum[k] += indirect_light * weight; + weight_accum[k] += weight; + } + } + } + + for (uint k = 0; k < 6; k++) { + if (weight_accum[k] > 0.0) { + light_accum[k] /= weight_accum[k]; + light_accum[k] *= albedo; + } + } + } + +#endif + { uint rgbe = process_voxels.data[voxel_index].light; @@ -153,18 +233,10 @@ void main() { uint aniso = process_voxels.data[voxel_index].light_aniso; for (uint i = 0; i < 6; i++) { float strength = ((aniso >> (i * 5)) & 0x1F) / float(0x1F); - light_accum[i] = l * strength; + light_accum[i] += l * strength; } } - const vec3 aniso_dir[6] = vec3[]( - vec3(1, 0, 0), - vec3(0, 1, 0), - vec3(0, 0, 1), - vec3(-1, 0, 0), - vec3(0, -1, 0), - vec3(0, 0, -1)); - // Raytrace light vec3 pos_to_uvw = 1.0 / params.grid_size; @@ -184,22 +256,26 @@ void main() { direction = normalize(rel_vec); light_distance = length(rel_vec); rel_vec.y /= params.y_mult; - attenuation = pow(clamp(1.0 - length(rel_vec) / lights.data[i].radius, 0.0, 1.0), lights.data[i].attenuation); + attenuation = get_omni_attenuation(light_distance, 1.0 / lights.data[i].radius, lights.data[i].attenuation); + } break; case LIGHT_TYPE_SPOT: { vec3 rel_vec = lights.data[i].position - position; direction = normalize(rel_vec); light_distance = length(rel_vec); rel_vec.y /= params.y_mult; - attenuation = pow(clamp(1.0 - length(rel_vec) / lights.data[i].radius, 0.0, 1.0), lights.data[i].attenuation); - - float angle = acos(dot(normalize(rel_vec), -lights.data[i].direction)); - if (angle > lights.data[i].spot_angle) { - attenuation = 0.0; - } else { - float d = clamp(angle / lights.data[i].spot_angle, 0, 1); - attenuation *= pow(1.0 - d, lights.data[i].spot_attenuation); + attenuation = get_omni_attenuation(light_distance, 1.0 / lights.data[i].radius, lights.data[i].attenuation); + + float cos_spot_angle = lights.data[i].cos_spot_angle; + float cos_angle = dot(-direction, lights.data[i].direction); + + if (cos_angle < cos_spot_angle) { + continue; } + + float scos = max(cos_angle, cos_spot_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cos_spot_angle)); + attenuation *= 1.0 - pow(spot_rim, lights.data[i].inv_spot_attenuation); } break; } @@ -282,65 +358,6 @@ void main() { } } - // Add indirect light - - if (params.multibounce) { - vec3 pos = (vec3(positioni) + vec3(0.5)) * float(params.probe_axis_size - 1) / params.grid_size; - ivec3 probe_base_pos = ivec3(pos); - - vec4 probe_accum[6] = vec4[](vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0)); - float weight_accum[6] = float[](0, 0, 0, 0, 0, 0); - - ivec3 tex_pos = ivec3(probe_base_pos.xy, int(params.cascade)); - tex_pos.x += probe_base_pos.z * int(params.probe_axis_size); - - tex_pos.xy = tex_pos.xy * (OCT_SIZE + 2) + ivec2(1); - - vec3 base_tex_posf = vec3(tex_pos); - vec2 tex_pixel_size = 1.0 / vec2(ivec2((OCT_SIZE + 2) * params.probe_axis_size * params.probe_axis_size, (OCT_SIZE + 2) * params.probe_axis_size)); - vec3 probe_uv_offset = (ivec3(OCT_SIZE + 2, OCT_SIZE + 2, (OCT_SIZE + 2) * params.probe_axis_size)) * tex_pixel_size.xyx; - - for (uint j = 0; j < 8; j++) { - ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); - ivec3 probe_posi = probe_base_pos; - probe_posi += offset; - - // Compute weight - - vec3 probe_pos = vec3(probe_posi); - vec3 probe_to_pos = pos - probe_pos; - vec3 probe_dir = normalize(-probe_to_pos); - - // Compute lightprobe texture position - - vec3 trilinear = vec3(1.0) - abs(probe_to_pos); - - for (uint k = 0; k < 6; k++) { - if (bool(valid_aniso & (1 << k))) { - vec3 n = aniso_dir[k]; - float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(n, probe_dir)); - - vec3 tex_posf = base_tex_posf + vec3(octahedron_encode(n) * float(OCT_SIZE), 0.0); - tex_posf.xy *= tex_pixel_size; - - vec3 pos_uvw = tex_posf; - pos_uvw.xy += vec2(offset.xy) * probe_uv_offset.xy; - pos_uvw.x += float(offset.z) * probe_uv_offset.z; - vec4 indirect_light = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0); - - probe_accum[k] += indirect_light * weight; - weight_accum[k] += weight; - } - } - } - - for (uint k = 0; k < 6; k++) { - if (weight_accum[k] > 0.0) { - light_accum[k] += probe_accum[k].rgb * albedo / weight_accum[k]; - } - } - } - // Store the light in the light texture float lumas[6]; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl index d516ab22c3..007e4c113a 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl @@ -39,8 +39,11 @@ layout(rgba32i, set = 0, binding = 13) uniform restrict iimage2D lightprobe_aver layout(rgba16f, set = 0, binding = 14) uniform restrict writeonly image2DArray lightprobe_ambient_texture; +#ifdef USE_CUBEMAP_ARRAY +layout(set = 1, binding = 0) uniform textureCubeArray sky_irradiance; +#else layout(set = 1, binding = 0) uniform textureCube sky_irradiance; - +#endif layout(set = 1, binding = 1) uniform sampler linear_sampler_mipmaps; #define HISTORY_BITS 10 @@ -136,12 +139,24 @@ uint rgbe_encode(vec3 color) { return (uint(sRed) & 0x1FF) | ((uint(sGreen) & 0x1FF) << 9) | ((uint(sBlue) & 0x1FF) << 18) | ((uint(exps) & 0x1F) << 27); } +struct SH { +#if (SH_SIZE == 16) + float c[48]; +#else + float c[28]; +#endif +}; + +shared SH sh_accum[64]; //8x8 + void main() { ivec2 pos = ivec2(gl_GlobalInvocationID.xy); if (any(greaterThanEqual(pos, params.image_size))) { //too large, do nothing return; } + uint probe_index = gl_LocalInvocationID.x + gl_LocalInvocationID.y * 8; + #ifdef MODE_PROCESS float probe_cell_size = float(params.grid_size.x / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; @@ -154,27 +169,9 @@ void main() { vec3 probe_pos = cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size; vec3 pos_to_uvw = 1.0 / params.grid_size; - vec4 probe_sh_accum[SH_SIZE] = vec4[]( - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0) -#if (SH_SIZE == 16) - , - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0) -#endif - ); + for (uint i = 0; i < SH_SIZE * 3; i++) { + sh_accum[probe_index].c[i] = 0.0; + } // quickly ensure each probe has a different "offset" for the vogel function, based on integer world position uvec3 h3 = hash3(uvec3(params.world_offset + probe_cell)); @@ -195,14 +192,12 @@ void main() { vec3 inv_dir = 1.0 / ray_dir; bool hit = false; - vec3 hit_normal; - vec3 hit_light; - vec3 hit_aniso0; - vec3 hit_aniso1; + uint hit_cascade; float bias = params.ray_bias; vec3 abs_ray_dir = abs(ray_dir); ray_pos += ray_dir * 1.0 / max(abs_ray_dir.x, max(abs_ray_dir.y, abs_ray_dir.z)) * bias / cascades.data[params.cascade].to_cell; + vec3 uvw; for (uint j = params.cascade; j < params.max_cascades; j++) { //convert to local bounds @@ -221,14 +216,12 @@ void main() { float advance = 0.0; - vec3 uvw; - while (advance < max_advance) { //read how much to advance from SDF uvw = (pos + ray_dir * advance) * pos_to_uvw; float distance = texture(sampler3D(sdf_cascades[j], linear_sampler), uvw).r * 255.0 - 1.0; - if (distance < 0.001) { + if (distance < 0.05) { //consider hit hit = true; break; @@ -238,17 +231,7 @@ void main() { } if (hit) { - const float EPSILON = 0.001; - hit_normal = normalize(vec3( - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r, - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r, - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r)); - - hit_light = texture(sampler3D(light_cascades[j], linear_sampler), uvw).rgb; - vec4 aniso0 = texture(sampler3D(aniso0_cascades[j], linear_sampler), uvw); - hit_aniso0 = aniso0.rgb; - hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[j], linear_sampler), uvw).rg); - + hit_cascade = j; break; } @@ -261,11 +244,32 @@ void main() { vec4 light; if (hit) { - //one liner magic - light.rgb = hit_light * (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); - light.a = 1.0; + //avoid reading different texture from different threads + for (uint j = params.cascade; j < params.max_cascades; j++) { + if (j == hit_cascade) { + const float EPSILON = 0.001; + vec3 hit_normal = normalize(vec3( + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r, + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r, + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r)); + + vec3 hit_light = texture(sampler3D(light_cascades[hit_cascade], linear_sampler), uvw).rgb; + vec4 aniso0 = texture(sampler3D(aniso0_cascades[hit_cascade], linear_sampler), uvw); + vec3 hit_aniso0 = aniso0.rgb; + vec3 hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[hit_cascade], linear_sampler), uvw).rg); + + //one liner magic + light.rgb = hit_light * (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); + light.a = 1.0; + } + } + } else if (params.sky_mode == SKY_MODE_SKY) { +#ifdef USE_CUBEMAP_ARRAY + light.rgb = textureLod(samplerCubeArray(sky_irradiance, linear_sampler_mipmaps), vec4(ray_dir, 0.0), 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates +#else light.rgb = textureLod(samplerCube(sky_irradiance, linear_sampler_mipmaps), ray_dir, 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates +#endif light.rgb *= params.sky_energy; light.a = 0.0; @@ -278,33 +282,33 @@ void main() { } vec3 ray_dir2 = ray_dir * ray_dir; - float c[SH_SIZE] = float[]( - - 0.282095, //l0 - 0.488603 * ray_dir.y, //l1n1 - 0.488603 * ray_dir.z, //l1n0 - 0.488603 * ray_dir.x, //l1p1 - 1.092548 * ray_dir.x * ray_dir.y, //l2n2 - 1.092548 * ray_dir.y * ray_dir.z, //l2n1 - 0.315392 * (3.0 * ray_dir2.z - 1.0), //l20 - 1.092548 * ray_dir.x * ray_dir.z, //l2p1 - 0.546274 * (ray_dir2.x - ray_dir2.y) //l2p2 + +#define SH_ACCUM(m_idx, m_value) \ + { \ + vec3 l = light.rgb * (m_value); \ + sh_accum[probe_index].c[m_idx * 3 + 0] += l.r; \ + sh_accum[probe_index].c[m_idx * 3 + 1] += l.g; \ + sh_accum[probe_index].c[m_idx * 3 + 2] += l.b; \ + } + SH_ACCUM(0, 0.282095); //l0 + SH_ACCUM(1, 0.488603 * ray_dir.y); //l1n1 + SH_ACCUM(2, 0.488603 * ray_dir.z); //l1n0 + SH_ACCUM(3, 0.488603 * ray_dir.x); //l1p1 + SH_ACCUM(4, 1.092548 * ray_dir.x * ray_dir.y); //l2n2 + SH_ACCUM(5, 1.092548 * ray_dir.y * ray_dir.z); //l2n1 + SH_ACCUM(6, 0.315392 * (3.0 * ray_dir2.z - 1.0)); //l20 + SH_ACCUM(7, 1.092548 * ray_dir.x * ray_dir.z); //l2p1 + SH_ACCUM(8, 0.546274 * (ray_dir2.x - ray_dir2.y)); //l2p2 #if (SH_SIZE == 16) - , - 0.590043 * ray_dir.y * (3.0f * ray_dir2.x - ray_dir2.y), - 2.890611 * ray_dir.y * ray_dir.x * ray_dir.z, - 0.646360 * ray_dir.y * (-1.0f + 5.0f * ray_dir2.z), - 0.373176 * (5.0f * ray_dir2.z * ray_dir.z - 3.0f * ray_dir.z), - 0.457045 * ray_dir.x * (-1.0f + 5.0f * ray_dir2.z), - 1.445305 * (ray_dir2.x - ray_dir2.y) * ray_dir.z, - 0.590043 * ray_dir.x * (ray_dir2.x - 3.0f * ray_dir2.y) + SH_ACCUM(9, 0.590043 * ray_dir.y * (3.0f * ray_dir2.x - ray_dir2.y)); + SH_ACCUM(10, 2.890611 * ray_dir.y * ray_dir.x * ray_dir.z); + SH_ACCUM(11, 0.646360 * ray_dir.y * (-1.0f + 5.0f * ray_dir2.z)); + SH_ACCUM(12, 0.373176 * (5.0f * ray_dir2.z * ray_dir.z - 3.0f * ray_dir.z)); + SH_ACCUM(13, 0.457045 * ray_dir.x * (-1.0f + 5.0f * ray_dir2.z)); + SH_ACCUM(14, 1.445305 * (ray_dir2.x - ray_dir2.y) * ray_dir.z); + SH_ACCUM(15, 0.590043 * ray_dir.x * (ray_dir2.x - 3.0f * ray_dir2.y)); #endif - ); - - for (uint j = 0; j < SH_SIZE; j++) { - probe_sh_accum[j] += light * c[j]; - } } for (uint i = 0; i < SH_SIZE; i++) { @@ -312,7 +316,7 @@ void main() { ivec3 prev_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(params.history_index)); ivec2 average_pos = prev_pos.xy; - vec4 value = probe_sh_accum[i] * 4.0 / float(params.ray_count); + vec4 value = vec4(sh_accum[probe_index].c[i * 3 + 0], sh_accum[probe_index].c[i * 3 + 1], sh_accum[probe_index].c[i * 3 + 2], 1.0) * 4.0 / float(params.ray_count); ivec4 ivalue = clamp(ivec4(value * float(1 << HISTORY_BITS)), -32768, 32767); //clamp to 16 bits, so higher values don't break average @@ -344,37 +348,11 @@ void main() { ivec2 oct_pos = (pos / OCT_SIZE) * (OCT_SIZE + 2) + ivec2(1); ivec2 local_pos = pos % OCT_SIZE; - //fill the spherical harmonic - vec4 sh[SH_SIZE]; - - for (uint i = 0; i < SH_SIZE; i++) { - // store in history texture - ivec2 average_pos = sh_pos + ivec2(0, i); - ivec4 average = imageLoad(lightprobe_average_texture, average_pos); - - sh[i] = (vec4(average) / float(params.history_size)) / float(1 << HISTORY_BITS); - } - //compute the octahedral normal for this texel vec3 normal = octahedron_encode(vec2(local_pos) / float(OCT_SIZE)); - /* + // read the spherical harmonic - const float c1 = 0.429043; - const float c2 = 0.511664; - const float c3 = 0.743125; - const float c4 = 0.886227; - const float c5 = 0.247708; - vec4 light = (c1 * sh[8] * (normal.x * normal.x - normal.y * normal.y) + - c3 * sh[6] * normal.z * normal.z + - c4 * sh[0] - - c5 * sh[6] + - 2.0 * c1 * sh[4] * normal.x * normal.y + - 2.0 * c1 * sh[7] * normal.x * normal.z + - 2.0 * c1 * sh[5] * normal.y * normal.z + - 2.0 * c2 * sh[3] * normal.x + - 2.0 * c2 * sh[1] * normal.y + - 2.0 * c2 * sh[2] * normal.z); -*/ + vec3 normal2 = normal * normal; float c[SH_SIZE] = float[]( @@ -426,7 +404,14 @@ void main() { vec3 radiance = vec3(0.0); for (uint i = 0; i < SH_SIZE; i++) { - vec3 m = sh[i].rgb * c[i] * 4.0; + // store in history texture + ivec2 average_pos = sh_pos + ivec2(0, i); + ivec4 average = imageLoad(lightprobe_average_texture, average_pos); + + vec4 sh = (vec4(average) / float(params.history_size)) / float(1 << HISTORY_BITS); + + vec3 m = sh.rgb * c[i] * 4.0; + irradiance += m * l_mult[i]; radiance += m; } @@ -515,13 +500,15 @@ void main() { //can't scroll, must look for position in parent cascade //to global coords - float probe_cell_size = float(params.grid_size.x / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; + float cell_to_probe = float(params.grid_size.x / float(params.probe_axis_size - 1)); + + float probe_cell_size = cell_to_probe / cascades.data[params.cascade].to_cell; vec3 probe_pos = cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size; //to parent local coords + float probe_cell_size_next = cell_to_probe / cascades.data[params.cascade + 1].to_cell; probe_pos -= cascades.data[params.cascade + 1].offset; - probe_pos *= cascades.data[params.cascade + 1].to_cell; - probe_pos = probe_pos * float(params.probe_axis_size - 1) / float(params.grid_size.x); + probe_pos /= probe_cell_size_next; ivec3 probe_posi = ivec3(probe_pos); //add up all light, no need to use occlusion here, since occlusion will do its work afterwards @@ -574,20 +561,28 @@ void main() { } } else { - // clear and let it re-raytrace, only for the last cascade, which happens very un-often - //scroll + //scroll at the edge of the highest cascade, just copy what is there, + //since its the closest we have anyway + for (uint j = 0; j < params.history_size; j++) { + ivec2 tex_pos; + tex_pos = probe_cell.xy; + tex_pos.x += probe_cell.z * int(params.probe_axis_size); + for (int i = 0; i < SH_SIZE; i++) { // copy from history texture + ivec3 src_pos = ivec3(tex_pos.x, tex_pos.y * SH_SIZE + i, int(j)); ivec3 dst_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(j)); - imageStore(lightprobe_history_scroll_texture, dst_pos, ivec4(0)); + ivec4 value = imageLoad(lightprobe_history_texture, dst_pos); + imageStore(lightprobe_history_scroll_texture, dst_pos, value); } } for (int i = 0; i < SH_SIZE; i++) { // copy from average texture - ivec2 dst_pos = ivec2(pos.x, pos.y * SH_SIZE + i); - imageStore(lightprobe_average_scroll_texture, dst_pos, ivec4(0)); + ivec2 spos = ivec2(pos.x, pos.y * SH_SIZE + i); + ivec4 average = imageLoad(lightprobe_average_texture, spos); + imageStore(lightprobe_average_scroll_texture, spos, average); } } diff --git a/servers/rendering/renderer_rd/shaders/shadow_reduce.glsl b/servers/rendering/renderer_rd/shaders/shadow_reduce.glsl deleted file mode 100644 index 29443ae7db..0000000000 --- a/servers/rendering/renderer_rd/shaders/shadow_reduce.glsl +++ /dev/null @@ -1,105 +0,0 @@ -#[compute] - -#version 450 - -VERSION_DEFINES - -#define BLOCK_SIZE 8 - -layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in; - -#ifdef MODE_REDUCE - -shared float tmp_data[BLOCK_SIZE * BLOCK_SIZE]; -const uint swizzle_table[BLOCK_SIZE] = uint[](0, 4, 2, 6, 1, 5, 3, 7); -const uint unswizzle_table[BLOCK_SIZE] = uint[](0, 0, 0, 1, 0, 2, 1, 3); - -#endif - -layout(r32f, set = 0, binding = 0) uniform restrict readonly image2D source_depth; -layout(r32f, set = 0, binding = 1) uniform restrict writeonly image2D dst_depth; - -layout(push_constant, binding = 1, std430) uniform Params { - ivec2 source_size; - ivec2 source_offset; - uint min_size; - uint gaussian_kernel_version; - ivec2 filter_dir; -} -params; - -void main() { -#ifdef MODE_REDUCE - - uvec2 pos = gl_LocalInvocationID.xy; - - ivec2 image_offset = params.source_offset; - ivec2 image_pos = image_offset + ivec2(gl_GlobalInvocationID.xy); - uint dst_t = swizzle_table[pos.y] * BLOCK_SIZE + swizzle_table[pos.x]; - tmp_data[dst_t] = imageLoad(source_depth, min(image_pos, params.source_size - ivec2(1))).r; - ivec2 image_size = params.source_size; - - uint t = pos.y * BLOCK_SIZE + pos.x; - - //neighbours - uint size = BLOCK_SIZE; - - do { - groupMemoryBarrier(); - barrier(); - - size >>= 1; - image_size >>= 1; - image_offset >>= 1; - - if (all(lessThan(pos, uvec2(size)))) { - uint nx = t + size; - uint ny = t + (BLOCK_SIZE * size); - uint nxy = ny + size; - - tmp_data[t] += tmp_data[nx]; - tmp_data[t] += tmp_data[ny]; - tmp_data[t] += tmp_data[nxy]; - tmp_data[t] /= 4.0; - } - - } while (size > params.min_size); - - if (all(lessThan(pos, uvec2(size)))) { - image_pos = ivec2(unswizzle_table[size + pos.x], unswizzle_table[size + pos.y]); - image_pos += image_offset + ivec2(gl_WorkGroupID.xy) * int(size); - - image_size = max(ivec2(1), image_size); //in case image size became 0 - - if (all(lessThan(image_pos, uvec2(image_size)))) { - imageStore(dst_depth, image_pos, vec4(tmp_data[t])); - } - } -#endif - -#ifdef MODE_FILTER - - ivec2 image_pos = params.source_offset + ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(image_pos, params.source_size))) { - return; - } - - ivec2 clamp_min = ivec2(params.source_offset); - ivec2 clamp_max = ivec2(params.source_size) - 1; - - //gaussian kernel, size 9, sigma 4 - const int kernel_size = 9; - const float gaussian_kernel[kernel_size * 3] = float[]( - 0.000229, 0.005977, 0.060598, 0.241732, 0.382928, 0.241732, 0.060598, 0.005977, 0.000229, - 0.028532, 0.067234, 0.124009, 0.179044, 0.20236, 0.179044, 0.124009, 0.067234, 0.028532, - 0.081812, 0.101701, 0.118804, 0.130417, 0.134535, 0.130417, 0.118804, 0.101701, 0.081812); - float accum = 0.0; - for (int i = 0; i < kernel_size; i++) { - ivec2 ofs = clamp(image_pos + params.filter_dir * (i - kernel_size / 2), clamp_min, clamp_max); - accum += imageLoad(source_depth, ofs).r * gaussian_kernel[params.gaussian_kernel_version + i]; - } - - imageStore(dst_depth, image_pos, vec4(accum)); - -#endif -} diff --git a/servers/rendering/renderer_rd/shaders/ssao.glsl b/servers/rendering/renderer_rd/shaders/ssao.glsl index 315ef8fa13..231f8f91ec 100644 --- a/servers/rendering/renderer_rd/shaders/ssao.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao.glsl @@ -88,7 +88,7 @@ counter; layout(rg8, set = 2, binding = 0) uniform restrict writeonly image2D dest_image; // This push_constant is full - 128 bytes - if you need to add more data, consider adding to the uniform buffer instead -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, binding = 3, std430) uniform Params { ivec2 screen_size; int pass; int quality; diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl index 13b162f0c9..e7ba8feb80 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl @@ -4,6 +4,15 @@ VERSION_DEFINES +/* Do not use subgroups here, seems there is not much advantage and causes glitches +#extension GL_KHR_shader_subgroup_ballot: enable +#extension GL_KHR_shader_subgroup_arithmetic: enable + +#if defined(GL_KHR_shader_subgroup_ballot) && defined(GL_KHR_shader_subgroup_arithmetic) +#define USE_SUBGROUPS +#endif +*/ + #if defined(MODE_FOG) || defined(MODE_FILTER) layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -23,22 +32,25 @@ layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; layout(set = 0, binding = 1) uniform texture2D shadow_atlas; layout(set = 0, binding = 2) uniform texture2D directional_shadow_atlas; -layout(set = 0, binding = 3, std430) restrict readonly buffer Lights { +layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { LightData data[]; } -lights; +omni_lights; -layout(set = 0, binding = 4, std140) uniform DirectionalLights { +layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { + LightData data[]; +} +spot_lights; + +layout(set = 0, binding = 5, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } directional_lights; -layout(set = 0, binding = 5) uniform utexture3D cluster_texture; - -layout(set = 0, binding = 6, std430) restrict readonly buffer ClusterData { - uint indices[]; +layout(set = 0, binding = 6, std430) buffer restrict readonly ClusterBuffer { + uint data[]; } -cluster_data; +cluster_buffer; layout(set = 0, binding = 7) uniform sampler linear_sampler; @@ -132,7 +144,7 @@ layout(set = 1, binding = 2) uniform texture3D sdfgi_occlusion_texture; #endif //SDFGI -layout(push_constant, binding = 0, std430) uniform Params { +layout(set = 0, binding = 14, std140) uniform Params { vec2 fog_frustum_size_begin; vec2 fog_frustum_size_end; @@ -150,12 +162,24 @@ layout(push_constant, binding = 0, std430) uniform Params { float detail_spread; float gi_inject; uint max_gi_probes; - uint pad; + uint cluster_type_size; + + vec2 screen_size; + uint cluster_shift; + uint cluster_width; + + uint max_cluster_element_count_div_32; + bool use_temporal_reprojection; + uint temporal_frame; + float temporal_blend; mat3x4 cam_rotation; + mat4 to_prev_view; } params; +layout(set = 0, binding = 15) uniform texture3D prev_density_texture; + float get_depth_at_pos(float cell_depth_size, int z) { float d = float(z) * cell_depth_size + cell_depth_size * 0.5; //center of voxels d = pow(d, params.detail_spread); @@ -169,6 +193,51 @@ vec3 hash3f(uvec3 x) { return vec3(x & 0xFFFFF) / vec3(float(0xFFFFF)); } +float get_omni_attenuation(float distance, float inv_range, float decay) { + float nd = distance * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(distance, 0.0001), -decay); +} + +void cluster_get_item_range(uint p_offset, out uint item_min, out uint item_max, out uint item_from, out uint item_to) { + uint item_min_max = cluster_buffer.data[p_offset]; + item_min = item_min_max & 0xFFFF; + item_max = item_min_max >> 16; + ; + + item_from = item_min >> 5; + item_to = (item_max == 0) ? 0 : ((item_max - 1) >> 5) + 1; //side effect of how it is stored, as item_max 0 means no elements +} + +uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { + int local_min = clamp(int(z_min) - int(i) * 32, 0, 31); + int mask_width = min(int(z_max) - int(z_min), 32 - local_min); + return bitfieldInsert(uint(0), uint(0xFFFFFFFF), local_min, mask_width); +} + +#define TEMPORAL_FRAMES 16 + +const vec3 halton_map[TEMPORAL_FRAMES] = vec3[]( + vec3(0.5, 0.33333333, 0.2), + vec3(0.25, 0.66666667, 0.4), + vec3(0.75, 0.11111111, 0.6), + vec3(0.125, 0.44444444, 0.8), + vec3(0.625, 0.77777778, 0.04), + vec3(0.375, 0.22222222, 0.24), + vec3(0.875, 0.55555556, 0.44), + vec3(0.0625, 0.88888889, 0.64), + vec3(0.5625, 0.03703704, 0.84), + vec3(0.3125, 0.37037037, 0.08), + vec3(0.8125, 0.7037037, 0.28), + vec3(0.1875, 0.14814815, 0.48), + vec3(0.6875, 0.48148148, 0.68), + vec3(0.4375, 0.81481481, 0.88), + vec3(0.9375, 0.25925926, 0.12), + vec3(0.03125, 0.59259259, 0.32)); + void main() { vec3 fog_cell_size = 1.0 / vec3(params.fog_volume_size); @@ -184,6 +253,12 @@ void main() { //posf += mix(vec3(0.0),vec3(1.0),0.3) * hash3f(uvec3(pos)) * 2.0 - 1.0; vec3 fog_unit_pos = posf * fog_cell_size + fog_cell_size * 0.5; //center of voxels + + uvec2 screen_pos = uvec2(fog_unit_pos.xy * params.screen_size); + uvec2 cluster_pos = screen_pos >> params.cluster_shift; + uint cluster_offset = (params.cluster_width * cluster_pos.y + cluster_pos.x) * (params.max_cluster_element_count_div_32 + 32); + //positions in screen are too spread apart, no hopes for optimizing with subgroups + fog_unit_pos.z = pow(fog_unit_pos.z, params.detail_spread); vec3 view_pos; @@ -191,6 +266,47 @@ void main() { view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; view_pos.y = -view_pos.y; + vec4 reprojected_density = vec4(0.0); + float reproject_amount = 0.0; + + if (params.use_temporal_reprojection) { + vec3 prev_view = (params.to_prev_view * vec4(view_pos, 1.0)).xyz; + //undo transform into prev view + prev_view.y = -prev_view.y; + //z back to unit size + prev_view.z /= -params.fog_frustum_end; + //xy back to unit size + prev_view.xy /= mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(prev_view.z)); + prev_view.xy = prev_view.xy * 0.5 + 0.5; + //z back to unspread value + prev_view.z = pow(prev_view.z, 1.0 / params.detail_spread); + + if (all(greaterThan(prev_view, vec3(0.0))) && all(lessThan(prev_view, vec3(1.0)))) { + //reprojectinon fits + + reprojected_density = textureLod(sampler3D(prev_density_texture, linear_sampler), prev_view, 0.0); + reproject_amount = params.temporal_blend; + + // Since we can reproject, now we must jitter the current view pos. + // This is done here because cells that can't reproject should not jitter. + + fog_unit_pos = posf * fog_cell_size + fog_cell_size * halton_map[params.temporal_frame]; //center of voxels, offset by halton table + + screen_pos = uvec2(fog_unit_pos.xy * params.screen_size); + cluster_pos = screen_pos >> params.cluster_shift; + cluster_offset = (params.cluster_width * cluster_pos.y + cluster_pos.x) * (params.max_cluster_element_count_div_32 + 32); + //positions in screen are too spread apart, no hopes for optimizing with subgroups + + fog_unit_pos.z = pow(fog_unit_pos.z, params.detail_spread); + + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; + view_pos.y = -view_pos.y; + } + } + + uint cluster_z = uint(clamp((abs(view_pos.z) / params.z_far) * 32.0, 0.0, 31.0)); + vec3 total_light = params.light_color; float total_density = params.base_density; @@ -257,108 +373,160 @@ void main() { //compute lights from cluster - vec3 cluster_pos; - cluster_pos.xy = fog_unit_pos.xy; - cluster_pos.z = clamp((abs(view_pos.z) - params.z_near) / (params.z_far - params.z_near), 0.0, 1.0); + { //omni lights - uvec4 cluster_cell = texture(usampler3D(cluster_texture, linear_sampler), cluster_pos); + uint cluster_omni_offset = cluster_offset; - uint omni_light_count = cluster_cell.x >> CLUSTER_COUNTER_SHIFT; - uint omni_light_pointer = cluster_cell.x & CLUSTER_POINTER_MASK; + uint item_min; + uint item_max; + uint item_from; + uint item_to; - for (uint i = 0; i < omni_light_count; i++) { - uint light_index = cluster_data.indices[omni_light_pointer + i]; + cluster_get_item_range(cluster_omni_offset + params.