diff options
Diffstat (limited to 'drivers/gles3/storage')
| -rw-r--r-- | drivers/gles3/storage/config.cpp | 34 | ||||
| -rw-r--r-- | drivers/gles3/storage/config.h | 12 | ||||
| -rw-r--r-- | drivers/gles3/storage/material_storage.cpp | 321 | ||||
| -rw-r--r-- | drivers/gles3/storage/material_storage.h | 66 | ||||
| -rw-r--r-- | drivers/gles3/storage/mesh_storage.cpp | 618 | ||||
| -rw-r--r-- | drivers/gles3/storage/mesh_storage.h | 52 | ||||
| -rw-r--r-- | drivers/gles3/storage/particles_storage.cpp | 1211 | ||||
| -rw-r--r-- | drivers/gles3/storage/particles_storage.h | 315 | ||||
| -rw-r--r-- | drivers/gles3/storage/render_scene_buffers_gles3.cpp | 40 | ||||
| -rw-r--r-- | drivers/gles3/storage/render_scene_buffers_gles3.h | 5 | ||||
| -rw-r--r-- | drivers/gles3/storage/texture_storage.cpp | 1009 | ||||
| -rw-r--r-- | drivers/gles3/storage/texture_storage.h | 145 | ||||
| -rw-r--r-- | drivers/gles3/storage/utilities.cpp | 241 | ||||
| -rw-r--r-- | drivers/gles3/storage/utilities.h | 87 |
14 files changed, 3605 insertions, 551 deletions
diff --git a/drivers/gles3/storage/config.cpp b/drivers/gles3/storage/config.cpp index 242c1ce0a9..9b496c0999 100644 --- a/drivers/gles3/storage/config.cpp +++ b/drivers/gles3/storage/config.cpp @@ -34,6 +34,15 @@ #include "core/config/project_settings.h" #include "core/templates/vector.h" +#ifdef ANDROID_ENABLED +#include <GLES3/gl3.h> +#include <GLES3/gl3ext.h> +#include <GLES3/gl3platform.h> + +#include <EGL/egl.h> +#include <EGL/eglext.h> +#endif + using namespace GLES3; #define _GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT 0x84FF @@ -62,7 +71,7 @@ Config::Config() { s3tc_supported = true; rgtc_supported = true; //RGTC - core since OpenGL version 3.0 #else - float_texture_supported = extensions.has("GL_ARB_texture_float") || extensions.has("GL_OES_texture_float"); + float_texture_supported = extensions.has("GL_EXT_color_buffer_float"); etc2_supported = true; #if defined(ANDROID_ENABLED) || defined(IOS_ENABLED) // Some Android devices report support for S3TC but we don't expect that and don't export the textures. @@ -75,28 +84,29 @@ Config::Config() { rgtc_supported = extensions.has("GL_EXT_texture_compression_rgtc") || extensions.has("GL_ARB_texture_compression_rgtc") || extensions.has("EXT_texture_compression_rgtc"); #endif -#ifdef GLES_OVER_GL - use_rgba_2d_shadows = false; -#else - use_rgba_2d_shadows = !(float_texture_supported && extensions.has("GL_EXT_texture_rg")); -#endif - glGetIntegerv(GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS, &max_vertex_texture_image_units); glGetIntegerv(GL_MAX_TEXTURE_IMAGE_UNITS, &max_texture_image_units); glGetIntegerv(GL_MAX_TEXTURE_SIZE, &max_texture_size); glGetIntegerv(GL_MAX_UNIFORM_BLOCK_SIZE, &max_uniform_buffer_size); + glGetIntegerv(GL_MAX_VIEWPORT_DIMS, max_viewport_size); glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &uniform_buffer_offset_alignment); - // the use skeleton software path should be used if either float texture is not supported, - // OR max_vertex_texture_image_units is zero - use_skeleton_software = (float_texture_supported == false) || (max_vertex_texture_image_units == 0); - support_anisotropic_filter = extensions.has("GL_EXT_texture_filter_anisotropic"); if (support_anisotropic_filter) { glGetFloatv(_GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT, &anisotropic_level); - anisotropic_level = MIN(float(1 << int(ProjectSettings::get_singleton()->get("rendering/textures/default_filters/anisotropic_filtering_level"))), anisotropic_level); + anisotropic_level = MIN(float(1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level"))), anisotropic_level); + } + + multiview_supported = extensions.has("GL_OVR_multiview2") || extensions.has("GL_OVR_multiview"); +#ifdef ANDROID_ENABLED + if (multiview_supported) { + eglFramebufferTextureMultiviewOVR = (PFNGLFRAMEBUFFERTEXTUREMULTIVIEWOVRPROC)eglGetProcAddress("glFramebufferTextureMultiviewOVR"); + if (eglFramebufferTextureMultiviewOVR == nullptr) { + multiview_supported = false; + } } +#endif force_vertex_shading = false; //GLOBAL_GET("rendering/quality/shading/force_vertex_shading"); use_nearest_mip_filter = GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter"); diff --git a/drivers/gles3/storage/config.h b/drivers/gles3/storage/config.h index fe18345775..87202fde84 100644 --- a/drivers/gles3/storage/config.h +++ b/drivers/gles3/storage/config.h @@ -44,6 +44,10 @@ #include OPENGL_INCLUDE_H #endif +#ifdef ANDROID_ENABLED +typedef void (*PFNGLFRAMEBUFFERTEXTUREMULTIVIEWOVRPROC)(GLenum, GLenum, GLuint, GLint, GLint, GLsizei); +#endif + namespace GLES3 { class Config { @@ -52,13 +56,12 @@ private: public: bool use_nearest_mip_filter = false; - bool use_skeleton_software = false; bool use_depth_prepass = true; - bool use_rgba_2d_shadows = false; int max_vertex_texture_image_units = 0; int max_texture_image_units = 0; int max_texture_size = 0; + int max_viewport_size[2] = { 0, 0 }; int max_uniform_buffer_size = 0; int max_renderable_elements = 0; int max_renderable_lights = 0; @@ -82,6 +85,11 @@ public: bool support_anisotropic_filter = false; float anisotropic_level = 0.0f; + bool multiview_supported = false; +#ifdef ANDROID_ENABLED + PFNGLFRAMEBUFFERTEXTUREMULTIVIEWOVRPROC eglFramebufferTextureMultiviewOVR = nullptr; +#endif + static Config *get_singleton() { return singleton; }; Config(); diff --git a/drivers/gles3/storage/material_storage.cpp b/drivers/gles3/storage/material_storage.cpp index 8e6009c943..d413c2b00e 100644 --- a/drivers/gles3/storage/material_storage.cpp +++ b/drivers/gles3/storage/material_storage.cpp @@ -34,6 +34,7 @@ #include "config.h" #include "material_storage.h" +#include "particles_storage.h" #include "texture_storage.h" #include "drivers/gles3/rasterizer_canvas_gles3.h" @@ -89,7 +90,7 @@ _FORCE_INLINE_ static void _fill_std140_variant_ubo_value(ShaderLanguage::DataTy gui[j + 3] = 0; // ignored } } else { - int v = value; + uint32_t v = value; gui[0] = v & 1 ? 1 : 0; gui[1] = v & 2 ? 1 : 0; } @@ -116,7 +117,7 @@ _FORCE_INLINE_ static void _fill_std140_variant_ubo_value(ShaderLanguage::DataTy gui[j + 3] = 0; // ignored } } else { - int v = value; + uint32_t v = value; gui[0] = (v & 1) ? 1 : 0; gui[1] = (v & 2) ? 1 : 0; gui[2] = (v & 4) ? 1 : 0; @@ -145,7 +146,7 @@ _FORCE_INLINE_ static void _fill_std140_variant_ubo_value(ShaderLanguage::DataTy } } } else { - int v = value; + uint32_t v = value; gui[0] = (v & 1) ? 1 : 0; gui[1] = (v & 2) ? 1 : 0; gui[2] = (v & 4) ? 1 : 0; @@ -728,7 +729,7 @@ _FORCE_INLINE_ static void _fill_std140_ubo_value(ShaderLanguage::DataType type, switch (type) { case ShaderLanguage::TYPE_BOOL: { uint32_t *gui = (uint32_t *)data; - *gui = value[0].boolean ? 1 : 0; + gui[0] = value[0].boolean ? 1 : 0; } break; case ShaderLanguage::TYPE_BVEC2: { uint32_t *gui = (uint32_t *)data; @@ -897,7 +898,9 @@ _FORCE_INLINE_ static void _fill_std140_ubo_empty(ShaderLanguage::DataType type, case ShaderLanguage::TYPE_BVEC3: case ShaderLanguage::TYPE_IVEC3: case ShaderLanguage::TYPE_UVEC3: - case ShaderLanguage::TYPE_VEC3: + case ShaderLanguage::TYPE_VEC3: { + memset(data, 0, 12 * p_array_size); + } break; case ShaderLanguage::TYPE_BVEC4: case ShaderLanguage::TYPE_IVEC4: case ShaderLanguage::TYPE_UVEC4: @@ -1340,13 +1343,13 @@ MaterialStorage::MaterialStorage() { shader_data_request_func[RS::SHADER_SPATIAL] = _create_scene_shader_func; shader_data_request_func[RS::SHADER_CANVAS_ITEM] = _create_canvas_shader_func; - shader_data_request_func[RS::SHADER_PARTICLES] = nullptr; + shader_data_request_func[RS::SHADER_PARTICLES] = _create_particles_shader_func; shader_data_request_func[RS::SHADER_SKY] = _create_sky_shader_func; shader_data_request_func[RS::SHADER_FOG] = nullptr; material_data_request_func[RS::SHADER_SPATIAL] = _create_scene_material_func; material_data_request_func[RS::SHADER_CANVAS_ITEM] = _create_canvas_material_func; - material_data_request_func[RS::SHADER_PARTICLES] = nullptr; + material_data_request_func[RS::SHADER_PARTICLES] = _create_particles_material_func; material_data_request_func[RS::SHADER_SKY] = _create_sky_material_func; material_data_request_func[RS::SHADER_FOG] = nullptr; @@ -1611,32 +1614,32 @@ MaterialStorage::MaterialStorage() { { // Setup Particles compiler - /* -ShaderCompiler::DefaultIdentifierActions actions; - actions.renames["COLOR"] = "PARTICLE.color"; - actions.renames["VELOCITY"] = "PARTICLE.velocity"; + ShaderCompiler::DefaultIdentifierActions actions; + + actions.renames["COLOR"] = "out_color"; + actions.renames["VELOCITY"] = "out_velocity_flags.xyz"; //actions.renames["MASS"] = "mass"; ? actions.renames["ACTIVE"] = "particle_active"; actions.renames["RESTART"] = "restart"; - actions.renames["CUSTOM"] = "PARTICLE.custom"; - for (int i = 0; i < ParticlesShader::MAX_USERDATAS; i++) { + actions.renames["CUSTOM"] = "out_custom"; + for (int i = 0; i < PARTICLES_MAX_USERDATAS; i++) { String udname = "USERDATA" + itos(i + 1); - actions.renames[udname] = "PARTICLE.userdata" + itos(i + 1); + actions.renames[udname] = "out_userdata" + itos(i + 1); actions.usage_defines[udname] = "#define USERDATA" + itos(i + 1) + "_USED\n"; } - actions.renames["TRANSFORM"] = "PARTICLE.xform"; - actions.renames["TIME"] = "frame_history.data[0].time"; + actions.renames["TRANSFORM"] = "xform"; + actions.renames["TIME"] = "time"; actions.renames["PI"] = _MKSTR(Math_PI); actions.renames["TAU"] = _MKSTR(Math_TAU); actions.renames["E"] = _MKSTR(Math_E); - actions.renames["LIFETIME"] = "params.lifetime"; + actions.renames["LIFETIME"] = "lifetime"; actions.renames["DELTA"] = "local_delta"; actions.renames["NUMBER"] = "particle_number"; actions.renames["INDEX"] = "index"; //actions.renames["GRAVITY"] = "current_gravity"; - actions.renames["EMISSION_TRANSFORM"] = "FRAME.emission_transform"; - actions.renames["RANDOM_SEED"] = "FRAME.random_seed"; + actions.renames["EMISSION_TRANSFORM"] = "emission_transform"; + actions.renames["RANDOM_SEED"] = "random_seed"; actions.renames["FLAG_EMIT_POSITION"] = "EMISSION_FLAG_HAS_POSITION"; actions.renames["FLAG_EMIT_ROT_SCALE"] = "EMISSION_FLAG_HAS_ROTATION_SCALE"; actions.renames["FLAG_EMIT_VELOCITY"] = "EMISSION_FLAG_HAS_VELOCITY"; @@ -1658,18 +1661,10 @@ ShaderCompiler::DefaultIdentifierActions actions; actions.render_mode_defines["keep_data"] = "#define ENABLE_KEEP_DATA\n"; actions.render_mode_defines["collision_use_scale"] = "#define USE_COLLISION_SCALE\n"; - actions.sampler_array_name = "material_samplers"; - actions.base_texture_binding_index = 1; - actions.texture_layout_set = 3; - actions.base_uniform_string = "material."; - actions.base_varying_index = 10; - actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; actions.default_repeat = ShaderLanguage::REPEAT_ENABLE; - actions.global_buffer_array_variable = "global_shader_uniforms.data"; - particles_shader.compiler.initialize(actions); - */ + shaders.compiler_particles.initialize(actions); } { @@ -2212,7 +2207,7 @@ void MaterialStorage::global_shader_parameters_load_settings(bool p_load_texture for (const PropertyInfo &E : settings) { if (E.name.begins_with("shader_globals/")) { StringName name = E.name.get_slice("/", 1); - Dictionary d = ProjectSettings::get_singleton()->get(E.name); + Dictionary d = GLOBAL_GET(E.name); ERR_CONTINUE(!d.has("type")); ERR_CONTINUE(!d.has("value")); @@ -2312,7 +2307,7 @@ void MaterialStorage::global_shader_parameters_instance_free(RID p_instance) { global_shader_uniforms.instance_buffer_pos.erase(p_instance); } -void MaterialStorage::global_shader_parameters_instance_update(RID p_instance, int p_index, const Variant &p_value) { +void MaterialStorage::global_shader_parameters_instance_update(RID p_instance, int p_index, const Variant &p_value, int p_flags_count) { if (!global_shader_uniforms.instance_buffer_pos.has(p_instance)) { return; //just not allocated, ignore } @@ -2322,7 +2317,9 @@ void MaterialStorage::global_shader_parameters_instance_update(RID p_instance, i return; //again, not allocated, ignore } ERR_FAIL_INDEX(p_index, ShaderLanguage::MAX_INSTANCE_UNIFORM_INDICES); - ERR_FAIL_COND_MSG(p_value.get_type() > Variant::COLOR, "Unsupported variant type for instance parameter: " + Variant::get_type_name(p_value.get_type())); //anything greater not supported + + Variant::Type value_type = p_value.get_type(); + ERR_FAIL_COND_MSG(p_value.get_type() > Variant::COLOR, "Unsupported variant type for instance parameter: " + Variant::get_type_name(value_type)); //anything greater not supported ShaderLanguage::DataType datatype_from_value[Variant::COLOR + 1] = { ShaderLanguage::TYPE_MAX, //nil @@ -2348,9 +2345,24 @@ void MaterialStorage::global_shader_parameters_instance_update(RID p_instance, i ShaderLanguage::TYPE_VEC4 //color }; - ShaderLanguage::DataType datatype = datatype_from_value[p_value.get_type()]; + ShaderLanguage::DataType datatype = ShaderLanguage::TYPE_MAX; + if (value_type == Variant::INT && p_flags_count > 0) { + switch (p_flags_count) { + case 1: + datatype = ShaderLanguage::TYPE_BVEC2; + break; + case 2: + datatype = ShaderLanguage::TYPE_BVEC3; + break; + case 3: + datatype = ShaderLanguage::TYPE_BVEC4; + break; + } + } else { + datatype = datatype_from_value[value_type]; + } - ERR_FAIL_COND_MSG(datatype == ShaderLanguage::TYPE_MAX, "Unsupported variant type for instance parameter: " + Variant::get_type_name(p_value.get_type())); //anything greater not supported + ERR_FAIL_COND_MSG(datatype == ShaderLanguage::TYPE_MAX, "Unsupported variant type for instance parameter: " + Variant::get_type_name(value_type)); //anything greater not supported pos += p_index; @@ -2451,8 +2463,8 @@ void MaterialStorage::shader_set_code(RID p_shader, const String &p_code) { RS::ShaderMode new_mode; if (mode_string == "canvas_item") { new_mode = RS::SHADER_CANVAS_ITEM; - //} else if (mode_string == "particles") { - // new_mode = RS::SHADER_PARTICLES; + } else if (mode_string == "particles") { + new_mode = RS::SHADER_PARTICLES; } else if (mode_string == "spatial") { new_mode = RS::SHADER_SPATIAL; } else if (mode_string == "sky") { @@ -2523,6 +2535,9 @@ void MaterialStorage::shader_set_path_hint(RID p_shader, const String &p_path) { ERR_FAIL_COND(!shader); shader->path_hint = p_path; + if (shader->data) { + shader->data->set_path_hint(p_path); + } } String MaterialStorage::shader_get_code(RID p_shader) const { @@ -2790,6 +2805,10 @@ void MaterialStorage::material_update_dependency(RID p_material, DependencyTrack /* Canvas Shader Data */ +void CanvasShaderData::set_path_hint(const String &p_path) { + path = p_path; +} + void CanvasShaderData::set_code(const String &p_code) { // compile the shader @@ -2988,7 +3007,7 @@ GLES3::ShaderData *GLES3::_create_canvas_shader_func() { } void CanvasMaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { - return update_parameters_internal(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size); + update_parameters_internal(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size); } void CanvasMaterialData::bind_uniforms() { @@ -3024,6 +3043,10 @@ GLES3::MaterialData *GLES3::_create_canvas_material_func(ShaderData *p_shader) { //////////////////////////////////////////////////////////////////////////////// // SKY SHADER +void SkyShaderData::set_path_hint(const String &p_path) { + path = p_path; +} + void SkyShaderData::set_code(const String &p_code) { //compile @@ -3232,7 +3255,7 @@ GLES3::ShaderData *GLES3::_create_sky_shader_func() { void SkyMaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { uniform_set_updated = true; - return update_parameters_internal(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size); + update_parameters_internal(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size); } SkyMaterialData::~SkyMaterialData() { @@ -3267,6 +3290,10 @@ void SkyMaterialData::bind_uniforms() { //////////////////////////////////////////////////////////////////////////////// // Scene SHADER +void SceneShaderData::set_path_hint(const String &p_path) { + path = p_path; +} + void SceneShaderData::set_code(const String &p_code) { //compile @@ -3339,6 +3366,9 @@ void SceneShaderData::set_code(const String &p_code) { actions.usage_flag_pointers["ALPHA"] = &uses_alpha; actions.usage_flag_pointers["ALPHA_SCISSOR_THRESHOLD"] = &uses_alpha_clip; + // Use alpha clip pipeline for alpha hash/dither. + // This prevents sorting issues inherent to alpha blending and allows such materials to cast shadows. + actions.usage_flag_pointers["ALPHA_HASH_SCALE"] = &uses_alpha_clip; actions.render_mode_flags["depth_prepass_alpha"] = &uses_depth_pre_pass; actions.usage_flag_pointers["SSS_STRENGTH"] = &uses_sss; @@ -3399,6 +3429,8 @@ void SceneShaderData::set_code(const String &p_code) { vertex_input_mask |= uses_bones << 9; vertex_input_mask |= uses_weights << 10; uses_screen_texture_mipmaps = gen_code.uses_screen_texture_mipmaps; + uses_vertex_time = gen_code.uses_vertex_time; + uses_fragment_time = gen_code.uses_fragment_time; #if 0 print_line("**compiling shader:"); @@ -3519,11 +3551,15 @@ bool SceneShaderData::is_parameter_texture(const StringName &p_param) const { } bool SceneShaderData::is_animated() const { - return false; + return (uses_fragment_time && uses_discard) || (uses_vertex_time && uses_vertex); } bool SceneShaderData::casts_shadows() const { - return false; + bool has_read_screen_alpha = uses_screen_texture || uses_depth_texture || uses_normal_texture; + bool has_base_alpha = (uses_alpha && !uses_alpha_clip) || has_read_screen_alpha; + bool has_alpha = has_base_alpha || uses_blend_alpha; + + return !has_alpha || (uses_depth_pre_pass && !