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); - vec3 light_pos = lights.data[i].position; - float d = distance(lights.data[i].position, view_pos) * lights.data[i].inv_radius; - vec3 shadow_attenuation = vec3(1.0); +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif - if (d < 1.0) { - vec2 attenuation_energy = unpackHalf2x16(lights.data[i].attenuation_energy); - vec4 color_specular = unpackUnorm4x8(lights.data[i].color_specular); + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_omni_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif - float attenuation = pow(max(1.0 - d, 0.0), attenuation_energy.x); + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + uint light_index = 32 * i + bit; - vec3 light = attenuation_energy.y * color_specular.rgb / M_PI; + //if (!bool(omni_omni_lights.data[light_index].mask & draw_call.layer_mask)) { + // continue; //not masked + //} - vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[i].shadow_color_enabled); + vec3 light_pos = omni_lights.data[light_index].position; + float d = distance(omni_lights.data[light_index].position, view_pos); + float shadow_attenuation = 1.0; - if (shadow_color_enabled.a > 0.5) { - //has shadow - vec4 v = vec4(view_pos, 1.0); + if (d * omni_lights.data[light_index].inv_radius < 1.0) { + float attenuation = get_omni_attenuation(d, omni_lights.data[light_index].inv_radius, omni_lights.data[light_index].attenuation); - vec4 splane = (lights.data[i].shadow_matrix * v); - float shadow_len = length(splane.xyz); //need to remember shadow len from here + vec3 light = omni_lights.data[light_index].color / M_PI; - splane.xyz = normalize(splane.xyz); - vec4 clamp_rect = lights.data[i].atlas_rect; + if (omni_lights.data[light_index].shadow_enabled) { + //has shadow + vec4 v = vec4(view_pos, 1.0); - if (splane.z >= 0.0) { - splane.z += 1.0; + vec4 splane = (omni_lights.data[light_index].shadow_matrix * v); + float shadow_len = length(splane.xyz); //need to remember shadow len from here - clamp_rect.y += clamp_rect.w; + splane.xyz = normalize(splane.xyz); + vec4 clamp_rect = omni_lights.data[light_index].atlas_rect; - } else { - splane.z = 1.0 - splane.z; - } + if (splane.z >= 0.0) { + splane.z += 1.0; + + clamp_rect.y += clamp_rect.w; - splane.xy /= splane.z; + } else { + splane.z = 1.0 - splane.z; + } - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = shadow_len * lights.data[i].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already + splane.xy /= splane.z; - float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; - float shadow = exp(min(0.0, (depth - splane.z)) / lights.data[i].inv_radius * lights.data[i].shadow_volumetric_fog_fade); + splane.xy = splane.xy * 0.5 + 0.5; + splane.z = shadow_len * omni_lights.data[light_index].inv_radius; + splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; + splane.w = 1.0; //needed? i think it should be 1 already - shadow_attenuation = mix(shadow_color_enabled.rgb, vec3(1.0), shadow); + float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; + + shadow_attenuation = exp(min(0.0, (depth - splane.z)) / omni_lights.data[light_index].inv_radius * omni_lights.data[light_index].shadow_volumetric_fog_fade); + } + total_light += light * attenuation * shadow_attenuation; + } } - total_light += light * attenuation * shadow_attenuation; } } - uint spot_light_count = cluster_cell.y >> CLUSTER_COUNTER_SHIFT; - uint spot_light_pointer = cluster_cell.y & CLUSTER_POINTER_MASK; + { //spot lights - for (uint i = 0; i < spot_light_count; i++) { - uint light_index = cluster_data.indices[spot_light_pointer + i]; + uint cluster_spot_offset = cluster_offset + params.cluster_type_size; - vec3 light_pos = lights.data[i].position; - vec3 light_rel_vec = lights.data[i].position - view_pos; - float d = length(light_rel_vec) * lights.data[i].inv_radius; - vec3 shadow_attenuation = vec3(1.0); + uint item_min; + uint item_max; + uint item_from; + uint item_to; - if (d < 1.0) { - vec2 attenuation_energy = unpackHalf2x16(lights.data[i].attenuation_energy); - vec4 color_specular = unpackUnorm4x8(lights.data[i].color_specular); + cluster_get_item_range(cluster_spot_offset + params.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); - float attenuation = pow(max(1.0 - d, 0.0), attenuation_energy.x); +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif - vec3 spot_dir = lights.data[i].direction; - vec2 spot_att_angle = unpackHalf2x16(lights.data[i].cone_attenuation_angle); - float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_att_angle.y); - float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_att_angle.y)); - attenuation *= 1.0 - pow(spot_rim, spot_att_angle.x); + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_spot_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif - vec3 light = attenuation_energy.y * color_specular.rgb / M_PI; + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif - vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[i].shadow_color_enabled); + //if (!bool(omni_lights.data[light_index].mask & draw_call.layer_mask)) { + // continue; //not masked + //} - if (shadow_color_enabled.a > 0.5) { - //has shadow - vec4 v = vec4(view_pos, 1.0); + uint light_index = 32 * i + bit; - vec4 splane = (lights.data[i].shadow_matrix * v); - splane /= splane.w; + vec3 light_pos = spot_lights.data[light_index].position; + vec3 light_rel_vec = spot_lights.data[light_index].position - view_pos; + float d = length(light_rel_vec); + float shadow_attenuation = 1.0; - float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; - float shadow = exp(min(0.0, (depth - splane.z)) / lights.data[i].inv_radius * lights.data[i].shadow_volumetric_fog_fade); + if (d * spot_lights.data[light_index].inv_radius < 1.0) { + float attenuation = get_omni_attenuation(d, spot_lights.data[light_index].inv_radius, spot_lights.data[light_index].attenuation); - shadow_attenuation = mix(shadow_color_enabled.rgb, vec3(1.0), shadow); - } + vec3 spot_dir = spot_lights.data[light_index].direction; + float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[light_index].cone_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[light_index].cone_angle)); + attenuation *= 1.0 - pow(spot_rim, spot_lights.data[light_index].cone_attenuation); + + vec3 light = spot_lights.data[light_index].color / M_PI; + + if (spot_lights.data[light_index].shadow_enabled) { + //has shadow + vec4 v = vec4(view_pos, 1.0); + + vec4 splane = (spot_lights.data[light_index].shadow_matrix * v); + splane /= splane.w; - total_light += light * attenuation * shadow_attenuation; + float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; + + shadow_attenuation = exp(min(0.0, (depth - splane.z)) / spot_lights.data[light_index].inv_radius * spot_lights.data[light_index].shadow_volumetric_fog_fade); + } + + total_light += light * attenuation * shadow_attenuation; + } + } } } @@ -461,7 +629,11 @@ void main() { #endif - imageStore(density_map, pos, vec4(total_light, total_density)); + vec4 final_density = vec4(total_light, total_density); + + final_density = mix(final_density, reprojected_density, reproject_amount); + + imageStore(density_map, pos, final_density); #endif #ifdef MODE_FOG diff --git a/servers/rendering/renderer_scene.h b/servers/rendering/renderer_scene.h index c483898fed..e8966414ab 100644 --- a/servers/rendering/renderer_scene.h +++ b/servers/rendering/renderer_scene.h @@ -95,7 +95,7 @@ public: virtual Variant instance_geometry_get_shader_parameter(RID p_instance, const StringName &p_parameter) const = 0; virtual Variant instance_geometry_get_shader_parameter_default_value(RID p_instance, const StringName &p_parameter) const = 0; - virtual void directional_shadow_atlas_set_size(int p_size) = 0; + virtual void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false) = 0; /* SKY API */ @@ -122,12 +122,10 @@ public: virtual void environment_glow_set_use_bicubic_upscale(bool p_enable) = 0; virtual void environment_glow_set_use_high_quality(bool p_enable) = 0; - virtual void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RS::EnvVolumetricFogShadowFilter p_shadow_filter) = 0; + virtual void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) = 0; virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth) = 0; virtual void environment_set_volumetric_fog_filter_active(bool p_enable) = 0; - virtual void environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size) = 0; - virtual void environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size) = 0; virtual void environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance) = 0; virtual void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) = 0; @@ -140,6 +138,7 @@ public: virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) = 0; virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) = 0; + virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) = 0; virtual void environment_set_tonemap(RID p_env, RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale) = 0; @@ -172,7 +171,7 @@ public: virtual void directional_shadow_quality_set(RS::ShadowQuality p_quality) = 0; virtual RID shadow_atlas_create() = 0; - virtual void shadow_atlas_set_size(RID p_atlas, int p_size) = 0; + virtual void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_use_16_bits = false) = 0; virtual void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) = 0; /* Render Buffers */ @@ -180,6 +179,8 @@ public: virtual RID render_buffers_create() = 0; virtual void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding) = 0; + virtual void gi_set_use_half_resolution(bool p_enable) = 0; + virtual void set_debug_draw_mode(RS::ViewportDebugDraw p_debug_draw) = 0; virtual TypedArray<Image> bake_render_uv2(RID p_base, const Vector<RID> &p_material_overrides, const Size2i &p_image_size) = 0; diff --git a/servers/rendering/renderer_scene_cull.cpp b/servers/rendering/renderer_scene_cull.cpp index 2e32c69cba..8067f9574c 100644 --- a/servers/rendering/renderer_scene_cull.cpp +++ b/servers/rendering/renderer_scene_cull.cpp @@ -135,7 +135,7 @@ void RendererSceneCull::_instance_pair(Instance *p_A, Instance *p_B) { idata.flags |= InstanceData::FLAG_GEOM_LIGHTING_DIRTY; } - } else if (self->pair_volumes_to_mesh && B->base_type == RS::INSTANCE_REFLECTION_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { + } else if (self->geometry_instance_pair_mask & (1 << RS::INSTANCE_REFLECTION_PROBE) && B->base_type == RS::INSTANCE_REFLECTION_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(B->base_data); InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data); @@ -147,7 +147,7 @@ void RendererSceneCull::_instance_pair(Instance *p_A, Instance *p_B) { idata.flags |= InstanceData::FLAG_GEOM_REFLECTION_DIRTY; } - } else if (self->pair_volumes_to_mesh && B->base_type == RS::INSTANCE_DECAL && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { + } else if (self->geometry_instance_pair_mask & (1 << RS::INSTANCE_DECAL) && B->base_type == RS::INSTANCE_DECAL && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { InstanceDecalData *decal = static_cast<InstanceDecalData *>(B->base_data); InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data); @@ -174,7 +174,7 @@ void RendererSceneCull::_instance_pair(Instance *p_A, Instance *p_B) { ((RendererSceneCull *)self)->_instance_queue_update(A, false, false); //need to update capture } - } else if (B->base_type == RS::INSTANCE_GI_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { + } else if (self->geometry_instance_pair_mask & (1 << RS::INSTANCE_GI_PROBE) && B->base_type == RS::INSTANCE_GI_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(B->base_data); InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data); @@ -195,7 +195,8 @@ void RendererSceneCull::_instance_pair(Instance *p_A, Instance *p_B) { InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(B->base_data); gi_probe->lights.insert(A); } else if (B->base_type == RS::INSTANCE_PARTICLES_COLLISION && A->base_type == RS::INSTANCE_PARTICLES) { - RSG::storage->particles_add_collision(A->base, B); + InstanceParticlesCollisionData *collision = static_cast<InstanceParticlesCollisionData *>(B->base_data); + RSG::storage->particles_add_collision(A->base, collision->instance); } } @@ -225,7 +226,7 @@ void RendererSceneCull::_instance_unpair(Instance *p_A, Instance *p_B) { idata.flags |= InstanceData::FLAG_GEOM_LIGHTING_DIRTY; } - } else if (self->pair_volumes_to_mesh && B->base_type == RS::INSTANCE_REFLECTION_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { + } else if (self->geometry_instance_pair_mask & (1 << RS::INSTANCE_REFLECTION_PROBE) && B->base_type == RS::INSTANCE_REFLECTION_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(B->base_data); InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data); @@ -237,7 +238,7 @@ void RendererSceneCull::_instance_unpair(Instance *p_A, Instance *p_B) { idata.flags |= InstanceData::FLAG_GEOM_REFLECTION_DIRTY; } - } else if (self->pair_volumes_to_mesh && B->base_type == RS::INSTANCE_DECAL && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { + } else if (self->geometry_instance_pair_mask & (1 << RS::INSTANCE_DECAL) && B->base_type == RS::INSTANCE_DECAL && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { InstanceDecalData *decal = static_cast<InstanceDecalData *>(B->base_data); InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data); @@ -264,7 +265,7 @@ void RendererSceneCull::_instance_unpair(Instance *p_A, Instance *p_B) { ((RendererSceneCull *)self)->_instance_queue_update(A, false, false); //need to update capture } - } else if (B->base_type == RS::INSTANCE_GI_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { + } else if (self->geometry_instance_pair_mask & (1 << RS::INSTANCE_GI_PROBE) && B->base_type == RS::INSTANCE_GI_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) { InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(B->base_data); InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data); @@ -284,7 +285,8 @@ void RendererSceneCull::_instance_unpair(Instance *p_A, Instance *p_B) { InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(B->base_data); gi_probe->lights.erase(A); } else if (B->base_type == RS::INSTANCE_PARTICLES_COLLISION && A->base_type == RS::INSTANCE_PARTICLES) { - RSG::storage->particles_remove_collision(A->base, B); + InstanceParticlesCollisionData *collision = static_cast<InstanceParticlesCollisionData *>(B->base_data); + RSG::storage->particles_remove_collision(A->base, collision->instance); } } @@ -386,6 +388,9 @@ void RendererSceneCull::_instance_update_mesh_instance(Instance *p_instance) { p_instance->mesh_instance = RID(); } + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(p_instance->base_data); + scene_render->geometry_instance_set_mesh_instance(geom->geometry_instance, p_instance->mesh_instance); + if (p_instance->scenario && p_instance->array_index >= 0) { InstanceData &idata = p_instance->scenario->instance_data[p_instance->array_index]; if (p_instance->mesh_instance.is_valid()) { @@ -421,10 +426,17 @@ void RendererSceneCull::instance_set_base(RID p_instance, RID p_base) { } switch (instance->base_type) { + case RS::INSTANCE_MESH: + case RS::INSTANCE_MULTIMESH: + case RS::INSTANCE_IMMEDIATE: + case RS::INSTANCE_PARTICLES: { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + scene_render->geometry_instance_free(geom->geometry_instance); + } break; case RS::INSTANCE_LIGHT: { InstanceLightData *light = static_cast<InstanceLightData *>(instance->base_data); - if (scenario && RSG::storage->light_get_type(instance->base) != RS::LIGHT_DIRECTIONAL && light->bake_mode == RS::LIGHT_BAKE_DYNAMIC) { + if (scenario && instance->visible && RSG::storage->light_get_type(instance->base) != RS::LIGHT_DIRECTIONAL && light->bake_mode == RS::LIGHT_BAKE_DYNAMIC) { scenario->dynamic_lights.erase(light->instance); } @@ -439,6 +451,10 @@ void RendererSceneCull::instance_set_base(RID p_instance, RID p_base) { } scene_render->free(light->instance); } break; + case RS::INSTANCE_PARTICLES_COLLISION: { + InstanceParticlesCollisionData *collision = static_cast<InstanceParticlesCollisionData *>(instance->base_data); + RSG::storage->free(collision->instance); + } break; case RS::INSTANCE_REFLECTION_PROBE: { InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(instance->base_data); scene_render->free(reflection_probe->instance); @@ -457,6 +473,7 @@ void RendererSceneCull::instance_set_base(RID p_instance, RID p_base) { while (lightmap_data->users.front()) { instance_geometry_set_lightmap(lightmap_data->users.front()->get()->self, RID(), Rect2(), 0); } + scene_render->free(lightmap_data->instance); } break; case RS::INSTANCE_GI_PROBE: { InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(instance->base_data); @@ -514,8 +531,29 @@ void RendererSceneCull::instance_set_base(RID p_instance, RID p_base) { case RS::INSTANCE_PARTICLES: { InstanceGeometryData *geom = memnew(InstanceGeometryData); instance->base_data = geom; + geom->geometry_instance = scene_render->geometry_instance_create(p_base); + + scene_render->geometry_instance_set_skeleton(geom->geometry_instance, instance->skeleton); + scene_render->geometry_instance_set_material_override(geom->geometry_instance, instance->material_override); + scene_render->geometry_instance_set_surface_materials(geom->geometry_instance, instance->materials); + scene_render->geometry_instance_set_transform(geom->geometry_instance, instance->transform, instance->aabb, instance->transformed_aabb); + scene_render->geometry_instance_set_layer_mask(geom->geometry_instance, instance->layer_mask); + scene_render->geometry_instance_set_lod_bias(geom->geometry_instance, instance->lod_bias); + scene_render->geometry_instance_set_use_baked_light(geom->geometry_instance, instance->baked_light); + scene_render->geometry_instance_set_use_dynamic_gi(geom->geometry_instance, instance->dynamic_gi); + scene_render->geometry_instance_set_cast_double_sided_shadows(geom->geometry_instance, instance->cast_shadows == RS::SHADOW_CASTING_SETTING_DOUBLE_SIDED); + scene_render->geometry_instance_set_use_lightmap(geom->geometry_instance, RID(), instance->lightmap_uv_scale, instance->lightmap_slice_index); + if (instance->lightmap_sh.size() == 9) { + scene_render->geometry_instance_set_lightmap_capture(geom->geometry_instance, instance->lightmap_sh.ptr()); + } } break; + case RS::INSTANCE_PARTICLES_COLLISION: { + InstanceParticlesCollisionData *collision = memnew(InstanceParticlesCollisionData); + collision->instance = RSG::storage->particles_collision_instance_create(p_base); + RSG::storage->particles_collision_instance_set_active(collision->instance, instance->visible); + instance->base_data = collision; + } break; case RS::INSTANCE_REFLECTION_PROBE: { InstanceReflectionProbeData *reflection_probe = memnew(InstanceReflectionProbeData); reflection_probe->owner = instance; @@ -533,7 +571,7 @@ void RendererSceneCull::instance_set_base(RID p_instance, RID p_base) { case RS::INSTANCE_LIGHTMAP: { InstanceLightmapData *lightmap_data = memnew(InstanceLightmapData); instance->base_data = lightmap_data; - //lightmap_data->instance = scene_render->lightmap_data_instance_create(p_base); + lightmap_data->instance = scene_render->lightmap_instance_create(p_base); } break; case RS::INSTANCE_GI_PROBE: { InstanceGIProbeData *gi_probe = memnew(InstanceGIProbeData); @@ -558,7 +596,7 @@ void RendererSceneCull::instance_set_base(RID p_instance, RID p_base) { } //forcefully update the dependency now, so if for some reason it gets removed, we can immediately clear it - RSG::storage->base_update_dependency(p_base, instance); + RSG::storage->base_update_dependency(p_base, &instance->dependency_tracker); } _instance_queue_update(instance, true, true); @@ -659,6 +697,11 @@ void RendererSceneCull::instance_set_layer_mask(RID p_instance, uint32_t p_mask) if (instance->scenario && instance->array_index >= 0) { instance->scenario->instance_data[instance->array_index].layer_mask = p_mask; } + + if ((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK && instance->base_data) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + scene_render->geometry_instance_set_layer_mask(geom->geometry_instance, p_mask); + } } void RendererSceneCull::instance_set_transform(RID p_instance, const Transform &p_transform) { @@ -739,6 +782,22 @@ void RendererSceneCull::instance_set_visible(RID p_instance, bool p_visible) { } else if (instance->indexer_id.is_valid()) { _unpair_instance(instance); } + + if (instance->base_type == RS::INSTANCE_LIGHT) { + InstanceLightData *light = static_cast<InstanceLightData *>(instance->base_data); + if (instance->scenario && RSG::storage->light_get_type(instance->base) != RS::LIGHT_DIRECTIONAL && light->bake_mode == RS::LIGHT_BAKE_DYNAMIC) { + if (p_visible) { + instance->scenario->dynamic_lights.push_back(light->instance); + } else { + instance->scenario->dynamic_lights.erase(light->instance); + } + } + } + + if (instance->base_type == RS::INSTANCE_PARTICLES_COLLISION) { + InstanceParticlesCollisionData *collision = static_cast<InstanceParticlesCollisionData *>(instance->base_data); + RSG::storage->particles_collision_instance_set_active(collision->instance, p_visible); + } } inline bool is_geometry_instance(RenderingServer::InstanceType p_type) { @@ -782,12 +841,17 @@ void RendererSceneCull::instance_attach_skeleton(RID p_instance, RID p_skeleton) if (p_skeleton.is_valid()) { //update the dependency now, so if cleared, we remove it - RSG::storage->skeleton_update_dependency(p_skeleton, instance); + RSG::storage->skeleton_update_dependency(p_skeleton, &instance->dependency_tracker); } - _instance_update_mesh_instance(instance); - _instance_queue_update(instance, true, true); + + if ((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK && instance->base_data) { + _instance_update_mesh_instance(instance); + + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + scene_render->geometry_instance_set_skeleton(geom->geometry_instance, p_skeleton); + } } void RendererSceneCull::instance_set_exterior(RID p_instance, bool p_enabled) { @@ -892,6 +956,11 @@ void RendererSceneCull::instance_geometry_set_flag(RID p_instance, RS::InstanceF } } + if ((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK && instance->base_data) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + scene_render->geometry_instance_set_use_baked_light(geom->geometry_instance, p_enabled); + } + } break; case RS::INSTANCE_FLAG_USE_DYNAMIC_GI: { if (p_enabled == instance->dynamic_gi) { @@ -907,6 +976,11 @@ void RendererSceneCull::instance_geometry_set_flag(RID p_instance, RS::InstanceF //once out of octree, can be changed instance->dynamic_gi = p_enabled; + if ((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK && instance->base_data) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + scene_render->geometry_instance_set_use_dynamic_gi(geom->geometry_instance, p_enabled); + } + } break; case RS::INSTANCE_FLAG_DRAW_NEXT_FRAME_IF_VISIBLE: { instance->redraw_if_visible = p_enabled; @@ -948,6 +1022,11 @@ void RendererSceneCull::instance_geometry_set_cast_shadows_setting(RID p_instanc } } + if ((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK && instance->base_data) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + scene_render->geometry_instance_set_cast_double_sided_shadows(geom->geometry_instance, instance->cast_shadows == RS::SHADOW_CASTING_SETTING_DOUBLE_SIDED); + } + _instance_queue_update(instance, false, true); } @@ -957,6 +1036,11 @@ void RendererSceneCull::instance_geometry_set_material_override(RID p_instance, instance->material_override = p_material; _instance_queue_update(instance, false, true); + + if ((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK && instance->base_data) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + scene_render->geometry_instance_set_material_override(geom->geometry_instance, p_material); + } } void RendererSceneCull::instance_geometry_set_draw_range(RID p_instance, float p_min, float p_max, float p_min_margin, float p_max_margin) { @@ -981,9 +1065,17 @@ void RendererSceneCull::instance_geometry_set_lightmap(RID p_instance, RID p_lig instance->lightmap_uv_scale = p_lightmap_uv_scale; instance->lightmap_slice_index = p_slice_index; + RID lightmap_instance_rid; + if (lightmap_instance) { InstanceLightmapData *lightmap_data = static_cast<InstanceLightmapData *>(lightmap_instance->base_data); lightmap_data->users.insert(instance); + lightmap_instance_rid = lightmap_data->instance; + } + + if ((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK && instance->base_data) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + scene_render->geometry_instance_set_use_lightmap(geom->geometry_instance, lightmap_instance_rid, p_lightmap_uv_scale, p_slice_index); } } @@ -992,16 +1084,21 @@ void RendererSceneCull::instance_geometry_set_lod_bias(RID p_instance, float p_l ERR_FAIL_COND(!instance); instance->lod_bias = p_lod_bias; + + if ((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK && instance->base_data) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + scene_render->geometry_instance_set_lod_bias(geom->geometry_instance, p_lod_bias); + } } void RendererSceneCull::instance_geometry_set_shader_parameter(RID p_instance, const StringName &p_parameter, const Variant &p_value) { Instance *instance = instance_owner.getornull(p_instance); ERR_FAIL_COND(!instance); - Map<StringName, RendererSceneRender::InstanceBase::InstanceShaderParameter>::Element *E = instance->instance_shader_parameters.find(p_parameter); + Map<StringName, Instance::InstanceShaderParameter>::Element *E = instance->instance_shader_parameters.find(p_parameter); if (!E) { - RendererSceneRender::InstanceBase::InstanceShaderParameter isp; + Instance::InstanceShaderParameter isp; isp.index = -1; isp.info = PropertyInfo(); isp.value = p_value; @@ -1042,7 +1139,7 @@ void RendererSceneCull::instance_geometry_get_shader_parameter_list(RID p_instan const_cast<RendererSceneCull *>(this)->update_dirty_instances(); Vector<StringName> names; - for (Map<StringName, RendererSceneRender::InstanceBase::InstanceShaderParameter>::Element *E = instance->instance_shader_parameters.front(); E; E = E->next()) { + for (Map<StringName, Instance::InstanceShaderParameter>::Element *E = instance->instance_shader_parameters.front(); E; E = E->next()) { names.push_back(E->key()); } names.sort_custom<StringName::AlphCompare>(); @@ -1064,13 +1161,13 @@ void RendererSceneCull::_update_instance(Instance *p_instance) { RS::LightBakeMode bake_mode = RSG::storage->light_get_bake_mode(p_instance->base); if (RSG::storage->light_get_type(p_instance->base) != RS::LIGHT_DIRECTIONAL && bake_mode != light->bake_mode) { - if (p_instance->scenario && light->bake_mode == RS::LIGHT_BAKE_DYNAMIC) { + if (p_instance->visible && p_instance->scenario && light->bake_mode == RS::LIGHT_BAKE_DYNAMIC) { p_instance->scenario->dynamic_lights.erase(light->instance); } light->bake_mode = bake_mode; - if (p_instance->scenario && light->bake_mode == RS::LIGHT_BAKE_DYNAMIC) { + if (p_instance->visible && p_instance->scenario && light->bake_mode == RS::LIGHT_BAKE_DYNAMIC) { p_instance->scenario->dynamic_lights.push_back(light->instance); } } @@ -1079,9 +1176,7 @@ void RendererSceneCull::_update_instance(Instance *p_instance) { if (light->max_sdfgi_cascade != max_sdfgi_cascade) { light->max_sdfgi_cascade = max_sdfgi_cascade; //should most likely make sdfgi dirty in scenario } - } - - if (p_instance->base_type == RS::INSTANCE_REFLECTION_PROBE) { + } else if (p_instance->base_type == RS::INSTANCE_REFLECTION_PROBE) { InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(p_instance->base_data); scene_render->reflection_probe_instance_set_transform(reflection_probe->instance, p_instance->transform); @@ -1090,35 +1185,49 @@ void RendererSceneCull::_update_instance(Instance *p_instance) { InstanceData &idata = p_instance->scenario->instance_data[p_instance->array_index]; idata.flags |= InstanceData::FLAG_REFLECTION_PROBE_DIRTY; } - } - - if (p_instance->base_type == RS::INSTANCE_DECAL) { + } else if (p_instance->base_type == RS::INSTANCE_DECAL) { InstanceDecalData *decal = static_cast<InstanceDecalData *>(p_instance->base_data); scene_render->decal_instance_set_transform(decal->instance, p_instance->transform); - } + } else if (p_instance->base_type == RS::INSTANCE_LIGHTMAP) { + InstanceLightmapData *lightmap = static_cast<InstanceLightmapData *>(p_instance->base_data); - if (p_instance->base_type == RS::INSTANCE_GI_PROBE) { + scene_render->lightmap_instance_set_transform(lightmap->instance, p_instance->transform); + } else if (p_instance->base_type == RS::INSTANCE_GI_PROBE) { InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(p_instance->base_data); scene_render->gi_probe_instance_set_transform_to_data(gi_probe->probe_instance, p_instance->transform); - } - - if (p_instance->base_type == RS::INSTANCE_PARTICLES) { + } else if (p_instance->base_type == RS::INSTANCE_PARTICLES) { RSG::storage->particles_set_emission_transform(p_instance->base, p_instance->transform); - } + } else if (p_instance->base_type == RS::INSTANCE_PARTICLES_COLLISION) { + InstanceParticlesCollisionData *collision = static_cast<InstanceParticlesCollisionData *>(p_instance->base_data); - if (p_instance->base_type == RS::INSTANCE_PARTICLES_COLLISION) { //remove materials no longer used and un-own them if (RSG::storage->particles_collision_is_heightfield(p_instance->base)) { heightfield_particle_colliders_update_list.insert(p_instance); } + RSG::storage->particles_collision_instance_set_transform(collision->instance, p_instance->transform); } if (p_instance->aabb.has_no_surface()) { return; } + if (p_instance->base_type == RS::INSTANCE_LIGHTMAP) { + //if this moved, update the captured objects + InstanceLightmapData *lightmap_data = static_cast<InstanceLightmapData *>(p_instance->base_data); + //erase dependencies, since no longer a lightmap + + for (Set<Instance *>::Element *E = lightmap_data->geometries.front(); E; E = E->next()) { + Instance *geom = E->get(); + _instance_queue_update(geom, true, false); + } + } + + AABB new_aabb; + new_aabb = p_instance->transform.xform(p_instance->aabb); + p_instance->transformed_aabb = new_aabb; + if ((1 << p_instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) { InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(p_instance->base_data); //make sure lights are updated if it casts shadow @@ -1137,30 +1246,15 @@ void RendererSceneCull::_update_instance(Instance *p_instance) { if (!p_instance->lightmap_sh.is_empty()) { p_instance->lightmap_sh.clear(); //don't need SH p_instance->lightmap_target_sh.clear(); //don't need SH + scene_render->geometry_instance_set_lightmap_capture(geom->geometry_instance, nullptr); } } - } - - if (p_instance->base_type == RS::INSTANCE_LIGHTMAP) { - //if this moved, update the captured objects - InstanceLightmapData *lightmap_data = static_cast<InstanceLightmapData *>(p_instance->base_data); - //erase dependencies, since no longer a lightmap - for (Set<Instance *>::Element *E = lightmap_data->geometries.front(); E; E = E->next()) { - Instance *geom = E->get(); - _instance_queue_update(geom, true, false); - } + scene_render->geometry_instance_set_transform(geom->geometry_instance, p_instance->transform, p_instance->aabb, p_instance->transformed_aabb); } - p_instance->mirror = p_instance->transform.basis.determinant() < 0.0; - - AABB new_aabb; - - new_aabb = p_instance->transform.xform(p_instance->aabb); - - p_instance->transformed_aabb = new_aabb; - - if (p_instance->scenario == nullptr || !p_instance->visible || Math::is_zero_approx(p_instance->transform.basis.determinant())) { + // note: we had to remove is equal approx check here, it meant that det == 0.000004 won't work, which is the case for some of our scenes. + if (p_instance->scenario == nullptr || !p_instance->visible || p_instance->transform.basis.determinant() == 0) { p_instance->prev_transformed_aabb = p_instance->transformed_aabb; return; } @@ -1195,17 +1289,26 @@ void RendererSceneCull::_update_instance(Instance *p_instance) { idata.flags = p_instance->base_type; //changing it means de-indexing, so this never needs to be changed later idata.base_rid = p_instance->base; switch (p_instance->base_type) { + case RS::INSTANCE_MESH: + case RS::INSTANCE_MULTIMESH: + case RS::INSTANCE_IMMEDIATE: + case RS::INSTANCE_PARTICLES: { + idata.instance_geometry = static_cast<InstanceGeometryData *>(p_instance->base_data)->geometry_instance; + } break; case RS::INSTANCE_LIGHT: { - idata.instance_data_rid = static_cast<InstanceLightData *>(p_instance->base_data)->instance; + idata.instance_data_rid = static_cast<InstanceLightData *>(p_instance->base_data)->instance.get_id(); } break; case RS::INSTANCE_REFLECTION_PROBE: { - idata.instance_data_rid = static_cast<InstanceReflectionProbeData *>(p_instance->base_data)->instance; + idata.instance_data_rid = static_cast<InstanceReflectionProbeData *>(p_instance->base_data)->instance.get_id(); } break; case RS::INSTANCE_DECAL: { - idata.instance_data_rid = static_cast<InstanceDecalData *>(p_instance->base_data)->instance; + idata.instance_data_rid = static_cast<InstanceDecalData *>(p_instance->base_data)->instance.get_id(); + } break; + case RS::INSTANCE_LIGHTMAP: { + idata.instance_data_rid = static_cast<InstanceLightmapData *>(p_instance->base_data)->instance.get_id(); } break; case RS::INSTANCE_GI_PROBE: { - idata.instance_data_rid = static_cast<InstanceGIProbeData *>(p_instance->base_data)->probe_instance; + idata.instance_data_rid = static_cast<InstanceGIProbeData *>(p_instance->base_data)->probe_instance.get_id(); } break; default: { } @@ -1258,10 +1361,8 @@ void RendererSceneCull::_update_instance(Instance *p_instance) { pair.pair_mask |= 1 << RS::INSTANCE_GI_PROBE; pair.pair_mask |= 1 << RS::INSTANCE_LIGHTMAP; - if (pair_volumes_to_mesh) { - pair.pair_mask |= 1 << RS::INSTANCE_DECAL; - pair.pair_mask |= 1 << RS::INSTANCE_REFLECTION_PROBE; - } + pair.pair_mask |= geometry_instance_pair_mask; + pair.bvh2 = &p_instance->scenario->indexers[Scenario::INDEXER_VOLUMES]; } else if (p_instance->base_type == RS::INSTANCE_LIGHT) { pair.pair_mask |= RS::INSTANCE_GEOMETRY_MASK; @@ -1271,7 +1372,10 @@ void RendererSceneCull::_update_instance(Instance *p_instance) { pair.pair_mask |= (1 << RS::INSTANCE_GI_PROBE); pair.bvh2 = &p_instance->scenario->indexers[Scenario::INDEXER_VOLUMES]; } - } else if (pair_volumes_to_mesh && (p_instance->base_type == RS::INSTANCE_REFLECTION_PROBE || p_instance->base_type == RS::INSTANCE_DECAL)) { + } else if (geometry_instance_pair_mask & (1 << RS::INSTANCE_REFLECTION_PROBE) && (p_instance->base_type == RS::INSTANCE_REFLECTION_PROBE)) { + pair.pair_mask = RS::INSTANCE_GEOMETRY_MASK; + pair.bvh = &p_instance->scenario->indexers[Scenario::INDEXER_GEOMETRY]; + } else if (geometry_instance_pair_mask & (1 << RS::INSTANCE_DECAL) && (p_instance->base_type == RS::INSTANCE_DECAL)) { pair.pair_mask = RS::INSTANCE_GEOMETRY_MASK; pair.bvh = &p_instance->scenario->indexers[Scenario::INDEXER_GEOMETRY]; } else if (p_instance->base_type == RS::INSTANCE_PARTICLES_COLLISION) { @@ -1325,10 +1429,12 @@ void RendererSceneCull::_unpair_instance(Instance *p_instance) { p_instance->array_index = -1; if ((1 << p_instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) { // Clear these now because the InstanceData containing the dirty flags is gone - p_instance->light_instances.clear(); - p_instance->reflection_probe_instances.clear(); - //p_instance->decal_instances.clear(); will implement later - p_instance->gi_probe_instances.clear(); + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(p_instance->base_data); + + scene_render->geometry_instance_pair_light_instances(geom->geometry_instance, nullptr, 0); + scene_render->geometry_instance_pair_reflection_probe_instances(geom->geometry_instance, nullptr, 0); + scene_render->geometry_instance_pair_decal_instances(geom->geometry_instance, nullptr, 0); + scene_render->geometry_instance_pair_gi_probe_instances(geom->geometry_instance, nullptr, 0); } } @@ -1486,6 +1592,8 @@ void RendererSceneCull::_update_instance_lightmap_captures(Instance *p_instance) } } } + + scene_render->geometry_instance_set_lightmap_capture(geom->geometry_instance, p_instance->lightmap_sh.ptr()); } void RendererSceneCull::_light_instance_setup_directional_shadow(int p_shadow_index, Instance *p_instance, const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect) { @@ -1798,6 +1906,9 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons RS::LightOmniShadowMode shadow_mode = RSG::storage->light_omni_get_shadow_mode(p_instance->base); if (shadow_mode == RS::LIGHT_OMNI_SHADOW_DUAL_PARABOLOID || !scene_render->light_instances_can_render_shadow_cube()) { + if (max_shadows_used + 2 > MAX_UPDATE_SHADOWS) { + return true; + } for (int i = 0; i < 2; i++) { //using this one ensures that raster deferred will have it RENDER_TIMESTAMP("Culling Shadow Paraboloid" + itos(i)); @@ -1814,7 +1925,6 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons planes.write[4] = light_transform.xform(Plane(Vector3(0, -1, z).normalized(), radius)); planes.write[5] = light_transform.xform(Plane(Vector3(0, 0, -z), 0)); - geometry_instances_to_shadow_render.clear(); instance_shadow_cull_result.clear(); Vector<Vector3> points = Geometry3D::compute_convex_mesh_points(&planes[0], planes.size()); @@ -1835,6 +1945,8 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons Plane near_plane(light_transform.origin, light_transform.basis.get_axis(2) * z); + RendererSceneRender::RenderShadowData &shadow_data = render_shadow_data[max_shadows_used++]; + for (int j = 0; j < (int)instance_shadow_cull_result.size(); j++) { Instance *instance = instance_shadow_cull_result[j]; if (!instance->visible || !((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) || !static_cast<InstanceGeometryData *>(instance->base_data)->can_cast_shadows) { @@ -1849,16 +1961,21 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons } } - geometry_instances_to_shadow_render.push_back(instance); + shadow_data.instances.push_back(static_cast<InstanceGeometryData *>(instance->base_data)->geometry_instance); } RSG::storage->update_mesh_instances(); scene_render->light_instance_set_shadow_transform(light->instance, CameraMatrix(), light_transform, radius, 0, i, 0); - scene_render->render_shadow(light->instance, p_shadow_atlas, i, geometry_instances_to_shadow_render); + shadow_data.light = light->instance; + shadow_data.pass = i; } } else { //shadow cube + if (max_shadows_used + 6 > MAX_UPDATE_SHADOWS) { + return true; + } + real_t radius = RSG::storage->light_get_param(p_instance->base, RS::LIGHT_PARAM_RANGE); CameraMatrix cm; cm.set_perspective(90, 1, 0.01, radius); @@ -1888,7 +2005,6 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons Vector<Plane> planes = cm.get_projection_planes(xform); - geometry_instances_to_shadow_render.clear(); instance_shadow_cull_result.clear(); Vector<Vector3> points = Geometry3D::compute_convex_mesh_points(&planes[0], planes.size()); @@ -1907,7 +2023,7 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons p_scenario->indexers[Scenario::INDEXER_GEOMETRY].convex_query(planes.ptr(), planes.size(), points.ptr(), points.size(), cull_convex); - Plane near_plane(xform.origin, -xform.basis.get_axis(2)); + RendererSceneRender::RenderShadowData &shadow_data = render_shadow_data[max_shadows_used++]; for (int j = 0; j < (int)instance_shadow_cull_result.size(); j++) { Instance *instance = instance_shadow_cull_result[j]; @@ -1922,22 +2038,28 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons } } - geometry_instances_to_shadow_render.push_back(instance); + shadow_data.instances.push_back(static_cast<InstanceGeometryData *>(instance->base_data)->geometry_instance); } RSG::storage->update_mesh_instances(); scene_render->light_instance_set_shadow_transform(light->instance, cm, xform, radius, 0, i, 0); - scene_render->render_shadow(light->instance, p_shadow_atlas, i, geometry_instances_to_shadow_render); + + shadow_data.light = light->instance; + shadow_data.pass = i; } //restore the regular DP matrix - scene_render->light_instance_set_shadow_transform(light->instance, CameraMatrix(), light_transform, radius, 0, 0, 0); + //scene_render->light_instance_set_shadow_transform(light->instance, CameraMatrix(), light_transform, radius, 0, 0, 0); } } break; case RS::LIGHT_SPOT: { RENDER_TIMESTAMP("Culling Spot Light"); + if (max_shadows_used + 1 > MAX_UPDATE_SHADOWS) { + return true; + } + real_t radius = RSG::storage->light_get_param(p_instance->base, RS::LIGHT_PARAM_RANGE); real_t angle = RSG::storage->light_get_param(p_instance->base, RS::LIGHT_PARAM_SPOT_ANGLE); @@ -1946,7 +2068,6 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons Vector<Plane> planes = cm.get_projection_planes(light_transform); - geometry_instances_to_shadow_render.clear(); instance_shadow_cull_result.clear(); Vector<Vector3> points = Geometry3D::compute_convex_mesh_points(&planes[0], planes.size()); @@ -1965,7 +2086,7 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons p_scenario->indexers[Scenario::INDEXER_GEOMETRY].convex_query(planes.ptr(), planes.size(), points.ptr(), points.size(), cull_convex); - Plane near_plane(light_transform.origin, -light_transform.basis.get_axis(2)); + RendererSceneRender::RenderShadowData &shadow_data = render_shadow_data[max_shadows_used++]; for (int j = 0; j < (int)instance_shadow_cull_result.size(); j++) { Instance *instance = instance_shadow_cull_result[j]; @@ -1980,13 +2101,14 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons RSG::storage->mesh_instance_check_for_update(instance->mesh_instance); } } - geometry_instances_to_shadow_render.push_back(instance); + shadow_data.instances.push_back(static_cast<InstanceGeometryData *>(instance->base_data)->geometry_instance); } RSG::storage->update_mesh_instances(); scene_render->light_instance_set_shadow_transform(light->instance, cm, light_transform, radius, 0, 0, 0); - scene_render->render_shadow(light->instance, p_shadow_atlas, 0, geometry_instances_to_shadow_render); + shadow_data.light = light->instance; + shadow_data.pass = 0; } break; } @@ -2039,14 +2161,13 @@ void RendererSceneCull::render_camera(RID p_render_buffers, RID p_camera, RID p_ RID environment = _render_get_environment(p_camera, p_scenario); - _prepare_scene(camera->transform, camera_matrix, ortho, camera->vaspect, p_render_buffers, environment, camera->visible_layers, p_scenario, p_shadow_atlas, RID(), p_screen_lod_threshold); - _render_scene(p_render_buffers, camera->transform, camera_matrix, ortho, environment, camera->effects, p_scenario, p_shadow_atlas, RID(), -1, p_screen_lod_threshold); + _render_scene(camera->transform, camera_matrix, ortho, camera->vaspect, p_render_buffers, environment, camera->effects, camera->visible_layers, p_scenario, p_shadow_atlas, RID(), -1, p_screen_lod_threshold); #endif } void RendererSceneCull::render_camera(RID p_render_buffers, Ref<XRInterface> &p_interface, XRInterface::Eyes p_eye, RID p_camera, RID p_scenario, Size2 p_viewport_size, float p_screen_lod_threshold, RID p_shadow_atlas) { // render for AR/VR interface - +#if 0 Camera *camera = camera_owner.getornull(p_camera); ERR_FAIL_COND(!camera); @@ -2126,9 +2247,226 @@ void RendererSceneCull::render_camera(RID p_render_buffers, Ref<XRInterface> &p_ // And render our scene... _render_scene(p_render_buffers, cam_transform, camera_matrix, false, environment, camera->effects, p_scenario, p_shadow_atlas, RID(), -1, p_screen_lod_threshold); +#endif }; -void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect, RID p_render_buffers, RID p_environment, uint32_t p_visible_layers, RID p_scenario, RID p_shadow_atlas, RID p_reflection_probe, float p_screen_lod_threshold, bool p_using_shadows) { +void RendererSceneCull::_frustum_cull_threaded(uint32_t p_thread, FrustumCullData *cull_data) { + uint32_t cull_total = cull_data->scenario->instance_data.size(); + uint32_t total_threads = RendererThreadPool::singleton->thread_work_pool.get_thread_count(); + uint32_t cull_from = p_thread * cull_total / total_threads; + uint32_t cull_to = (p_thread + 1 == total_threads) ? cull_total : ((p_thread + 1) * cull_total / total_threads); + + _frustum_cull(*cull_data, frustum_cull_result_threads[p_thread], cull_from, cull_to); +} + +void RendererSceneCull::_frustum_cull(FrustumCullData &cull_data, FrustumCullResult &cull_result, uint64_t p_from, uint64_t p_to) { + uint64_t frame_number = RSG::rasterizer->get_frame_number(); + float lightmap_probe_update_speed = RSG::storage->lightmap_get_probe_capture_update_speed() * RSG::rasterizer->get_frame_delta_time(); + + uint32_t sdfgi_last_light_index = 0xFFFFFFFF; + uint32_t sdfgi_last_light_cascade = 0xFFFFFFFF; + + RID instance_pair_buffer[MAX_INSTANCE_PAIRS]; + + for (uint64_t i = p_from; i < p_to; i++) { + bool mesh_visible = false; + + if (cull_data.scenario->instance_aabbs[i].in_frustum(cull_data.cull->frustum)) { + InstanceData &idata = cull_data.scenario->instance_data[i]; + uint32_t base_type = idata.flags & InstanceData::FLAG_BASE_TYPE_MASK; + + if ((cull_data.visible_layers & idata.layer_mask) == 0) { + //failure + } else if (base_type == RS::INSTANCE_LIGHT) { + cull_result.lights.push_back(idata.instance); + cull_result.light_instances.push_back(RID::from_uint64(idata.instance_data_rid)); + if (cull_data.shadow_atlas.is_valid() && RSG::storage->light_has_shadow(idata.base_rid)) { + scene_render->light_instance_mark_visible(RID::from_uint64(idata.instance_data_rid)); //mark it visible for shadow allocation later + } + + } else if (base_type == RS::INSTANCE_REFLECTION_PROBE) { + if (cull_data.render_reflection_probe != idata.instance) { + //avoid entering The Matrix + + if ((idata.flags & InstanceData::FLAG_REFLECTION_PROBE_DIRTY) || scene_render->reflection_probe_instance_needs_redraw(RID::from_uint64(idata.instance_data_rid))) { + InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(idata.instance->base_data); + cull_data.cull->lock.lock(); + if (!reflection_probe->update_list.in_list()) { + reflection_probe->render_step = 0; + reflection_probe_render_list.add_last(&reflection_probe->update_list); + } + cull_data.cull->lock.unlock(); + + idata.flags &= ~uint32_t(InstanceData::FLAG_REFLECTION_PROBE_DIRTY); + } + + if (scene_render->reflection_probe_instance_has_reflection(RID::from_uint64(idata.instance_data_rid))) { + cull_result.reflections.push_back(RID::from_uint64(idata.instance_data_rid)); + } + } + } else if (base_type == RS::INSTANCE_DECAL) { + cull_result.decals.push_back(RID::from_uint64(idata.instance_data_rid)); + + } else if (base_type == RS::INSTANCE_GI_PROBE) { + InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(idata.instance->base_data); + cull_data.cull->lock.lock(); + if (!gi_probe->update_element.in_list()) { + gi_probe_update_list.add(&gi_probe->update_element); + } + cull_data.cull->lock.unlock(); + cull_result.gi_probes.push_back(RID::from_uint64(idata.instance_data_rid)); + + } else if (base_type == RS::INSTANCE_LIGHTMAP) { + cull_result.gi_probes.push_back(RID::from_uint64(idata.instance_data_rid)); + } else if (((1 << base_type) & RS::INSTANCE_GEOMETRY_MASK) && !