(depth_draw == DEPTH_DRAW_DISABLED || depth_test == DEPTH_TEST_DISABLED)); } Variant SceneShaderData::get_default_parameter(const StringName &p_parameter) const { @@ -3564,7 +3600,7 @@ void SceneMaterialData::set_next_pass(RID p_pass) { } void SceneMaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { - return update_parameters_internal(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size); + update_parameters_internal(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size); } SceneMaterialData::~SceneMaterialData() { @@ -3597,4 +3633,209 @@ void SceneMaterialData::bind_uniforms() { } } +/* Particles SHADER */ + +void ParticlesShaderData::set_path_hint(const String &p_path) { + path = p_path; +} + +void ParticlesShaderData::set_code(const String &p_code) { + //compile + + code = p_code; + valid = false; + ubo_size = 0; + uniforms.clear(); + uses_collision = false; + + if (code.is_empty()) { + return; //just invalid, but no error + } + + ShaderCompiler::GeneratedCode gen_code; + ShaderCompiler::IdentifierActions actions; + actions.entry_point_stages["start"] = ShaderCompiler::STAGE_VERTEX; + actions.entry_point_stages["process"] = ShaderCompiler::STAGE_VERTEX; + + actions.usage_flag_pointers["COLLIDED"] = &uses_collision; + + userdata_count = 0; + for (uint32_t i = 0; i < PARTICLES_MAX_USERDATAS; i++) { + userdatas_used[i] = false; + actions.usage_flag_pointers["USERDATA" + itos(i + 1)] = &userdatas_used[i]; + } + + actions.uniforms = &uniforms; + + Error err = MaterialStorage::get_singleton()->shaders.compiler_particles.compile(RS::SHADER_PARTICLES, code, &actions, path, gen_code); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); + + if (version.is_null()) { + version = MaterialStorage::get_singleton()->shaders.particles_process_shader.version_create(); + } + + for (uint32_t i = 0; i < PARTICLES_MAX_USERDATAS; i++) { + if (userdatas_used[i]) { + userdata_count++; + } + } + + Vector<StringName> texture_uniform_names; + for (int i = 0; i < gen_code.texture_uniforms.size(); i++) { + texture_uniform_names.push_back(gen_code.texture_uniforms[i].name); + } + + MaterialStorage::get_singleton()->shaders.particles_process_shader.version_set_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompiler::STAGE_VERTEX], gen_code.stage_globals[ShaderCompiler::STAGE_FRAGMENT], gen_code.defines, texture_uniform_names); + ERR_FAIL_COND(!MaterialStorage::get_singleton()->shaders.particles_process_shader.version_is_valid(version)); + + ubo_size = gen_code.uniform_total_size; + ubo_offsets = gen_code.uniform_offsets; + texture_uniforms = gen_code.texture_uniforms; + + valid = true; +} + +void ParticlesShaderData::set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index) { + if (!p_texture.is_valid()) { + if (default_texture_params.has(p_name) && default_texture_params[p_name].has(p_index)) { + default_texture_params[p_name].erase(p_index); + + if (default_texture_params[p_name].is_empty()) { + default_texture_params.erase(p_name); + } + } + } else { + if (!default_texture_params.has(p_name)) { + default_texture_params[p_name] = HashMap<int, RID>(); + } + default_texture_params[p_name][p_index] = p_texture; + } +} + +void ParticlesShaderData::get_shader_uniform_list(List<PropertyInfo> *p_param_list) const { + HashMap<int, StringName> order; + + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL || E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + if (E.value.texture_order >= 0) { + order[E.value.texture_order + 100000] = E.key; + } else { + order[E.value.order] = E.key; + } + } + + String last_group; + for (const KeyValue<int, StringName> &E : order) { + String group = uniforms[E.value].group; + if (!uniforms[E.value].subgroup.is_empty()) { + group += "::" + uniforms[E.value].subgroup; + } + + if (group != last_group) { + PropertyInfo pi; + pi.usage = PROPERTY_USAGE_GROUP; + pi.name = group; + p_param_list->push_back(pi); + + last_group = group; + } + + PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E.value]); + pi.name = E.value; + p_param_list->push_back(pi); + } +} + +void ParticlesShaderData::get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const { + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + RendererMaterialStorage::InstanceShaderParam p; + p.info = ShaderLanguage::uniform_to_property_info(E.value); + p.info.name = E.key; //supply name + p.index = E.value.instance_index; + p.default_value = ShaderLanguage::constant_value_to_variant(E.value.default_value, E.value.type, E.value.array_size, E.value.hint); + p_param_list->push_back(p); + } +} + +bool ParticlesShaderData::is_parameter_texture(const StringName &p_param) const { + if (!uniforms.has(p_param)) { + return false; + } + + return uniforms[p_param].texture_order >= 0; +} + +bool ParticlesShaderData::is_animated() const { + return false; +} + +bool ParticlesShaderData::casts_shadows() const { + return false; +} + +Variant ParticlesShaderData::get_default_parameter(const StringName &p_parameter) const { + if (uniforms.has(p_parameter)) { + ShaderLanguage::ShaderNode::Uniform uniform = uniforms[p_parameter]; + Vector<ShaderLanguage::ConstantNode::Value> default_value = uniform.default_value; + return ShaderLanguage::constant_value_to_variant(default_value, uniform.type, uniform.array_size, uniform.hint); + } + return Variant(); +} + +RS::ShaderNativeSourceCode ParticlesShaderData::get_native_source_code() const { + return MaterialStorage::get_singleton()->shaders.particles_process_shader.version_get_native_source_code(version); +} + +ParticlesShaderData::~ParticlesShaderData() { + if (version.is_valid()) { + MaterialStorage::get_singleton()->shaders.particles_process_shader.version_free(version); + } +} + +GLES3::ShaderData *GLES3::_create_particles_shader_func() { + ParticlesShaderData *shader_data = memnew(ParticlesShaderData); + return shader_data; +} + +void ParticleProcessMaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + update_parameters_internal(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size); +} + +ParticleProcessMaterialData::~ParticleProcessMaterialData() { +} + +GLES3::MaterialData *GLES3::_create_particles_material_func(ShaderData *p_shader) { + ParticleProcessMaterialData *material_data = memnew(ParticleProcessMaterialData); + material_data->shader_data = static_cast<ParticlesShaderData *>(p_shader); + //update will happen later anyway so do nothing. + return material_data; +} + +void ParticleProcessMaterialData::bind_uniforms() { + // Bind Material Uniforms + glBindBufferBase(GL_UNIFORM_BUFFER, GLES3::PARTICLES_MATERIAL_UNIFORM_LOCATION, uniform_buffer); + + RID *textures = texture_cache.ptrw(); + ShaderCompiler::GeneratedCode::Texture *texture_uniforms = shader_data->texture_uniforms.ptrw(); + for (int ti = 0; ti < texture_cache.size(); ti++) { + Texture *texture = TextureStorage::get_singleton()->get_texture(textures[ti]); + glActiveTexture(GL_TEXTURE1 + ti); // Start at GL_TEXTURE1 becuase texture slot 0 is reserved for the heightmap texture. + glBindTexture(target_from_type[texture_uniforms[ti].type], texture->tex_id); + + // Set sampler state here as the same texture can be used in multiple places with different flags + // Need to convert sampler state from ShaderLanguage::Texture* to RS::CanvasItemTexture* + RS::CanvasItemTextureFilter filter = RS::CanvasItemTextureFilter((int(texture_uniforms[ti].filter) + 1) % RS::CANVAS_ITEM_TEXTURE_FILTER_MAX); + RS::CanvasItemTextureRepeat repeat = RS::CanvasItemTextureRepeat((int(texture_uniforms[ti].repeat) + 1) % RS::CANVAS_ITEM_TEXTURE_REPEAT_MIRROR); + texture->gl_set_filter(filter); + texture->gl_set_repeat(repeat); + } +} + #endif // !GLES3_ENABLED diff --git a/drivers/gles3/storage/material_storage.h b/drivers/gles3/storage/material_storage.h index 6504c7748c..8ae5e5eb9c 100644 --- a/drivers/gles3/storage/material_storage.h +++ b/drivers/gles3/storage/material_storage.h @@ -44,6 +44,7 @@ #include "../shaders/canvas.glsl.gen.h" #include "../shaders/cubemap_filter.glsl.gen.h" +#include "../shaders/particles.glsl.gen.h" #include "../shaders/scene.glsl.gen.h" #include "../shaders/sky.glsl.gen.h" @@ -53,6 +54,7 @@ namespace GLES3 { struct ShaderData { virtual void set_code(const String &p_Code) = 0; + virtual void set_path_hint(const String &p_hint) = 0; virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index) = 0; virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const = 0; @@ -165,6 +167,7 @@ struct CanvasShaderData : public ShaderData { bool uses_time = false; virtual void set_code(const String &p_Code); + virtual void set_path_hint(const String &p_hint); virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; virtual void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; @@ -216,6 +219,7 @@ struct SkyShaderData : public ShaderData { bool uses_light; virtual void set_code(const String &p_Code); + virtual void set_path_hint(const String &p_hint); virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; virtual void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; @@ -318,6 +322,8 @@ struct SceneShaderData : public ShaderData { bool uses_depth_texture; bool uses_normal_texture; bool uses_time; + bool uses_vertex_time; + bool uses_fragment_time; bool writes_modelview_or_projection; bool uses_world_coordinates; bool uses_tangent; @@ -337,6 +343,7 @@ struct SceneShaderData : public ShaderData { uint32_t index = 0; virtual void set_code(const String &p_Code); + virtual void set_path_hint(const String &p_hint); virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; virtual void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; @@ -368,6 +375,62 @@ struct SceneMaterialData : public MaterialData { MaterialData *_create_scene_material_func(ShaderData *p_shader); +/* Particle Shader */ + +enum { + PARTICLES_MAX_USERDATAS = 6 +}; + +struct ParticlesShaderData : public ShaderData { + bool valid = false; + RID version; + bool uses_collision = false; + + HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompiler::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size = 0; + + String path; + String code; + HashMap<StringName, HashMap<int, RID>> default_texture_params; + + bool uses_time = false; + + bool userdatas_used[PARTICLES_MAX_USERDATAS] = {}; + uint32_t userdata_count = 0; + + virtual void set_code(const String &p_Code); + virtual void set_path_hint(const String &p_hint); + virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); + virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; + virtual void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; + virtual bool is_parameter_texture(const StringName &p_param) const; + 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(); +}; + +ShaderData *_create_particles_shader_func(); + +struct ParticleProcessMaterialData : public MaterialData { + ParticlesShaderData *shader_data = nullptr; + RID uniform_set; + + virtual void set_render_priority(int p_priority) {} + virtual void set_next_pass(RID p_pass) {} + virtual void update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual void bind_uniforms(); + virtual ~ParticleProcessMaterialData(); +}; + +MaterialData *_create_particles_material_func(ShaderData *p_shader); + /* Global shader uniform structs */ struct GlobalShaderUniforms { enum { @@ -504,6 +567,7 @@ public: SkyShaderGLES3 sky_shader; SceneShaderGLES3 scene_shader; CubemapFilterShaderGLES3 cubemap_filter_shader; + ParticlesShaderGLES3 particles_process_shader; ShaderCompiler compiler_canvas; ShaderCompiler compiler_scene; @@ -530,7 +594,7 @@ public: virtual int32_t global_shader_parameters_instance_allocate(RID p_instance) override; virtual void global_shader_parameters_instance_free(RID p_instance) override; - virtual void global_shader_parameters_instance_update(RID p_instance, int p_index, const Variant &p_value) override; + virtual void global_shader_parameters_instance_update(RID p_instance, int p_index, const Variant &p_value, int p_flags_count = 0) override; GLuint global_shader_parameters_get_uniform_buffer() const; diff --git a/drivers/gles3/storage/mesh_storage.cpp b/drivers/gles3/storage/mesh_storage.cpp index 22d84eba93..285f32f1a5 100644 --- a/drivers/gles3/storage/mesh_storage.cpp +++ b/drivers/gles3/storage/mesh_storage.cpp @@ -44,10 +44,16 @@ MeshStorage *MeshStorage::get_singleton() { MeshStorage::MeshStorage() { singleton = this; + + { + skeleton_shader.shader.initialize(); + skeleton_shader.shader_version = skeleton_shader.shader.version_create(); + } } MeshStorage::~MeshStorage() { singleton = nullptr; + skeleton_shader.shader.version_free(skeleton_shader.shader_version); } /* MESH API */ @@ -87,7 +93,6 @@ void MeshStorage::mesh_set_blend_shape_count(RID p_mesh, int p_blend_shape_count ERR_FAIL_COND(!mesh); ERR_FAIL_COND(mesh->surface_count > 0); //surfaces already exist - WARN_PRINT_ONCE("blend shapes not supported by GLES3 renderer yet"); mesh->blend_shape_count = p_blend_shape_count; } @@ -111,7 +116,6 @@ void MeshStorage::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_surface) uint32_t attrib_stride = 0; uint32_t skin_stride = 0; - // TODO: I think this should be <=, but it is copied from RendererRD, will have to verify later for (int i = 0; i < RS::ARRAY_WEIGHTS; i++) { if ((p_surface.format & (1 << i))) { switch (i) { @@ -245,8 +249,77 @@ void MeshStorage::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_surface) s->aabb = p_surface.aabb; s->bone_aabbs = p_surface.bone_aabbs; //only really useful for returning them. - if (mesh->blend_shape_count > 0) { - //s->blend_shape_buffer = RD::get_singleton()->storage_buffer_create(p_surface.blend_shape_data.size(), p_surface.blend_shape_data); + if (p_surface.skin_data.size() || mesh->blend_shape_count > 0) { + // Size must match the size of the vertex array. + int size = p_surface.vertex_data.size(); + int vertex_size = 0; + int stride = 0; + int normal_offset = 0; + int tangent_offset = 0; + if ((p_surface.format & (1 << RS::ARRAY_VERTEX))) { + if (p_surface.format & RS::ARRAY_FLAG_USE_2D_VERTICES) { + vertex_size = 2; + } else { + vertex_size = 3; + } + stride = sizeof(float) * vertex_size; + } + if ((p_surface.format & (1 << RS::ARRAY_NORMAL))) { + normal_offset = stride; + stride += sizeof(uint16_t) * 2; + } + if ((p_surface.format & (1 << RS::ARRAY_TANGENT))) { + tangent_offset = stride; + stride += sizeof(uint16_t) * 2; + } + + if (mesh->blend_shape_count > 0) { + // Blend shapes are passed as one large array, for OpenGL, we need to split each of them into their own buffer + s->blend_shapes = memnew_arr(Mesh::Surface::BlendShape, mesh->blend_shape_count); + + for (uint32_t i = 0; i < mesh->blend_shape_count; i++) { + glGenVertexArrays(1, &s->blend_shapes[i].vertex_array); + glBindVertexArray(s->blend_shapes[i].vertex_array); + glGenBuffers(1, &s->blend_shapes[i].vertex_buffer); + glBindBuffer(GL_ARRAY_BUFFER, s->blend_shapes[i].vertex_buffer); + glBufferData(GL_ARRAY_BUFFER, size, p_surface.blend_shape_data.ptr() + i * size, (s->format & RS::ARRAY_FLAG_USE_DYNAMIC_UPDATE) ? GL_DYNAMIC_DRAW : GL_STATIC_DRAW); + + if ((p_surface.format & (1 << RS::ARRAY_VERTEX))) { + glEnableVertexAttribArray(RS::ARRAY_VERTEX + 3); + glVertexAttribPointer(RS::ARRAY_VERTEX + 3, vertex_size, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(0)); + } + if ((p_surface.format & (1 << RS::ARRAY_NORMAL))) { + glEnableVertexAttribArray(RS::ARRAY_NORMAL + 3); + glVertexAttribPointer(RS::ARRAY_NORMAL + 3, 2, GL_UNSIGNED_SHORT, GL_TRUE, stride, CAST_INT_TO_UCHAR_PTR(normal_offset)); + } + if ((p_surface.format & (1 << RS::ARRAY_TANGENT))) { + glEnableVertexAttribArray(RS::ARRAY_TANGENT + 3); + glVertexAttribPointer(RS::ARRAY_TANGENT + 3, 2, GL_UNSIGNED_SHORT, GL_TRUE, stride, CAST_INT_TO_UCHAR_PTR(tangent_offset)); + } + } + glBindVertexArray(0); + glBindBuffer(GL_ARRAY_BUFFER, 0); + } + + // Create a vertex array to use for skeleton/blend shapes. + glGenVertexArrays(1, &s->skeleton_vertex_array); + glBindVertexArray(s->skeleton_vertex_array); + glBindBuffer(GL_ARRAY_BUFFER, s->vertex_buffer); + + if ((p_surface.format & (1 << RS::ARRAY_VERTEX))) { + glEnableVertexAttribArray(RS::ARRAY_VERTEX); + glVertexAttribPointer(RS::ARRAY_VERTEX, vertex_size, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(0)); + } + if ((p_surface.format & (1 << RS::ARRAY_NORMAL))) { + glEnableVertexAttribArray(RS::ARRAY_NORMAL); + glVertexAttribPointer(RS::ARRAY_NORMAL, 2, GL_UNSIGNED_SHORT, GL_TRUE, stride, CAST_INT_TO_UCHAR_PTR(normal_offset)); + } + if ((p_surface.format & (1 << RS::ARRAY_TANGENT))) { + glEnableVertexAttribArray(RS::ARRAY_TANGENT); + glVertexAttribPointer(RS::ARRAY_TANGENT, 2, GL_UNSIGNED_SHORT, GL_TRUE, stride, CAST_INT_TO_UCHAR_PTR(tangent_offset)); + } + glBindVertexArray(0); + glBindBuffer(GL_ARRAY_BUFFER, 0); } if (mesh->surface_count == 0) { @@ -309,12 +382,48 @@ RS::BlendShapeMode MeshStorage::mesh_get_blend_shape_mode(RID p_mesh) const { } void MeshStorage::mesh_surface_update_vertex_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + + uint64_t data_size = p_data.size(); + ERR_FAIL_COND(p_offset + data_size > mesh->surfaces[p_surface]->vertex_buffer_size); + const uint8_t *r = p_data.ptr(); + + glBindBuffer(GL_ARRAY_BUFFER, mesh->surfaces[p_surface]->vertex_buffer); + glBufferSubData(GL_ARRAY_BUFFER, p_offset, data_size, r); + glBindBuffer(GL_ARRAY_BUFFER, 0); } void MeshStorage::mesh_surface_update_attribute_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + + uint64_t data_size = p_data.size(); + ERR_FAIL_COND(p_offset + data_size > mesh->surfaces[p_surface]->attribute_buffer_size); + const uint8_t *r = p_data.ptr(); + + glBindBuffer(GL_ARRAY_BUFFER, mesh->surfaces[p_surface]->attribute_buffer); + glBufferSubData(GL_ARRAY_BUFFER, p_offset, data_size, r); + glBindBuffer(GL_ARRAY_BUFFER, 0); } void MeshStorage::mesh_surface_update_skin_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + + uint64_t data_size = p_data.size(); + ERR_FAIL_COND(p_offset + data_size > mesh->surfaces[p_surface]->skin_buffer_size); + const uint8_t *r = p_data.ptr(); + + glBindBuffer(GL_ARRAY_BUFFER, mesh->surfaces[p_surface]->skin_buffer); + glBufferSubData(GL_ARRAY_BUFFER, p_offset, data_size, r); + glBindBuffer(GL_ARRAY_BUFFER, 0); } void MeshStorage::mesh_surface_set_material(RID p_mesh, int p_surface, RID p_material) { @@ -373,7 +482,13 @@ RS::SurfaceData MeshStorage::mesh_get_surface(RID p_mesh, int p_surface) const { } sd.bone_aabbs = s.bone_aabbs; - glBindBuffer(GL_ARRAY_BUFFER, 0); + + if (mesh->blend_shape_count) { + sd.blend_shape_data = Vector<uint8_t>(); + for (uint32_t i = 0; i < mesh->blend_shape_count; i++) { + sd.blend_shape_data.append_array(Utilities::buffer_get_data(GL_ARRAY_BUFFER, s.blend_shapes[i].vertex_buffer, s.vertex_buffer_size)); + } + } return sd; } @@ -569,6 +684,24 @@ void MeshStorage::mesh_clear(RID p_mesh) { memdelete_arr(s.lods); } + if (mesh->blend_shape_count) { + for (uint32_t j = 0; j < mesh->blend_shape_count; j++) { + if (s.blend_shapes[j].vertex_buffer != 0) { + glDeleteBuffers(1, &s.blend_shapes[j].vertex_buffer); + s.blend_shapes[j].vertex_buffer = 0; + } + if (s.blend_shapes[j].vertex_array != 0) { + glDeleteVertexArrays(1, &s.blend_shapes[j].vertex_array); + s.blend_shapes[j].vertex_array = 0; + } + } + memdelete_arr(s.blend_shapes); + } + if (s.skeleton_vertex_array != 0) { + glDeleteVertexArrays(1, &s.skeleton_vertex_array); + s.skeleton_vertex_array = 0; + } + memdelete(mesh->surfaces[i]); } if (mesh->surfaces) { @@ -624,15 +757,15 @@ void MeshStorage::_mesh_surface_generate_version_for_input_mask(Mesh::Surface::V case RS::ARRAY_NORMAL: { attribs[i].offset = vertex_stride; attribs[i].size = 2; - attribs[i].type = GL_UNSIGNED_SHORT; - vertex_stride += sizeof(uint16_t) * 2; + attribs[i].type = (mis ? GL_FLOAT : GL_UNSIGNED_SHORT); + vertex_stride += sizeof(uint16_t) * 2 * (mis ? 2 : 1); attribs[i].normalized = GL_TRUE; } break; case RS::ARRAY_TANGENT: { attribs[i].offset = vertex_stride; attribs[i].size = 2; - attribs[i].type = GL_UNSIGNED_SHORT; - vertex_stride += sizeof(uint16_t) * 2; + attribs[i].type = (mis ? GL_FLOAT : GL_UNSIGNED_SHORT); + vertex_stride += sizeof(uint16_t) * 2 * (mis ? 2 : 1); attribs[i].normalized = GL_TRUE; } break; case RS::ARRAY_COLOR: { @@ -677,7 +810,7 @@ void MeshStorage::_mesh_surface_generate_version_for_input_mask(Mesh::Surface::V attribs[i].offset = skin_stride; attribs[i].size = 4; attribs[i].type = GL_UNSIGNED_SHORT; - attributes_stride += 4 * sizeof(uint16_t); + skin_stride += 4 * sizeof(uint16_t); attribs[i].normalized = GL_FALSE; attribs[i].integer = true; } break; @@ -685,7 +818,7 @@ void MeshStorage::_mesh_surface_generate_version_for_input_mask(Mesh::Surface::V attribs[i].offset = skin_stride; attribs[i].size = 4; attribs[i].type = GL_UNSIGNED_SHORT; - attributes_stride += 4 * sizeof(uint16_t); + skin_stride += 4 * sizeof(uint16_t); attribs[i].normalized = GL_TRUE; } break; } @@ -776,7 +909,7 @@ void MeshStorage::mesh_instance_set_blend_shape_weight(RID p_mesh_instance, int ERR_FAIL_COND(!mi); ERR_FAIL_INDEX(p_shape, (int)mi->blend_weights.size()); mi->blend_weights[p_shape] = p_weight; - mi->weights_dirty = true; + mi->dirty = true; } void MeshStorage::_mesh_instance_clear(MeshInstance *mi) { @@ -788,38 +921,65 @@ void MeshStorage::_mesh_instance_clear(MeshInstance *mi) { } memfree(mi->surfaces[i].versions); } + + if (mi->surfaces[i].vertex_buffers[0] != 0) { + glDeleteBuffers(2, mi->surfaces[i].vertex_buffers); + mi->surfaces[i].vertex_buffers[0] = 0; + mi->surfaces[i].vertex_buffers[1] = 0; + } + if (mi->surfaces[i].vertex_buffer != 0) { glDeleteBuffers(1, &mi->surfaces[i].vertex_buffer); mi->surfaces[i].vertex_buffer = 0; } } mi->surfaces.clear(); - - if (mi->blend_weights_buffer != 0) { - glDeleteBuffers(1, &mi->blend_weights_buffer); - mi->blend_weights_buffer = 0; - } mi->blend_weights.clear(); - mi->weights_dirty = false; mi->skeleton_version = 0; } void MeshStorage::_mesh_instance_add_surface(MeshInstance *mi, Mesh *mesh, uint32_t p_surface) { - if (mesh->blend_shape_count > 0 && mi->blend_weights_buffer == 0) { + if (mesh->blend_shape_count > 0) { mi->blend_weights.resize(mesh->blend_shape_count); for (uint32_t i = 0; i < mi->blend_weights.size(); i++) { - mi->blend_weights[i] = 0; + mi->blend_weights[i] = 0.0; } - // Todo allocate buffer for blend_weights and copy data to it - //mi->blend_weights_buffer = RD::get_singleton()->storage_buffer_create(sizeof(float) * mi->blend_weights.size(), mi->blend_weights.to_byte_array()); - - mi->weights_dirty = true; } MeshInstance::Surface s; - if (mesh->blend_shape_count > 0 || (mesh->surfaces[p_surface]->format & RS::ARRAY_FORMAT_BONES)) { - //surface warrants transform - //s.vertex_buffer = RD::get_singleton()->vertex_buffer_create(mesh->surfaces[p_surface]->vertex_buffer_size, Vector<uint8_t>(), true); + if ((mesh->blend_shape_count > 0 || (mesh->surfaces[p_surface]->format & RS::ARRAY_FORMAT_BONES)) && mesh->surfaces[p_surface]->vertex_buffer_size > 0) { + // Cache surface properties + s.format_cache = mesh->surfaces[p_surface]->format; + if ((s.format_cache & (1 << RS::ARRAY_VERTEX))) { + if (s.format_cache & RS::ARRAY_FLAG_USE_2D_VERTICES) { + s.vertex_size_cache = 2; + } else { + s.vertex_size_cache = 3; + } + s.vertex_stride_cache = sizeof(float) * s.vertex_size_cache; + } + if ((s.format_cache & (1 << RS::ARRAY_NORMAL))) { + s.vertex_normal_offset_cache = s.vertex_stride_cache; + s.vertex_stride_cache += sizeof(uint32_t) * 2; + } + if ((s.format_cache & (1 << RS::ARRAY_TANGENT))) { + s.vertex_tangent_offset_cache = s.vertex_stride_cache; + s.vertex_stride_cache += sizeof(uint32_t) * 2; + } + + // Buffer to be used for rendering. Final output of skeleton and blend shapes. + glGenBuffers(1, &s.vertex_buffer); + glBindBuffer(GL_ARRAY_BUFFER, s.vertex_buffer); + glBufferData(GL_ARRAY_BUFFER, s.vertex_stride_cache * mesh->surfaces[p_surface]->vertex_count, nullptr, GL_DYNAMIC_DRAW); + if (mesh->blend_shape_count > 0) { + // Ping-Pong buffers for processing blendshapes. + glGenBuffers(2, s.vertex_buffers); + for (uint32_t i = 0; i < 2; i++) { + glBindBuffer(GL_ARRAY_BUFFER, s.vertex_buffers[i]); + glBufferData(GL_ARRAY_BUFFER, s.vertex_stride_cache * mesh->surfaces[p_surface]->vertex_count, nullptr, GL_DYNAMIC_DRAW); + } + } + glBindBuffer(GL_ARRAY_BUFFER, 0); //unbind } mi->surfaces.push_back(s); @@ -831,11 +991,6 @@ void MeshStorage::mesh_instance_check_for_update(RID p_mesh_instance) { bool needs_update = mi->dirty; - if (mi->weights_dirty && !mi->weight_update_list.in_list()) { - dirty_mesh_instance_weights.add(&mi->weight_update_list); - needs_update = true; - } - if (mi->array_update_list.in_list()) { return; } @@ -852,22 +1007,223 @@ void MeshStorage::mesh_instance_check_for_update(RID p_mesh_instance) { } } -void MeshStorage::update_mesh_instances() { - while (dirty_mesh_instance_weights.first()) { - MeshInstance *mi = dirty_mesh_instance_weights.first()->self(); +void MeshStorage::_blend_shape_bind_mesh_instance_buffer(MeshInstance *p_mi, uint32_t p_surface) { + glBindBuffer(GL_ARRAY_BUFFER, p_mi->surfaces[p_surface].vertex_buffers[0]); - if (mi->blend_weights_buffer != 0) { - //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; + if ((p_mi->surfaces[p_surface].format_cache & (1 << RS::ARRAY_VERTEX))) { + glEnableVertexAttribArray(RS::ARRAY_VERTEX); + glVertexAttribPointer(RS::ARRAY_VERTEX, p_mi->surfaces[p_surface].vertex_size_cache, GL_FLOAT, GL_FALSE, p_mi->surfaces[p_surface].vertex_stride_cache, CAST_INT_TO_UCHAR_PTR(0)); + } else { + glDisableVertexAttribArray(RS::ARRAY_VERTEX); + } + if ((p_mi->surfaces[p_surface].format_cache & (1 << RS::ARRAY_NORMAL))) { + glEnableVertexAttribArray(RS::ARRAY_NORMAL); + glVertexAttribIPointer(RS::ARRAY_NORMAL, 2, GL_UNSIGNED_INT, p_mi->surfaces[p_surface].vertex_stride_cache, CAST_INT_TO_UCHAR_PTR(p_mi->surfaces[p_surface].vertex_normal_offset_cache)); + } else { + glDisableVertexAttribArray(RS::ARRAY_NORMAL); + } + if ((p_mi->surfaces[p_surface].format_cache & (1 << RS::ARRAY_TANGENT))) { + glEnableVertexAttribArray(RS::ARRAY_TANGENT); + glVertexAttribIPointer(RS::ARRAY_TANGENT, 2, GL_UNSIGNED_INT, p_mi->surfaces[p_surface].vertex_stride_cache, CAST_INT_TO_UCHAR_PTR(p_mi->surfaces[p_surface].vertex_tangent_offset_cache)); + } else { + glDisableVertexAttribArray(RS::ARRAY_TANGENT); + } +} + +void MeshStorage::_compute_skeleton(MeshInstance *p_mi, Skeleton *p_sk, uint32_t p_surface) { + glBindBuffer(GL_ARRAY_BUFFER, 0); + + // Add in the bones and weights. + glBindBuffer(GL_ARRAY_BUFFER, p_mi->mesh->surfaces[p_surface]->skin_buffer); + + bool use_8_weights = p_mi->surfaces[p_surface].format_cache & RS::ARRAY_FLAG_USE_8_BONE_WEIGHTS; + int skin_stride = sizeof(int16_t) * (use_8_weights ? 16 : 8); + glEnableVertexAttribArray(RS::ARRAY_BONES); + glVertexAttribIPointer(RS::ARRAY_BONES, 4, GL_UNSIGNED_SHORT, skin_stride, CAST_INT_TO_UCHAR_PTR(0)); + if (use_8_weights) { + glEnableVertexAttribArray(11); + glVertexAttribIPointer(11, 4, GL_UNSIGNED_SHORT, skin_stride, CAST_INT_TO_UCHAR_PTR(4 * sizeof(uint16_t))); + glEnableVertexAttribArray(12); + glVertexAttribPointer(12, 4, GL_UNSIGNED_SHORT, GL_TRUE, skin_stride, CAST_INT_TO_UCHAR_PTR(8 * sizeof(uint16_t))); + glEnableVertexAttribArray(13); + glVertexAttribPointer(13, 4, GL_UNSIGNED_SHORT, GL_TRUE, skin_stride, CAST_INT_TO_UCHAR_PTR(12 * sizeof(uint16_t))); + } else { + glEnableVertexAttribArray(RS::ARRAY_WEIGHTS); + glVertexAttribPointer(RS::ARRAY_WEIGHTS, 4, GL_UNSIGNED_SHORT, GL_TRUE, skin_stride, CAST_INT_TO_UCHAR_PTR(4 * sizeof(uint16_t))); } + + glBindBufferBase(GL_TRANSFORM_FEEDBACK_BUFFER, 0, p_mi->surfaces[p_surface].vertex_buffer); + glActiveTexture(GL_TEXTURE0); + glBindTexture(GL_TEXTURE_2D, p_sk->transforms_texture); + + glBeginTransformFeedback(GL_POINTS); + glDrawArrays(GL_POINTS, 0, p_mi->mesh->surfaces[p_surface]->vertex_count); + glEndTransformFeedback(); + + glDisableVertexAttribArray(RS::ARRAY_BONES); + glDisableVertexAttribArray(RS::ARRAY_WEIGHTS); + glDisableVertexAttribArray(RS::ARRAY_BONES + 2); + glDisableVertexAttribArray(RS::ARRAY_WEIGHTS + 2); + glBindVertexArray(0); + glBindBuffer(GL_TRANSFORM_FEEDBACK_BUFFER, 0); +} + +void MeshStorage::update_mesh_instances() { if (dirty_mesh_instance_arrays.first() == nullptr) { return; //nothing to do } + glEnable(GL_RASTERIZER_DISCARD); // Process skeletons and blend shapes using transform feedback - // TODO: Implement when working on skeletons and blend shapes + while (dirty_mesh_instance_arrays.first()) { + MeshInstance *mi = dirty_mesh_instance_arrays.first()->self(); + + Skeleton *sk = skeleton_owner.get_or_null(mi->skeleton); + + // Precompute base weight if using blend shapes. + float base_weight = 1.0; + if (mi->mesh->blend_shape_count && mi->mesh->blend_shape_mode == RS::BLEND_SHAPE_MODE_NORMALIZED) { + for (uint32_t i = 0; i < mi->mesh->blend_shape_count; i++) { + base_weight -= mi->blend_weights[i]; + } + } + + for (uint32_t i = 0; i < mi->surfaces.size(); i++) { + if (mi->surfaces[i].vertex_buffer == 0 || mi->mesh->surfaces[i]->skeleton_vertex_array == 0) { + continue; + } + + bool array_is_2d = mi->surfaces[i].format_cache & RS::ARRAY_FLAG_USE_2D_VERTICES; + bool can_use_skeleton = sk != nullptr && sk->use_2d == array_is_2d && (mi->surfaces[i].format_cache & RS::ARRAY_FORMAT_BONES); + bool use_8_weights = mi->surfaces[i].format_cache & RS::ARRAY_FLAG_USE_8_BONE_WEIGHTS; + + // Always process blend shapes first. + if (mi->mesh->blend_shape_count) { + SkeletonShaderGLES3::ShaderVariant variant = SkeletonShaderGLES3::MODE_BASE_PASS; + uint64_t specialization = 0; + specialization |= array_is_2d ? SkeletonShaderGLES3::MODE_2D : 0; + specialization |= SkeletonShaderGLES3::USE_BLEND_SHAPES; + if (!array_is_2d) { + if ((mi->surfaces[i].format_cache & (1 << RS::ARRAY_NORMAL))) { + specialization |= SkeletonShaderGLES3::USE_NORMAL; + } + if ((mi->surfaces[i].format_cache & (1 << RS::ARRAY_TANGENT))) { + specialization |= SkeletonShaderGLES3::USE_TANGENT; + } + } + + bool success = skeleton_shader.shader.version_bind_shader(skeleton_shader.shader_version, variant, specialization); + if (!success) { + continue; + } + + skeleton_shader.shader.version_set_uniform(SkeletonShaderGLES3::BLEND_WEIGHT, base_weight, skeleton_shader.shader_version, variant, specialization); + skeleton_shader.shader.version_set_uniform(SkeletonShaderGLES3::BLEND_SHAPE_COUNT, float(mi->mesh->blend_shape_count), skeleton_shader.shader_version, variant, specialization); + + glBindBuffer(GL_ARRAY_BUFFER, 0); + glBindVertexArray(mi->mesh->surfaces[i]->skeleton_vertex_array); + glBindBufferBase(GL_TRANSFORM_FEEDBACK_BUFFER, 0, mi->surfaces[i].vertex_buffers[0]); + glBeginTransformFeedback(GL_POINTS); + glDrawArrays(GL_POINTS, 0, mi->mesh->surfaces[i]->vertex_count); + glEndTransformFeedback(); + + variant = SkeletonShaderGLES3::MODE_BLEND_PASS; + success = skeleton_shader.shader.version_bind_shader(skeleton_shader.shader_version, variant, specialization); + if (!success) { + continue; + } + + //Do the last blend shape separately, as it can be combined with the skeleton pass. + for (uint32_t bs = 0; bs < mi->mesh->blend_shape_count - 1; bs++) { + float weight = mi->blend_weights[bs]; + + if (Math::is_zero_approx(weight)) { + //not bother with this one + continue; + } + skeleton_shader.shader.version_set_uniform(SkeletonShaderGLES3::BLEND_WEIGHT, weight, skeleton_shader.shader_version, variant, specialization); + skeleton_shader.shader.version_set_uniform(SkeletonShaderGLES3::BLEND_SHAPE_COUNT, float(mi->mesh->blend_shape_count), skeleton_shader.shader_version, variant, specialization); + + glBindVertexArray(mi->mesh->surfaces[i]->blend_shapes[bs].vertex_array); + _blend_shape_bind_mesh_instance_buffer(mi, i); + glBindBufferBase(GL_TRANSFORM_FEEDBACK_BUFFER, 0, mi->surfaces[i].vertex_buffers[1]); + + glBeginTransformFeedback(GL_POINTS); + glDrawArrays(GL_POINTS, 0, mi->mesh->surfaces[i]->vertex_count); + glEndTransformFeedback(); + + SWAP(mi->surfaces[i].vertex_buffers[0], mi->surfaces[i].vertex_buffers[1]); + } + uint32_t bs = mi->mesh->blend_shape_count - 1; + + float weight = mi->blend_weights[bs]; + + glBindVertexArray(mi->mesh->surfaces[i]->blend_shapes[bs].vertex_array); + _blend_shape_bind_mesh_instance_buffer(mi, i); + + specialization |= can_use_skeleton ? SkeletonShaderGLES3::USE_SKELETON : 0; + specialization |= (can_use_skeleton && use_8_weights) ? SkeletonShaderGLES3::USE_EIGHT_WEIGHTS : 0; + specialization |= SkeletonShaderGLES3::FINAL_PASS; + success = skeleton_shader.shader.version_bind_shader(skeleton_shader.shader_version, variant, specialization); + if (!success) { + continue; + } + + skeleton_shader.shader.version_set_uniform(SkeletonShaderGLES3::BLEND_WEIGHT, weight, skeleton_shader.shader_version, variant, specialization); + skeleton_shader.shader.version_set_uniform(SkeletonShaderGLES3::BLEND_SHAPE_COUNT, float(mi->mesh->blend_shape_count), skeleton_shader.shader_version, variant, specialization); + + if (can_use_skeleton) { + // Do last blendshape in the same pass as the Skeleton. + _compute_skeleton(mi, sk, i); + can_use_skeleton = false; + } else { + // Do last blendshape by itself and prepare vertex data for use by the renderer. + glBindBufferBase(GL_TRANSFORM_FEEDBACK_BUFFER, 0, mi->surfaces[i].vertex_buffer); + + glBeginTransformFeedback(GL_POINTS); + glDrawArrays(GL_POINTS, 0, mi->mesh->surfaces[i]->vertex_count); + glEndTransformFeedback(); + } + + glBindVertexArray(0); + glBindBuffer(GL_TRANSFORM_FEEDBACK_BUFFER, 