(idata.flags & InstanceData::FLAG_CAST_SHADOWS_ONLY)) { + bool keep = true; + + if (idata.flags & InstanceData::FLAG_REDRAW_IF_VISIBLE) { + RenderingServerDefault::redraw_request(); + } + + if (base_type == RS::INSTANCE_MESH) { + mesh_visible = true; + } else if (base_type == RS::INSTANCE_PARTICLES) { + //particles visible? process them + if (RSG::storage->particles_is_inactive(idata.base_rid)) { + //but if nothing is going on, don't do it. + keep = false; + } else { + cull_data.cull->lock.lock(); + RSG::storage->particles_request_process(idata.base_rid); + cull_data.cull->lock.unlock(); + RSG::storage->particles_set_view_axis(idata.base_rid, -cull_data.cam_transform.basis.get_axis(2).normalized()); + //particles visible? request redraw + RenderingServerDefault::redraw_request(); + } + } + + if (geometry_instance_pair_mask & (1 << RS::INSTANCE_LIGHT) && (idata.flags & InstanceData::FLAG_GEOM_LIGHTING_DIRTY)) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data); + uint32_t idx = 0; + + for (Set<Instance *>::Element *E = geom->lights.front(); E; E = E->next()) { + InstanceLightData *light = static_cast<InstanceLightData *>(E->get()->base_data); + instance_pair_buffer[idx++] = light->instance; + if (idx == MAX_INSTANCE_PAIRS) { + break; + } + } + + scene_render->geometry_instance_pair_light_instances(geom->geometry_instance, instance_pair_buffer, idx); + idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_LIGHTING_DIRTY); + } + + if (geometry_instance_pair_mask & (1 << RS::INSTANCE_REFLECTION_PROBE) && (idata.flags & InstanceData::FLAG_GEOM_REFLECTION_DIRTY)) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data); + uint32_t idx = 0; + + for (Set<Instance *>::Element *E = geom->reflection_probes.front(); E; E = E->next()) { + InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(E->get()->base_data); + + instance_pair_buffer[idx++] = reflection_probe->instance; + if (idx == MAX_INSTANCE_PAIRS) { + break; + } + } + + scene_render->geometry_instance_pair_reflection_probe_instances(geom->geometry_instance, instance_pair_buffer, idx); + idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_REFLECTION_DIRTY); + } + + if (geometry_instance_pair_mask & (1 << RS::INSTANCE_DECAL) && (idata.flags & InstanceData::FLAG_GEOM_DECAL_DIRTY)) { + //InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data); + //todo for GLES3 + idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_DECAL_DIRTY); + /*for (Set<Instance *>::Element *E = geom->dec.front(); E; E = E->next()) { + InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(E->get()->base_data); + + instance_pair_buffer[idx++] = reflection_probe->instance; + if (idx==MAX_INSTANCE_PAIRS) { + break; + } + }*/ + //scene_render->geometry_instance_pair_decal_instances(geom->geometry_instance, light_instances, idx); + } + + if (idata.flags & InstanceData::FLAG_GEOM_GI_PROBE_DIRTY) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data); + uint32_t idx = 0; + for (Set<Instance *>::Element *E = geom->gi_probes.front(); E; E = E->next()) { + InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(E->get()->base_data); + + instance_pair_buffer[idx++] = gi_probe->probe_instance; + if (idx == MAX_INSTANCE_PAIRS) { + break; + } + } + + scene_render->geometry_instance_pair_gi_probe_instances(geom->geometry_instance, instance_pair_buffer, idx); + idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_GI_PROBE_DIRTY); + } + + if ((idata.flags & InstanceData::FLAG_LIGHTMAP_CAPTURE) && idata.instance->last_frame_pass != frame_number && !idata.instance->lightmap_target_sh.is_empty() && !idata.instance->lightmap_sh.is_empty()) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data); + Color *sh = idata.instance->lightmap_sh.ptrw(); + const Color *target_sh = idata.instance->lightmap_target_sh.ptr(); + for (uint32_t j = 0; j < 9; j++) { + sh[j] = sh[j].lerp(target_sh[j], MIN(1.0, lightmap_probe_update_speed)); + } + scene_render->geometry_instance_set_lightmap_capture(geom->geometry_instance, sh); + idata.instance->last_frame_pass = frame_number; + } + + if (keep) { + cull_result.geometry_instances.push_back(idata.instance_geometry); + } + } + } + + for (uint32_t j = 0; j < cull_data.cull->shadow_count; j++) { + for (uint32_t k = 0; k < cull_data.cull->shadows[j].cascade_count; k++) { + if (cull_data.scenario->instance_aabbs[i].in_frustum(cull_data.cull->shadows[j].cascades[k].frustum)) { + InstanceData &idata = cull_data.scenario->instance_data[i]; + uint32_t base_type = idata.flags & InstanceData::FLAG_BASE_TYPE_MASK; + + if (((1 << base_type) & RS::INSTANCE_GEOMETRY_MASK) && idata.flags & InstanceData::FLAG_CAST_SHADOWS) { + cull_result.directional_shadows[j].cascade_geometry_instances[k].push_back(idata.instance_geometry); + mesh_visible = true; + } + } + } + } + + for (uint32_t j = 0; j < cull_data.cull->sdfgi.region_count; j++) { + if (cull_data.scenario->instance_aabbs[i].in_aabb(cull_data.cull->sdfgi.region_aabb[j])) { + InstanceData &idata = cull_data.scenario->instance_data[i]; + uint32_t base_type = idata.flags & InstanceData::FLAG_BASE_TYPE_MASK; + + if (base_type == RS::INSTANCE_LIGHT) { + InstanceLightData *instance_light = (InstanceLightData *)idata.instance->base_data; + if (instance_light->bake_mode == RS::LIGHT_BAKE_STATIC && cull_data.cull->sdfgi.region_cascade[j] <= instance_light->max_sdfgi_cascade) { + if (sdfgi_last_light_index != i || sdfgi_last_light_cascade != cull_data.cull->sdfgi.region_cascade[j]) { + sdfgi_last_light_index = i; + sdfgi_last_light_cascade = cull_data.cull->sdfgi.region_cascade[j]; + cull_result.sdfgi_cascade_lights[sdfgi_last_light_cascade].push_back(instance_light->instance); + } + } + } else if ((1 << base_type) & RS::INSTANCE_GEOMETRY_MASK) { + if (idata.flags & InstanceData::FLAG_USES_BAKED_LIGHT) { + cull_result.sdfgi_region_geometry_instances[j].push_back(idata.instance_geometry); + mesh_visible = true; + } + } + } + } + + if (mesh_visible && cull_data.scenario->instance_data[i].flags & InstanceData::FLAG_USES_MESH_INSTANCE) { + cull_result.mesh_instances.push_back(cull_data.scenario->instance_data[i].instance->mesh_instance); + } + } +} + +void RendererSceneCull::_render_scene(const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect, RID p_render_buffers, RID p_environment, RID p_force_camera_effects, uint32_t p_visible_layers, RID p_scenario, RID p_shadow_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, bool p_using_shadows) { // Note, in stereo rendering: // - p_cam_transform will be a transform in the middle of our two eyes // - p_cam_projection is a wider frustrum that encompasses both eyes @@ -2142,6 +2480,7 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca scene_render->set_scene_pass(render_pass); if (p_render_buffers.is_valid()) { + //no rendering code here, this is only to set up what needs to be done, request regions, etc. scene_render->sdfgi_update(p_render_buffers, p_environment, p_cam_transform.origin); //update conditions for SDFGI (whether its used or not) } @@ -2153,9 +2492,6 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca Plane near_plane(p_cam_transform.origin, -p_cam_transform.basis.get_axis(2).normalized()); - uint64_t frame_number = RSG::rasterizer->get_frame_number(); - float lightmap_probe_update_speed = RSG::storage->lightmap_get_probe_capture_update_speed() * RSG::rasterizer->get_frame_delta_time(); - /* STEP 2 - CULL */ cull.frustum = Frustum(planes); @@ -2163,13 +2499,6 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca Vector<RID> directional_lights; // directional lights { - //reset shadows - for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) { - for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) { - cull.shadows[i].cascades[j].cull_result.clear(); - } - } - cull.shadow_count = 0; Vector<Instance *> lights_with_shadow; @@ -2206,18 +2535,7 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca { //sdfgi cull.sdfgi.region_count = 0; - for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) { - cull.sdfgi.region_cull_result[i].clear(); - } - - for (int i = 0; i < SDFGI_MAX_CASCADES; i++) { - cull.sdfgi.cascade_lights[i].clear(); - } - if (p_render_buffers.is_valid()) { - for (int i = 0; i < SDFGI_MAX_CASCADES; i++) { - cull.sdfgi.cascade_lights[i].clear(); - } cull.sdfgi.cascade_light_count = 0; uint32_t prev_cascade = 0xFFFFFFFF; @@ -2239,209 +2557,53 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca } } - { - //pre-clear results - geometry_instances_to_render.clear(); - light_cull_result.clear(); - lightmap_cull_result.clear(); - reflection_probe_instance_cull_result.clear(); - light_instance_cull_result.clear(); - gi_probe_instance_cull_result.clear(); - lightmap_cull_result.clear(); - decal_instance_cull_result.clear(); - mesh_instance_cull_result.clear(); - } + frustum_cull_result.clear(); { - uint64_t cull_count = scenario->instance_data.size(); - uint32_t sdfgi_last_light_index = 0xFFFFFFFF; - uint32_t sdfgi_last_light_cascade = 0xFFFFFFFF; - - for (uint64_t i = 0; i < cull_count; i++) { - bool mesh_visible = false; - - if (scenario->instance_aabbs[i].in_frustum(cull.frustum)) { - InstanceData &idata = scenario->instance_data[i]; - uint32_t base_type = idata.flags & InstanceData::FLAG_BASE_TYPE_MASK; - - if ((p_visible_layers & idata.layer_mask) == 0) { - //failure - } else if (base_type == RS::INSTANCE_LIGHT) { - light_cull_result.push_back(idata.instance); - light_instance_cull_result.push_back(idata.instance_data_rid); - if (p_shadow_atlas.is_valid() && RSG::storage->light_has_shadow(idata.base_rid)) { - scene_render->light_instance_mark_visible(idata.instance_data_rid); //mark it visible for shadow allocation later - } - - } else if (base_type == RS::INSTANCE_REFLECTION_PROBE) { - if (render_reflection_probe != idata.instance) { - //avoid entering The Matrix - - if ((idata.flags & InstanceData::FLAG_REFLECTION_PROBE_DIRTY) || scene_render->reflection_probe_instance_needs_redraw(idata.instance_data_rid)) { - InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(idata.instance->base_data); - cull.lock.lock(); - if (!reflection_probe->update_list.in_list()) { - reflection_probe->render_step = 0; - reflection_probe_render_list.add_last(&reflection_probe->update_list); - } - cull.lock.unlock(); - - idata.flags &= ~uint32_t(InstanceData::FLAG_REFLECTION_PROBE_DIRTY); - } - - if (scene_render->reflection_probe_instance_has_reflection(idata.instance_data_rid)) { - reflection_probe_instance_cull_result.push_back(idata.instance_data_rid); - } - } - } else if (base_type == RS::INSTANCE_DECAL) { - decal_instance_cull_result.push_back(idata.instance_data_rid); - - } else if (base_type == RS::INSTANCE_GI_PROBE) { - InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(idata.instance->base_data); - cull.lock.lock(); - if (!gi_probe->update_element.in_list()) { - gi_probe_update_list.add(&gi_probe->update_element); - } - cull.lock.unlock(); - gi_probe_instance_cull_result.push_back(idata.instance_data_rid); - - } else if (base_type == RS::INSTANCE_LIGHTMAP) { - lightmap_cull_result.push_back(idata.instance); - } else if (((1 << base_type) & RS::INSTANCE_GEOMETRY_MASK) && !(idata.flags & InstanceData::FLAG_CAST_SHADOWS_ONLY)) { - bool keep = true; - - if (idata.flags & InstanceData::FLAG_REDRAW_IF_VISIBLE) { - RenderingServerDefault::redraw_request(); - } - - if (base_type == RS::INSTANCE_MESH) { - mesh_visible = true; - } else if (base_type == RS::INSTANCE_PARTICLES) { - //particles visible? process them - if (RSG::storage->particles_is_inactive(idata.base_rid)) { - //but if nothing is going on, don't do it. - keep = false; - } else { - cull.lock.lock(); - RSG::storage->particles_request_process(idata.base_rid); - cull.lock.unlock(); - RSG::storage->particles_set_view_axis(idata.base_rid, -p_cam_transform.basis.get_axis(2).normalized()); - //particles visible? request redraw - RenderingServerDefault::redraw_request(); - } - } - - if (pair_volumes_to_mesh && (idata.flags & InstanceData::FLAG_GEOM_LIGHTING_DIRTY)) { - InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data); - int l = 0; - //only called when lights AABB enter/exit this geometry - idata.instance->light_instances.resize(geom->lights.size()); - - for (Set<Instance *>::Element *E = geom->lights.front(); E; E = E->next()) { - InstanceLightData *light = static_cast<InstanceLightData *>(E->get()->base_data); - - idata.instance->light_instances.write[l++] = light->instance; - } - - idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_LIGHTING_DIRTY); - } - - if (pair_volumes_to_mesh && (idata.flags & InstanceData::FLAG_GEOM_REFLECTION_DIRTY)) { - InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data); - int l = 0; - //only called when reflection probe AABB enter/exit this geometry - idata.instance->reflection_probe_instances.resize(geom->reflection_probes.size()); - - for (Set<Instance *>::Element *E = geom->reflection_probes.front(); E; E = E->next()) { - InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(E->get()->base_data); - - idata.instance->reflection_probe_instances.write[l++] = reflection_probe->instance; - } - - idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_REFLECTION_DIRTY); - } - - if (pair_volumes_to_mesh && (idata.flags & InstanceData::FLAG_GEOM_DECAL_DIRTY)) { - //InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data); - //todo for GLES3 - idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_DECAL_DIRTY); - } - - if (idata.flags & InstanceData::FLAG_GEOM_GI_PROBE_DIRTY) { - InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data); - int l = 0; - //only called when reflection probe AABB enter/exit this geometry - idata.instance->gi_probe_instances.resize(geom->gi_probes.size()); - - for (Set<Instance *>::Element *E = geom->gi_probes.front(); E; E = E->next()) { - InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(E->get()->base_data); - - idata.instance->gi_probe_instances.write[l++] = gi_probe->probe_instance; - } - - idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_GI_PROBE_DIRTY); - } - - if ((idata.flags & InstanceData::FLAG_LIGHTMAP_CAPTURE) && idata.instance->last_frame_pass != frame_number && !idata.instance->lightmap_target_sh.is_empty() && !idata.instance->lightmap_sh.is_empty()) { - Color *sh = idata.instance->lightmap_sh.ptrw(); - const Color *target_sh = idata.instance->lightmap_target_sh.ptr(); - for (uint32_t j = 0; j < 9; j++) { - sh[j] = sh[j].lerp(target_sh[j], MIN(1.0, lightmap_probe_update_speed)); - } - idata.instance->last_frame_pass = frame_number; - } - - if (keep) { - geometry_instances_to_render.push_back(idata.instance); - } - } - } - - for (uint32_t j = 0; j < cull.shadow_count; j++) { - for (uint32_t k = 0; k < cull.shadows[j].cascade_count; k++) { - if (scenario->instance_aabbs[i].in_frustum(cull.shadows[j].cascades[k].frustum)) { - InstanceData &idata = scenario->instance_data[i]; - uint32_t base_type = idata.flags & InstanceData::FLAG_BASE_TYPE_MASK; - - if (((1 << base_type) & RS::INSTANCE_GEOMETRY_MASK) && idata.flags & InstanceData::FLAG_CAST_SHADOWS) { - cull.shadows[j].cascades[k].cull_result.push_back(idata.instance); - mesh_visible = true; - } - } - } + uint64_t cull_from = 0; + uint64_t cull_to = scenario->instance_data.size(); + + FrustumCullData cull_data; + + //prepare for eventual thread usage + cull_data.cull = &cull; + cull_data.scenario = scenario; + cull_data.shadow_atlas = p_shadow_atlas; + cull_data.cam_transform = p_cam_transform; + cull_data.visible_layers = p_visible_layers; + cull_data.render_reflection_probe = render_reflection_probe; +//#define DEBUG_CULL_TIME +#ifdef DEBUG_CULL_TIME + uint64_t time_from = OS::get_singleton()->get_ticks_usec(); +#endif + if (cull_to > thread_cull_threshold) { + //multiple threads + for (uint32_t i = 0; i < frustum_cull_result_threads.size(); i++) { + frustum_cull_result_threads[i].clear(); } - for (uint32_t j = 0; j < cull.sdfgi.region_count; j++) { - if (scenario->instance_aabbs[i].in_aabb(cull.sdfgi.region_aabb[j])) { - InstanceData &idata = scenario->instance_data[i]; - uint32_t base_type = idata.flags & InstanceData::FLAG_BASE_TYPE_MASK; + RendererThreadPool::singleton->thread_work_pool.do_work(frustum_cull_result_threads.size(), this, &RendererSceneCull::_frustum_cull_threaded, &cull_data); - if (base_type == RS::INSTANCE_LIGHT) { - InstanceLightData *instance_light = (InstanceLightData *)idata.instance->base_data; - if (instance_light->bake_mode == RS::LIGHT_BAKE_STATIC && cull.sdfgi.region_cascade[j] <= instance_light->max_sdfgi_cascade) { - if (sdfgi_last_light_index != i || sdfgi_last_light_cascade != cull.sdfgi.region_cascade[j]) { - sdfgi_last_light_index = i; - sdfgi_last_light_cascade = cull.sdfgi.region_cascade[j]; - cull.sdfgi.cascade_lights[sdfgi_last_light_cascade].push_back(instance_light->instance); - } - } - } else if ((1 << base_type) & RS::INSTANCE_GEOMETRY_MASK) { - if (idata.flags & InstanceData::FLAG_USES_BAKED_LIGHT) { - cull.sdfgi.region_cull_result[j].push_back(idata.instance); - mesh_visible = true; - } - } - } + for (uint32_t i = 0; i < frustum_cull_result_threads.size(); i++) { + frustum_cull_result.append_from(frustum_cull_result_threads[i]); } - if (mesh_visible && scenario->instance_data[i].flags & InstanceData::FLAG_USES_MESH_INSTANCE) { - mesh_instance_cull_result.push_back(scenario->instance_data[i].instance->mesh_instance); - } + } else { + //single threaded + _frustum_cull(cull_data, frustum_cull_result, cull_from, cull_to); } - if (mesh_instance_cull_result.size()) { - for (uint64_t i = 0; i < mesh_instance_cull_result.size(); i++) { - RSG::storage->mesh_instance_check_for_update(mesh_instance_cull_result[i]); +#ifdef DEBUG_CULL_TIME + static float time_avg = 0; + static uint32_t time_count = 0; + time_avg += double(OS::get_singleton()->get_ticks_usec() - time_from) / 1000.0; + time_count++; + print_line("time taken: " + rtos(time_avg / time_count)); +#endif + + if (frustum_cull_result.mesh_instances.size()) { + for (uint64_t i = 0; i < frustum_cull_result.mesh_instances.size(); i++) { + RSG::storage->mesh_instance_check_for_update(frustum_cull_result.mesh_instances[i]); } RSG::storage->update_mesh_instances(); } @@ -2449,64 +2611,30 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca //render shadows - for (uint32_t i = 0; i < cull.shadow_count; i++) { - for (uint32_t j = 0; j < cull.shadows[i].cascade_count; j++) { - const Cull::Shadow::Cascade &c = cull.shadows[i].cascades[j]; - // print_line("shadow " + itos(i) + " cascade " + itos(j) + " elements: " + itos(c.cull_result.size())); - scene_render->light_instance_set_shadow_transform(cull.shadows[i].light_instance, c.projection, c.transform, c.zfar, c.split, j, c.shadow_texel_size, c.bias_scale, c.range_begin, c.uv_scale); - scene_render->render_shadow(cull.shadows[i].light_instance, p_shadow_atlas, j, c.cull_result, near_plane, p_cam_projection.get_lod_multiplier(), p_screen_lod_threshold); - } - } + max_shadows_used = 0; - //render SDFGI + if (p_using_shadows) { //setup shadow maps - { - if (cull.sdfgi.region_count > 0) { - //update regions - for (uint32_t i = 0; i < cull.sdfgi.region_count; i++) { - scene_render->render_sdfgi(p_render_buffers, i, cull.sdfgi.region_cull_result[i]); - } - //check if static lights were culled - bool static_lights_culled = false; - for (uint32_t i = 0; i < cull.sdfgi.cascade_light_count; i++) { - if (cull.sdfgi.cascade_lights[i].size()) { - static_lights_culled = true; - break; - } - } + // Directional Shadows - if (static_lights_culled) { - scene_render->render_sdfgi_static_lights(p_render_buffers, cull.sdfgi.cascade_light_count, cull.sdfgi.cascade_light_index, cull.sdfgi.cascade_lights); + for (uint32_t i = 0; i < cull.shadow_count; i++) { + for (uint32_t j = 0; j < cull.shadows[i].cascade_count; j++) { + const Cull::Shadow::Cascade &c = cull.shadows[i].cascades[j]; + // print_line("shadow " + itos(i) + " cascade " + itos(j) + " elements: " + itos(c.cull_result.size())); + scene_render->light_instance_set_shadow_transform(cull.shadows[i].light_instance, c.projection, c.transform, c.zfar, c.split, j, c.shadow_texel_size, c.bias_scale, c.range_begin, c.uv_scale); + if (max_shadows_used == MAX_UPDATE_SHADOWS) { + continue; + } + render_shadow_data[max_shadows_used].light = cull.shadows[i].light_instance; + render_shadow_data[max_shadows_used].pass = j; + render_shadow_data[max_shadows_used].instances.merge_unordered(frustum_cull_result.directional_shadows[i].cascade_geometry_instances[j]); + max_shadows_used++; } } - if (p_render_buffers.is_valid()) { - scene_render->sdfgi_update_probes(p_render_buffers, p_environment, directional_lights, scenario->dynamic_lights.ptr(), scenario->dynamic_lights.size()); - } - } - - //light_samplers_culled=0; - - /* - print_line("OT: "+rtos( (OS::get_singleton()->get_ticks_usec()-t)/1000.0)); - print_line("OTO: "+itos(p_scenario->octree.get_octant_count())); - print_line("OTE: "+itos(p_scenario->octree.get_elem_count())); - print_line("OTP: "+itos(p_scenario->octree.get_pair_count())); - */ - - /* STEP 3 - PROCESS PORTALS, VALIDATE ROOMS */ - //removed, will replace with culling - - /* STEP 4 - REMOVE FURTHER CULLED OBJECTS, ADD LIGHTS */ - - /* STEP 5 - PROCESS POSITIONAL LIGHTS */ - - if (p_using_shadows) { //setup shadow maps - - //SortArray<Instance*,_InstanceLightsort> sorter; - //sorter.sort(light_cull_result,light_cull_count); - for (uint32_t i = 0; i < (uint32_t)light_cull_result.size(); i++) { - Instance *ins = light_cull_result[i]; + // Positional Shadowss + for (uint32_t i = 0; i < (uint32_t)frustum_cull_result.lights.size(); i++) { + Instance *ins = frustum_cull_result.lights[i]; if (!p_shadow_atlas.is_valid() || !RSG::storage->light_has_shadow(ins->base)) { continue; @@ -2591,19 +2719,78 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca bool redraw = scene_render->shadow_atlas_update_light(p_shadow_atlas, light->instance, coverage, light->last_version); - if (redraw) { + if (redraw && max_shadows_used < MAX_UPDATE_SHADOWS) { //must redraw! RENDER_TIMESTAMP(">Rendering Light " + itos(i)); light->shadow_dirty = _light_instance_update_shadow(ins, p_cam_transform, p_cam_projection, p_cam_orthogonal, p_cam_vaspect, p_shadow_atlas, scenario, p_screen_lod_threshold); RENDER_TIMESTAMP("<Rendering Light " + itos(i)); + } else { + light->shadow_dirty = redraw; + } + } + } + + //render SDFGI + + { + sdfgi_update_data.update_static = false; + + if (cull.sdfgi.region_count > 0) { + //update regions + for (uint32_t i = 0; i < cull.sdfgi.region_count; i++) { + render_sdfgi_data[i].instances.merge_unordered(frustum_cull_result.sdfgi_region_geometry_instances[i]); + render_sdfgi_data[i].region = i; + } + //check if static lights were culled + bool static_lights_culled = false; + for (uint32_t i = 0; i < cull.sdfgi.cascade_light_count; i++) { + if (frustum_cull_result.sdfgi_cascade_lights[i].size()) { + static_lights_culled = true; + break; + } + } + + if (static_lights_culled) { + sdfgi_update_data.static_cascade_count = cull.sdfgi.cascade_light_count; + sdfgi_update_data.static_cascade_indices = cull.sdfgi.cascade_light_index; + sdfgi_update_data.static_positional_lights = frustum_cull_result.sdfgi_cascade_lights; + sdfgi_update_data.update_static = true; } } + + if (p_render_buffers.is_valid()) { + sdfgi_update_data.directional_lights = &directional_lights; + sdfgi_update_data.positional_light_instances = scenario->dynamic_lights.ptr(); + sdfgi_update_data.positional_light_count = scenario->dynamic_lights.size(); + } } //append the directional lights to the lights culled for (int i = 0; i < directional_lights.size(); i++) { - light_instance_cull_result.push_back(directional_lights[i]); + frustum_cull_result.light_instances.push_back(directional_lights[i]); } + + RID camera_effects; + if (p_force_camera_effects.is_valid()) { + camera_effects = p_force_camera_effects; + } else { + camera_effects = scenario->camera_effects; + } + /* PROCESS GEOMETRY AND DRAW SCENE */ + + RENDER_TIMESTAMP("Render Scene "); + scene_render->render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_orthogonal, frustum_cull_result.geometry_instances, frustum_cull_result.light_instances, frustum_cull_result.reflections, frustum_cull_result.gi_probes, frustum_cull_result.decals, frustum_cull_result.lightmaps, p_environment, camera_effects, p_shadow_atlas, p_reflection_probe.is_valid() ? RID() : scenario->reflection_atlas, p_reflection_probe, p_reflection_probe_pass, p_screen_lod_threshold, render_shadow_data, max_shadows_used, render_sdfgi_data, cull.sdfgi.region_count, &sdfgi_update_data); + + for (uint32_t i = 0; i < max_shadows_used; i++) { + render_shadow_data[i].instances.clear(); + } + max_shadows_used = 0; + + for (uint32_t i = 0; i < cull.sdfgi.region_count; i++) { + render_sdfgi_data[i].instances.clear(); + } + + // virtual void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold,const RenderShadowData *p_render_shadows,int p_render_shadow_count,const RenderSDFGIData *p_render_sdfgi_regions,int p_render_sdfgi_region_count,const RenderSDFGIStaticLightData *p_render_sdfgi_static_lights=nullptr) = 0; } RID RendererSceneCull::_render_get_environment(RID p_camera, RID p_scenario) { @@ -2627,21 +2814,6 @@ RID RendererSceneCull::_render_get_environment(RID p_camera, RID p_scenario) { return RID(); } -void RendererSceneCull::_render_scene(RID p_render_buffers, const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, RID p_environment, RID p_force_camera_effects, RID p_scenario, RID p_shadow_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) { - Scenario *scenario = scenario_owner.getornull(p_scenario); - - RID camera_effects; - if (p_force_camera_effects.is_valid()) { - camera_effects = p_force_camera_effects; - } else { - camera_effects = scenario->camera_effects; - } - /* PROCESS GEOMETRY AND DRAW SCENE */ - - RENDER_TIMESTAMP("Render Scene "); - scene_render->render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_orthogonal, geometry_instances_to_render, light_instance_cull_result, reflection_probe_instance_cull_result, gi_probe_instance_cull_result, decal_instance_cull_result, lightmap_cull_result, p_environment, camera_effects, p_shadow_atlas, p_reflection_probe.is_valid() ? RID() : scenario->reflection_atlas, p_reflection_probe, p_reflection_probe_pass, p_screen_lod_threshold); -} - void RendererSceneCull::render_empty_scene(RID p_render_buffers, RID p_scenario, RID p_shadow_atlas) { #ifndef _3D_DISABLED @@ -2654,7 +2826,7 @@ void RendererSceneCull::render_empty_scene(RID p_render_buffers, RID p_scenario, environment = scenario->fallback_environment; } RENDER_TIMESTAMP("Render Empty Scene "); - scene_render->render_scene(p_render_buffers, Transform(), CameraMatrix(), true, PagedArray<RendererSceneRender::InstanceBase *>(), PagedArray<RID>(), PagedArray<RID>(), PagedArray<RID>(), PagedArray<RID>(), PagedArray<RendererSceneRender::InstanceBase *>(), RID(), RID(), p_shadow_atlas, scenario->reflection_atlas, RID(), 0, 0); + scene_render->render_scene(p_render_buffers, Transform(), CameraMatrix(), true, PagedArray<RendererSceneRender::GeometryInstance *>(), PagedArray<RID>(), PagedArray<RID>(), PagedArray<RID>(), PagedArray<RID>(), PagedArray<RID>(), RID(), RID(), p_shadow_atlas, scenario->reflection_atlas, RID(), 0, 0, nullptr, 0, nullptr, 0, nullptr); #endif } @@ -2717,8 +2889,7 @@ bool RendererSceneCull::_render_reflection_probe_step(Instance *p_instance, int } RENDER_TIMESTAMP("Render Reflection Probe, Step " + itos(p_step)); - _prepare_scene(xform, cm, false, false, RID(), RID(), RSG::storage->reflection_probe_get_cull_mask(p_instance->base), p_instance->scenario->self, shadow_atlas, reflection_probe->instance, lod_threshold, use_shadows); - _render_scene(RID(), xform, cm, false, RID(), RID(), p_instance->scenario->self, shadow_atlas, reflection_probe->instance, p_step, lod_threshold); + _render_scene(xform, cm, false, false, RID(), RID(), RID(), RSG::storage->reflection_probe_get_cull_mask(p_instance->base), p_instance->scenario->self, shadow_atlas, reflection_probe->instance, p_step, lod_threshold, use_shadows); } else { //do roughness postprocess step until it believes it's done @@ -2929,7 +3100,9 @@ void RendererSceneCull::render_probes() { update_lights = true; } - geometry_instances_to_render.clear(); + frustum_cull_result.geometry_instances.clear(); + + RID instance_pair_buffer[MAX_INSTANCE_PAIRS]; for (Set<Instance *>::Element *E = probe->dynamic_geometries.front(); E; E = E->next()) { Instance *ins = E->get(); @@ -2939,24 +3112,25 @@ void RendererSceneCull::render_probes() { InstanceGeometryData *geom = (InstanceGeometryData *)ins->base_data; if (ins->scenario && ins->array_index >= 0 && (ins->scenario->instance_data[ins->array_index].flags & InstanceData::FLAG_GEOM_GI_PROBE_DIRTY)) { - //giprobes may be dirty, so update - int l = 0; - //only called when reflection probe AABB enter/exit this geometry - ins->gi_probe_instances.resize(geom->gi_probes.size()); - + uint32_t idx = 0; for (Set<Instance *>::Element *F = geom->gi_probes.front(); F; F = F->next()) { InstanceGIProbeData *gi_probe2 = static_cast<InstanceGIProbeData *>(F->get()->base_data); - ins->gi_probe_instances.write[l++] = gi_probe2->probe_instance; + instance_pair_buffer[idx++] = gi_probe2->probe_instance; + if (idx == MAX_INSTANCE_PAIRS) { + break; + } } + scene_render->geometry_instance_pair_gi_probe_instances(geom->geometry_instance, instance_pair_buffer, idx); + ins->scenario->instance_data[ins->array_index].flags &= ~uint32_t(InstanceData::FLAG_GEOM_GI_PROBE_DIRTY); } - geometry_instances_to_render.push_back(E->get()); + frustum_cull_result.geometry_instances.push_back(geom->geometry_instance); } - scene_render->gi_probe_update(probe->probe_instance, update_lights, probe->light_instances, geometry_instances_to_render); + scene_render->gi_probe_update(probe->probe_instance, update_lights, probe->light_instances, frustum_cull_result.geometry_instances); gi_probe_update_list.remove(gi_probe); @@ -2971,7 +3145,7 @@ void RendererSceneCull::render_particle_colliders() { if (hfpc->scenario && hfpc->base_type == RS::INSTANCE_PARTICLES_COLLISION && RSG::storage->particles_collision_is_heightfield(hfpc->base)) { //update heightfield instance_cull_result.clear(); - geometry_instances_to_render.clear(); + frustum_cull_result.geometry_instances.clear(); struct CullAABB { PagedArray<Instance *> *result; @@ -2992,16 +3166,17 @@ void RendererSceneCull::render_particle_colliders() { if (!instance || !((1 << instance->base_type) & (RS::INSTANCE_GEOMETRY_MASK & (~(1 << RS::INSTANCE_PARTICLES))))) { //all but particles to avoid self collision continue; } - geometry_instances_to_render.push_back(instance); + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(instance->base_data); + frustum_cull_result.geometry_instances.push_back(geom->geometry_instance); } - scene_render->render_particle_collider_heightfield(hfpc->base, hfpc->transform, geometry_instances_to_render); + scene_render->render_particle_collider_heightfield(hfpc->base, hfpc->transform, frustum_cull_result.geometry_instances); } heightfield_particle_colliders_update_list.erase(heightfield_particle_colliders_update_list.front()); } } -void RendererSceneCull::_update_instance_shader_parameters_from_material(Map<StringName, RendererSceneRender::InstanceBase::InstanceShaderParameter> &isparams, const Map<StringName, RendererSceneRender::InstanceBase::InstanceShaderParameter> &existing_isparams, RID p_material) { +void RendererSceneCull::_update_instance_shader_parameters_from_material(Map<StringName, Instance::InstanceShaderParameter> &isparams, const Map<StringName, Instance::InstanceShaderParameter> &existing_isparams, RID p_material) { List<RendererStorage::InstanceShaderParam> plist; RSG::storage->material_get_instance_shader_parameters(p_material, &plist); for (List<RendererStorage::InstanceShaderParam>::Element *E = plist.front(); E; E = E->next()) { @@ -3016,7 +3191,7 @@ void RendererSceneCull::_update_instance_shader_parameters_from_material(Map<Str continue; //first one found always has priority } - RendererSceneRender::InstanceBase::InstanceShaderParameter isp; + Instance::InstanceShaderParameter isp; isp.index = E->get().index; isp.info = E->get().info; isp.default_value = E->get().default_value; @@ -3035,14 +3210,14 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) { } if (p_instance->update_dependencies) { - p_instance->instance_increase_version(); + p_instance->dependency_tracker.update_begin(); if (p_instance->base.is_valid()) { - RSG::storage->base_update_dependency(p_instance->base, p_instance); + RSG::storage->base_update_dependency(p_instance->base, &p_instance->dependency_tracker); } if (p_instance->material_override.is_valid()) { - RSG::storage->material_update_dependency(p_instance->material_override, p_instance); + RSG::storage->material_update_dependency(p_instance->material_override, &p_instance->dependency_tracker); } if (p_instance->base_type == RS::INSTANCE_MESH) { @@ -3059,7 +3234,7 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) { bool can_cast_shadows = true; bool is_animated = false; - Map<StringName, RendererSceneRender::InstanceBase::InstanceShaderParameter> isparams; + Map<StringName, Instance::InstanceShaderParameter> isparams; if (p_instance->cast_shadows == RS::SHADOW_CASTING_SETTING_OFF) { can_cast_shadows = false; @@ -3094,7 +3269,7 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) { _update_instance_shader_parameters_from_material(isparams, p_instance->instance_shader_parameters, mat); - RSG::storage->material_update_dependency(mat, p_instance); + RSG::storage->material_update_dependency(mat, &p_instance->dependency_tracker); } } @@ -3125,7 +3300,7 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) { _update_instance_shader_parameters_from_material(isparams, p_instance->instance_shader_parameters, mat); - RSG::storage->material_update_dependency(mat, p_instance); + RSG::storage->material_update_dependency(mat, &p_instance->dependency_tracker); } } @@ -3133,7 +3308,7 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) { can_cast_shadows = false; } - RSG::storage->base_update_dependency(mesh, p_instance); + RSG::storage->base_update_dependency(mesh, &p_instance->dependency_tracker); } } else if (p_instance->base_type == RS::INSTANCE_IMMEDIATE) { RID mat = RSG::storage->immediate_get_material(p_instance->base); @@ -3151,7 +3326,7 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) { } if (mat.is_valid()) { - RSG::storage->material_update_dependency(mat, p_instance); + RSG::storage->material_update_dependency(mat, &p_instance->dependency_tracker); } } else if (p_instance->base_type == RS::INSTANCE_PARTICLES) { @@ -3182,7 +3357,7 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) { _update_instance_shader_parameters_from_material(isparams, p_instance->instance_shader_parameters, mat); - RSG::storage->material_update_dependency(mat, p_instance); + RSG::storage->material_update_dependency(mat, &p_instance->dependency_tracker); } } } @@ -3210,7 +3385,9 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) { p_instance->instance_allocated_shader_parameters = (p_instance->instance_shader_parameters.size() > 0); if (p_instance->instance_allocated_shader_parameters) { p_instance->instance_allocated_shader_parameters_offset = RSG::storage->global_variables_instance_allocate(p_instance->self); - for (Map<StringName, RendererSceneRender::InstanceBase::InstanceShaderParameter>::Element *E = p_instance->instance_shader_parameters.front(); E; E = E->next()) { + scene_render->geometry_instance_set_instance_shader_parameters_offset(geom->geometry_instance, p_instance->instance_allocated_shader_parameters_offset); + + for (Map<StringName, Instance::InstanceShaderParameter>::Element *E = p_instance->instance_shader_parameters.front(); E; E = E->next()) { if (E->get().value.get_type() != Variant::NIL) { RSG::storage->global_variables_instance_update(p_instance->self, E->get().index, E->get().value); } @@ -3218,15 +3395,21 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) { } else { RSG::storage->global_variables_instance_free(p_instance->self); p_instance->instance_allocated_shader_parameters_offset = -1; + scene_render->geometry_instance_set_instance_shader_parameters_offset(geom->geometry_instance, -1); } } } if (p_instance->skeleton.is_valid()) { - RSG::storage->skeleton_update_dependency(p_instance->skeleton, p_instance); + RSG::storage->skeleton_update_dependency(p_instance->skeleton, &p_instance->dependency_tracker); } - p_instance->clean_up_dependencies(); + p_instance->dependency_tracker.update_end(); + + if ((1 << p_instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) { + InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(p_instance->base_data); + scene_render->geometry_instance_set_surface_materials(geom->geometry_instance, p_instance->materials); + } } _instance_update_list.remove(&p_instance->update_item); @@ -3322,70 +3505,50 @@ TypedArray<Image> RendererSceneCull::bake_render_uv2(RID p_base, const Vector<RI RendererSceneCull *RendererSceneCull::singleton = nullptr; +void RendererSceneCull::set_scene_render(RendererSceneRender *p_scene_render) { + scene_render = p_scene_render; + geometry_instance_pair_mask = scene_render->geometry_instance_get_pair_mask(); +} + RendererSceneCull::RendererSceneCull() { render_pass = 1; singleton = this; - pair_volumes_to_mesh = false; instance_cull_result.set_page_pool(&instance_cull_page_pool); - mesh_instance_cull_result.set_page_pool(&rid_cull_page_pool); instance_shadow_cull_result.set_page_pool(&instance_cull_page_pool); - instance_sdfgi_cull_result.set_page_pool(&instance_cull_page_pool); - light_cull_result.set_page_pool(&instance_cull_page_pool); - geometry_instances_to_render.set_page_pool(&base_instance_cull_page_pool); - geometry_instances_to_shadow_render.set_page_pool(&base_instance_cull_page_pool); - lightmap_cull_result.set_page_pool(&base_instance_cull_page_pool); - - reflection_probe_instance_cull_result.set_page_pool(&rid_cull_page_pool); - light_instance_cull_result.set_page_pool(&rid_cull_page_pool); - gi_probe_instance_cull_result.set_page_pool(&rid_cull_page_pool); - decal_instance_cull_result.set_page_pool(&rid_cull_page_pool); - - for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) { - for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) { - cull.shadows[i].cascades[j].cull_result.set_page_pool(&base_instance_cull_page_pool); - } + for (uint32_t i = 0; i < MAX_UPDATE_SHADOWS; i++) { + render_shadow_data[i].instances.set_page_pool(&geometry_instance_cull_page_pool); } - - for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) { - cull.sdfgi.region_cull_result[i].set_page_pool(&base_instance_cull_page_pool); + for (uint32_t i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) { + render_sdfgi_data[i].instances.set_page_pool(&geometry_instance_cull_page_pool); } - for (int i = 0; i < SDFGI_MAX_CASCADES; i++) { - cull.sdfgi.cascade_lights[i].set_page_pool(&rid_cull_page_pool); + frustum_cull_result.init(&rid_cull_page_pool, &geometry_instance_cull_page_pool, &instance_cull_page_pool); + frustum_cull_result_threads.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); + for (uint32_t i = 0; i < frustum_cull_result_threads.size(); i++) { + frustum_cull_result_threads[i].init(&rid_cull_page_pool, &geometry_instance_cull_page_pool, &instance_cull_page_pool); } indexer_update_iterations = GLOBAL_GET("rendering/spatial_indexer/update_iterations_per_frame"); + thread_cull_threshold = GLOBAL_GET("rendering/spatial_indexer/threaded_cull_minimum_instances"); + thread_cull_threshold = MAX(thread_cull_threshold, (uint32_t)RendererThreadPool::singleton->thread_work_pool.get_thread_count()); //make sure there is at least one thread per CPU } RendererSceneCull::~RendererSceneCull() { instance_cull_result.reset(); - mesh_instance_cull_result.reset(); instance_shadow_cull_result.reset(); - instance_sdfgi_cull_result.reset(); - light_cull_result.reset(); - - geometry_instances_to_render.reset(); - geometry_instances_to_shadow_render.reset(); - lightmap_cull_result.reset(); - - reflection_probe_instance_cull_result.reset(); - light_instance_cull_result.reset(); - gi_probe_instance_cull_result.reset(); - decal_instance_cull_result.reset(); - for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) { - for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) { - cull.shadows[i].cascades[j].cull_result.reset(); - } + for (uint32_t i = 0; i < MAX_UPDATE_SHADOWS; i++) { + render_shadow_data[i].instances.reset(); } - - for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) { - cull.sdfgi.region_cull_result[i].reset(); + for (uint32_t i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) { + render_sdfgi_data[i].instances.reset(); } - for (int i = 0; i < SDFGI_MAX_CASCADES; i++) { - cull.sdfgi.cascade_lights[i].reset(); + frustum_cull_result.reset(); + for (uint32_t i = 0; i < frustum_cull_result_threads.size(); i++) { + frustum_cull_result_threads[i].reset(); } + frustum_cull_result_threads.clear(); } diff --git a/servers/rendering/renderer_scene_cull.h b/servers/rendering/renderer_scene_cull.h index 85b4c53c59..d6d730af15 100644 --- a/servers/rendering/renderer_scene_cull.h +++ b/servers/rendering/renderer_scene_cull.h @@ -53,7 +53,9 @@ public: enum { SDFGI_MAX_CASCADES = 8, - SDFGI_MAX_REGIONS_PER_CASCADE = 3 + SDFGI_MAX_REGIONS_PER_CASCADE = 3, + MAX_INSTANCE_PAIRS = 32, + MAX_UPDATE_SHADOWS = 512 }; uint64_t render_pass; @@ -249,7 +251,10 @@ public: uint32_t flags = 0; uint32_t layer_mask = 0; //for fast layer-mask discard RID base_rid; - RID instance_data_rid; + union { + uint64_t instance_data_rid; + RendererSceneRender::GeometryInstance *instance_geometry; + }; Instance *instance = nullptr; }; @@ -296,7 +301,7 @@ public: static void _instance_pair(Instance *p_A, Instance *p_B); static void _instance_unpair(Instance *p_A, Instance *p_B); - static void _instance_update_mesh_instance(Instance *p_instance); + void _instance_update_mesh_instance(Instance *p_instance); virtual RID scenario_create(); @@ -325,7 +330,55 @@ public: virtual ~InstanceBaseData() {} }; - struct Instance : RendererSceneRender::InstanceBase { + struct Instance { + RS::InstanceType base_type; + RID base; + + RID skeleton; + RID material_override; + + RID mesh_instance; //only used for meshes and when skeleton/blendshapes exist + + Transform transform; + + float lod_bias; + + Vector<RID> materials; + + RS::ShadowCastingSetting cast_shadows; + + uint32_t layer_mask; + //fit in 32 bits + bool mirror : 8; + bool receive_shadows : 8; + bool visible : 8; + bool baked_light : 2; //this flag is only to know if it actually did use baked light + bool dynamic_gi : 2; //same above for dynamic objects + bool redraw_if_visible : 4; + + Instance *lightmap; + Rect2 lightmap_uv_scale; + int lightmap_slice_index; + uint32_t lightmap_cull_index; + Vector<Color> lightmap_sh; //spherical harmonic + + AABB aabb; + AABB transformed_aabb; + AABB prev_transformed_aabb; + + struct InstanceShaderParameter { + int32_t index = -1; + Variant value; + Variant default_value; + PropertyInfo info; + }; + + Map<StringName, InstanceShaderParameter> instance_shader_parameters; + bool instance_allocated_shader_parameters = false; + int32_t instance_allocated_shader_parameters_offset = -1; + + // + RID self; //scenario stuff DynamicBVH::ID indexer_id; @@ -360,23 +413,61 @@ public: SelfList<InstancePair>::List pairs; uint64_t pair_check; - virtual void dependency_deleted(RID p_dependency) { - if (p_dependency == base) { - singleton->instance_set_base(self, RID()); - } else if (p_dependency == skeleton) { - singleton->instance_attach_skeleton(self, RID()); - } else { - singleton->_instance_queue_update(this, false, true); + RendererStorage::DependencyTracker dependency_tracker; + + static void dependency_changed(RendererStorage::DependencyChangedNotification p_notification, RendererStorage::DependencyTracker *tracker) { + Instance *instance = (Instance *)tracker->userdata; + switch (p_notification) { + case RendererStorage::DEPENDENCY_CHANGED_SKELETON_DATA: + case RendererStorage::DEPENDENCY_CHANGED_AABB: { + singleton->_instance_queue_update(instance, true, false); + + } break; + case RendererStorage::DEPENDENCY_CHANGED_MATERIAL: { + singleton->_instance_queue_update(instance, false, true); + } break; + case RendererStorage::DEPENDENCY_CHANGED_MESH: + case RendererStorage::DEPENDENCY_CHANGED_MULTIMESH: + case RendererStorage::DEPENDENCY_CHANGED_DECAL: + case RendererStorage::DEPENDENCY_CHANGED_LIGHT: + case RendererStorage::DEPENDENCY_CHANGED_REFLECTION_PROBE: { + singleton->_instance_queue_update(instance, true, true); + } break; + case RendererStorage::DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES: + case RendererStorage::DEPENDENCY_CHANGED_SKELETON_BONES: { + //ignored + } break; } } - virtual void dependency_changed(bool p_aabb, bool p_dependencies) { - singleton->_instance_queue_update(this, p_aabb, p_dependencies); + static void dependency_deleted(const RID &p_dependency, RendererStorage::DependencyTracker *tracker) { + Instance *instance = (Instance *)tracker->userdata; + + if (p_dependency == instance->base) { + singleton->instance_set_base(instance->self, RID()); + } else if (p_dependency == instance->skeleton) { + singleton->instance_attach_skeleton(instance->self, RID()); + } else { + singleton->_instance_queue_update(instance, false, true); + } } Instance() : scenario_item(this), update_item(this) { + base_type = RS::INSTANCE_NONE; + cast_shadows = RS::SHADOW_CASTING_SETTING_ON; + receive_shadows = true; + visible = true; + layer_mask = 1; + baked_light = false; + dynamic_gi = false; + redraw_if_visible = false; + lightmap_slice_index = 0; + lightmap = nullptr; + lightmap_cull_index = 0; + lod_bias = 1.0; + scenario = nullptr; update_aabb = false; @@ -399,6 +490,10 @@ public: pair_check = 0; array_index = -1; + + dependency_tracker.userdata = this; + dependency_tracker.changed_callback = dependency_changed; + dependency_tracker.deleted_callback = dependency_deleted; } ~Instance() { @@ -415,6 +510,7 @@ public: void _instance_queue_update(Instance *p_instance, bool p_update_aabb, bool p_update_dependencies = false); struct InstanceGeometryData : public InstanceBaseData { + RendererSceneRender::GeometryInstance *geometry_instance = nullptr; Set<Instance *> lights; bool can_cast_shadows; bool material_is_animated; @@ -458,6 +554,10 @@ public: SelfList<InstanceReflectionProbeData>::List reflection_probe_render_list; + struct InstanceParticlesCollisionData : public InstanceBaseData { + RID instance; + }; + struct InstanceLightData : public InstanceBaseData { RID instance; uint64_t last_version; @@ -523,6 +623,7 @@ public: SelfList<InstanceGIProbeData>::List gi_probe_update_list; struct InstanceLightmapData : public InstanceBaseData { + RID instance; Set<Instance *> geometries; Set<Instance *> users; @@ -588,38 +689,144 @@ public: } }; - struct CullResult { - PagedArray<Instance *> *result; - _FORCE_INLINE_ bool operator()(void *p_data) { - Instance *p_instance = (Instance *)p_data; - result->push_back(p_instance); - return false; - } - }; - Set<Instance *> heightfield_particle_colliders_update_list; PagedArrayPool<Instance *> instance_cull_page_pool; - PagedArrayPool<RendererSceneRender::InstanceBase *> base_instance_cull_page_pool; + PagedArrayPool<RendererSceneRender::GeometryInstance *> geometry_instance_cull_page_pool; PagedArrayPool<RID> rid_cull_page_pool; PagedArray<Instance *> instance_cull_result; - PagedArray<RID> mesh_instance_cull_result; - PagedArray<RendererSceneRender::InstanceBase *> geometry_instances_to_render; PagedArray<Instance *> instance_shadow_cull_result; - PagedArray<RendererSceneRender::InstanceBase *> geometry_instances_to_shadow_render; - PagedArray<Instance *> instance_sdfgi_cull_result; - PagedArray<Instance *> light_cull_result; - PagedArray<RendererSceneRender::InstanceBase *> lightmap_cull_result; - PagedArray<RID> reflection_probe_instance_cull_result; - PagedArray<RID> light_instance_cull_result; - PagedArray<RID> gi_probe_instance_cull_result; - PagedArray<RID> decal_instance_cull_result; + struct FrustumCullResult { + PagedArray<RendererSceneRender::GeometryInstance *> geometry_instances; + PagedArray<Instance *> lights; + PagedArray<RID> light_instances; + PagedArray<RID> lightmaps; + PagedArray<RID> reflections; + PagedArray<RID> decals; + PagedArray<RID> gi_probes; + PagedArray<RID> mesh_instances; + + struct DirectionalShadow { + PagedArray<RendererSceneRender::GeometryInstance *> cascade_geometry_instances[RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES]; + } directional_shadows[RendererSceneRender::MAX_DIRECTIONAL_LIGHTS]; + + PagedArray<RendererSceneRender::GeometryInstance *> sdfgi_region_geometry_instances[SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE]; + PagedArray<RID> sdfgi_cascade_lights[SDFGI_MAX_CASCADES]; + + void clear() { + geometry_instances.clear(); + lights.clear(); + light_instances.clear(); + lightmaps.clear(); + reflections.clear(); + decals.clear(); + gi_probes.clear(); + mesh_instances.clear(); + for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) { + for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) { + directional_shadows[i].cascade_geometry_instances[j].clear(); + } + } + + for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) { + sdfgi_region_geometry_instances[i].clear(); + } + + for (int i = 0; i < SDFGI_MAX_CASCADES; i++) { + sdfgi_cascade_lights[i].clear(); + } + } + + void reset() { + geometry_instances.reset(); + lights.reset(); + light_instances.reset(); + lightmaps.reset(); + reflections.reset(); + decals.reset(); + gi_probes.reset(); + mesh_instances.reset(); + for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) { + for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) { + directional_shadows[i].cascade_geometry_instances[j].reset(); + } + } + + for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) { + sdfgi_region_geometry_instances[i].reset(); + } + + for (int i = 0; i < SDFGI_MAX_CASCADES; i++) { + sdfgi_cascade_lights[i].reset(); + } + } + + void append_from(FrustumCullResult &p_cull_result) { + geometry_instances.merge_unordered(p_cull_result.geometry_instances); + lights.merge_unordered(p_cull_result.lights); + light_instances.merge_unordered(p_cull_result.light_instances); + lightmaps.merge_unordered(p_cull_result.lightmaps); + reflections.merge_unordered(p_cull_result.reflections); + decals.merge_unordered(p_cull_result.decals); + gi_probes.merge_unordered(p_cull_result.gi_probes); + mesh_instances.merge_unordered(p_cull_result.mesh_instances); + + for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) { + for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) { + directional_shadows[i].cascade_geometry_instances[j].merge_unordered(p_cull_result.directional_shadows[i].cascade_geometry_instances[j]); + } + } + + for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) { + sdfgi_region_geometry_instances[i].merge_unordered(p_cull_result.sdfgi_region_geometry_instances[i]); + } + + for (int i = 0; i < SDFGI_MAX_CASCADES; i++) { + sdfgi_cascade_lights[i].merge_unordered(p_cull_result.sdfgi_cascade_lights[i]); + } + } + + void init(PagedArrayPool<RID> *p_rid_pool, PagedArrayPool<RendererSceneRender::GeometryInstance *> *p_geometry_instance_pool, PagedArrayPool<Instance *> *p_instance_pool) { + geometry_instances.set_page_pool(p_geometry_instance_pool); + light_instances.set_page_pool(p_rid_pool); + lights.set_page_pool(p_instance_pool); + lightmaps.set_page_pool(p_rid_pool); + reflections.set_page_pool(p_rid_pool); + decals.set_page_pool(p_rid_pool); + gi_probes.set_page_pool(p_rid_pool); + mesh_instances.set_page_pool(p_rid_pool); + for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) { + for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) { + directional_shadows[i].cascade_geometry_instances[j].set_page_pool(p_geometry_instance_pool); + } + } + + for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) { + sdfgi_region_geometry_instances[i].set_page_pool(p_geometry_instance_pool); + } + + for (int i = 0; i < SDFGI_MAX_CASCADES; i++) { + sdfgi_cascade_lights[i].set_page_pool(p_rid_pool); + } + } + }; + + FrustumCullResult frustum_cull_result; + LocalVector<FrustumCullResult> frustum_cull_result_threads; + + RendererSceneRender::RenderShadowData render_shadow_data[MAX_UPDATE_SHADOWS]; + uint32_t max_shadows_used = 0; + + RendererSceneRender::RenderSDFGIData render_sdfgi_data[SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE]; + RendererSceneRender::RenderSDFGIUpdateData sdfgi_update_data; + + uint32_t