0); + } + + // This branch should only execute when Skeleton is run by itself. + if (can_use_skeleton) { + SkeletonShaderGLES3::ShaderVariant variant = SkeletonShaderGLES3::MODE_BASE_PASS; + uint64_t specialization = 0; + specialization |= array_is_2d ? SkeletonShaderGLES3::MODE_2D : 0; + specialization |= SkeletonShaderGLES3::USE_SKELETON; + specialization |= SkeletonShaderGLES3::FINAL_PASS; + specialization |= use_8_weights ? SkeletonShaderGLES3::USE_EIGHT_WEIGHTS : 0; + if (!array_is_2d) { + if ((mi->surfaces[i].format_cache & (1 << RS::ARRAY_NORMAL))) { + specialization |= SkeletonShaderGLES3::USE_NORMAL; + } + if ((mi->surfaces[i].format_cache & (1 << RS::ARRAY_TANGENT))) { + specialization |= SkeletonShaderGLES3::USE_TANGENT; + } + } + + bool success = skeleton_shader.shader.version_bind_shader(skeleton_shader.shader_version, variant, specialization); + if (!success) { + continue; + } + + glBindVertexArray(mi->mesh->surfaces[i]->skeleton_vertex_array); + _compute_skeleton(mi, sk, i); + } + } + mi->dirty = false; + if (sk) { + mi->skeleton_version = sk->version; + } + dirty_mesh_instance_arrays.remove(&mi->array_update_list); + } + glDisable(GL_RASTERIZER_DISCARD); + glBindBuffer(GL_ARRAY_BUFFER, 0); + glBindBufferBase(GL_TRANSFORM_FEEDBACK_BUFFER, 0, 0); } /* MULTIMESH API */ @@ -968,7 +1324,7 @@ void MeshStorage::multimesh_set_mesh(RID p_multimesh, RID p_mesh) { #define MULTIMESH_DIRTY_REGION_SIZE 512 void MeshStorage::_multimesh_make_local(MultiMesh *multimesh) const { - if (multimesh->data_cache.size() > 0) { + if (multimesh->data_cache.size() > 0 || multimesh->instances == 0) { return; //already local } ERR_FAIL_COND(multimesh->data_cache.size() > 0); @@ -1385,7 +1741,7 @@ Vector<float> MeshStorage::multimesh_get_buffer(RID p_multimesh) const { MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); ERR_FAIL_COND_V(!multimesh, Vector<float>()); Vector<float> ret; - if (multimesh->buffer == 0) { + if (multimesh->buffer == 0 || multimesh->instances == 0) { return Vector<float>(); } else if (multimesh->data_cache.size()) { ret = multimesh->data_cache; @@ -1538,45 +1894,207 @@ void MeshStorage::_update_dirty_multimeshes() { /* SKELETON API */ RID MeshStorage::skeleton_allocate() { - return RID(); + return skeleton_owner.allocate_rid(); } void MeshStorage::skeleton_initialize(RID p_rid) { + skeleton_owner.initialize_rid(p_rid, Skeleton()); } void MeshStorage::skeleton_free(RID p_rid) { + _update_dirty_skeletons(); + skeleton_allocate_data(p_rid, 0); + Skeleton *skeleton = skeleton_owner.get_or_null(p_rid); + skeleton->dependency.deleted_notify(p_rid); + skeleton_owner.free(p_rid); +} + +void MeshStorage::_skeleton_make_dirty(Skeleton *skeleton) { + if (!skeleton->dirty) { + skeleton->dirty = true; + skeleton->dirty_list = skeleton_dirty_list; + skeleton_dirty_list = skeleton; + } } void MeshStorage::skeleton_allocate_data(RID p_skeleton, int p_bones, bool p_2d_skeleton) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + ERR_FAIL_COND(!skeleton); + ERR_FAIL_COND(p_bones < 0); + + if (skeleton->size == p_bones && skeleton->use_2d == p_2d_skeleton) { + return; + } + + skeleton->size = p_bones; + skeleton->use_2d = p_2d_skeleton; + skeleton->height = (p_bones * (p_2d_skeleton ? 2 : 3)) / 256; + if ((p_bones * (p_2d_skeleton ? 2 : 3)) % 256) { + skeleton->height++; + } + + if (skeleton->transforms_texture != 0) { + glDeleteTextures(1, &skeleton->transforms_texture); + skeleton->transforms_texture = 0; + skeleton->data.clear(); + } + + if (skeleton->size) { + skeleton->data.resize(256 * skeleton->height * 4); + glGenTextures(1, &skeleton->transforms_texture); + glBindTexture(GL_TEXTURE_2D, skeleton->transforms_texture); + glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA32F, 256, skeleton->height, 0, GL_RGBA, GL_FLOAT, nullptr); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glBindTexture(GL_TEXTURE_2D, 0); + + memset(skeleton->data.ptrw(), 0, skeleton->data.size() * sizeof(float)); + + _skeleton_make_dirty(skeleton); + } + + skeleton->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_SKELETON_DATA); } void MeshStorage::skeleton_set_base_transform_2d(RID p_skeleton, const Transform2D &p_base_transform) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_NULL(skeleton); + ERR_FAIL_COND(!skeleton->use_2d); + + skeleton->base_transform_2d = p_base_transform; } int MeshStorage::skeleton_get_bone_count(RID p_skeleton) const { - return 0; + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + ERR_FAIL_COND_V(!skeleton, 0); + + return skeleton->size; } void MeshStorage::skeleton_bone_set_transform(RID p_skeleton, int p_bone, const Transform3D &p_transform) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_COND(!skeleton); + ERR_FAIL_INDEX(p_bone, skeleton->size); + ERR_FAIL_COND(skeleton->use_2d); + + float *dataptr = skeleton->data.ptrw() + p_bone * 12; + + dataptr[0] = p_transform.basis.rows[0][0]; + dataptr[1] = p_transform.basis.rows[0][1]; + dataptr[2] = p_transform.basis.rows[0][2]; + dataptr[3] = p_transform.origin.x; + dataptr[4] = p_transform.basis.rows[1][0]; + dataptr[5] = p_transform.basis.rows[1][1]; + dataptr[6] = p_transform.basis.rows[1][2]; + dataptr[7] = p_transform.origin.y; + dataptr[8] = p_transform.basis.rows[2][0]; + dataptr[9] = p_transform.basis.rows[2][1]; + dataptr[10] = p_transform.basis.rows[2][2]; + dataptr[11] = p_transform.origin.z; + + _skeleton_make_dirty(skeleton); } Transform3D MeshStorage::skeleton_bone_get_transform(RID p_skeleton, int p_bone) const { - return Transform3D(); + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_COND_V(!skeleton, Transform3D()); + ERR_FAIL_INDEX_V(p_bone, skeleton->size, Transform3D()); + ERR_FAIL_COND_V(skeleton->use_2d, Transform3D()); + + const float *dataptr = skeleton->data.ptr() + p_bone * 12; + + Transform3D t; + + t.basis.rows[0][0] = dataptr[0]; + t.basis.rows[0][1] = dataptr[1]; + t.basis.rows[0][2] = dataptr[2]; + t.origin.x = dataptr[3]; + t.basis.rows[1][0] = dataptr[4]; + t.basis.rows[1][1] = dataptr[5]; + t.basis.rows[1][2] = dataptr[6]; + t.origin.y = dataptr[7]; + t.basis.rows[2][0] = dataptr[8]; + t.basis.rows[2][1] = dataptr[9]; + t.basis.rows[2][2] = dataptr[10]; + t.origin.z = dataptr[11]; + + return t; } void MeshStorage::skeleton_bone_set_transform_2d(RID p_skeleton, int p_bone, const Transform2D &p_transform) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_COND(!skeleton); + ERR_FAIL_INDEX(p_bone, skeleton->size); + ERR_FAIL_COND(!skeleton->use_2d); + + float *dataptr = skeleton->data.ptrw() + p_bone * 8; + + dataptr[0] = p_transform.columns[0][0]; + dataptr[1] = p_transform.columns[1][0]; + dataptr[2] = 0; + dataptr[3] = p_transform.columns[2][0]; + dataptr[4] = p_transform.columns[0][1]; + dataptr[5] = p_transform.columns[1][1]; + dataptr[6] = 0; + dataptr[7] = p_transform.columns[2][1]; + + _skeleton_make_dirty(skeleton); } Transform2D MeshStorage::skeleton_bone_get_transform_2d(RID p_skeleton, int p_bone) const { - return Transform2D(); + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_COND_V(!skeleton, Transform2D()); + ERR_FAIL_INDEX_V(p_bone, skeleton->size, Transform2D()); + ERR_FAIL_COND_V(!skeleton->use_2d, Transform2D()); + + const float *dataptr = skeleton->data.ptr() + p_bone * 8; + + Transform2D t; + t.columns[0][0] = dataptr[0]; + t.columns[1][0] = dataptr[1]; + t.columns[2][0] = dataptr[3]; + t.columns[0][1] = dataptr[4]; + t.columns[1][1] = dataptr[5]; + t.columns[2][1] = dataptr[7]; + + return t; } -void MeshStorage::skeleton_update_dependency(RID p_base, DependencyTracker *p_instance) { +void MeshStorage::_update_dirty_skeletons() { + while (skeleton_dirty_list) { + Skeleton *skeleton = skeleton_dirty_list; + + if (skeleton->size) { + glBindTexture(GL_TEXTURE_2D, skeleton->transforms_texture); + glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA32F, 256, skeleton->height, 0, GL_RGBA, GL_FLOAT, skeleton->data.ptr()); + glBindTexture(GL_TEXTURE_2D, 0); + } + + skeleton_dirty_list = skeleton->dirty_list; + + skeleton->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_SKELETON_BONES); + + skeleton->version++; + + skeleton->dirty = false; + skeleton->dirty_list = nullptr; + } + + skeleton_dirty_list = nullptr; } -/* OCCLUDER */ +void MeshStorage::skeleton_update_dependency(RID p_skeleton, DependencyTracker *p_instance) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + ERR_FAIL_COND(!skeleton); -void MeshStorage::occluder_set_mesh(RID p_occluder, const PackedVector3Array &p_vertices, const PackedInt32Array &p_indices) { + p_instance->update_dependency(&skeleton->dependency); } #endif // GLES3_ENABLED diff --git a/drivers/gles3/storage/mesh_storage.h b/drivers/gles3/storage/mesh_storage.h index 74f5800795..0f30814928 100644 --- a/drivers/gles3/storage/mesh_storage.h +++ b/drivers/gles3/storage/mesh_storage.h @@ -33,6 +33,7 @@ #ifdef GLES3_ENABLED +#include "../shaders/skeleton.glsl.gen.h" #include "core/templates/local_vector.h" #include "core/templates/rid_owner.h" #include "core/templates/self_list.h" @@ -102,7 +103,13 @@ struct Mesh { Vector<AABB> bone_aabbs; - GLuint blend_shape_buffer = 0; + struct BlendShape { + GLuint vertex_buffer = 0; + GLuint vertex_array = 0; + }; + + BlendShape *blend_shapes = nullptr; + GLuint skeleton_vertex_array = 0; RID material; }; @@ -136,7 +143,14 @@ struct MeshInstance { Mesh *mesh = nullptr; RID skeleton; struct Surface { + GLuint vertex_buffers[2] = { 0, 0 }; + GLuint vertex_arrays[2] = { 0, 0 }; GLuint vertex_buffer = 0; + int vertex_stride_cache = 0; + int vertex_size_cache = 0; + int vertex_normal_offset_cache = 0; + int vertex_tangent_offset_cache = 0; + uint32_t format_cache = 0; Mesh::Surface::Version *versions = nullptr; //allocated on demand uint32_t version_count = 0; @@ -144,7 +158,6 @@ struct MeshInstance { LocalVector<Surface> surfaces; LocalVector<float> blend_weights; - GLuint blend_weights_buffer = 0; List<MeshInstance *>::Element *I = nullptr; //used to erase itself uint64_t skeleton_version = 0; bool dirty = false; @@ -186,13 +199,15 @@ struct MultiMesh { struct Skeleton { bool use_2d = false; int size = 0; + int height = 0; Vector<float> data; - GLuint buffer = 0; bool dirty = false; Skeleton *dirty_list = nullptr; Transform2D base_transform_2d; + GLuint transforms_texture = 0; + uint64_t version = 1; Dependency dependency; @@ -202,6 +217,11 @@ class MeshStorage : public RendererMeshStorage { private: static MeshStorage *singleton; + struct { + SkeletonShaderGLES3 shader; + RID shader_version; + } skeleton_shader; + /* Mesh */ mutable RID_Owner<Mesh, true> mesh_owner; @@ -214,6 +234,7 @@ private: void _mesh_instance_clear(MeshInstance *mi); void _mesh_instance_add_surface(MeshInstance *mi, Mesh *mesh, uint32_t p_surface); + void _blend_shape_bind_mesh_instance_buffer(MeshInstance *p_mi, uint32_t p_surface); SelfList<MeshInstance>::List dirty_mesh_instance_weights; SelfList<MeshInstance>::List dirty_mesh_instance_arrays; @@ -232,9 +253,10 @@ private: mutable RID_Owner<Skeleton, true> skeleton_owner; - Skeleton *skeleton_dirty_list = nullptr; - _FORCE_INLINE_ void _skeleton_make_dirty(Skeleton *skeleton); + void _compute_skeleton(MeshInstance *p_mi, Skeleton *p_sk, uint32_t p_surface); + + Skeleton *skeleton_dirty_list = nullptr; public: static MeshStorage *get_singleton(); @@ -325,13 +347,13 @@ public: return s->index_count ? s->index_count : s->vertex_count; } - _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_mesh_lod_threshold, uint32_t *r_index_count = nullptr) const { + _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_mesh_lod_threshold, uint32_t &r_index_count) const { Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + ERR_FAIL_COND_V(!s, 0); int32_t current_lod = -1; - if (r_index_count) { - *r_index_count = s->index_count; - } + r_index_count = s->index_count; + for (uint32_t i = 0; i < s->lod_count; i++) { float screen_size = s->lods[i].edge_length * p_model_scale / p_distance_threshold; if (screen_size > p_mesh_lod_threshold) { @@ -342,9 +364,7 @@ public: if (current_lod == -1) { return 0; } else { - if (r_index_count) { - *r_index_count = s->lods[current_lod].index_count; - } + r_index_count = s->lods[current_lod].index_count; return current_lod + 1; } } @@ -406,6 +426,8 @@ public: virtual void mesh_instance_check_for_update(RID p_mesh_instance) override; virtual void update_mesh_instances() override; + // TODO: considering hashing versions with multimesh buffer RID. + // Doing so would allow us to avoid specifying multimesh buffer pointers every frame and may improve performance. _FORCE_INLINE_ void mesh_instance_surface_get_vertex_arrays_and_format(RID p_mesh_instance, uint32_t p_surface_index, uint32_t p_input_mask, GLuint &r_vertex_array_gl) { MeshInstance *mi = mesh_instance_owner.get_or_null(p_mesh_instance); ERR_FAIL_COND(!mi); @@ -534,9 +556,11 @@ public: virtual void skeleton_update_dependency(RID p_base, DependencyTracker *p_instance) override; - /* OCCLUDER */ + void _update_dirty_skeletons(); - void occluder_set_mesh(RID p_occluder, const PackedVector3Array &p_vertices, const PackedInt32Array &p_indices); + _FORCE_INLINE_ bool skeleton_is_valid(RID p_skeleton) { + return skeleton_owner.get_or_null(p_skeleton) != nullptr; + } }; } // namespace GLES3 diff --git a/drivers/gles3/storage/particles_storage.cpp b/drivers/gles3/storage/particles_storage.cpp index 9ed9fedd5a..1a0d97df01 100644 --- a/drivers/gles3/storage/particles_storage.cpp +++ b/drivers/gles3/storage/particles_storage.cpp @@ -31,6 +31,12 @@ #ifdef GLES3_ENABLED #include "particles_storage.h" +#include "material_storage.h" +#include "mesh_storage.h" +#include "texture_storage.h" +#include "utilities.h" + +#include "servers/rendering/rendering_server_default.h" using namespace GLES3; @@ -42,213 +48,1338 @@ ParticlesStorage *ParticlesStorage::get_singleton() { ParticlesStorage::ParticlesStorage() { singleton = this; + GLES3::MaterialStorage *material_storage = GLES3::MaterialStorage::get_singleton(); + + { + String global_defines; + global_defines += "#define MAX_GLOBAL_SHADER_UNIFORMS 256\n"; // TODO: this is arbitrary for now + material_storage->shaders.particles_process_shader.initialize(global_defines); + } + { + // default material and shader for particles shader + particles_shader.default_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(particles_shader.default_shader); + material_storage->shader_set_code(particles_shader.default_shader, R"( +// Default particles shader. + +shader_type particles; + +void process() { + COLOR = vec4(1.0); +} +)"); + particles_shader.default_material = material_storage->material_allocate(); + material_storage->material_initialize(particles_shader.default_material); + material_storage->material_set_shader(particles_shader.default_material, particles_shader.default_shader); + } + { + particles_shader.copy_shader.initialize(); + particles_shader.copy_shader_version = particles_shader.copy_shader.version_create(); + } } ParticlesStorage::~ParticlesStorage() { singleton = nullptr; + GLES3::MaterialStorage *material_storage = GLES3::MaterialStorage::get_singleton(); + + material_storage->material_free(particles_shader.default_material); + material_storage->shader_free(particles_shader.default_shader); + particles_shader.copy_shader.version_free(particles_shader.copy_shader_version); } /* PARTICLES */ RID ParticlesStorage::particles_allocate() { - return RID(); + return particles_owner.allocate_rid(); } void ParticlesStorage::particles_initialize(RID p_rid) { + particles_owner.initialize_rid(p_rid, Particles()); } void ParticlesStorage::particles_free(RID p_rid) { + update_particles(); + Particles *particles = particles_owner.get_or_null(p_rid); + particles->dependency.deleted_notify(p_rid); + _particles_free_data(particles); + particles_owner.free(p_rid); } void ParticlesStorage::particles_set_mode(RID p_particles, RS::ParticlesMode p_mode) { -} + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + if (particles->mode == p_mode) { + return; + } -void ParticlesStorage::particles_emit(RID p_particles, const Transform3D &p_transform, const Vector3 &p_velocity, const Color &p_color, const Color &p_custom, uint32_t p_emit_flags) { + _particles_free_data(particles); + + particles->mode = p_mode; } void ParticlesStorage::particles_set_emitting(RID p_particles, bool p_emitting) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->emitting = p_emitting; +} + +bool ParticlesStorage::particles_get_emitting(RID p_particles) { + ERR_FAIL_COND_V_MSG(RSG::threaded, false, "This function should never be used with threaded rendering, as it stalls the renderer."); + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, false); + + return particles->emitting; +} + +void ParticlesStorage::_particles_free_data(Particles *particles) { + particles->userdata_count = 0; + particles->instance_buffer_size_cache = 0; + particles->instance_buffer_stride_cache = 0; + particles->num_attrib_arrays_cache = 0; + particles->process_buffer_stride_cache = 0; + + if (particles->front_process_buffer != 0) { + glDeleteVertexArrays(1, &particles->front_vertex_array); + glDeleteBuffers(1, &particles->front_process_buffer); + glDeleteBuffers(1, &particles->front_instance_buffer); + particles->front_vertex_array = 0; + particles->front_process_buffer = 0; + particles->front_instance_buffer = 0; + + glDeleteVertexArrays(1, &particles->back_vertex_array); + glDeleteBuffers(1, &particles->back_process_buffer); + glDeleteBuffers(1, &particles->back_instance_buffer); + particles->back_vertex_array = 0; + particles->back_process_buffer = 0; + particles->back_instance_buffer = 0; + } + + if (particles->sort_buffer != 0) { + glDeleteBuffers(1, &particles->last_frame_buffer); + glDeleteBuffers(1, &particles->sort_buffer); + particles->last_frame_buffer = 0; + particles->sort_buffer = 0; + particles->sort_buffer_filled = false; + particles->last_frame_buffer_filled = false; + } + + if (particles->frame_params_ubo != 0) { + glDeleteBuffers(1, &particles->frame_params_ubo); + particles->frame_params_ubo = 0; + } } void ParticlesStorage::particles_set_amount(RID p_particles, int p_amount) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + if (particles->amount == p_amount) { + return; + } + + _particles_free_data(particles); + + particles->amount = p_amount; + + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); } void ParticlesStorage::particles_set_lifetime(RID p_particles, double p_lifetime) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->lifetime = p_lifetime; } void ParticlesStorage::particles_set_one_shot(RID p_particles, bool p_one_shot) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->one_shot = p_one_shot; } void ParticlesStorage::particles_set_pre_process_time(RID p_particles, double p_time) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->pre_process_time = p_time; } - void ParticlesStorage::particles_set_explosiveness_ratio(RID p_particles, real_t p_ratio) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->explosiveness = p_ratio; } - void ParticlesStorage::particles_set_randomness_ratio(RID p_particles, real_t p_ratio) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->randomness = p_ratio; } void ParticlesStorage::particles_set_custom_aabb(RID p_particles, const AABB &p_aabb) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->custom_aabb = p_aabb; + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); } void ParticlesStorage::particles_set_speed_scale(RID p_particles, double p_scale) { -} - -void ParticlesStorage::particles_set_use_local_coordinates(RID p_particles, bool p_enable) { -} + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); -void ParticlesStorage::particles_set_process_material(RID p_particles, RID p_material) { + particles->speed_scale = p_scale; } +void ParticlesStorage::particles_set_use_local_coordinates(RID p_particles, bool p_enable) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); -RID ParticlesStorage::particles_get_process_material(RID p_particles) const { - return RID(); + particles->use_local_coords = p_enable; + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); } void ParticlesStorage::particles_set_fixed_fps(RID p_particles, int p_fps) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->fixed_fps = p_fps; + + _particles_free_data(particles); + + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); } void ParticlesStorage::particles_set_interpolate(RID p_particles, bool p_enable) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->interpolate = p_enable; } void ParticlesStorage::particles_set_fractional_delta(RID p_particles, bool p_enable) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->fractional_delta = p_enable; } -void ParticlesStorage::particles_set_subemitter(RID p_particles, RID p_subemitter_particles) { +void ParticlesStorage::particles_set_trails(RID p_particles, bool p_enable, double p_length) { + if (p_enable) { + WARN_PRINT_ONCE("The OpenGL 3 renderer does not support particle trails"); + } } -void ParticlesStorage::particles_set_view_axis(RID p_particles, const Vector3 &p_axis, const Vector3 &p_up_axis) { +void ParticlesStorage::particles_set_trail_bind_poses(RID p_particles, const Vector<Transform3D> &p_bind_poses) { + if (p_bind_poses.size() != 0) { + WARN_PRINT_ONCE("The OpenGL 3 renderer does not support particle trails"); + } } void ParticlesStorage::particles_set_collision_base_size(RID p_particles, real_t p_size) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->collision_base_size = p_size; } void ParticlesStorage::particles_set_transform_align(RID p_particles, RS::ParticlesTransformAlign p_transform_align) { -} + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); -void ParticlesStorage::particles_set_trails(RID p_particles, bool p_enable, double p_length) { + particles->transform_align = p_transform_align; } -void ParticlesStorage::particles_set_trail_bind_poses(RID p_particles, const Vector<Transform3D> &p_bind_poses) { +void ParticlesStorage::particles_set_process_material(RID p_particles, RID p_material) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->process_material = p_material; + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); //the instance buffer may have changed } -void ParticlesStorage::particles_restart(RID p_particles) { +RID ParticlesStorage::particles_get_process_material(RID p_particles) const { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, RID()); + + return particles->process_material; } void ParticlesStorage::particles_set_draw_order(RID p_particles, RS::ParticlesDrawOrder p_order) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->draw_order = p_order; } -void ParticlesStorage::particles_set_draw_passes(RID p_particles, int p_count) { +void ParticlesStorage::particles_set_draw_passes(RID p_particles, int p_passes) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->draw_passes.resize(p_passes); } void ParticlesStorage::particles_set_draw_pass_mesh(RID p_particles, int p_pass, RID p_mesh) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + ERR_FAIL_INDEX(p_pass, particles->draw_passes.size()); + particles->draw_passes.write[p_pass] = p_mesh; + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); +} + +void ParticlesStorage::particles_restart(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->restart_request = true; +} + +void ParticlesStorage::particles_set_subemitter(RID p_particles, RID p_subemitter_particles) { + if (p_subemitter_particles.is_valid()) { + WARN_PRINT_ONCE("The OpenGL 3 renderer does not support particle sub emitters"); + } +} + +void ParticlesStorage::particles_emit(RID p_particles, const Transform3D &p_transform, const Vector3 &p_velocity, const Color &p_color, const Color &p_custom, uint32_t p_emit_flags) { + WARN_PRINT_ONCE("The OpenGL 3 renderer does not support manually emitting particles"); } void ParticlesStorage::particles_request_process(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + if (!particles->dirty) { + particles->dirty = true; + particles->update_list = particle_update_list; + particle_update_list = particles; + } } AABB ParticlesStorage::particles_get_current_aabb(RID p_particles) { - return AABB(); + if (RSG::threaded) { + WARN_PRINT_ONCE("Calling this function with threaded rendering enabled stalls the renderer, use with care."); + } + + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, AABB()); + + int total_amount = particles->amount; + + // If available, read from the sort buffer which should be 2 frames out of date. + // This will help alleviate GPU stalls. + GLuint read_buffer = particles->sort_buffer_filled ? particles->sort_buffer : particles->back_instance_buffer; + + Vector<uint8_t> buffer = Utilities::buffer_get_data(GL_ARRAY_BUFFER, read_buffer, total_amount * sizeof(ParticleInstanceData3D)); + ERR_FAIL_COND_V(buffer.size() != (int)(total_amount * sizeof(ParticleInstanceData3D)), AABB()); + + Transform3D inv = particles->emission_transform.affine_inverse(); + + AABB aabb; + if (buffer.size()) { + bool first = true; + + const uint8_t *data_ptr = (const uint8_t *)buffer.ptr(); + uint32_t particle_data_size = sizeof(ParticleInstanceData3D) + sizeof(float) * particles->userdata_count; + + for (int i = 0; i < total_amount; i++) { + const ParticleInstanceData3D &particle_data = *(const ParticleInstanceData3D *)&data_ptr[particle_data_size * i]; + // If scale is 0.0, we assume the particle is inactive. + if (particle_data.xform[0] > 0.0) { + Vector3 pos = Vector3(particle_data.xform[3], particle_data.xform[7], particle_data.xform[11]); + if (!particles->use_local_coords) { + pos = inv.xform(pos); + } + if (first) { + aabb.position = pos; + first = false; + } else { + aabb.expand_to(pos); + } + } + } + } + + float longest_axis_size = 0; + for (int i = 0; i < particles->draw_passes.size(); i++) { + if (particles->draw_passes[i].is_valid()) { + AABB maabb = MeshStorage::get_singleton()->mesh_get_aabb(particles->draw_passes[i], RID()); + longest_axis_size = MAX(maabb.get_longest_axis_size(), longest_axis_size); + } + } + + aabb.grow_by(longest_axis_size); + + return aabb; } AABB ParticlesStorage::particles_get_aabb(RID p_particles) const { - return AABB(); + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, AABB()); + + return particles->custom_aabb; } void ParticlesStorage::particles_set_emission_transform(RID p_particles, const Transform3D &p_transform) { -} + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); -bool ParticlesStorage::particles_get_emitting(RID p_particles) { - return false; + particles->emission_transform = p_transform; } int ParticlesStorage::particles_get_draw_passes(RID p_particles) const { - return 0; -} + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, 0); -RID ParticlesStorage::particles_get_draw_pass_mesh(RID p_particles, int p_pass) const { - return RID(); + return particles->draw_passes.size(); } -void ParticlesStorage::particles_add_collision(RID p_particles, RID p_instance) { +RID ParticlesStorage::particles_get_draw_pass_mesh(RID p_particles, int p_pass) const { + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, RID()); + ERR_FAIL_INDEX_V(p_pass, particles->draw_passes.size(), RID()); + + return particles->draw_passes[p_pass]; } -void ParticlesStorage::particles_remove_collision(RID p_particles, RID p_instance) { +void ParticlesStorage::particles_add_collision(RID p_particles, RID p_particles_collision_instance) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->collisions.insert(p_particles_collision_instance); +} + +void ParticlesStorage::particles_remove_collision(RID p_particles, RID p_particles_collision_instance) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->collisions.erase(p_particles_collision_instance); +} + +void ParticlesStorage::particles_set_canvas_sdf_collision(RID p_particles, bool p_enable, const Transform2D &p_xform, const Rect2 &p_to_screen, GLuint p_texture) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->has_sdf_collision = p_enable; + particles->sdf_collision_transform = p_xform; + particles->sdf_collision_to_screen = p_to_screen; + particles->sdf_collision_texture = p_texture; +} + +// Does one step of processing particles by reading from back_process_buffer and writing to front_process_buffer. +void ParticlesStorage::_particles_process(Particles *p_particles, double p_delta) { + GLES3::TextureStorage *texture_storage = GLES3::TextureStorage::get_singleton(); + GLES3::MaterialStorage *material_storage = GLES3::MaterialStorage::get_singleton(); + + double new_phase = Math::fmod(p_particles->phase + (p_delta / p_particles->lifetime) * p_particles->speed_scale, 1.0); + + //update current frame + ParticlesFrameParams frame_params; + + if (p_particles->clear) { + p_particles->cycle_number = 0; + p_particles->random_seed = Math::rand(); + } else if (new_phase < p_particles->phase) { + if (p_particles->one_shot) { + p_particles->emitting = false; + } + p_particles->cycle_number++; + } + + frame_params.emitting = p_particles->emitting; + frame_params.system_phase = new_phase; + frame_params.prev_system_phase = p_particles->phase; + + p_particles->phase = new_phase; + + frame_params.time = RSG::rasterizer->get_total_time(); + frame_params.delta = p_delta * p_particles->speed_scale; + frame_params.random_seed = p_particles->random_seed; + frame_params.explosiveness = p_particles->explosiveness; + frame_params.randomness = p_particles->randomness; + + if (p_particles->use_local_coords) { + GLES3::MaterialStorage::store_transform(Transform3D(), frame_params.emission_transform); + } else { + GLES3::MaterialStorage::store_transform(p_particles->emission_transform, frame_params.emission_transform); + } + + frame_params.cycle = p_particles->cycle_number; + frame_params.frame = p_particles->frame_counter++; + frame_params.pad0 = 0; + frame_params.pad1 = 0; + frame_params.pad2 = 0; + + { //collision and attractors + + frame_params.collider_count = 0; + frame_params.attractor_count = 0; + frame_params.particle_size = p_particles->collision_base_size; + + GLuint collision_heightmap_texture = 0; + + Transform3D to_particles; + if (p_particles->use_local_coords) { + to_particles = p_particles->emission_transform.affine_inverse(); + } + + if (p_particles->has_sdf_collision && p_particles->sdf_collision_texture != 0) { + //2D collision + + Transform2D xform = p_particles->sdf_collision_transform; //will use dotproduct manually so invert beforehand + Transform2D revert = xform.affine_inverse(); + frame_params.collider_count = 1; + frame_params.colliders[0].transform[0] = xform.columns[0][0]; + frame_params.colliders[0].transform[1] = xform.columns[0][1]; + frame_params.colliders[0].transform[2] = 0; + frame_params.colliders[0].transform[3] = xform.columns[2][0]; + + frame_params.colliders[0].transform[4] = xform.columns[1][0]; + frame_params.colliders[0].transform[5] = xform.columns[1][1]; + frame_params.colliders[0].transform[6] = 0; + frame_params.colliders[0].transform[7] = xform.columns[2][1]; + + frame_params.colliders[0].transform[8] = revert.columns[0][0]; + frame_params.colliders[0].transform[9] = revert.columns[0][1]; + frame_params.colliders[0].transform[10] = 0; + frame_params.colliders[0].transform[11] = revert.columns[2][0]; + + frame_params.colliders[0].transform[12] = revert.columns[1][0]; + frame_params.colliders[0].transform[13] = revert.columns[1][1]; + frame_params.colliders[0].transform[14] = 0; + frame_params.colliders[0].transform[15] = revert.columns[2][1]; + + frame_params.colliders[0].extents[0] = p_particles->sdf_collision_to_screen.size.x; + frame_params.colliders[0].extents[1] = p_particles->sdf_collision_to_screen.size.y; + frame_params.colliders[0].extents[2] = p_particles->sdf_collision_to_screen.position.x; + frame_params.colliders[0].scale = p_particles->sdf_collision_to_screen.position.y; + frame_params.colliders[0].type = ParticlesFrameParams::COLLISION_TYPE_2D_SDF; + + collision_heightmap_texture = p_particles->sdf_collision_texture; + } + + for (const RID &E : p_particles->collisions) { + ParticlesCollisionInstance *pci = particles_collision_instance_owner.get_or_null(E); + if (!pci || !pci->active) { + continue; + } + ParticlesCollision *pc = particles_collision_owner.get_or_null(pci->collision); + ERR_CONTINUE(!pc); + + Transform3D to_collider = pci->transform; + if (p_particles->use_local_coords) { + to_collider = to_particles * to_collider; + } + Vector3 scale = to_collider.basis.get_scale(); + to_collider.basis.orthonormalize(); + + if (pc->type <= RS::PARTICLES_COLLISION_TYPE_VECTOR_FIELD_ATTRACT) { + //attractor + if (frame_params.attractor_count >= ParticlesFrameParams::MAX_ATTRACTORS) { + continue; + } + + ParticlesFrameParams::Attractor &attr = frame_params.attractors[frame_params.attractor_count]; + + GLES3::MaterialStorage::store_transform(to_collider, attr.transform); + attr.strength = pc->attractor_strength; + attr.attenuation = pc->attractor_attenuation; + attr.directionality = pc->attractor_directionality; + + switch (pc->type) { + case RS::PARTICLES_COLLISION_TYPE_SPHERE_ATTRACT: { + attr.type = ParticlesFrameParams::ATTRACTOR_TYPE_SPHERE; + float radius = pc->radius; + radius *= (scale.x + scale.y + scale.z) / 3.0; + attr.extents[0] = radius; + attr.extents[1] = radius; + attr.extents[2] = radius; + } break; + case RS::PARTICLES_COLLISION_TYPE_BOX_ATTRACT: { + attr.type = ParticlesFrameParams::ATTRACTOR_TYPE_BOX; + Vector3 extents = pc->extents * scale; + attr.extents[0] = extents.x; + attr.extents[1] = extents.y; + attr.extents[2] = extents.z; + } break; + case RS::PARTICLES_COLLISION_TYPE_VECTOR_FIELD_ATTRACT: { + WARN_PRINT_ONCE("Vector field particle attractors are not available in the OpenGL2 renderer."); + } break; + default: { + } + } + + frame_params.attractor_count++; + } else { + //collider + if (frame_params.collider_count >= ParticlesFrameParams::MAX_COLLIDERS) { + continue; + } + + ParticlesFrameParams::Collider &col = frame_params.colliders[frame_params.collider_count]; + + GLES3::MaterialStorage::store_transform(to_collider, col.transform); + switch (pc->type) { + case RS::PARTICLES_COLLISION_TYPE_SPHERE_COLLIDE: { + col.type = ParticlesFrameParams::COLLISION_TYPE_SPHERE; + float radius = pc->radius; + radius *= (scale.x + scale.y + scale.z) / 3.0; + col.extents[0] = radius; + col.extents[1] = radius; + col.extents[2] = radius; + } break; + case RS::PARTICLES_COLLISION_TYPE_BOX_COLLIDE: { + col.type = ParticlesFrameParams::COLLISION_TYPE_BOX; + Vector3 extents = pc->extents * scale; + col.extents[0] = extents.x; + col.extents[1] = extents.y; + col.extents[2] = extents.z; + } break; + case RS::PARTICLES_COLLISION_TYPE_SDF_COLLIDE: { + WARN_PRINT_ONCE("SDF Particle Colliders are not available in the OpenGL 3 renderer."); + } break; + case RS::PARTICLES_COLLISION_TYPE_HEIGHTFIELD_COLLIDE: { + if (collision_heightmap_texture != 0) { //already taken + continue; + } + + col.type = ParticlesFrameParams::COLLISION_TYPE_HEIGHT_FIELD; + Vector3 extents = pc->extents * scale; + col.extents[0] = extents.x; + col.extents[1] = extents.y; + col.extents[2] = extents.z; + collision_heightmap_texture = pc->heightfield_texture; + } break; + default: { + } + } + + frame_params.collider_count++; + } + } + + // Bind heightmap or SDF texture. + GLuint heightmap = collision_heightmap_texture; + if (heightmap == 0) { + GLES3::Texture *tex = texture_storage->get_texture(texture_storage->texture_gl_get_default(GLES3::DEFAULT_GL_TEXTURE_BLACK)); + heightmap = tex->tex_id; + } + glActiveTexture(GL_TEXTURE0); + glBindTexture(GL_TEXTURE_2D, heightmap); + } + + if (p_particles->frame_params_ubo == 0) { + glGenBuffers(1, &p_particles->frame_params_ubo); + } + // Update per-frame UBO. + glBindBufferBase(GL_UNIFORM_BUFFER, PARTICLES_FRAME_UNIFORM_LOCATION, p_particles->frame_params_ubo); + glBufferData(GL_UNIFORM_BUFFER, sizeof(ParticlesFrameParams), &frame_params, GL_STREAM_DRAW); + + // Get shader and set shader uniforms; + ParticleProcessMaterialData *m = static_cast<ParticleProcessMaterialData *>(material_storage->material_get_data(p_particles->process_material, RS::SHADER_PARTICLES)); + if (!m) { + m = static_cast<ParticleProcessMaterialData *>(material_storage->material_get_data(particles_shader.default_material, RS::SHADER_PARTICLES)); + } + + ERR_FAIL_COND(!m); + + ParticlesShaderGLES3::ShaderVariant variant = ParticlesShaderGLES3::MODE_DEFAULT; + + uint32_t specialization = 0; + for (uint32_t i = 0; i < p_particles->userdata_count; i++) { + specialization |= (1 << i); + } + + if (p_particles->mode == RS::ParticlesMode::PARTICLES_MODE_3D) { + specialization |= ParticlesShaderGLES3::MODE_3D; + } + + RID version = particles_shader.default_shader_version; + if (m->shader_data->version.is_valid() && m->shader_data->valid) { + // Bind material uniform buffer and textures. + m->bind_uniforms(); + version = m->shader_data->version; + } + + bool success = material_storage->shaders.particles_process_shader.version_bind_shader(version, variant, specialization); + if (!success) { + return; + } + + material_storage->shaders.particles_process_shader.version_set_uniform(ParticlesShaderGLES3::LIFETIME, p_particles->lifetime, version, variant, specialization); + material_storage->shaders.particles_process_shader.version_set_uniform(ParticlesShaderGLES3::CLEAR, p_particles->clear, version, variant, specialization); + material_storage->shaders.particles_process_shader.version_set_uniform(ParticlesShaderGLES3::TOTAL_PARTICLES, uint32_t(p_particles->amount), version, variant, specialization); + material_storage->shaders.particles_process_shader.version_set_uniform(ParticlesShaderGLES3::USE_FRACTIONAL_DELTA, p_particles->fractional_delta, version, variant, specialization); + + p_particles->clear = false; + + p_particles->has_collision_cache = m->shader_data->uses_collision; + + glBindVertexArray(p_particles->back_vertex_array); + + glBindBufferBase(GL_TRANSFORM_FEEDBACK_BUFFER, 0, p_particles->front_process_buffer); + + glBeginTransformFeedback(GL_POINTS); + glDrawArrays(GL_POINTS, 0, p_particles->amount); + glEndTransformFeedback(); + + glBindBufferBase(GL_TRANSFORM_FEEDBACK_BUFFER, 0, 0); + glBindVertexArray(0); + + SWAP(p_particles->front_process_buffer, p_particles->back_process_buffer); + SWAP(p_particles->front_vertex_array, p_particles->back_vertex_array); } -void ParticlesStorage::particles_set_canvas_sdf_collision(RID p_particles, bool p_enable, const Transform2D &p_xform, const Rect2 &p_to_screen, RID p_texture) { +void ParticlesStorage::particles_set_view_axis(RID p_particles, const Vector3 &p_axis, const Vector3 &p_up_axis) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + if (particles->draw_order != RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY) { + return; + } + + if (particles->front_process_buffer == 0) { + return; //particles have not processed yet + } + + Vector3 axis = -p_axis; // cameras look to z negative + + if (particles->use_local_coords) { + axis = particles->emission_transform.basis.xform_inv(axis).normalized(); + } + + // Sort will be done on CPU since we don't have compute shaders. + // If the sort_buffer has valid data + // Use a buffer that is 2 frames out of date to avoid stalls. + if (particles->draw_order == RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH && particles->sort_buffer_filled) { + glBindBuffer(GL_ARRAY_BUFFER, particles->sort_buffer); + + ParticleInstanceData3D *particle_array; +#ifndef __EMSCRIPTEN__ + particle_array = static_cast<ParticleInstanceData3D *>(glMapBufferRange(GL_ARRAY_BUFFER, 0, particles->amount * sizeof(ParticleInstanceData3D), GL_MAP_READ_BIT | GL_MAP_WRITE_BIT)); + ERR_FAIL_NULL(particle_array); +#else + LocalVector<ParticleInstanceData3D> particle_vector; + particle_vector.resize(particles->amount); + particle_array = particle_vector.ptr(); + glGetBufferSubData(GL_ARRAY_BUFFER, 0, particles->amount * sizeof(ParticleInstanceData3D), particle_array); +#endif + SortArray<ParticleInstanceData3D, ParticlesViewSort> sorter; + sorter.compare.z_dir = axis; + sorter.sort(particle_array, particles->amount); + +#ifndef __EMSCRIPTEN__ + glUnmapBuffer(GL_ARRAY_BUFFER); +#else + glBufferSubData(GL_ARRAY_BUFFER, 0, particles->amount * sizeof(ParticleInstanceData3D), particle_vector.ptr()); +#endif + } + + glEnable(GL_RASTERIZER_DISCARD); + _particles_update_instance_buffer(particles, axis, p_up_axis); + glDisable(GL_RASTERIZER_DISCARD); +} + +void ParticlesStorage::_particles_update_buffers(Particles *particles) { + GLES3::MaterialStorage *material_storage = GLES3::MaterialStorage::get_singleton(); + uint32_t userdata_count = 0; + + if (particles->process_material.is_valid()) { + GLES3::ParticleProcessMaterialData *material_data = static_cast<GLES3::ParticleProcessMaterialData *>(material_storage->material_get_data(particles->process_material, RS::SHADER_PARTICLES)); + if (material_data && material_data->shader_data->version.is_valid() && material_data->shader_data->valid) { + userdata_count = material_data->shader_data->userdata_count; + } + } + + if (userdata_count != particles->userdata_count) { + // Mismatch userdata, re-create buffers. + _particles_free_data(particles); + } + + if (particles->amount > 0 && particles->front_process_buffer == 0) { + int total_amount = particles->amount; + + particles->userdata_count = userdata_count; + + uint32_t xform_size = particles->mode == RS::PARTICLES_MODE_2D ? 2 : 3; + particles->instance_buffer_stride_cache = sizeof(float) * 4 * (xform_size + 1); + particles->instance_buffer_size_cache = particles->instance_buffer_stride_cache * total_amount; + particles->num_attrib_arrays_cache = 5 + userdata_count + (xform_size - 2); + particles->process_buffer_stride_cache = sizeof(float) * 4 * particles->num_attrib_arrays_cache; + + int process_data_amount = 4 * particles->num_attrib_arrays_cache * total_amount; + float *data = memnew_arr(float, process_data_amount); + + for (int i = 0; i < process_data_amount; i++) { + data[i] = 0; + } + + { + glGenVertexArrays(1, &particles->front_vertex_array); + glBindVertexArray(particles->front_vertex_array); + glGenBuffers(1, &particles->front_process_buffer); + glGenBuffers(1, &particles->front_instance_buffer); + + glBindBuffer(GL_ARRAY_BUFFER, particles->front_process_buffer); + glBufferData(GL_ARRAY_BUFFER, particles->process_buffer_stride_cache * total_amount, data, GL_DYNAMIC_COPY); + + for (uint32_t j = 0; j < particles->num_attrib_arrays_cache; j++) { + glEnableVertexAttribArray(j); + glVertexAttribPointer(j, 4, GL_FLOAT, GL_FALSE, particles->process_buffer_stride_cache, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * j)); + } + glBindVertexArray(0); + + glBindBuffer(GL_ARRAY_BUFFER, particles->front_instance_buffer); + glBufferData(GL_ARRAY_BUFFER, particles->instance_buffer_size_cache, nullptr, GL_DYNAMIC_COPY); + } + + { + glGenVertexArrays(1, &particles->back_vertex_array); + glBindVertexArray(particles->back_vertex_array); + glGenBuffers(1, &particles->back_process_buffer); + glGenBuffers(1, &particles->back_instance_buffer); + + glBindBuffer(GL_ARRAY_BUFFER, particles->back_process_buffer); + glBufferData(GL_ARRAY_BUFFER, particles->process_buffer_stride_cache * total_amount, data, GL_DYNAMIC_COPY); + + for (uint32_t j = 0; j < particles->num_attrib_arrays_cache; j++) { + glEnableVertexAttribArray(j); + glVertexAttribPointer(j, 4, GL_FLOAT, GL_FALSE, particles->process_buffer_stride_cache, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * j)); + } + glBindVertexArray(0); + + glBindBuffer(GL_ARRAY_BUFFER, particles->back_instance_buffer); + glBufferData(GL_ARRAY_BUFFER, particles->instance_buffer_size_cache, nullptr, GL_DYNAMIC_COPY); + } + glBindBuffer(GL_ARRAY_BUFFER, 0); + + memdelete_arr(data); + } +} + +void ParticlesStorage::_particles_allocate_history_buffers(Particles *particles) { + if (particles->sort_buffer == 0) { + glGenBuffers(1, &particles->last_frame_buffer); + glBindBuffer(GL_ARRAY_BUFFER, particles->last_frame_buffer); + glBufferData(GL_ARRAY_BUFFER, particles->instance_buffer_size_cache, nullptr, GL_DYNAMIC_READ); + + glGenBuffers(1, &particles->sort_buffer); + glBindBuffer(GL_ARRAY_BUFFER, particles->sort_buffer); + glBufferData(GL_ARRAY_BUFFER, particles->instance_buffer_size_cache, nullptr, GL_DYNAMIC_READ); + particles->sort_buffer_filled = false; + particles->last_frame_buffer_filled = false; + glBindBuffer(GL_ARRAY_BUFFER, 0); + } +} +void ParticlesStorage::_particles_update_instance_buffer(Particles *particles, const Vector3 &p_axis, const Vector3 &p_up_axis) { + ParticlesCopyShaderGLES3::ShaderVariant variant = ParticlesCopyShaderGLES3::MODE_DEFAULT; + + uint64_t specialization = 0; + if (particles->mode == RS::ParticlesMode::PARTICLES_MODE_3D) { + specialization |= ParticlesCopyShaderGLES3::MODE_3D; + } + + bool success = particles_shader.copy_shader.version_bind_shader(particles_shader.copy_shader_version, variant, specialization); + if (!success) { + return; + } + + // Affect 2D only. + if (particles->use_local_coords) { + // In local mode, particle positions are calculated locally (relative to the node position) + // and they're also drawn locally. + // It works as expected, so we just pass an identity transform. + particles_shader.copy_shader.version_set_uniform(ParticlesCopyShaderGLES3::INV_EMISSION_TRANSFORM, Transform3D(), particles_shader.copy_shader_version, variant, specialization); + } else { + // In global mode, particle positions are calculated globally (relative to the canvas origin) + // but they're drawn locally. + // So, we need to pass the inverse of the emission transform to bring the + // particles to local coordinates before drawing. + Transform3D inv = particles->emission_transform.affine_inverse(); + particles_shader.copy_shader.version_set_uniform(ParticlesCopyShaderGLES3::INV_EMISSION_TRANSFORM, inv, particles_shader.copy_shader_version, variant, specialization); + } + + particles_shader.copy_shader.version_set_uniform(ParticlesCopyShaderGLES3::FRAME_REMAINDER, particles->interpolate ? particles->frame_remainder : 0.0, particles_shader.copy_shader_version, variant, specialization); + particles_shader.copy_shader.version_set_uniform(ParticlesCopyShaderGLES3::ALIGN_MODE, uint32_t(particles->transform_align), particles_shader.copy_shader_version, variant, specialization); + particles_shader.copy_shader.version_set_uniform(ParticlesCopyShaderGLES3::ALIGN_UP, p_up_axis, particles_shader.copy_shader_version, variant, specialization); + particles_shader.copy_shader.version_set_uniform(ParticlesCopyShaderGLES3::SORT_DIRECTION, p_axis, particles_shader.copy_shader_version, variant, specialization); + + glBindVertexArray(particles->back_vertex_array); + glBindBufferRange(GL_TRANSFORM_FEEDBACK_BUFFER, 0, particles->front_instance_buffer, 0, particles->instance_buffer_size_cache); + glBeginTransformFeedback(GL_POINTS); + + if (particles->draw_order == RS::PARTICLES_DRAW_ORDER_LIFETIME) { + uint32_t lifetime_split = MIN(particles->amount * particles->phase, particles->amount - 1); + uint32_t stride = particles->process_buffer_stride_cache; + + glBindBuffer(GL_ARRAY_BUFFER, particles->back_process_buffer); + + // Offset VBO so you render starting at the newest particle. + if (particles->amount - lifetime_split > 0) { + glEnableVertexAttribArray(0); // Color. + glVertexAttribPointer(0, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * lifetime_split + sizeof(float) * 4 * 0)); + glEnableVertexAttribArray(1); // .xyz: velocity. .z: flags. + glVertexAttribPointer(1, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * lifetime_split + sizeof(float) * 4 * 1)); + glEnableVertexAttribArray(2); // Custom. + glVertexAttribPointer(2, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * lifetime_split + sizeof(float) * 4 * 2)); + glEnableVertexAttribArray(3); // Xform1. + glVertexAttribPointer(3, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * lifetime_split + sizeof(float) * 4 * 3)); + glEnableVertexAttribArray(4); // Xform2. + glVertexAttribPointer(4, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * lifetime_split + sizeof(float) * 4 * 4)); + if (particles->mode == RS::PARTICLES_MODE_3D) { + glEnableVertexAttribArray(5); // Xform3. + glVertexAttribPointer(5, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(stride * lifetime_split + sizeof(float) * 4 * 5)); + } + + uint32_t to_draw = particles->amount - lifetime_split; + glDrawArrays(GL_POINTS, 0, to_draw); + } + + // Then render from index 0 up intil the newest particle. + if (lifetime_split > 0) { + glEndTransformFeedback(); + // Now output to the second portion of the instance buffer. + glBindBufferRange(GL_TRANSFORM_FEEDBACK_BUFFER, 0, particles->front_instance_buffer, particles->instance_buffer_stride_cache * (particles->amount - lifetime_split), particles->instance_buffer_stride_cache * (lifetime_split)); + glBeginTransformFeedback(GL_POINTS); + // Reset back to normal. + for (uint32_t j = 0; j < particles->num_attrib_arrays_cache; j++) { + glEnableVertexAttribArray(j); + glVertexAttribPointer(j, 4, GL_FLOAT, GL_FALSE, stride, CAST_INT_TO_UCHAR_PTR(sizeof(float) * 4 * j)); + } + + glDrawArrays(GL_POINTS, 0, lifetime_split); + } + } else { + glDrawArrays(GL_POINTS, 0, particles->amount); + } + + glEndTransformFeedback(); + glBindBufferRange(GL_TRANSFORM_FEEDBACK_BUFFER, 0, 0, 0, 0); + glBindVertexArray(0); + glBindBuffer(GL_ARRAY_BUFFER, 0); } void ParticlesStorage::update_particles() { + glEnable(GL_RASTERIZER_DISCARD); + + GLuint global_buffer = GLES3::MaterialStorage::get_singleton()->global_shader_parameters_get_uniform_buffer(); + + glBindBufferBase(GL_UNIFORM_BUFFER, PARTICLES_GLOBALS_UNIFORM_LOCATION, global_buffer); + glBindBuffer(GL_UNIFORM_BUFFER, 0); + + while (particle_update_list) { + // Use transform feedback to process particles. + + Particles *particles = particle_update_list; + + particle_update_list = particles->update_list; + particles->update_list = nullptr; + particles->dirty = false; + + _particles_update_buffers(particles); + + if (particles->restart_request) { + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + particles->restart_request = false; + } + + if (particles->inactive && !particles->emitting) { + //go next + continue; + } + + if (particles->emitting) { + if (particles->inactive) { + //restart system from scratch + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + } + particles->inactive = false; + particles->inactive_time = 0; + } else { + particles->inactive_time += particles->speed_scale * RSG::rasterizer->get_frame_delta_time(); + if (particles->inactive_time > particles->lifetime * 1.2) { + particles->inactive = true; + continue; + } + } + + // Copy the instance buffer that was last used into the last_frame buffer. + // sort_buffer should now be 2 frames out of date. + if (particles->draw_order == RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH || particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME) { + _particles_allocate_history_buffers(particles); + SWAP(particles->last_frame_buffer, particles->sort_buffer); + + glBindBuffer(GL_COPY_READ_BUFFER, particles->back_instance_buffer); + glBindBuffer(GL_COPY_WRITE_BUFFER, particles->last_frame_buffer); + glCopyBufferSubData(GL_COPY_READ_BUFFER, GL_COPY_WRITE_BUFFER, 0, 0, particles->instance_buffer_size_cache); + + // Last frame's last_frame turned into this frame's sort buffer. + particles->sort_buffer_filled = particles->last_frame_buffer_filled; + particles->sort_buffer_phase = particles->last_frame_phase; + particles->last_frame_buffer_filled = true; + particles->last_frame_phase = particles->phase; + glBindBuffer(GL_COPY_READ_BUFFER, 0); + glBindBuffer(GL_COPY_WRITE_BUFFER, 0); + } + + int fixed_fps = 0; + if (particles->fixed_fps > 0) { + fixed_fps = particles->fixed_fps; + } + + bool zero_time_scale = Engine::get_singleton()->get_time_scale() <= 0.0; + + if (particles->clear && particles->pre_process_time > 0.0) { + double frame_time; + if (fixed_fps > 0) { + frame_time = 1.0 / fixed_fps; + } else { + frame_time = 1.0 / 30.0; + } + + double todo = particles->pre_process_time; + + while (todo >= 0) { + _particles_process(particles, frame_time); + todo -= frame_time; + } + } + + if (fixed_fps > 0) { + double frame_time; + double decr; + if (zero_time_scale) { + frame_time = 0.0; + decr = 1.0 / fixed_fps; + } else { + frame_time = 1.0 / fixed_fps; + decr = frame_time; + } + double delta = RSG::rasterizer->get_frame_delta_time(); + if (delta > 0.1) { //avoid recursive stalls if fps goes below 10 + delta = 0.1; + } else if (delta <= 0.0) { //unlikely but.. + delta = 0.001; + } + double todo = particles->frame_remainder + delta; + + while (todo >= frame_time) { + _particles_process(particles, frame_time); + todo -= decr; + } + + particles->frame_remainder = todo; + + } else { + if (zero_time_scale) { + _particles_process(particles, 0.0); + } else { + _particles_process(particles, RSG::rasterizer->get_frame_delta_time()); + } + } + + // Copy particles to instance buffer and pack Color/Custom. + // We don't have camera information here, so don't copy here if we need camera information for view depth or align mode. + if (particles->draw_order != RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY) { + _particles_update_instance_buffer(particles, Vector3(0.0, 0.0, 0.0), Vector3(0.0, 0.0, 0.0)); + + if (particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME && particles->sort_buffer_filled) { + if (particles->mode == RS::ParticlesMode::PARTICLES_MODE_2D) { + _particles_reverse_lifetime_sort<ParticleInstanceData2D>(particles); + } else { + _particles_reverse_lifetime_sort<ParticleInstanceData3D>(particles); + } + } + } + + SWAP(particles->front_instance_buffer, particles->back_instance_buffer); + + // At the end of update, the back_buffer contains the most up-to-date-information to read from. + + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); + } + + glDisable(GL_RASTERIZER_DISCARD); +} + +template <typename ParticleInstanceData> +void ParticlesStorage::_particles_reverse_lifetime_sort(Particles *particles) { + glBindBuffer(GL_ARRAY_BUFFER, particles->sort_buffer); + + ParticleInstanceData *particle_array; + uint32_t buffer_size = particles->amount * sizeof(ParticleInstanceData); +#ifndef __EMSCRIPTEN__ + particle_array = static_cast<ParticleInstanceData *>(glMapBufferRange(GL_ARRAY_BUFFER, 0, buffer_size, GL_MAP_READ_BIT | GL_MAP_WRITE_BIT)); + + ERR_FAIL_NULL(particle_array); +#else + LocalVector<ParticleInstanceData> particle_vector; + particle_vector.resize(particles->amount); + particle_array = particle_vector.ptr(); + glGetBufferSubData(GL_ARRAY_BUFFER, 0, buffer_size, particle_array); +#endif + + uint32_t lifetime_split = MIN(particles->amount * particles->sort_buffer_phase, particles->amount - 1); + + for (uint32_t i = 0; i < lifetime_split / 2; i++) { + SWAP(particle_array[i], particle_array[lifetime_split - i]); + } + + for (uint32_t i = 0; i < (particles->amount - lifetime_split) / 2; i++) { + SWAP(particle_array[lifetime_split + i + 1], particle_array[particles->amount - 1 - i]); + } + +#ifndef __EMSCRIPTEN__ + glUnmapBuffer(GL_ARRAY_BUFFER); +#else + glBufferSubData(GL_ARRAY_BUFFER, 0, buffer_size, particle_vector.ptr()); +#endif + glBindBuffer(GL_ARRAY_BUFFER, 0); +} + +Dependency *ParticlesStorage::particles_get_dependency(RID p_particles) const { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_NULL_V(particles, nullptr); + + return &particles->dependency; } bool ParticlesStorage::particles_is_inactive(RID p_particles) const { - return false; + ERR_FAIL_COND_V_MSG(RSG::threaded, false, "This function should never be used with threaded rendering, as it stalls the renderer."); + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, false); + return !particles->emitting && particles->inactive; } -/* PARTICLES COLLISION */ +/* PARTICLES COLLISION API */ RID ParticlesStorage::particles_collision_allocate() { - return RID(); + return particles_collision_owner.allocate_rid(); } - void ParticlesStorage::particles_collision_initialize(RID p_rid) { + particles_collision_owner.initialize_rid(p_rid, ParticlesCollision()); } void ParticlesStorage::particles_collision_free(RID p_rid) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_rid); + + if (particles_collision->heightfield_texture != 0) { + glDeleteTextures(1, &particles_collision->heightfield_texture); + particles_collision->heightfield_texture = 0; + glDeleteFramebuffers(1, &particles_collision->heightfield_fb); + particles_collision->heightfield_fb = 0; + } + particles_collision->dependency.deleted_notify(p_rid); + particles_collision_owner.free(p_rid); +} + +GLuint ParticlesStorage::particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND_V(!particles_collision, 0); + ERR_FAIL_COND_V(particles_collision->type != RS::PARTICLES_COLLISION_TYPE_HEIGHTFIELD_COLLIDE, 0); + + if (particles_collision->heightfield_texture == 0) { + //create + const int resolutions[RS::PARTICLES_COLLISION_HEIGHTFIELD_RESOLUTION_MAX] = { 256, 512, 1024, 2048, 4096, 8192 }; + Size2i size; + if (particles_collision->extents.x > particles_collision->extents.z) { + size.x = resolutions[particles_collision->heightfield_resolution]; + size.y = int32_t(particles_collision->extents.z / particles_collision->extents.x * size.x); + } else { + size.y = resolutions[particles_collision->heightfield_resolution]; + size.x = int32_t(particles_collision->extents.x / particles_collision->extents.z * size.y); + } + + glGenTextures(1, &particles_collision->heightfield_texture); + glActiveTexture(GL_TEXTURE0); + glBindTexture(GL_TEXTURE_2D, particles_collision->heightfield_texture); + glTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT32F, size.x, size.y, 0, GL_DEPTH_COMPONENT, GL_FLOAT, nullptr); + + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, 0); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 1); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + + glGenFramebuffers(1, &particles_collision->heightfield_fb); + glBindFramebuffer(GL_FRAMEBUFFER, particles_collision->heightfield_fb); + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, particles_collision->heightfield_texture, 0); +#ifdef DEBUG_ENABLED + GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER); + if (status != GL_FRAMEBUFFER_COMPLETE) { + WARN_PRINT("Could create heightmap texture status: " + GLES3::TextureStorage::get_singleton()->get_framebuffer_error(status)); + } +#endif + particles_collision->heightfield_fb_size = size; + + glBindTexture(GL_TEXTURE_2D, 0); + glBindFramebuffer(GL_FRAMEBUFFER, 0); + } + + return particles_collision->heightfield_fb; } void ParticlesStorage::particles_collision_set_collision_type(RID p_particles_collision, RS::ParticlesCollisionType p_type) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + if (p_type == particles_collision->type) { + return; + } + + if (particles_collision->heightfield_texture != 0) { + glDeleteTextures(1, &particles_collision->heightfield_texture); + particles_collision->heightfield_texture = 0; + glDeleteFramebuffers(1, &particles_collision->heightfield_fb); + particles_collision->heightfield_fb = 0; + } + + particles_collision->type = p_type; + particles_collision->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); } void ParticlesStorage::particles_collision_set_cull_mask(RID p_particles_collision, uint32_t p_cull_mask) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + particles_collision->cull_mask = p_cull_mask; } void ParticlesStorage::particles_collision_set_sphere_radius(RID p_particles_collision, real_t p_radius) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->radius = p_radius; + particles_collision->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); } void ParticlesStorage::particles_collision_set_box_extents(RID p_particles_collision, const Vector3 &p_extents) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->extents = p_extents; + particles_collision->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); } void ParticlesStorage::particles_collision_set_attractor_strength(RID p_particles_collision, real_t p_strength) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->attractor_strength = p_strength; } void ParticlesStorage::particles_collision_set_attractor_directionality(RID p_particles_collision, real_t p_directionality) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->attractor_directionality = p_directionality; } void ParticlesStorage::particles_collision_set_attractor_attenuation(RID p_particles_collision, real_t p_curve) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->attractor_attenuation = p_curve; } void ParticlesStorage::particles_collision_set_field_texture(RID p_particles_collision, RID p_texture) { + WARN_PRINT_ONCE("The OpenGL 3 renderer does not support SDF collisions in 3D particle shaders"); } void ParticlesStorage::particles_collision_height_field_update(RID p_particles_collision) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + particles_collision->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); } void ParticlesStorage::particles_collision_set_height_field_resolution(RID p_particles_collision, RS::ParticlesCollisionHeightfieldResolution p_resolution) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + ERR_FAIL_INDEX(p_resolution, RS::PARTICLES_COLLISION_HEIGHTFIELD_RESOLUTION_MAX); + + if (particles_collision->heightfield_resolution == p_resolution) { + return; + } + + particles_collision->heightfield_resolution = p_resolution; + + if (particles_collision->heightfield_texture != 0) { + glDeleteTextures(1, &particles_collision->heightfield_texture); + particles_collision->heightfield_texture = 0; + glDeleteFramebuffers(1, &particles_collision->heightfield_fb); + particles_collision->heightfield_fb = 0; + } } AABB ParticlesStorage::particles_collision_get_aabb(RID p_particles_collision) const { - return AABB(); + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND_V(!particles_collision, AABB()); + + switch (particles_collision->type) { + case RS::PARTICLES_COLLISION_TYPE_SPHERE_ATTRACT: + case RS::PARTICLES_COLLISION_TYPE_SPHERE_COLLIDE: { + AABB aabb; + aabb.position = -Vector3(1, 1, 1) * particles_collision->radius; + aabb.size = Vector3(2, 2, 2) * particles_collision->radius; + return aabb; + } + default: { + AABB aabb; + aabb.position = -particles_collision->extents; + aabb.size = particles_collision->extents * 2; + return aabb; + } + } +} + +Vector3 ParticlesStorage::particles_collision_get_extents(RID p_particles_collision) const { + const ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND_V(!particles_collision, Vector3()); + return particles_collision->extents; } bool ParticlesStorage::particles_collision_is_heightfield(RID p_particles_collision) const { - return false; + const ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND_V(!particles_collision, false); + return particles_collision->type == RS::PARTICLES_COLLISION_TYPE_HEIGHTFIELD_COLLIDE; } -RID ParticlesStorage::particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const { - return RID(); +Dependency *ParticlesStorage::particles_collision_get_dependency(RID p_particles_collision) const { + ParticlesCollision *pc = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_NULL_V(pc, nullptr); + + return &pc->dependency; } +/* Particles collision instance */ + RID ParticlesStorage::particles_collision_instance_create(RID p_collision) { - return RID(); + ParticlesCollisionInstance pci; + pci.collision = p_collision; + return particles_collision_instance_owner.make_rid(pci); } void ParticlesStorage::particles_collision_instance_free(RID p_rid) { + particles_collision_instance_owner.free(p_rid); } void ParticlesStorage::particles_collision_instance_set_transform(RID p_collision_instance, const Transform3D &p_transform) { + ParticlesCollisionInstance *pci = particles_collision_instance_owner.get_or_null(p_collision_instance); + ERR_FAIL_COND(!pci); + pci->transform = p_transform; } void ParticlesStorage::particles_collision_instance_set_active(RID p_collision_instance, bool p_active) { + ParticlesCollisionInstance *pci = particles_collision_instance_owner.get_or_null(p_collision_instance); + ERR_FAIL_COND(!pci); + pci->active = p_active; } #endif // GLES3_ENABLED diff --git a/drivers/gles3/storage/particles_storage.h b/drivers/gles3/storage/particles_storage.h index 84d1f94d8c..434718006e 100644 --- a/drivers/gles3/storage/particles_storage.h +++ b/drivers/gles3/storage/particles_storage.h @@ -33,25 +33,283 @@ #ifdef GLES3_ENABLED +#include "../shaders/particles_copy.glsl.gen.h" #include "core/templates/local_vector.h" #include "core/templates/rid_owner.h" #include "core/templates/self_list.h" #include "servers/rendering/storage/particles_storage.h" +#include "servers/rendering/storage/utilities.h" + +#include "platform_config.h" +#ifndef OPENGL_INCLUDE_H +#include <GLES3/gl3.h> +#else +#include OPENGL_INCLUDE_H +#endif namespace GLES3 { +enum ParticlesUniformLocation { + PARTICLES_FRAME_UNIFORM_LOCATION, + PARTICLES_GLOBALS_UNIFORM_LOCATION, + PARTICLES_MATERIAL_UNIFORM_LOCATION, +}; + class ParticlesStorage : public RendererParticlesStorage { private: static ParticlesStorage *singleton; + /* PARTICLES */ + + struct ParticleInstanceData3D { + float xform[12]; + float color[2]; // Color and custom are packed together into one vec4; + float custom[2]; + }; + + struct ParticleInstanceData2D { + float xform[8]; + float color[2]; // Color and custom are packed together into one vec4; + float custom[2]; + }; + + struct ParticlesViewSort { + Vector3 z_dir; + bool operator()(const ParticleInstanceData3D &p_a, const ParticleInstanceData3D &p_b) const { + return z_dir.dot(Vector3(p_a.xform[3], p_a.xform[7], p_a.xform[11])) < z_dir.dot(Vector3(p_b.xform[3], p_b.xform[7], p_b.xform[11])); + } + }; + + struct ParticlesFrameParams { + enum { + MAX_ATTRACTORS = 32, + MAX_COLLIDERS = 32, + MAX_3D_TEXTURES = 0 // GLES3 renderer doesn't support using 3D textures for flow field or collisions. + }; + + enum AttractorType { + ATTRACTOR_TYPE_SPHERE, + ATTRACTOR_TYPE_BOX, + ATTRACTOR_TYPE_VECTOR_FIELD, + }; + + struct Attractor { + float transform[16]; + float extents[4]; // Extents or radius. w-channel is padding. + + uint32_t type; + float strength; + float attenuation; + float directionality; + }; + + enum CollisionType { + COLLISION_TYPE_SPHERE, + COLLISION_TYPE_BOX, + COLLISION_TYPE_SDF, + COLLISION_TYPE_HEIGHT_FIELD, + COLLISION_TYPE_2D_SDF, + + }; + + struct Collider { + float transform[16]; + float extents[4]; // Extents or radius. w-channel is padding. + + uint32_t type; + float scale; + float pad0; + float pad1; + }; + + uint32_t emitting; + uint32_t cycle; + float system_phase; + float prev_system_phase; + + float explosiveness; + float randomness; + float time; + float delta; + + float particle_size; + float pad0; + float pad1; + float pad2; + + uint32_t random_seed; + uint32_t attractor_count; + uint32_t collider_count; + uint32_t frame; + + float emission_transform[16]; + + Attractor attractors[MAX_ATTRACTORS]; + Collider colliders[MAX_COLLIDERS]; + }; + + struct Particles { + RS::ParticlesMode mode = RS::PARTICLES_MODE_3D; + bool inactive = true; + double inactive_time = 0.0; + bool emitting = false; + bool one_shot = false; + int amount = 0; + double lifetime = 1.0; + double pre_process_time = 0.0; + real_t explosiveness = 0.0; + real_t randomness = 0.0; + bool restart_request = false; + AABB custom_aabb = AABB(Vector3(-4, -4, -4), Vector3(8, 8, 8)); + bool use_local_coords = false; + bool has_collision_cache = false; + + bool has_sdf_collision = false; + Transform2D sdf_collision_transform; + Rect2 sdf_collision_to_screen; + GLuint sdf_collision_texture = 0; + + RID process_material; + uint32_t frame_counter = 0; + RS::ParticlesTransformAlign transform_align = RS::PARTICLES_TRANSFORM_ALIGN_DISABLED; + + RS::ParticlesDrawOrder draw_order = RS::PARTICLES_DRAW_ORDER_INDEX; + + Vector<RID> draw_passes; + + GLuint frame_params_ubo = 0; + + // We may process particles multiple times each frame (if they have a fixed FPS higher