thread_cull_threshold = 200; RID_PtrOwner<Instance> instance_owner; - bool pair_volumes_to_mesh; // used in traditional forward, unnecesary on clustered + uint32_t geometry_instance_pair_mask; // used in traditional forward, unnecesary on clustered virtual RID instance_create(); @@ -653,7 +860,7 @@ public: virtual void instance_geometry_set_lightmap(RID p_instance, RID p_lightmap, const Rect2 &p_lightmap_uv_scale, int p_slice_index); virtual void instance_geometry_set_lod_bias(RID p_instance, float p_lod_bias); - void _update_instance_shader_parameters_from_material(Map<StringName, RendererSceneRender::InstanceBase::InstanceShaderParameter> &isparams, const Map<StringName, RendererSceneRender::InstanceBase::InstanceShaderParameter> &existing_isparams, RID p_material); + void _update_instance_shader_parameters_from_material(Map<StringName, Instance::InstanceShaderParameter> &isparams, const Map<StringName, Instance::InstanceShaderParameter> &existing_isparams, RID p_material); virtual void instance_geometry_set_shader_parameter(RID p_instance, const StringName &p_parameter, const Variant &p_value); virtual void instance_geometry_get_shader_parameter_list(RID p_instance, List<PropertyInfo> *p_parameters) const; @@ -687,8 +894,6 @@ public: real_t range_begin; Vector2 uv_scale; - PagedArray<RendererSceneRender::InstanceBase *> cull_result; - } cascades[RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES]; //max 4 cascades uint32_t cascade_count; @@ -698,12 +903,10 @@ public: struct SDFGI { //have arrays here because SDFGI functions expects this, plus regions can have areas - PagedArray<RendererSceneRender::InstanceBase *> region_cull_result[SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE]; AABB region_aabb[SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE]; //max 3 regions per cascade uint32_t region_cascade[SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE]; //max 3 regions per cascade uint32_t region_count = 0; - PagedArray<RID> cascade_lights[SDFGI_MAX_CASCADES]; uint32_t cascade_light_index[SDFGI_MAX_CASCADES]; uint32_t cascade_light_count = 0; @@ -714,9 +917,20 @@ public: Frustum frustum; } cull; + struct FrustumCullData { + Cull *cull; + Scenario *scenario; + RID shadow_atlas; + Transform cam_transform; + uint32_t visible_layers; + Instance *render_reflection_probe; + }; + + void _frustum_cull_threaded(uint32_t p_thread, FrustumCullData *cull_data); + void _frustum_cull(FrustumCullData &cull_data, FrustumCullResult &cull_result, uint64_t p_from, uint64_t p_to); + bool _render_reflection_probe_step(Instance *p_instance, int p_step); - void _prepare_scene(const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect, RID p_render_buffers, RID p_environment, uint32_t p_visible_layers, RID p_scenario, RID p_shadow_atlas, RID p_reflection_probe, float p_screen_lod_threshold, bool p_using_shadows = true); - void _render_scene(RID p_render_buffers, const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, RID p_environment, RID p_force_camera_effects, RID p_scenario, RID p_shadow_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold); + void _render_scene(const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect, RID p_render_buffers, RID p_environment, RID p_force_camera_effects, uint32_t p_visible_layers, RID p_scenario, RID p_shadow_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, bool p_using_shadows = true); void render_empty_scene(RID p_render_buffers, RID p_scenario, RID p_shadow_atlas); void render_camera(RID p_render_buffers, RID p_camera, RID p_scenario, Size2 p_viewport_size, float p_screen_lod_threshold, RID p_shadow_atlas); @@ -738,7 +952,7 @@ public: #define PASSBASE scene_render - PASS1(directional_shadow_atlas_set_size, int) + PASS2(directional_shadow_atlas_set_size, int, bool) PASS1(gi_probe_set_quality, RS::GIProbeQuality) /* SKY API */ @@ -777,16 +991,15 @@ public: PASS7(environment_set_adjustment, RID, bool, float, float, float, bool, RID) PASS9(environment_set_fog, RID, bool, const Color &, float, float, float, float, float, float) - PASS9(environment_set_volumetric_fog, RID, bool, float, const Color &, float, float, float, float, RS::EnvVolumetricFogShadowFilter) + PASS10(environment_set_volumetric_fog, RID, bool, float, const Color &, float, float, float, float, bool, float) PASS2(environment_set_volumetric_fog_volume_size, int, int) PASS1(environment_set_volumetric_fog_filter_active, bool) - PASS1(environment_set_volumetric_fog_directional_shadow_shrink_size, int) - PASS1(environment_set_volumetric_fog_positional_shadow_shrink_size, int) PASS11(environment_set_sdfgi, RID, bool, RS::EnvironmentSDFGICascades, float, RS::EnvironmentSDFGIYScale, bool, bool, bool, float, float, float) PASS1(environment_set_sdfgi_ray_count, RS::EnvironmentSDFGIRayCount) PASS1(environment_set_sdfgi_frames_to_converge, RS::EnvironmentSDFGIFramesToConverge) + PASS1(environment_set_sdfgi_frames_to_update_light, RS::EnvironmentSDFGIFramesToUpdateLight) PASS1RC(RS::EnvironmentBG, environment_get_background, RID) PASS1RC(int, environment_get_canvas_max_layer, RID) @@ -816,10 +1029,11 @@ public: PASS0R(RID, render_buffers_create) PASS7(render_buffers_configure, RID, RID, int, int, RS::ViewportMSAA, RS::ViewportScreenSpaceAA, bool) + PASS1(gi_set_use_half_resolution, bool) /* Shadow Atlas */ PASS0R(RID, shadow_atlas_create) - PASS2(shadow_atlas_set_size, RID, int) + PASS3(shadow_atlas_set_size, RID, int, bool) PASS3(shadow_atlas_set_quadrant_subdivision, RID, int, int) PASS1(set_debug_draw_mode, RS::ViewportDebugDraw) @@ -828,6 +1042,8 @@ public: bool free(RID p_rid); + void set_scene_render(RendererSceneRender *p_scene_render); + RendererSceneCull(); virtual ~RendererSceneCull(); }; diff --git a/servers/rendering/renderer_scene_render.h b/servers/rendering/renderer_scene_render.h index 805d3dcfce..72fcdd3758 100644 --- a/servers/rendering/renderer_scene_render.h +++ b/servers/rendering/renderer_scene_render.h @@ -41,27 +41,52 @@ public: MAX_DIRECTIONAL_LIGHTS = 8, MAX_DIRECTIONAL_LIGHT_CASCADES = 4 }; + + struct GeometryInstance { + virtual ~GeometryInstance() {} + }; + + virtual GeometryInstance *geometry_instance_create(RID p_base) = 0; + virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) = 0; + virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) = 0; + virtual void geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_material) = 0; + virtual void geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance) = 0; + virtual void geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabbb) = 0; + virtual void geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) = 0; + virtual void geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias) = 0; + virtual void geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable) = 0; + virtual void geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable) = 0; + virtual void geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) = 0; + virtual void geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) = 0; + virtual void geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) = 0; + virtual void geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) = 0; + + virtual uint32_t geometry_instance_get_pair_mask() = 0; + virtual void geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) = 0; + virtual void geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) = 0; + virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) = 0; + virtual void geometry_instance_pair_gi_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_gi_probe_instances, uint32_t p_gi_probe_instance_count) = 0; + + virtual void geometry_instance_free(GeometryInstance *p_geometry_instance) = 0; + /* SHADOW ATLAS API */ virtual RID shadow_atlas_create() = 0; - virtual void shadow_atlas_set_size(RID p_atlas, int p_size) = 0; + virtual void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits = false) = 0; virtual void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) = 0; virtual bool shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version) = 0; - virtual void directional_shadow_atlas_set_size(int p_size) = 0; + virtual void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false) = 0; virtual int get_directional_light_shadow_size(RID p_light_intance) = 0; virtual void set_directional_shadow_count(int p_count) = 0; /* SDFGI UPDATE */ - struct InstanceBase; - virtual void sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position) = 0; virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const = 0; virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const = 0; virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const = 0; - virtual void sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count) = 0; /* SKY API */ @@ -92,12 +117,9 @@ public: virtual void environment_glow_set_use_bicubic_upscale(bool p_enable) = 0; virtual void environment_glow_set_use_high_quality(bool p_enable) = 0; - virtual void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RS::EnvVolumetricFogShadowFilter p_shadow_filter) = 0; - + virtual void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) = 0; virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth) = 0; virtual void environment_set_volumetric_fog_filter_active(bool p_enable) = 0; - virtual void environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size) = 0; - virtual void environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size) = 0; virtual void environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance) = 0; virtual void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) = 0; @@ -110,6 +132,7 @@ public: virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) = 0; virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) = 0; + virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) = 0; virtual void environment_set_tonemap(RID p_env, RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale) = 0; @@ -134,83 +157,6 @@ public: virtual void shadows_quality_set(RS::ShadowQuality p_quality) = 0; virtual void directional_shadow_quality_set(RS::ShadowQuality p_quality) = 0; - struct InstanceBase : public RendererStorage::InstanceBaseDependency { - RS::InstanceType base_type; - RID base; - - RID skeleton; - RID material_override; - - RID mesh_instance; //only used for meshes and when skeleton/blendshapes exist - - Transform transform; - - float lod_bias; - - int depth_layer; - uint32_t layer_mask; - - //RID sampled_light; - - Vector<RID> materials; - Vector<RID> light_instances; - Vector<RID> reflection_probe_instances; - Vector<RID> gi_probe_instances; - - RS::ShadowCastingSetting cast_shadows; - - //fit in 32 bits - bool mirror : 8; - bool receive_shadows : 8; - bool visible : 8; - bool baked_light : 2; //this flag is only to know if it actually did use baked light - bool dynamic_gi : 2; //this flag is only to know if it actually did use baked light - bool redraw_if_visible : 4; - - float depth; //used for sorting - - InstanceBase *lightmap; - Rect2 lightmap_uv_scale; - int lightmap_slice_index; - uint32_t lightmap_cull_index; - Vector<Color> lightmap_sh; //spherical harmonic - - AABB aabb; - AABB transformed_aabb; - AABB prev_transformed_aabb; - - struct InstanceShaderParameter { - int32_t index = -1; - Variant value; - Variant default_value; - PropertyInfo info; - }; - - Map<StringName, InstanceShaderParameter> instance_shader_parameters; - bool instance_allocated_shader_parameters = false; - int32_t instance_allocated_shader_parameters_offset = -1; - - InstanceBase() { - base_type = RS::INSTANCE_NONE; - cast_shadows = RS::SHADOW_CASTING_SETTING_ON; - receive_shadows = true; - visible = true; - depth_layer = 0; - layer_mask = 1; - instance_version = 0; - baked_light = false; - dynamic_gi = false; - redraw_if_visible = false; - lightmap_slice_index = 0; - lightmap = nullptr; - lightmap_cull_index = 0; - lod_bias = 1.0; - } - - virtual ~InstanceBase() { - } - }; - virtual RID light_instance_create(RID p_light) = 0; virtual void light_instance_set_transform(RID p_light_instance, const Transform &p_transform) = 0; virtual void light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb) = 0; @@ -235,20 +181,42 @@ public: virtual RID decal_instance_create(RID p_decal) = 0; virtual void decal_instance_set_transform(RID p_decal, const Transform &p_transform) = 0; + virtual RID lightmap_instance_create(RID p_lightmap) = 0; + virtual void lightmap_instance_set_transform(RID p_lightmap, const Transform &p_transform) = 0; + virtual RID gi_probe_instance_create(RID p_gi_probe) = 0; virtual void gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform) = 0; virtual bool gi_probe_needs_update(RID p_probe) const = 0; - virtual void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::InstanceBase *> &p_dynamic_objects) = 0; + virtual void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<GeometryInstance *> &p_dynamic_objects) = 0; virtual void gi_probe_set_quality(RS::GIProbeQuality) = 0; - virtual void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) = 0; + struct RenderShadowData { + RID light; + int pass = 0; + PagedArray<GeometryInstance *> instances; + }; + + struct RenderSDFGIData { + int region = 0; + PagedArray<GeometryInstance *> instances; + }; + + struct RenderSDFGIUpdateData { + bool update_static = false; + uint32_t static_cascade_count; + uint32_t *static_cascade_indices; + PagedArray<RID> *static_positional_lights; + + const Vector<RID> *directional_lights; + const RID *positional_light_instances; + uint32_t positional_light_count; + }; + + virtual void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data = nullptr) = 0; - virtual void render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<InstanceBase *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0) = 0; - virtual void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; - virtual void render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<InstanceBase *> &p_instances) = 0; - virtual void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_lights) = 0; - virtual void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<InstanceBase *> &p_instances) = 0; + virtual void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; + virtual void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances) = 0; virtual void set_scene_pass(uint64_t p_pass) = 0; virtual void set_time(double p_time, double p_step) = 0; @@ -256,6 +224,7 @@ public: virtual RID render_buffers_create() = 0; virtual void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding) = 0; + virtual void gi_set_use_half_resolution(bool p_enable) = 0; virtual void screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit) = 0; virtual bool screen_space_roughness_limiter_is_active() const = 0; diff --git a/servers/rendering/renderer_storage.cpp b/servers/rendering/renderer_storage.cpp index 2edf62df56..a402ecc668 100644 --- a/servers/rendering/renderer_storage.cpp +++ b/servers/rendering/renderer_storage.cpp @@ -32,28 +32,31 @@ RendererStorage *RendererStorage::base_singleton = nullptr; -void RendererStorage::InstanceDependency::instance_notify_changed(bool p_aabb, bool p_dependencies) { - for (Map<InstanceBaseDependency *, uint32_t>::Element *E = instances.front(); E; E = E->next()) { - E->key()->dependency_changed(p_aabb, p_dependencies); +void RendererStorage::Dependency::changed_notify(DependencyChangedNotification p_notification) { + for (Map<DependencyTracker *, uint32_t>::Element *E = instances.front(); E; E = E->next()) { + if (E->key()->changed_callback) { + E->key()->changed_callback(p_notification, E->key()); + } } } -void RendererStorage::InstanceDependency::instance_notify_deleted(RID p_deleted) { - for (Map<InstanceBaseDependency *, uint32_t>::Element *E = instances.front(); E; E = E->next()) { - E->key()->dependency_deleted(p_deleted); +void RendererStorage::Dependency::deleted_notify(const RID &p_rid) { + for (Map<DependencyTracker *, uint32_t>::Element *E = instances.front(); E; E = E->next()) { + if (E->key()->deleted_callback) { + E->key()->deleted_callback(p_rid, E->key()); + } } - for (Map<InstanceBaseDependency *, uint32_t>::Element *E = instances.front(); E; E = E->next()) { + for (Map<DependencyTracker *, uint32_t>::Element *E = instances.front(); E; E = E->next()) { E->key()->dependencies.erase(this); } - instances.clear(); } -RendererStorage::InstanceDependency::~InstanceDependency() { +RendererStorage::Dependency::~Dependency() { #ifdef DEBUG_ENABLED if (instances.size()) { WARN_PRINT("Leaked instance dependency: Bug - did not call instance_notify_deleted when freeing."); - for (Map<InstanceBaseDependency *, uint32_t>::Element *E = instances.front(); E; E = E->next()) { + for (Map<DependencyTracker *, uint32_t>::Element *E = instances.front(); E; E = E->next()) { E->key()->dependencies.erase(this); } } diff --git a/servers/rendering/renderer_storage.h b/servers/rendering/renderer_storage.h index 835bf32863..f015b50eee 100644 --- a/servers/rendering/renderer_storage.h +++ b/servers/rendering/renderer_storage.h @@ -37,43 +37,59 @@ class RendererStorage { Color default_clear_color; public: - struct InstanceBaseDependency; + enum DependencyChangedNotification { + DEPENDENCY_CHANGED_AABB, + DEPENDENCY_CHANGED_MATERIAL, + DEPENDENCY_CHANGED_MESH, + DEPENDENCY_CHANGED_MULTIMESH, + DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES, + DEPENDENCY_CHANGED_DECAL, + DEPENDENCY_CHANGED_SKELETON_DATA, + DEPENDENCY_CHANGED_SKELETON_BONES, + DEPENDENCY_CHANGED_LIGHT, + DEPENDENCY_CHANGED_REFLECTION_PROBE, + }; + + struct DependencyTracker; - struct InstanceDependency { - void instance_notify_changed(bool p_aabb, bool p_dependencies); - void instance_notify_deleted(RID p_deleted); +protected: + struct Dependency { + void changed_notify(DependencyChangedNotification p_notification); + void deleted_notify(const RID &p_rid); - ~InstanceDependency(); + ~Dependency(); private: - friend struct InstanceBaseDependency; - Map<InstanceBaseDependency *, uint32_t> instances; + friend struct DependencyTracker; + Map<DependencyTracker *, uint32_t> instances; }; - struct InstanceBaseDependency { - uint32_t instance_version; - Set<InstanceDependency *> dependencies; +public: + struct DependencyTracker { + void *userdata = nullptr; + typedef void (*ChangedCallback)(DependencyChangedNotification, DependencyTracker *); + typedef void (*DeletedCallback)(const RID &, DependencyTracker *); - virtual void dependency_deleted(RID p_dependency) {} - virtual void dependency_changed(bool p_aabb, bool p_dependencies) {} + ChangedCallback changed_callback = nullptr; + DeletedCallback deleted_callback = nullptr; - void instance_increase_version() { + void update_begin() { // call before updating dependencies instance_version++; } - void update_dependency(InstanceDependency *p_dependency) { + void update_dependency(Dependency *p_dependency) { //called internally, can't be used directly, use update functions in Storage dependencies.insert(p_dependency); p_dependency->instances[this] = instance_version; } - void clean_up_dependencies() { - List<Pair<InstanceDependency *, Map<InstanceBaseDependency *, uint32_t>::Element *>> to_clean_up; - for (Set<InstanceDependency *>::Element *E = dependencies.front(); E; E = E->next()) { - InstanceDependency *dep = E->get(); - Map<InstanceBaseDependency *, uint32_t>::Element *F = dep->instances.find(this); + void update_end() { //call after updating dependencies + List<Pair<Dependency *, Map<DependencyTracker *, uint32_t>::Element *>> to_clean_up; + for (Set<Dependency *>::Element *E = dependencies.front(); E; E = E->next()) { + Dependency *dep = E->get(); + Map<DependencyTracker *, uint32_t>::Element *F = dep->instances.find(this); ERR_CONTINUE(!F); if (F->get() != instance_version) { - Pair<InstanceDependency *, Map<InstanceBaseDependency *, uint32_t>::Element *> p; + Pair<Dependency *, Map<DependencyTracker *, uint32_t>::Element *> p; p.first = dep; p.second = F; to_clean_up.push_back(p); @@ -82,19 +98,25 @@ public: while (to_clean_up.size()) { to_clean_up.front()->get().first->instances.erase(to_clean_up.front()->get().second); + dependencies.erase(to_clean_up.front()->get().first); to_clean_up.pop_front(); } } - void clear_dependencies() { - for (Set<InstanceDependency *>::Element *E = dependencies.front(); E; E = E->next()) { - InstanceDependency *dep = E->get(); + void clear() { // clear all dependencies + for (Set<Dependency *>::Element *E = dependencies.front(); E; E = E->next()) { + Dependency *dep = E->get(); dep->instances.erase(this); } dependencies.clear(); } - virtual ~InstanceBaseDependency() { clear_dependencies(); } + ~DependencyTracker() { clear(); } + + private: + friend struct Dependency; + uint32_t instance_version = 0; + Set<Dependency *> dependencies; }; /* TEXTURE API */ @@ -158,6 +180,8 @@ public: virtual RID shader_get_default_texture_param(RID p_shader, const StringName &p_name) const = 0; virtual Variant shader_get_param_default(RID p_material, const StringName &p_param) const = 0; + virtual RS::ShaderNativeSourceCode shader_get_native_source_code(RID p_shader) const = 0; + /* COMMON MATERIAL API */ virtual RID material_create() = 0; @@ -181,7 +205,7 @@ public: virtual void material_get_instance_shader_parameters(RID p_material, List<InstanceShaderParam> *r_parameters) = 0; - virtual void material_update_dependency(RID p_material, InstanceBaseDependency *p_instance) = 0; + virtual void material_update_dependency(RID p_material, DependencyTracker *p_instance) = 0; /* MESH API */ @@ -211,6 +235,8 @@ public: virtual AABB mesh_get_aabb(RID p_mesh, RID p_skeleton = RID()) = 0; + virtual void mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh) = 0; + virtual void mesh_clear(RID p_mesh) = 0; virtual bool mesh_needs_instance(RID p_mesh, bool p_has_skeleton) = 0; @@ -349,8 +375,8 @@ public: virtual bool reflection_probe_renders_shadows(RID p_probe) const = 0; virtual float reflection_probe_get_lod_threshold(RID p_probe) const = 0; - virtual void base_update_dependency(RID p_base, InstanceBaseDependency *p_instance) = 0; - virtual void skeleton_update_dependency(RID p_base, InstanceBaseDependency *p_instance) = 0; + virtual void base_update_dependency(RID p_base, DependencyTracker *p_instance) = 0; + virtual void skeleton_update_dependency(RID p_base, DependencyTracker *p_instance) = 0; /* DECAL API */ @@ -474,8 +500,8 @@ public: virtual void particles_set_view_axis(RID p_particles, const Vector3 &p_axis) = 0; - virtual void particles_add_collision(RID p_particles, InstanceBaseDependency *p_instance) = 0; - virtual void particles_remove_collision(RID p_particles, InstanceBaseDependency *p_instance) = 0; + virtual void particles_add_collision(RID p_particles, RID p_particles_collision_instance) = 0; + virtual void particles_remove_collision(RID p_particles, RID p_particles_collision_instance) = 0; virtual void update_particles() = 0; @@ -496,6 +522,11 @@ public: virtual bool particles_collision_is_heightfield(RID p_particles_collision) const = 0; virtual RID particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const = 0; + //used from 2D and 3D + virtual RID particles_collision_instance_create(RID p_collision) = 0; + virtual void particles_collision_instance_set_transform(RID p_collision_instance, const Transform &p_transform) = 0; + virtual void particles_collision_instance_set_active(RID p_collision_instance, bool p_active) = 0; + /* GLOBAL VARIABLES */ virtual void global_variable_add(const StringName &p_name, RS::GlobalVariableType p_type, const Variant &p_value) = 0; diff --git a/servers/rendering/renderer_thread_pool.cpp b/servers/rendering/renderer_thread_pool.cpp new file mode 100644 index 0000000000..98050dd508 --- /dev/null +++ b/servers/rendering/renderer_thread_pool.cpp @@ -0,0 +1,42 @@ +/*************************************************************************/ +/* renderer_thread_pool.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "renderer_thread_pool.h" + +RendererThreadPool *RendererThreadPool::singleton = nullptr; + +RendererThreadPool::RendererThreadPool() { + singleton = this; + thread_work_pool.init(); +} + +RendererThreadPool::~RendererThreadPool() { + thread_work_pool.finish(); +} diff --git a/servers/rendering/renderer_thread_pool.h b/servers/rendering/renderer_thread_pool.h new file mode 100644 index 0000000000..ae25415a0d --- /dev/null +++ b/servers/rendering/renderer_thread_pool.h @@ -0,0 +1,45 @@ +/*************************************************************************/ +/* renderer_thread_pool.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDERERTHREADPOOL_H +#define RENDERERTHREADPOOL_H + +#include "core/templates/thread_work_pool.h" + +class RendererThreadPool { +public: + ThreadWorkPool thread_work_pool; + + static RendererThreadPool *singleton; + RendererThreadPool(); + ~RendererThreadPool(); +}; + +#endif // RENDERERTHREADPOOL_H diff --git a/servers/rendering/renderer_viewport.cpp b/servers/rendering/renderer_viewport.cpp index 9956e4050b..d52da5b331 100644 --- a/servers/rendering/renderer_viewport.cpp +++ b/servers/rendering/renderer_viewport.cpp @@ -831,13 +831,14 @@ void RendererViewport::viewport_set_canvas_stacking(RID p_viewport, RID p_canvas viewport->canvas_map[p_canvas].sublayer = p_sublayer; } -void RendererViewport::viewport_set_shadow_atlas_size(RID p_viewport, int p_size) { +void RendererViewport::viewport_set_shadow_atlas_size(RID p_viewport, int p_size, bool p_16_bits) { Viewport *viewport = viewport_owner.getornull(p_viewport); ERR_FAIL_COND(!viewport); viewport->shadow_atlas_size = p_size; + viewport->shadow_atlas_16_bits = p_16_bits; - RSG::scene->shadow_atlas_set_size(viewport->shadow_atlas, viewport->shadow_atlas_size); + RSG::scene->shadow_atlas_set_size(viewport->shadow_atlas, viewport->shadow_atlas_size, viewport->shadow_atlas_16_bits); } void RendererViewport::viewport_set_shadow_atlas_quadrant_subdivision(RID p_viewport, int p_quadrant, int p_subdiv) { diff --git a/servers/rendering/renderer_viewport.h b/servers/rendering/renderer_viewport.h index c3ff52a836..979cbb095b 100644 --- a/servers/rendering/renderer_viewport.h +++ b/servers/rendering/renderer_viewport.h @@ -81,6 +81,7 @@ public: RID shadow_atlas; int shadow_atlas_size; + bool shadow_atlas_16_bits = false; bool sdf_active; @@ -217,7 +218,7 @@ public: void viewport_set_global_canvas_transform(RID p_viewport, const Transform2D &p_transform); void viewport_set_canvas_stacking(RID p_viewport, RID p_canvas, int p_layer, int p_sublayer); - void viewport_set_shadow_atlas_size(RID p_viewport, int p_size); + void viewport_set_shadow_atlas_size(RID p_viewport, int p_size, bool p_16_bits = false); void viewport_set_shadow_atlas_quadrant_subdivision(RID p_viewport, int p_quadrant, int p_subdiv); void viewport_set_msaa(RID p_viewport, RS::ViewportMSAA p_msaa); diff --git a/servers/rendering/rendering_device.cpp b/servers/rendering/rendering_device.cpp index 73c86a0a1d..4b0eafe369 100644 --- a/servers/rendering/rendering_device.cpp +++ b/servers/rendering/rendering_device.cpp @@ -174,8 +174,8 @@ RID RenderingDevice::_uniform_set_create(const Array &p_uniforms, RID p_shader, return uniform_set_create(uniforms, p_shader, p_shader_set); } -Error RenderingDevice::_buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const Vector<uint8_t> &p_data, bool p_sync_with_draw) { - return buffer_update(p_buffer, p_offset, p_size, p_data.ptr(), p_sync_with_draw); +Error RenderingDevice::_buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const Vector<uint8_t> &p_data, uint32_t p_post_barrier) { + return buffer_update(p_buffer, p_offset, p_size, p_data.ptr(), p_post_barrier); } RID RenderingDevice::_render_pipeline_create(RID p_shader, FramebufferFormatID p_framebuffer_format, VertexFormatID p_vertex_format, RenderPrimitive p_render_primitive, const Ref<RDPipelineRasterizationState> &p_rasterization_state, const Ref<RDPipelineMultisampleState> &p_multisample_state, const Ref<RDPipelineDepthStencilState> &p_depth_stencil_state, const Ref<RDPipelineColorBlendState> &p_blend_state, int p_dynamic_state_flags) { @@ -240,16 +240,12 @@ void RenderingDevice::_compute_list_set_push_constant(ComputeListID p_list, cons compute_list_set_push_constant(p_list, p_data.ptr(), p_data_size); } -void RenderingDevice::compute_list_dispatch_threads(ComputeListID p_list, uint32_t p_x_threads, uint32_t p_y_threads, uint32_t p_z_threads, uint32_t p_x_local_group, uint32_t p_y_local_group, uint32_t p_z_local_group) { - compute_list_dispatch(p_list, (p_x_threads - 1) / p_x_local_group + 1, (p_y_threads - 1) / p_y_local_group + 1, (p_z_threads - 1) / p_z_local_group + 1); -} - void RenderingDevice::_bind_methods() { ClassDB::bind_method(D_METHOD("texture_create", "format", "view", "data"), &RenderingDevice::_texture_create, DEFVAL(Array())); ClassDB::bind_method(D_METHOD("texture_create_shared", "view", "with_texture"), &RenderingDevice::_texture_create_shared); ClassDB::bind_method(D_METHOD("texture_create_shared_from_slice", "view", "with_texture", "layer", "mipmap", "slice_type"), &RenderingDevice::_texture_create_shared_from_slice, DEFVAL(TEXTURE_SLICE_2D)); - ClassDB::bind_method(D_METHOD("texture_update", "texture", "layer", "data", "sync_with_draw"), &RenderingDevice::texture_update, DEFVAL(false)); + ClassDB::bind_method(D_METHOD("texture_update", "texture", "layer", "data", "post_barrier"), &RenderingDevice::texture_update, DEFVAL(BARRIER_MASK_ALL)); ClassDB::bind_method(D_METHOD("texture_get_data", "texture", "layer"), &RenderingDevice::texture_get_data); ClassDB::bind_method(D_METHOD("texture_is_format_supported_for_usage", "format", "usage_flags"), &RenderingDevice::texture_is_format_supported_for_usage); @@ -257,15 +253,15 @@ void RenderingDevice::_bind_methods() { ClassDB::bind_method(D_METHOD("texture_is_shared", "texture"), &RenderingDevice::texture_is_shared); ClassDB::bind_method(D_METHOD("texture_is_valid", "texture"), &RenderingDevice::texture_is_valid); - ClassDB::bind_method(D_METHOD("texture_copy", "from_texture", "to_texture", "from_pos", "to_pos", "size", "src_mipmap", "dst_mipmap", "src_layer", "dst_layer", "sync_with_draw"), &RenderingDevice::texture_copy, DEFVAL(false)); - ClassDB::bind_method(D_METHOD("texture_clear", "texture", "color", "base_mipmap", "mipmap_count", "base_layer", "layer_count", "sync_with_draw"), &RenderingDevice::texture_clear, DEFVAL(false)); - ClassDB::bind_method(D_METHOD("texture_resolve_multisample", "from_texture", "to_texture", "sync_with_draw"), &RenderingDevice::texture_resolve_multisample, DEFVAL(false)); + ClassDB::bind_method(D_METHOD("texture_copy", "from_texture", "to_texture", "from_pos", "to_pos", "size", "src_mipmap", "dst_mipmap", "src_layer", "dst_layer", "post_barrier"), &RenderingDevice::texture_copy, DEFVAL(BARRIER_MASK_ALL)); + ClassDB::bind_method(D_METHOD("texture_clear", "texture", "color", "base_mipmap", "mipmap_count", "base_layer", "layer_count", "post_barrier"), &RenderingDevice::texture_clear, DEFVAL(BARRIER_MASK_ALL)); + ClassDB::bind_method(D_METHOD("texture_resolve_multisample", "from_texture", "to_texture", "post_barrier"), &RenderingDevice::texture_resolve_multisample, DEFVAL(BARRIER_MASK_ALL)); ClassDB::bind_method(D_METHOD("framebuffer_format_create", "attachments"), &RenderingDevice::_framebuffer_format_create); - ClassDB::bind_method(D_METHOD("framebuffer_format_create_empty", "size"), &RenderingDevice::framebuffer_format_create_empty); + ClassDB::bind_method(D_METHOD("framebuffer_format_create_empty", "samples"), &RenderingDevice::framebuffer_format_create_empty, DEFVAL(TEXTURE_SAMPLES_1)); ClassDB::bind_method(D_METHOD("framebuffer_format_get_texture_samples", "format"), &RenderingDevice::framebuffer_format_get_texture_samples); ClassDB::bind_method(D_METHOD("framebuffer_create", "textures", "validate_with_format"), &RenderingDevice::_framebuffer_create, DEFVAL(INVALID_FORMAT_ID)); - ClassDB::bind_method(D_METHOD("framebuffer_create_empty", "size", "validate_with_format"), &RenderingDevice::framebuffer_create_empty, DEFVAL(INVALID_FORMAT_ID)); + ClassDB::bind_method(D_METHOD("framebuffer_create_empty", "size", "samples", "validate_with_format"), &RenderingDevice::framebuffer_create_empty, DEFVAL(TEXTURE_SAMPLES_1), DEFVAL(INVALID_FORMAT_ID)); ClassDB::bind_method(D_METHOD("framebuffer_get_format", "framebuffer"), &RenderingDevice::framebuffer_get_format); ClassDB::bind_method(D_METHOD("sampler_create", "state"), &RenderingDevice::_sampler_create); @@ -287,7 +283,8 @@ void RenderingDevice::_bind_methods() { ClassDB::bind_method(D_METHOD("uniform_set_create", "uniforms", "shader", "shader_set"), &RenderingDevice::_uniform_set_create); ClassDB::bind_method(D_METHOD("uniform_set_is_valid", "uniform_set"), &RenderingDevice::uniform_set_is_valid); - ClassDB::bind_method(D_METHOD("buffer_update", "buffer", "offset", "size_bytes", "data", "sync_with_draw"), &RenderingDevice::_buffer_update, DEFVAL(true)); + ClassDB::bind_method(D_METHOD("buffer_update", "buffer", "offset", "size_bytes", "data", "post_barrier"), &RenderingDevice::_buffer_update, DEFVAL(BARRIER_MASK_ALL)); + ClassDB::bind_method(D_METHOD("buffer_clear", "buffer", "offset", "size_bytes", "post_barrier"), &RenderingDevice::buffer_clear, DEFVAL(BARRIER_MASK_ALL)); ClassDB::bind_method(D_METHOD("buffer_get_data", "buffer"), &RenderingDevice::buffer_get_data); ClassDB::bind_method(D_METHOD("render_pipeline_create", "shader", "framebuffer_format", "vertex_format", "primitive", "rasterization_state", "multisample_state", "stencil_state", "color_blend_state", "dynamic_state_flags"), &RenderingDevice::_render_pipeline_create, DEFVAL(0)); @@ -316,19 +313,19 @@ void RenderingDevice::_bind_methods() { ClassDB::bind_method(D_METHOD("draw_list_enable_scissor", "draw_list", "rect"), &RenderingDevice::draw_list_enable_scissor, DEFVAL(Rect2i())); ClassDB::bind_method(D_METHOD("draw_list_disable_scissor", "draw_list"), &RenderingDevice::draw_list_disable_scissor); - ClassDB::bind_method(D_METHOD("draw_list_end"), &RenderingDevice::draw_list_end); + ClassDB::bind_method(D_METHOD("draw_list_end", "post_barrier"), &RenderingDevice::draw_list_end, DEFVAL(BARRIER_MASK_ALL)); - ClassDB::bind_method(D_METHOD("compute_list_begin"), &RenderingDevice::compute_list_begin); + ClassDB::bind_method(D_METHOD("compute_list_begin", "allow_draw_overlap"), &RenderingDevice::compute_list_begin, DEFVAL(false)); ClassDB::bind_method(D_METHOD("compute_list_bind_compute_pipeline", "compute_list", "compute_pipeline"), &RenderingDevice::compute_list_bind_compute_pipeline); ClassDB::bind_method(D_METHOD("compute_list_set_push_constant", "compute_list", "buffer", "size_bytes"), &RenderingDevice::_compute_list_set_push_constant); ClassDB::bind_method(D_METHOD("compute_list_bind_uniform_set", "compute_list", "uniform_set", "set_index"), &RenderingDevice::compute_list_bind_uniform_set); ClassDB::bind_method(D_METHOD("compute_list_dispatch", "compute_list", "x_groups", "y_groups", "z_groups"), &RenderingDevice::compute_list_dispatch); ClassDB::bind_method(D_METHOD("compute_list_add_barrier", "compute_list"), &RenderingDevice::compute_list_add_barrier); - ClassDB::bind_method(D_METHOD("compute_list_end"), &RenderingDevice::compute_list_end); + ClassDB::bind_method(D_METHOD("compute_list_end", "post_barrier"), &RenderingDevice::compute_list_end, DEFVAL(BARRIER_MASK_ALL)); ClassDB::bind_method(D_METHOD("free", "rid"), &RenderingDevice::free); - ClassDB::bind_method(D_METHOD("capture_timestamp", "name", "sync_to_draw"), &RenderingDevice::capture_timestamp); + ClassDB::bind_method(D_METHOD("capture_timestamp", "name"), &RenderingDevice::capture_timestamp); ClassDB::bind_method(D_METHOD("get_captured_timestamps_count"), &RenderingDevice::get_captured_timestamps_count); ClassDB::bind_method(D_METHOD("get_captured_timestamps_frame"), &RenderingDevice::get_captured_timestamps_frame); ClassDB::bind_method(D_METHOD("get_captured_timestamp_gpu_time", "index"), &RenderingDevice::get_captured_timestamp_gpu_time); @@ -340,8 +337,27 @@ void RenderingDevice::_bind_methods() { ClassDB::bind_method(D_METHOD("submit"), &RenderingDevice::submit); ClassDB::bind_method(D_METHOD("sync"), &RenderingDevice::sync); + ClassDB::bind_method(D_METHOD("barrier", "from", "to"), &RenderingDevice::barrier, DEFVAL(BARRIER_MASK_ALL), DEFVAL(BARRIER_MASK_ALL)); + ClassDB::bind_method(D_METHOD("full_barrier"), &RenderingDevice::full_barrier); + ClassDB::bind_method(D_METHOD("create_local_device"), &RenderingDevice::create_local_device); + ClassDB::bind_method(D_METHOD("set_resource_name", "id", "name"), &RenderingDevice::set_resource_name); + + ClassDB::bind_method(D_METHOD("draw_command_begin_label", "name", "color"), &RenderingDevice::draw_command_begin_label); + ClassDB::bind_method(D_METHOD("draw_command_insert_label", "name", "color"), &RenderingDevice::draw_command_insert_label); + ClassDB::bind_method(D_METHOD("draw_command_end_label"), &RenderingDevice::draw_command_end_label); + + ClassDB::bind_method(D_METHOD("get_device_vendor_name"), &RenderingDevice::get_device_vendor_name); + ClassDB::bind_method(D_METHOD("get_device_name"), &RenderingDevice::get_device_name); + ClassDB::bind_method(D_METHOD("get_device_pipeline_cache_uuid"), &RenderingDevice::get_device_pipeline_cache_uuid); + + BIND_CONSTANT(BARRIER_MASK_RASTER); + BIND_CONSTANT(BARRIER_MASK_COMPUTE); + BIND_CONSTANT(BARRIER_MASK_TRANSFER); + BIND_CONSTANT(BARRIER_MASK_ALL); + BIND_CONSTANT(BARRIER_MASK_NO_BARRIER); + BIND_ENUM_CONSTANT(DATA_FORMAT_R4G4_UNORM_PACK8); BIND_ENUM_CONSTANT(DATA_FORMAT_R4G4B4A4_UNORM_PACK16); BIND_ENUM_CONSTANT(DATA_FORMAT_B4G4R4A4_UNORM_PACK16); @@ -744,6 +760,8 @@ void RenderingDevice::_bind_methods() { BIND_ENUM_CONSTANT(DYNAMIC_STATE_STENCIL_REFERENCE); BIND_ENUM_CONSTANT(INITIAL_ACTION_CLEAR); //start rendering and clear the framebuffer (supply params) + BIND_ENUM_CONSTANT(INITIAL_ACTION_CLEAR_REGION); //start rendering and clear the framebuffer (supply params) + BIND_ENUM_CONSTANT(INITIAL_ACTION_CLEAR_REGION_CONTINUE); //continue rendering and clear the framebuffer (supply params) BIND_ENUM_CONSTANT(INITIAL_ACTION_KEEP); //start rendering); but keep attached color texture contents (depth will be cleared) BIND_ENUM_CONSTANT(INITIAL_ACTION_DROP); //start rendering); ignore what is there); just write above it BIND_ENUM_CONSTANT(INITIAL_ACTION_CONTINUE); //continue rendering (framebuffer must have been left in "continue" state as final action previously) diff --git a/servers/rendering/rendering_device.h b/servers/rendering/rendering_device.h index 787805ea6a..9fbf58d131 100644 --- a/servers/rendering/rendering_device.h +++ b/servers/rendering/rendering_device.h @@ -336,6 +336,18 @@ public: }; /*****************/ + /**** BARRIER ****/ + /*****************/ + + enum BarrierMask { + BARRIER_MASK_RASTER = 1, + BARRIER_MASK_COMPUTE = 2, + BARRIER_MASK_TRANSFER = 4, + BARRIER_MASK_NO_BARRIER = 8, + BARRIER_MASK_ALL = BARRIER_MASK_RASTER | BARRIER_MASK_COMPUTE | BARRIER_MASK_TRANSFER + }; + + /*****************/ /**** TEXTURE ****/ /*****************/ @@ -438,16 +450,16 @@ public: virtual RID texture_create_shared_from_slice(const TextureView &p_view, RID p_with_texture, uint32_t p_layer, uint32_t p_mipmap, TextureSliceType p_slice_type = TEXTURE_SLICE_2D) = 0; - virtual Error texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, bool p_sync_with_draw = false) = 0; //this function can be used from any thread and it takes effect at the beginning of the frame, unless sync with draw is used, which is used to mix updates with draw calls + virtual Error texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, uint32_t p_post_barrier = BARRIER_MASK_ALL) = 0; virtual Vector<uint8_t> texture_get_data(RID p_texture, uint32_t p_layer) = 0; // CPU textures will return immediately, while GPU textures will most likely force a flush virtual bool texture_is_format_supported_for_usage(DataFormat p_format, uint32_t p_usage) const = 0; virtual bool texture_is_shared(RID p_texture) = 0; virtual bool texture_is_valid(RID p_texture) = 0; - virtual Error texture_copy(RID p_from_texture, RID p_to_texture, const Vector3 &p_from, const Vector3 &p_to, const Vector3 &p_size, uint32_t p_src_mipmap, uint32_t p_dst_mipmap, uint32_t p_src_layer, uint32_t p_dst_layer, bool p_sync_with_draw = false) = 0; - virtual Error texture_clear(RID p_texture, const Color &p_color, uint32_t p_base_mipmap, uint32_t p_mipmaps, uint32_t p_base_layer, uint32_t p_layers, bool p_sync_with_draw = false) = 0; - virtual Error texture_resolve_multisample(RID p_from_texture, RID p_to_texture, bool p_sync_with_draw = false) = 0; + virtual Error texture_copy(RID p_from_texture, RID p_to_texture, const Vector3 &p_from, const Vector3 &p_to, const Vector3 &p_size, uint32_t p_src_mipmap, uint32_t p_dst_mipmap, uint32_t p_src_layer, uint32_t p_dst_layer, uint32_t p_post_barrier = BARRIER_MASK_ALL) = 0; + virtual Error texture_clear(RID p_texture, const Color &p_color, uint32_t p_base_mipmap, uint32_t p_mipmaps, uint32_t p_base_layer, uint32_t p_layers, uint32_t p_post_barrier = BARRIER_MASK_ALL) = 0; + virtual Error texture_resolve_multisample(RID p_from_texture, RID p_to_texture, uint32_t p_post_barrier = BARRIER_MASK_ALL) = 0; /*********************/ /**** FRAMEBUFFER ****/ @@ -468,11 +480,11 @@ public: // This ID is warranted to be unique for the same formats, does not need to be freed virtual FramebufferFormatID framebuffer_format_create(const Vector<AttachmentFormat> &p_format) = 0; - virtual FramebufferFormatID framebuffer_format_create_empty(const Size2i &p_size) = 0; + virtual FramebufferFormatID framebuffer_format_create_empty(TextureSamples p_samples = TEXTURE_SAMPLES_1) = 0; virtual TextureSamples framebuffer_format_get_texture_samples(FramebufferFormatID p_format) = 0; virtual RID framebuffer_create(const Vector<RID> &p_texture_attachments, FramebufferFormatID p_format_check = INVALID_ID) = 0; - virtual RID framebuffer_create_empty(const Size2i &p_size, FramebufferFormatID p_format_check = INVALID_ID) = 0; + virtual RID framebuffer_create_empty(const Size2i &p_size, TextureSamples p_samples = TEXTURE_SAMPLES_1, FramebufferFormatID p_format_check = INVALID_ID) = 0; virtual FramebufferFormatID framebuffer_get_format(RID p_framebuffer) = 0; @@ -649,7 +661,8 @@ public: virtual RID uniform_set_create(const Vector<Uniform> &p_uniforms, RID p_shader, uint32_t p_shader_set) = 0; virtual bool uniform_set_is_valid(RID p_uniform_set) = 0; - virtual Error buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const void *p_data, bool p_sync_with_draw = false) = 0; //this function can be used from any thread and it takes effect at the beginning of the frame, unless sync with draw is used, which is used to mix updates with draw calls + virtual Error buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const void *p_data, uint32_t p_post_barrier = BARRIER_MASK_ALL) = 0; + virtual Error buffer_clear(RID p_buffer, uint32_t p_offset, uint32_t p_size, uint32_t p_post_barrier = BARRIER_MASK_ALL) = 0; virtual Vector<uint8_t> buffer_get_data(RID p_buffer) = 0; //this causes stall, only use to retrieve large buffers for saving /*************************/ @@ -930,7 +943,9 @@ public: /********************/ enum InitialAction { - INITIAL_ACTION_CLEAR, //start rendering and clear the framebuffer (supply params) + INITIAL_ACTION_CLEAR, //start rendering and clear the whole framebuffer (region or not) (supply params) + INITIAL_ACTION_CLEAR_REGION, //start rendering and clear the framebuffer in the specified region (supply params) + INITIAL_ACTION_CLEAR_REGION_CONTINUE, //countinue rendering and clear the framebuffer in the specified region (supply params) INITIAL_ACTION_KEEP, //start rendering, but keep attached color texture contents (depth will be cleared) INITIAL_ACTION_DROP, //start rendering, ignore what is there, just write above it INITIAL_ACTION_CONTINUE, //continue rendering (framebuffer must have been left in "continue" state as final action previously) @@ -962,7 +977,7 @@ public: virtual void draw_list_enable_scissor(DrawListID p_list, const Rect2 &p_rect) = 0; virtual void draw_list_disable_scissor(DrawListID p_list) = 0; - virtual void draw_list_end() = 0; + virtual void draw_list_end(uint32_t p_post_barrier = BARRIER_MASK_ALL) = 0; /***********************/ /**** COMPUTE LISTS ****/ @@ -970,17 +985,18 @@ public: typedef int64_t ComputeListID; - virtual ComputeListID compute_list_begin() = 0; + virtual ComputeListID compute_list_begin(bool p_allow_draw_overlap = false) = 0; virtual void compute_list_bind_compute_pipeline(ComputeListID p_list, RID p_compute_pipeline) = 0; virtual void compute_list_bind_uniform_set(ComputeListID p_list, RID p_uniform_set, uint32_t p_index) = 0; virtual void compute_list_set_push_constant(ComputeListID p_list, const void *p_data, uint32_t p_data_size) = 0; virtual void compute_list_dispatch(ComputeListID p_list, uint32_t p_x_groups, uint32_t p_y_groups, uint32_t p_z_groups) = 0; - virtual void compute_list_dispatch_threads(ComputeListID p_list, uint32_t p_x_threads, uint32_t p_y_threads, uint32_t p_z_threads, uint32_t p_x_local_group, uint32_t p_y_local_group, uint32_t p_z_local_group); + virtual void compute_list_dispatch_threads(ComputeListID p_list, uint32_t p_x_threads, uint32_t p_y_threads, uint32_t p_z_threads) = 0; virtual void compute_list_dispatch_indirect(ComputeListID p_list, RID p_buffer, uint32_t p_offset) = 0; virtual void compute_list_add_barrier(ComputeListID p_list) = 0; - virtual void compute_list_end() = 0; + virtual void compute_list_end(uint32_t p_post_barrier = BARRIER_MASK_ALL) = 0; + virtual void barrier(uint32_t p_from = BARRIER_MASK_ALL, uint32_t p_to = BARRIER_MASK_ALL) = 0; virtual void full_barrier() = 0; /***************/ @@ -993,7 +1009,7 @@ public: /**** Timing ****/ /****************/ - virtual void capture_timestamp(const String &p_name, bool p_sync_to_draw) = 0; + virtual void capture_timestamp(const String &p_name) = 0; virtual uint32_t get_captured_timestamps_count() const = 0; virtual uint64_t get_captured_timestamps_frame() const = 0; virtual uint64_t get_captured_timestamp_gpu_time(uint32_t p_index) const = 0; @@ -1058,6 +1074,16 @@ public: virtual RenderingDevice *create_local_device() = 0; + virtual void set_resource_name(RID p_id, const String p_name) = 0; + + virtual void draw_command_begin_label(String p_label_name, const Color p_color = Color(1, 1, 1, 1)) = 0; + virtual void draw_command_insert_label(String p_label_name, const Color p_color = Color(1, 1, 1, 1)) = 0; + virtual void draw_command_end_label() = 0; + + virtual String get_device_vendor_name() const = 0; + virtual String get_device_name() const = 0; + virtual String get_device_pipeline_cache_uuid() const = 0; + static RenderingDevice *get_singleton(); RenderingDevice(); @@ -1077,7 +1103,7 @@ protected: RID _uniform_set_create(const Array &p_uniforms, RID p_shader, uint32_t p_shader_set); - Error _buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const Vector<uint8_t> &p_data, bool p_sync_with_draw = false); + Error _buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const Vector<uint8_t> &p_data, uint32_t p_post_barrier = BARRIER_MASK_ALL); RID _render_pipeline_create(RID p_shader, FramebufferFormatID p_framebuffer_format, VertexFormatID p_vertex_format, RenderPrimitive p_render_primitive, const Ref<RDPipelineRasterizationState> &p_rasterization_state, const Ref<RDPipelineMultisampleState> &p_multisample_state, const Ref<RDPipelineDepthStencilState> &p_depth_stencil_state, const Ref<RDPipelineColorBlendState> &p_blend_state, int p_dynamic_state_flags = 0); diff --git a/servers/rendering/rendering_server_default.cpp b/servers/rendering/rendering_server_default.cpp index fb5db8de60..360b333454 100644 --- a/servers/rendering/rendering_server_default.cpp +++ b/servers/rendering/rendering_server_default.cpp @@ -162,6 +162,51 @@ void RenderingServerDefault::draw(bool p_swap_buffers, double frame_step) { } frame_profile_frame = RSG::storage->get_captured_timestamps_frame(); + + if (print_gpu_profile) { + if (print_frame_profile_ticks_from == 0) { + print_frame_profile_ticks_from = OS::get_singleton()->get_ticks_usec(); + } + float total_time = 0.0; + + for (int i = 0; i < frame_profile.size() - 1; i++) { + String name = frame_profile[i].name; + if (name[0] == '<' || name[0] == '>') { + continue; + } + + float time = frame_profile[i + 1].gpu_msec - frame_profile[i].gpu_msec; + + if (name[0] != '<' && name[0] != '>') { + if (print_gpu_profile_task_time.has(name)) { + print_gpu_profile_task_time[name] += time; + } else { + print_gpu_profile_task_time[name] = time; + } + } + } + + if (frame_profile.size()) { + total_time = frame_profile[frame_profile.size() - 1].gpu_msec; + } + + uint64_t ticks_elapsed = OS::get_singleton()->get_ticks_usec() - print_frame_profile_ticks_from; + print_frame_profile_frame_count++; + if (ticks_elapsed > 1000000) { + print_line("GPU PROFILE (total " + rtos(total_time) + "ms): "); + + float print_threshold = 0.01; + for (OrderedHashMap<String, float>::Element E = print_gpu_profile_task_time.front(); E; E = E.next()) { + float time = E.value() / float(print_frame_profile_frame_count); + if (time > print_threshold) { + print_line("\t-" + E.key() + ": " + rtos(time) + "ms"); + } + } + print_gpu_profile_task_time.clear(); + print_frame_profile_ticks_from = OS::get_singleton()->get_ticks_usec(); + print_frame_profile_frame_count = 0; + } + } } float RenderingServerDefault::get_frame_setup_time_cpu() const { @@ -232,6 +277,11 @@ void RenderingServerDefault::sdfgi_set_debug_probe_select(const Vector3 &p_posit RSG::scene->sdfgi_set_debug_probe_select(p_position, p_dir); } +void RenderingServerDefault::set_print_gpu_profile(bool p_enable) { + RSG::storage->capturing_timestamps = p_enable; + print_gpu_profile = p_enable; +} + RID RenderingServerDefault::get_test_cube() { if (!test_cube.is_valid()) { test_cube = _make_test_cube(); @@ -267,7 +317,7 @@ RenderingServerDefault::RenderingServerDefault() { RSG::rasterizer = RendererCompositor::create(); RSG::storage = RSG::rasterizer->get_storage(); RSG::canvas_render = RSG::rasterizer->get_canvas(); - sr->scene_render = RSG::rasterizer->get_scene(); + sr->set_scene_render(RSG::rasterizer->get_scene()); frame_profile_frame = 0; diff --git a/servers/rendering/rendering_server_default.h b/servers/rendering/rendering_server_default.h index 3ccb5c4bbc..823d28c669 100644 --- a/servers/rendering/rendering_server_default.h +++ b/servers/rendering/rendering_server_default.h @@ -32,6 +32,7 @@ #define RENDERING_SERVER_DEFAULT_H #include "core/math/octree.h" +#include "core/templates/ordered_hash_map.h" #include "renderer_canvas_cull.h" #include "renderer_scene_cull.h" #include "renderer_viewport.h" @@ -74,6 +75,12 @@ class RenderingServerDefault : public RenderingServer { float frame_setup_time = 0; + //for printing + bool print_gpu_profile = false; + OrderedHashMap<String, float> print_gpu_profile_task_time; + uint64_t print_frame_profile_ticks_from = 0; + uint32_t print_frame_profile_frame_count = 0; + public: //if editor is redrawing when it shouldn't, enable this and put a breakpoint in _changes_changed() //#define DEBUG_CHANGES @@ -219,6 +226,8 @@ public: BIND2RC(RID, shader_get_default_texture_param, RID, const StringName &) BIND2RC(Variant, shader_get_param_default, RID, const StringName &) + BIND1RC(ShaderNativeSourceCode, shader_get_native_source_code, RID) + /* COMMON MATERIAL API */ BIND0R(RID, material_create) @@ -265,6 +274,8 @@ public: BIND2(mesh_set_custom_aabb, RID, const AABB &) BIND1RC(AABB, mesh_get_custom_aabb, RID) + BIND2(mesh_set_shadow_mesh, RID, RID) + BIND1(mesh_clear, RID) /* MULTIMESH API */ @@ -542,7 +553,7 @@ public: BIND2(viewport_set_global_canvas_transform, RID, const Transform2D &) BIND4(viewport_set_canvas_stacking, RID, RID, int, int) - BIND2(viewport_set_shadow_atlas_size, RID, int) + BIND3(viewport_set_shadow_atlas_size, RID, int, bool) BIND3(viewport_set_sdf_oversize_and_scale, RID, ViewportSDFOversize, ViewportSDFScale) BIND3(viewport_set_shadow_atlas_quadrant_subdivision, RID, int, int) BIND2(viewport_set_msaa, RID, ViewportMSAA) @@ -563,7 +574,7 @@ public: //from now on, calls forwarded to this singleton #define BINDBASE RSG::scene - BIND1(directional_shadow_atlas_set_size, int) + BIND2(directional_shadow_atlas_set_size, int, bool) BIND1(gi_probe_set_quality, GIProbeQuality) /* SKY API */ @@ -604,16 +615,15 @@ public: BIND7(environment_set_adjustment, RID, bool, float, float, float, bool, RID) BIND9(environment_set_fog, RID, bool, const Color &, float, float, float, float, float, float) - BIND9(environment_set_volumetric_fog, RID, bool, float, const Color &, float, float, float, float, EnvVolumetricFogShadowFilter) + BIND10(environment_set_volumetric_fog, RID, bool, float, const Color &, float, float, float, float, bool, float) BIND2(environment_set_volumetric_fog_volume_size, int, int) BIND1(environment_set_volumetric_fog_filter_active, bool) - BIND1(environment_set_volumetric_fog_directional_shadow_shrink_size, int) - BIND1(environment_set_volumetric_fog_positional_shadow_shrink_size, int) BIND11(environment_set_sdfgi, RID, bool, EnvironmentSDFGICascades, float, EnvironmentSDFGIYScale, bool, bool, bool, float, float, float) BIND1(environment_set_sdfgi_ray_count, EnvironmentSDFGIRayCount) BIND1(environment_set_sdfgi_frames_to_converge, EnvironmentSDFGIFramesToConverge) + BIND1(environment_set_sdfgi_frames_to_update_light, EnvironmentSDFGIFramesToUpdateLight) BIND3R(Ref<Image>, environment_bake_panorama, RID, bool, const Size2i &) @@ -686,6 +696,8 @@ public: BIND3R(TypedArray<Image>, bake_render_uv2, RID, const Vector<RID> &, const Size2i &) + BIND1(gi_set_use_half_resolution, bool) + #undef BINDBASE //from now on, calls forwarded to this singleton #define BINDBASE RSG::canvas @@ -863,6 +875,8 @@ public: virtual void sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir); + virtual void set_print_gpu_profile(bool p_enable); + RenderingServerDefault(); ~RenderingServerDefault(); diff --git a/servers/rendering/rendering_server_wrap_mt.cpp b/servers/rendering/rendering_server_wrap_mt.cpp index 3572c4dc78..9b8d35e5b3 100644 --- a/servers/rendering/rendering_server_wrap_mt.cpp +++ b/servers/rendering/rendering_server_wrap_mt.cpp @@ -97,7 +97,7 @@ void RenderingServerWrapMT::init() { print_verbose("RenderingServerWrapMT: Creating render thread"); DisplayServer::get_singleton()->release_rendering_thread(); if (create_thread) { - thread = Thread::create(_thread_callback, this); + thread.start(_thread_callback, this); print_verbose("RenderingServerWrapMT: Starting render thread"); } while (!draw_thread_up) { @@ -136,12 +136,9 @@ void RenderingServerWrapMT::finish() { canvas_light_occluder_free_cached_ids(); canvas_occluder_polygon_free_cached_ids(); - if (thread) { + if (create_thread) { command_queue.push(this, &RenderingServerWrapMT::thread_exit); - Thread::wait_to_finish(thread); - memdelete(thread); - - thread = nullptr; + thread.wait_to_finish(); } else { rendering_server->finish(); } @@ -160,7 +157,6 @@ RenderingServerWrapMT::RenderingServerWrapMT(RenderingServer *p_contained, bool rendering_server = p_contained; create_thread = p_create_thread; - thread = nullptr; draw_pending = 0; draw_thread_up = false; pool_max_size = GLOBAL_GET("memory/limits/multithreaded_server/rid_pool_prealloc"); diff --git a/servers/rendering/rendering_server_wrap_mt.h b/servers/rendering/rendering_server_wrap_mt.h index 29ee846f31..81e202780e 100644 --- a/servers/rendering/rendering_server_wrap_mt.h +++ b/servers/rendering/rendering_server_wrap_mt.h @@ -46,7 +46,7 @@ class RenderingServerWrapMT : public RenderingServer { Thread::ID server_thread; volatile bool exit; - Thread *thread; + Thread thread; volatile bool draw_thread_up; bool create_thread; @@ -129,6 +129,8 @@ public: FUNC2RC(RID, shader_get_default_texture_param, RID, const StringName &) FUNC2RC(Variant, shader_get_param_default, RID, const StringName &) + FUNC1RC(ShaderNativeSourceCode, shader_get_native_source_code, RID) + /* COMMON MATERIAL API */ FUNCRID(material) @@ -170,6 +172,7 @@ public: FUNC2(mesh_set_custom_aabb, RID, const AABB &) FUNC1RC(AABB, mesh_get_custom_aabb, RID) + FUNC2(mesh_set_shadow_mesh, RID, RID) FUNC1(mesh_clear, RID) /* MULTIMESH API */ @@ -443,7 +446,7 @@ public: FUNC2(viewport_set_global_canvas_transform, RID, const Transform2D &) FUNC4(viewport_set_canvas_stacking, RID, RID, int, int) - FUNC2(viewport_set_shadow_atlas_size, RID, int) + FUNC3(viewport_set_shadow_atlas_size, RID, int, bool) FUNC3(viewport_set_sdf_oversize_and_scale, RID, ViewportSDFOversize, ViewportSDFScale) FUNC3(viewport_set_shadow_atlas_quadrant_subdivision, RID, int, int) @@ -468,7 +471,7 @@ public: return rendering_server->viewport_get_measured_render_time_gpu(p_viewport); } - FUNC1(directional_shadow_atlas_set_size, int) + FUNC2(directional_shadow_atlas_set_size, int, bool) /* SKY API */ @@ -505,6 +508,7 @@ public: FUNC11(environment_set_sdfgi, RID, bool, EnvironmentSDFGICascades, float, EnvironmentSDFGIYScale, bool, bool, bool, float, float, float) FUNC1(environment_set_sdfgi_ray_count, EnvironmentSDFGIRayCount) FUNC1(environment_set_sdfgi_frames_to_converge, EnvironmentSDFGIFramesToConverge) + FUNC1(environment_set_sdfgi_frames_to_update_light, EnvironmentSDFGIFramesToUpdateLight) FUNC11(environment_set_glow, RID, bool, Vector<float>, float, float, float, float, EnvironmentGlowBlendMode, float, float, float) FUNC1(environment_glow_set_use_bicubic_upscale, bool) @@ -516,12 +520,10 @@ public: FUNC9(environment_set_fog, RID, bool, const Color &, float, float, float, float, float, float) - FUNC9(environment_set_volumetric_fog, RID, bool, float, const Color &, float, float, float, float, EnvVolumetricFogShadowFilter) + FUNC10(environment_set_volumetric_fog, RID, bool, float, const Color &, float, float, float, float, bool, float) FUNC2(environment_set_volumetric_fog_volume_size, int, int) FUNC1(environment_set_volumetric_fog_filter_active, bool) - FUNC1(environment_set_volumetric_fog_directional_shadow_shrink_size, int) - FUNC1(environment_set_volumetric_fog_positional_shadow_shrink_size, int) FUNC3R(Ref<Image>, environment_bake_panorama, RID, bool, const Size2i &) @@ -742,6 +744,8 @@ public: return rendering_server->get_video_adapter_vendor(); } + FUNC1(gi_set_use_half_resolution, bool) + FUNC4(set_boot_image, const Ref<Image> &, const Color &, bool, bool) FUNC1(set_default_clear_color, const Color &) @@ -784,6 +788,10 @@ public: rendering_server->sdfgi_set_debug_probe_select(p_position, p_dir); } + virtual void set_print_gpu_profile(bool p_enable) { + rendering_server->set_print_gpu_profile(p_enable); + } + RenderingServerWrapMT(RenderingServer *p_contained, bool p_create_thread); ~RenderingServerWrapMT(); diff --git a/servers/rendering/shader_language.cpp b/servers/rendering/shader_language.cpp index 0cb9220bb3..e6415c0258 100644 --- a/servers/rendering/shader_language.cpp +++ b/servers/rendering/shader_language.cpp @@ -643,7 +643,7 @@ ShaderLanguage::Token ShaderLanguage::_get_token() { } if (hexa_found) { - tk.constant = (double)str.hex_to_int(true); + tk.constant = (double)str.hex_to_int(); } else { tk.constant = str.to_float(); } @@ -6305,7 +6305,9 @@ Error ShaderLanguage::_parse_shader(const Map<StringName, FunctionInfo> &p_funct } uniform2.texture_order = -1; - uniform2.order = uniforms++; + if (uniform_scope != ShaderNode::Uniform::SCOPE_INSTANCE) { + uniform2.order = uniforms++; + } } uniform2.type = type; uniform2.scope = uniform_scope; diff --git a/servers/rendering/shader_types.cpp b/servers/rendering/shader_types.cpp index c1fa4a8ca7..e99b8504bb 100644 --- a/servers/rendering/shader_types.cpp +++ b/servers/rendering/shader_types.cpp @@ -347,7 +347,7 @@ ShaderTypes::ShaderTypes() { emit_vertex_func.arguments.push_back(ShaderLanguage::StageFunctionInfo::Argument("custom", ShaderLanguage::TYPE_VEC4)); emit_vertex_func.arguments.push_back(ShaderLanguage::StageFunctionInfo::Argument("flags", ShaderLanguage::TYPE_UINT)); emit_vertex_func.return_type = ShaderLanguage::TYPE_BOOL; //whether it could emit - shader_modes[RS::SHADER_PARTICLES].functions["compute"].stage_functions["emit_particle"] = emit_vertex_func; + shader_modes[RS::SHADER_PARTICLES].functions["compute"].stage_functions["emit_subparticle"] = emit_vertex_func; } shader_modes[RS::SHADER_PARTICLES].modes.push_back("collision_use_scale"); diff --git a/servers/rendering_server.cpp b/servers/rendering_server.cpp index 758a9a34cd..41c88aa3be 100644 --- a/servers/rendering_server.cpp +++ b/servers/rendering_server.cpp @@ -242,22 +242,24 @@ RID RenderingServer::_make_test_cube() { RID RenderingServer::make_sphere_mesh(int p_lats, int p_lons, float p_radius) { Vector<Vector3> vertices; Vector<Vector3> normals; + const double lat_step = Math_TAU / p_lats; + const double lon_step = Math_TAU / p_lons; for (int i = 1; i <= p_lats; i++) { - double lat0 = Math_PI * (-0.5 + (double)(i - 1) / p_lats); + double lat0 = lat_step * (i - 1) - Math_TAU / 4; double z0 = Math::sin(lat0); double zr0 = Math::cos(lat0); - double lat1 = Math_PI * (-0.5 + (double)i / p_lats); + double lat1 = lat_step * i - Math_TAU / 4; double z1 = Math::sin(lat1); double zr1 = Math::cos(lat1); for (int j = p_lons; j >= 1; j--) { - double lng0 = 2 * Math_PI * (double)(j - 1) / p_lons; + double lng0 = lon_step * (j - 1); double x0 = Math::cos(lng0); double y0 = Math::sin(lng0); - double lng1 = 2 * Math_PI * (double)(j) / p_lons; + double lng1 = lon_step * j; double x1 = Math::cos(lng1); double y1 = Math::sin(lng1); @@ -1661,7 +1663,7 @@ void RenderingServer::_bind_methods() { ClassDB::bind_method(D_METHOD("viewport_set_transparent_background", "viewport", "enabled"), &RenderingServer::viewport_set_transparent_background); ClassDB::bind_method(D_METHOD("viewport_set_global_canvas_transform", "viewport", "transform"), &RenderingServer::viewport_set_global_canvas_transform); ClassDB::bind_method(D_METHOD("viewport_set_canvas_stacking", "viewport", "canvas", "layer", "sublayer"), &RenderingServer::viewport_set_canvas_stacking); - ClassDB::bind_method(D_METHOD("viewport_set_shadow_atlas_size", "viewport", "size"), &RenderingServer::viewport_set_shadow_atlas_size); + ClassDB::bind_method(D_METHOD("viewport_set_shadow_atlas_size", "viewport", "size", "use_16_bits"), &RenderingServer::viewport_set_shadow_atlas_size, DEFVAL(false)); ClassDB::bind_method(D_METHOD("viewport_set_shadow_atlas_quadrant_subdivision", "viewport", "quadrant", "subdivision"), &RenderingServer::viewport_set_shadow_atlas_quadrant_subdivision); ClassDB::bind_method(D_METHOD("viewport_set_msaa", "viewport", "msaa"), &RenderingServer::viewport_set_msaa); ClassDB::bind_method(D_METHOD("viewport_set_use_debanding", "viewport", "enable"), &RenderingServer::viewport_set_use_debanding); @@ -2253,6 +2255,8 @@ void RenderingServer::set_render_loop_enabled(bool p_enabled) { RenderingServer::RenderingServer() { //ERR_FAIL_COND(singleton); + + thread_pool = memnew(RendererThreadPool); singleton = this; GLOBAL_DEF_RST("rendering/vram_compression/import_bptc", false); @@ -2270,6 +2274,7 @@ RenderingServer::RenderingServer() { GLOBAL_DEF("rendering/quality/directional_shadow/soft_shadow_quality", 2); GLOBAL_DEF("rendering/quality/directional_shadow/soft_shadow_quality.mobile", 0); ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/directional_shadow/soft_shadow_quality", PropertyInfo(Variant::INT, "rendering/quality/directional_shadow/soft_shadow_quality", PROPERTY_HINT_ENUM, "Hard (Fastest),Soft Low (Fast),Soft Medium (Average),Soft High (Slow),Soft Ultra (Slowest)")); + GLOBAL_DEF("rendering/quality/directional_shadow/16_bits", true); GLOBAL_DEF("rendering/quality/shadows/soft_shadow_quality", 2); GLOBAL_DEF("rendering/quality/shadows/soft_shadow_quality.mobile", 0); @@ -2280,18 +2285,6 @@ RenderingServer::RenderingServer() { GLOBAL_DEF("rendering/quality/rd_renderer/use_low_end_renderer", false); GLOBAL_DEF("rendering/quality/rd_renderer/use_low_end_renderer.mobile", true); - GLOBAL_DEF("rendering/quality/shadow_atlas/size", 4096); - GLOBAL_DEF("rendering/quality/shadow_atlas/size.mobile", 2048); - ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/shadow_atlas/size", PropertyInfo(Variant::INT, "rendering/quality/shadow_atlas/size", PROPERTY_HINT_RANGE, "256,16384")); - GLOBAL_DEF("rendering/quality/shadow_atlas/quadrant_0_subdiv", 1); - GLOBAL_DEF("rendering/quality/shadow_atlas/quadrant_1_subdiv", 2); - GLOBAL_DEF("rendering/quality/shadow_atlas/quadrant_2_subdiv", 3); - GLOBAL_DEF("rendering/quality/shadow_atlas/quadrant_3_subdiv", 4); - ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/shadow_atlas/quadrant_0_subdiv", PropertyInfo(Variant::INT, "rendering/quality/shadow_atlas/quadrant_0_subdiv", PROPERTY_HINT_ENUM, "Disabled,1 Shadow,4 Shadows,16 Shadows,64 Shadows,256 Shadows,1024 Shadows")); - ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/shadow_atlas/quadrant_1_subdiv", PropertyInfo(Variant::INT, "rendering/quality/shadow_atlas/quadrant_1_subdiv", PROPERTY_HINT_ENUM, "Disabled,1 Shadow,4 Shadows,16 Shadows,64 Shadows,256 Shadows,1024 Shadows")); - ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/shadow_atlas/quadrant_2_subdiv", PropertyInfo(Variant::INT, "rendering/quality/shadow_atlas/quadrant_2_subdiv", PROPERTY_HINT_ENUM, "Disabled,1 Shadow,4 Shadows,16 Shadows,64 Shadows,256 Shadows,1024 Shadows")); - ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/shadow_atlas/quadrant_3_subdiv", PropertyInfo(Variant::INT, "rendering/quality/shadow_atlas/quadrant_3_subdiv", PROPERTY_HINT_ENUM, "Disabled,1 Shadow,4 Shadows,16 Shadows,64 Shadows,256 Shadows,1024 Shadows")); - GLOBAL_DEF("rendering/quality/reflections/roughness_layers", 8); GLOBAL_DEF("rendering/quality/reflections/texture_array_reflections", true); GLOBAL_DEF("rendering/quality/reflections/texture_array_reflections.mobile", false); @@ -2302,6 +2295,8 @@ RenderingServer::RenderingServer() { GLOBAL_DEF("rendering/quality/reflection_atlas/reflection_size.mobile", 128); GLOBAL_DEF("rendering/quality/reflection_atlas/reflection_count", 64); + GLOBAL_DEF("rendering/quality/gi/use_half_resolution", false); + GLOBAL_DEF("rendering/quality/gi_probes/anisotropic", false); GLOBAL_DEF("rendering/quality/gi_probes/quality", 1); ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/gi_probes/quality", PropertyInfo(Variant::INT, "rendering/quality/gi_probes/quality", PROPERTY_HINT_ENUM, "Low (4 Cones - Fast),High (6 Cones - Slow)")); @@ -2365,26 +2360,32 @@ RenderingServer::RenderingServer() { GLOBAL_DEF("rendering/lightmapper/probe_capture_update_speed", 15); ProjectSettings::get_singleton()->set_custom_property_info("rendering/lightmapper/probe_capture_update_speed", PropertyInfo(Variant::FLOAT, "rendering/lightmapper/probe_capture_update_speed", PROPERTY_HINT_RANGE, "0.001,256,0.001")); - GLOBAL_DEF("rendering/sdfgi/probe_ray_count", 2); + GLOBAL_DEF("rendering/sdfgi/probe_ray_count", 1); ProjectSettings::get_singleton()->set_custom_property_info("rendering/sdfgi/probe_ray_count", PropertyInfo(Variant::INT, "rendering/sdfgi/probe_ray_count", PROPERTY_HINT_ENUM, "8 (Fastest),16,32,64,96,128 (Slowest)")); - GLOBAL_DEF("rendering/sdfgi/frames_to_converge", 1); + GLOBAL_DEF("rendering/sdfgi/frames_to_converge", 4); ProjectSettings::get_singleton()->set_custom_property_info("rendering/sdfgi/frames_to_converge", PropertyInfo(Variant::INT, "rendering/sdfgi/frames_to_converge", PROPERTY_HINT_ENUM, "5 (Less Latency but Lower Quality),10,15,20,25,30 (More Latency but Higher Quality)")); + GLOBAL_DEF("rendering/sdfgi/frames_to_update_lights", 2); + ProjectSettings::get_singleton()->set_custom_property_info("rendering/sdfgi/frames_to_update_lights", PropertyInfo(Variant::INT, "rendering/sdfgi/frames_to_update_lights", PROPERTY_HINT_ENUM, "1 (Slower),2,4,8,16 (Faster)")); GLOBAL_DEF("rendering/volumetric_fog/volume_size", 64); ProjectSettings::get_singleton()->set_custom_property_info("rendering/volumetric_fog/volume_size", PropertyInfo(Variant::INT, "rendering/volumetric_fog/volume_size", PROPERTY_HINT_RANGE, "16,512,1")); GLOBAL_DEF("rendering/volumetric_fog/volume_depth", 128); ProjectSettings::get_singleton()->set_custom_property_info("rendering/volumetric_fog/volume_depth", PropertyInfo(Variant::INT, "rendering/volumetric_fog/volume_depth", PROPERTY_HINT_RANGE, "16,512,1")); - GLOBAL_DEF("rendering/volumetric_fog/use_filter", 0); + GLOBAL_DEF("rendering/volumetric_fog/use_filter", 1); ProjectSettings::get_singleton()->set_custom_property_info("rendering/volumetric_fog/use_filter", PropertyInfo(Variant::INT, "rendering/volumetric_fog/use_filter", PROPERTY_HINT_ENUM, "No (Faster),Yes (Higher Quality)")); - GLOBAL_DEF("rendering/volumetric_fog/directional_shadow_shrink", 512); - ProjectSettings::get_singleton()->set_custom_property_info("rendering/volumetric_fog/directional_shadow_shrink", PropertyInfo(Variant::INT, "rendering/volumetric_fog/directional_shadow_shrink", PROPERTY_HINT_RANGE, "32,2048,1")); - GLOBAL_DEF("rendering/volumetric_fog/positional_shadow_shrink", 512); - ProjectSettings::get_singleton()->set_custom_property_info("rendering/volumetric_fog/positional_shadow_shrink", PropertyInfo(Variant::INT, "rendering/volumetric_fog/positional_shadow_shrink", PROPERTY_HINT_RANGE, "32,2048,1")); GLOBAL_DEF("rendering/spatial_indexer/update_iterations_per_frame", 10); ProjectSettings::get_singleton()->set_custom_property_info("rendering/spatial_indexer/update_iterations_per_frame", PropertyInfo(Variant::INT, "rendering/spatial_indexer/update_iterations_per_frame", PROPERTY_HINT_RANGE, "0,1024,1")); + GLOBAL_DEF("rendering/spatial_indexer/threaded_cull_minimum_instances", 1000); + ProjectSettings::get_singleton()->set_custom_property_info("rendering/spatial_indexer/threaded_cull_minimum_instances", PropertyInfo(Variant::INT, "rendering/spatial_indexer/threaded_cull_minimum_instances", PROPERTY_HINT_RANGE, "32,65536,1")); + GLOBAL_DEF("rendering/forward_renderer/threaded_render_minimum_instances", 500); + ProjectSettings::get_singleton()->set_custom_property_info("rendering/forward_renderer/threaded_render_minimum_instances", PropertyInfo(Variant::INT, "rendering/forward_renderer/threaded_render_minimum_instances", PROPERTY_HINT_RANGE, "32,65536,1")); + + GLOBAL_DEF("rendering/cluster_builder/max_clustered_elements", 512); + ProjectSettings::get_singleton()->set_custom_property_info("rendering/cluster_builder/max_clustered_elements", PropertyInfo(Variant::FLOAT, "rendering/cluster_builder/max_clustered_elements", PROPERTY_HINT_RANGE, "32,8192,1")); } RenderingServer::~RenderingServer() { + memdelete(thread_pool); singleton = nullptr; } diff --git a/servers/rendering_server.h b/servers/rendering_server.h index 7db2924612..5defa1f667 100644 --- a/servers/rendering_server.h +++ b/servers/rendering_server.h @@ -39,6 +39,7 @@ #include "core/variant/typed_array.h" #include "core/variant/variant.h" #include "servers/display_server.h" +#include "servers/rendering/renderer_thread_pool.h" #include "servers/rendering/rendering_device.h" #include "servers/rendering/shader_language.h" @@ -52,6 +53,8 @@ class RenderingServer : public Object { Array _get_array_from_surface(uint32_t p_format, Vector<uint8_t> p_vertex_data, Vector<uint8_t> p_attrib_data, Vector<uint8_t> p_skin_data, int p_vertex_len, Vector<uint8_t> p_index_data, int p_index_len) const; + RendererThreadPool *thread_pool = nullptr; + protected: RID _make_test_cube(); void _free_internal_rids(); @@ -176,6 +179,19 @@ public: virtual void shader_set_default_texture_param(RID p_shader, const StringName &p_name, RID p_texture) = 0; virtual RID shader_get_default_texture_param(RID p_shader, const StringName &p_name) const = 0; + struct ShaderNativeSourceCode { + struct Version { + struct Stage { + String name; + String code; + }; + Vector<Stage> stages; + }; + Vector<Version> versions; + }; + + virtual ShaderNativeSourceCode shader_get_native_source_code(RID p_shader) const = 0; + /* COMMON MATERIAL API */ enum { @@ -338,6 +354,8 @@ public: virtual void mesh_set_custom_aabb(RID p_mesh, const AABB &p_aabb) = 0; virtual AABB mesh_get_custom_aabb(RID p_mesh) const = 0; + virtual void mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh) = 0; + virtual void mesh_clear(RID p_mesh) = 0; /* MULTIMESH API */ @@ -782,7 +800,7 @@ public: virtual void viewport_set_sdf_oversize_and_scale(RID p_viewport, ViewportSDFOversize p_oversize, ViewportSDFScale p_scale) = 0; - virtual void viewport_set_shadow_atlas_size(RID p_viewport, int p_size) = 0; + virtual void viewport_set_shadow_atlas_size(RID p_viewport, int p_size, bool p_16_bits = false) = 0; virtual void viewport_set_shadow_atlas_quadrant_subdivision(RID p_viewport, int p_quadrant, int p_subdiv) = 0; enum ViewportMSAA { @@ -840,7 +858,10 @@ public: VIEWPORT_DEBUG_DRAW_SDFGI_PROBES, VIEWPORT_DEBUG_DRAW_GI_BUFFER, VIEWPORT_DEBUG_DRAW_DISABLE_LOD, - + VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS, + VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS, + VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS, + VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES, }; virtual void viewport_set_debug_draw(RID p_viewport, ViewportDebugDraw p_draw) = 0; @@ -849,7 +870,7 @@ public: virtual float viewport_get_measured_render_time_cpu(RID p_viewport) const = 0; virtual float viewport_get_measured_render_time_gpu(RID p_viewport) const = 0; - virtual void directional_shadow_atlas_set_size(int p_size) = 0; + virtual void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false) = 0; /* SKY API */ @@ -967,6 +988,7 @@ public: virtual void environment_set_sdfgi(RID p_env, bool p_enable, EnvironmentSDFGICascades p_cascades, float p_min_cell_size, EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) = 0; enum EnvironmentSDFGIRayCount { + ENV_SDFGI_RAY_COUNT_4, ENV_SDFGI_RAY_COUNT_8, ENV_SDFGI_RAY_COUNT_16, ENV_SDFGI_RAY_COUNT_32, @@ -990,20 +1012,22 @@ public: virtual void environment_set_sdfgi_frames_to_converge(EnvironmentSDFGIFramesToConverge p_frames) = 0; - virtual void environment_set_fog(RID p_env, bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_aerial_perspective) = 0; - - enum EnvVolumetricFogShadowFilter { - ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED, - ENV_VOLUMETRIC_FOG_SHADOW_FILTER_LOW, - ENV_VOLUMETRIC_FOG_SHADOW_FILTER_MEDIUM, - ENV_VOLUMETRIC_FOG_SHADOW_FILTER_HIGH, + enum EnvironmentSDFGIFramesToUpdateLight { + ENV_SDFGI_UPDATE_LIGHT_IN_1_FRAME, + ENV_SDFGI_UPDATE_LIGHT_IN_2_FRAMES, + ENV_SDFGI_UPDATE_LIGHT_IN_4_FRAMES, + ENV_SDFGI_UPDATE_LIGHT_IN_8_FRAMES, + ENV_SDFGI_UPDATE_LIGHT_IN_16_FRAMES, + ENV_SDFGI_UPDATE_LIGHT_MAX, }; - virtual void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, EnvVolumetricFogShadowFilter p_shadow_filter) = 0; + virtual void environment_set_sdfgi_frames_to_update_light(EnvironmentSDFGIFramesToUpdateLight p_update) = 0; + + virtual void environment_set_fog(RID p_env, bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_aerial_perspective) = 0; + + virtual void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) = 0; virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth) = 0; virtual void environment_set_volumetric_fog_filter_active(bool p_enable) = 0; - virtual void environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size) = 0; - virtual void environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size) = 0; virtual Ref<Image> environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size) = 0; @@ -1411,6 +1435,8 @@ public: virtual float get_frame_setup_time_cpu() const = 0; + virtual void gi_set_use_half_resolution(bool p_enable) = 0; + /* TESTING */ virtual RID get_test_cube() = 0; @@ -1443,6 +1469,8 @@ public: virtual bool is_low_end() const = 0; + virtual void set_print_gpu_profile(bool p_enable) = 0; + RenderingDevice *create_local_rendering_device() const; bool is_render_loop_enabled() const; |