than the game FPS). + // Unfortunately, this means we can't just use a round-robin system of 3 buffers. + // To ensure the sort buffer is accurate, we copy the last frame instance buffer just before processing. + + // Transform Feedback buffer and VAO for rendering. + // Each frame we render to this one. + GLuint front_vertex_array = 0; // Binds process buffer. Used for processing. + GLuint front_process_buffer = 0; // Transform + color + custom data + userdata + velocity + flags. Only needed for processing. + GLuint front_instance_buffer = 0; // Transform + color + custom data. In packed format needed for rendering. + + // VAO for transform feedback, contains last frame's data. + // Read from this one for particles process and then copy to last frame buffer. + GLuint back_vertex_array = 0; // Binds process buffer. Used for processing. + GLuint back_process_buffer = 0; // Transform + color + custom data + userdata + velocity + flags. Only needed for processing. + GLuint back_instance_buffer = 0; // Transform + color + custom data. In packed format needed for rendering. + + uint32_t instance_buffer_size_cache = 0; + uint32_t instance_buffer_stride_cache = 0; + uint32_t num_attrib_arrays_cache = 0; + uint32_t process_buffer_stride_cache = 0; + + // Only ever copied to, holds last frame's instance data, then swaps with sort_buffer. + GLuint last_frame_buffer = 0; + bool last_frame_buffer_filled = false; + float last_frame_phase = 0.0; + + // The frame-before-last's instance buffer. + // Use this to copy data back for sorting or computing AABB. + GLuint sort_buffer = 0; + bool sort_buffer_filled = false; + float sort_buffer_phase = 0.0; + + uint32_t userdata_count = 0; + + bool dirty = false; + Particles *update_list = nullptr; + + double phase = 0.0; + double prev_phase = 0.0; + uint64_t prev_ticks = 0; + uint32_t random_seed = 0; + + uint32_t cycle_number = 0; + + double speed_scale = 1.0; + + int fixed_fps = 30; + bool interpolate = true; + bool fractional_delta = false; + double frame_remainder = 0; + real_t collision_base_size = 0.01; + + bool clear = true; + + Transform3D emission_transform; + + HashSet<RID> collisions; + + Dependency dependency; + + double trail_length = 1.0; + bool trails_enabled = false; + + Particles() { + } + }; + + void _particles_process(Particles *p_particles, double p_delta); + void _particles_free_data(Particles *particles); + void _particles_update_buffers(Particles *particles); + void _particles_allocate_history_buffers(Particles *particles); + void _particles_update_instance_buffer(Particles *particles, const Vector3 &p_axis, const Vector3 &p_up_axis); + + template <typename T> + void _particles_reverse_lifetime_sort(Particles *particles); + + struct ParticlesShader { + RID default_shader; + RID default_material; + RID default_shader_version; + + ParticlesCopyShaderGLES3 copy_shader; + RID copy_shader_version; + } particles_shader; + + Particles *particle_update_list = nullptr; + + mutable RID_Owner<Particles, true> particles_owner; + + /* Particles Collision */ + + struct ParticlesCollision { + RS::ParticlesCollisionType type = RS::PARTICLES_COLLISION_TYPE_SPHERE_ATTRACT; + uint32_t cull_mask = 0xFFFFFFFF; + float radius = 1.0; + Vector3 extents = Vector3(1, 1, 1); + float attractor_strength = 1.0; + float attractor_attenuation = 1.0; + float attractor_directionality = 0.0; + GLuint field_texture = 0; + GLuint heightfield_texture = 0; + GLuint heightfield_fb = 0; + Size2i heightfield_fb_size; + + RS::ParticlesCollisionHeightfieldResolution heightfield_resolution = RS::PARTICLES_COLLISION_HEIGHTFIELD_RESOLUTION_1024; + + Dependency dependency; + }; + + struct ParticlesCollisionInstance { + RID collision; + Transform3D transform; + bool active = false; + }; + + mutable RID_Owner<ParticlesCollision, true> particles_collision_owner; + + mutable RID_Owner<ParticlesCollisionInstance> particles_collision_instance_owner; + public: static ParticlesStorage *get_singleton(); ParticlesStorage(); virtual ~ParticlesStorage(); + bool free(RID p_rid); + /* PARTICLES */ + bool owns_particles(RID p_rid) { return particles_owner.owns(p_rid); } + virtual RID particles_allocate() override; virtual void particles_initialize(RID p_rid) override; virtual void particles_free(RID p_rid) override; @@ -102,12 +360,51 @@ public: virtual void particles_add_collision(RID p_particles, RID p_instance) override; virtual void particles_remove_collision(RID p_particles, RID p_instance) override; - virtual void particles_set_canvas_sdf_collision(RID p_particles, bool p_enable, const Transform2D &p_xform, const Rect2 &p_to_screen, RID p_texture) override; + void particles_set_canvas_sdf_collision(RID p_particles, bool p_enable, const Transform2D &p_xform, const Rect2 &p_to_screen, GLuint p_texture); virtual void update_particles() override; virtual bool particles_is_inactive(RID p_particles) const override; + _FORCE_INLINE_ RS::ParticlesMode particles_get_mode(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, RS::PARTICLES_MODE_2D); + return particles->mode; + } + + _FORCE_INLINE_ uint32_t particles_get_amount(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, 0); + + return particles->amount; + } + + _FORCE_INLINE_ GLuint particles_get_gl_buffer(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + + if ((particles->draw_order == RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH || particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME) && particles->sort_buffer_filled) { + return particles->sort_buffer; + } + return particles->back_instance_buffer; + } + + _FORCE_INLINE_ bool particles_has_collision(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, 0); + + return particles->has_collision_cache; + } + + _FORCE_INLINE_ uint32_t particles_is_using_local_coords(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, false); + + return particles->use_local_coords; + } + + Dependency *particles_get_dependency(RID p_particles) const; + /* PARTICLES COLLISION */ + bool owns_particles_collision(RID p_rid) { return particles_collision_owner.owns(p_rid); } virtual RID particles_collision_allocate() override; virtual void particles_collision_initialize(RID p_rid) override; @@ -124,8 +421,22 @@ public: virtual void particles_collision_height_field_update(RID p_particles_collision) override; virtual void particles_collision_set_height_field_resolution(RID p_particles_collision, RS::ParticlesCollisionHeightfieldResolution p_resolution) override; virtual AABB particles_collision_get_aabb(RID p_particles_collision) const override; + Vector3 particles_collision_get_extents(RID p_particles_collision) const; virtual bool particles_collision_is_heightfield(RID p_particles_collision) const override; - virtual RID particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const override; + GLuint particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const; + + _FORCE_INLINE_ Size2i particles_collision_get_heightfield_size(RID p_particles_collision) const { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND_V(!particles_collision, Size2i()); + ERR_FAIL_COND_V(particles_collision->type != RS::PARTICLES_COLLISION_TYPE_HEIGHTFIELD_COLLIDE, Size2i()); + + return particles_collision->heightfield_fb_size; + } + + Dependency *particles_collision_get_dependency(RID p_particles) const; + + /* PARTICLES COLLISION INSTANCE*/ + bool owns_particles_collision_instance(RID p_rid) { return particles_collision_instance_owner.owns(p_rid); } virtual RID particles_collision_instance_create(RID p_collision) override; virtual void particles_collision_instance_free(RID p_rid) override; diff --git a/drivers/gles3/storage/render_scene_buffers_gles3.cpp b/drivers/gles3/storage/render_scene_buffers_gles3.cpp index 9123984dc7..e0e78de728 100644 --- a/drivers/gles3/storage/render_scene_buffers_gles3.cpp +++ b/drivers/gles3/storage/render_scene_buffers_gles3.cpp @@ -50,54 +50,16 @@ void RenderSceneBuffersGLES3::configure(RID p_render_target, const Size2i p_inte //msaa = p_msaa; //screen_space_aa = p_screen_space_aa; //use_debanding = p_use_debanding; - //view_count = p_view_count; + view_count = p_view_count; free_render_buffer_data(); GLES3::RenderTarget *rt = texture_storage->get_render_target(p_render_target); is_transparent = rt->is_transparent; - - // framebuffer - glGenFramebuffers(1, &framebuffer); - glBindFramebuffer(GL_FRAMEBUFFER, framebuffer); - - glBindTexture(GL_TEXTURE_2D, rt->color); - glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, rt->color, 0); - - glGenTextures(1, &depth_texture); - glBindTexture(GL_TEXTURE_2D, depth_texture); - - glTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT24, rt->size.x, rt->size.y, 0, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, nullptr); - - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); - - glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, depth_texture, 0); - - GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER); - - glBindTexture(GL_TEXTURE_2D, 0); - glBindFramebuffer(GL_FRAMEBUFFER, texture_storage->system_fbo); - - if (status != GL_FRAMEBUFFER_COMPLETE) { - free_render_buffer_data(); - WARN_PRINT("Could not create 3D renderbuffer, status: " + texture_storage->get_framebuffer_error(status)); - return; - } } void RenderSceneBuffersGLES3::free_render_buffer_data() { - if (depth_texture) { - glDeleteTextures(1, &depth_texture); - depth_texture = 0; - } - if (framebuffer) { - glDeleteFramebuffers(1, &framebuffer); - framebuffer = 0; - } } #endif // GLES3_ENABLED diff --git a/drivers/gles3/storage/render_scene_buffers_gles3.h b/drivers/gles3/storage/render_scene_buffers_gles3.h index ad0d2032b0..092c14e1b8 100644 --- a/drivers/gles3/storage/render_scene_buffers_gles3.h +++ b/drivers/gles3/storage/render_scene_buffers_gles3.h @@ -56,14 +56,11 @@ public: RS::ViewportMSAA msaa = RS::VIEWPORT_MSAA_DISABLED; //RS::ViewportScreenSpaceAA screen_space_aa = RS::VIEWPORT_SCREEN_SPACE_AA_DISABLED; //bool use_debanding = false; - //uint32_t view_count = 1; + uint32_t view_count = 1; bool is_transparent = false; RID render_target; - GLuint internal_texture = 0; // Used for rendering when post effects are enabled - GLuint depth_texture = 0; // Main depth texture - GLuint framebuffer = 0; // Main framebuffer, contains internal_texture and depth_texture or render_target->color and depth_texture //built-in textures used for ping pong image processing and blurring struct Blur { diff --git a/drivers/gles3/storage/texture_storage.cpp b/drivers/gles3/storage/texture_storage.cpp index 442bd69b55..99908d197a 100644 --- a/drivers/gles3/storage/texture_storage.cpp +++ b/drivers/gles3/storage/texture_storage.cpp @@ -34,6 +34,10 @@ #include "config.h" #include "drivers/gles3/effects/copy_effects.h" +#ifdef ANDROID_ENABLED +#define glFramebufferTextureMultiviewOVR GLES3::Config::get_singleton()->eglFramebufferTextureMultiviewOVR +#endif + using namespace GLES3; TextureStorage *TextureStorage::singleton = nullptr; @@ -59,9 +63,7 @@ TextureStorage::TextureStorage() { { //create default textures { // White Textures - Ref<Image> image; - image.instantiate(); - image->create(4, 4, true, Image::FORMAT_RGBA8); + Ref<Image> image = Image::create_empty(4, 4, true, Image::FORMAT_RGBA8); image->fill(Color(1, 1, 1, 1)); image->generate_mipmaps(); @@ -90,9 +92,7 @@ TextureStorage::TextureStorage() { } { // black - Ref<Image> image; - image.instantiate(); - image->create(4, 4, true, Image::FORMAT_RGBA8); + Ref<Image> image = Image::create_empty(4, 4, true, Image::FORMAT_RGBA8); image->fill(Color(0, 0, 0, 1)); image->generate_mipmaps(); @@ -116,9 +116,7 @@ TextureStorage::TextureStorage() { } { // transparent black - Ref<Image> image; - image.instantiate(); - image->create(4, 4, true, Image::FORMAT_RGBA8); + Ref<Image> image = Image::create_empty(4, 4, true, Image::FORMAT_RGBA8); image->fill(Color(0, 0, 0, 0)); image->generate_mipmaps(); @@ -127,9 +125,7 @@ TextureStorage::TextureStorage() { } { - Ref<Image> image; - image.instantiate(); - image->create(4, 4, true, Image::FORMAT_RGBA8); + Ref<Image> image = Image::create_empty(4, 4, true, Image::FORMAT_RGBA8); image->fill(Color(0.5, 0.5, 1, 1)); image->generate_mipmaps(); @@ -138,9 +134,7 @@ TextureStorage::TextureStorage() { } { - Ref<Image> image; - image.instantiate(); - image->create(4, 4, true, Image::FORMAT_RGBA8); + Ref<Image> image = Image::create_empty(4, 4, true, Image::FORMAT_RGBA8); image->fill(Color(1.0, 0.5, 1, 1)); image->generate_mipmaps(); @@ -197,6 +191,27 @@ TextureStorage::TextureStorage() { glBindTexture(GL_TEXTURE_2D, 0); + { // Atlas Texture initialize. + uint8_t pixel_data[4 * 4 * 4]; + for (int i = 0; i < 16; i++) { + pixel_data[i * 4 + 0] = 0; + pixel_data[i * 4 + 1] = 0; + pixel_data[i * 4 + 2] = 0; + pixel_data[i * 4 + 3] = 255; + } + + glGenTextures(1, &texture_atlas.texture); + glBindTexture(GL_TEXTURE_2D, texture_atlas.texture); + glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, 4, 4, 0, GL_RGBA, GL_UNSIGNED_BYTE, pixel_data); + } + + glBindTexture(GL_TEXTURE_2D, 0); + + { + sdf_shader.shader.initialize(); + sdf_shader.shader_version = sdf_shader.shader.version_create(); + } + #ifdef GLES_OVER_GL glEnable(GL_PROGRAM_POINT_SIZE); #endif @@ -207,6 +222,12 @@ TextureStorage::~TextureStorage() { for (int i = 0; i < DEFAULT_GL_TEXTURE_MAX; i++) { texture_free(default_gl_textures[i]); } + + glDeleteTextures(1, &texture_atlas.texture); + texture_atlas.texture = 0; + glDeleteFramebuffers(1, &texture_atlas.framebuffer); + texture_atlas.framebuffer = 0; + sdf_shader.shader.version_free(sdf_shader.shader_version); } //TODO, move back to storage @@ -230,6 +251,8 @@ void TextureStorage::canvas_texture_free(RID p_rid) { void TextureStorage::canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture) { CanvasTexture *ct = canvas_texture_owner.get_or_null(p_canvas_texture); + ERR_FAIL_NULL(ct); + switch (p_channel) { case RS::CANVAS_TEXTURE_CHANNEL_DIFFUSE: { ct->diffuse = p_texture; @@ -245,6 +268,8 @@ void TextureStorage::canvas_texture_set_channel(RID p_canvas_texture, RS::Canvas void TextureStorage::canvas_texture_set_shading_parameters(RID p_canvas_texture, const Color &p_specular_color, float p_shininess) { CanvasTexture *ct = canvas_texture_owner.get_or_null(p_canvas_texture); + ERR_FAIL_NULL(ct); + ct->specular_color.r = p_specular_color.r; ct->specular_color.g = p_specular_color.g; ct->specular_color.b = p_specular_color.b; @@ -253,61 +278,16 @@ void TextureStorage::canvas_texture_set_shading_parameters(RID p_canvas_texture, void TextureStorage::canvas_texture_set_texture_filter(RID p_canvas_texture, RS::CanvasItemTextureFilter p_filter) { CanvasTexture *ct = canvas_texture_owner.get_or_null(p_canvas_texture); + ERR_FAIL_NULL(ct); + ct->texture_filter = p_filter; } void TextureStorage::canvas_texture_set_texture_repeat(RID p_canvas_texture, RS::CanvasItemTextureRepeat p_repeat) { CanvasTexture *ct = canvas_texture_owner.get_or_null(p_canvas_texture); - ct->texture_repeat = p_repeat; -} - -/* CANVAS SHADOW */ - -RID TextureStorage::canvas_light_shadow_buffer_create(int p_width) { - Config *config = Config::get_singleton(); - CanvasLightShadow *cls = memnew(CanvasLightShadow); - - if (p_width > config->max_texture_size) { - p_width = config->max_texture_size; - } - - cls->size = p_width; - cls->height = 16; + ERR_FAIL_NULL(ct); - glActiveTexture(GL_TEXTURE0); - - glGenFramebuffers(1, &cls->fbo); - glBindFramebuffer(GL_FRAMEBUFFER, cls->fbo); - - glGenRenderbuffers(1, &cls->depth); - glBindRenderbuffer(GL_RENDERBUFFER, cls->depth); - glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH_COMPONENT16, cls->size, cls->height); - glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, cls->depth); - - glGenTextures(1, &cls->distance); - glBindTexture(GL_TEXTURE_2D, cls->distance); - if (config->use_rgba_2d_shadows) { - glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, cls->size, cls->height, 0, GL_RGBA, GL_UNSIGNED_BYTE, nullptr); - } else { - glTexImage2D(GL_TEXTURE_2D, 0, GL_R32F, cls->size, cls->height, 0, GL_RED, GL_FLOAT, nullptr); - } - - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); - glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, cls->distance, 0); - - GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER); - //printf("errnum: %x\n",status); - glBindFramebuffer(GL_FRAMEBUFFER, GLES3::TextureStorage::system_fbo); - - if (status != GL_FRAMEBUFFER_COMPLETE) { - memdelete(cls); - ERR_FAIL_COND_V(status != GL_FRAMEBUFFER_COMPLETE, RID()); - } - - return canvas_light_shadow_owner.make_rid(cls); + ct->texture_repeat = p_repeat; } /* Texture API */ @@ -642,7 +622,9 @@ void TextureStorage::texture_free(RID p_texture) { } if (t->tex_id != 0) { - glDeleteTextures(1, &t->tex_id); + if (!t->is_external) { + glDeleteTextures(1, &t->tex_id); + } t->tex_id = 0; } @@ -653,7 +635,7 @@ void TextureStorage::texture_free(RID p_texture) { } } - //decal_atlas_remove_texture(p_texture); + texture_atlas_remove_texture(p_texture); for (int i = 0; i < t->proxies.size(); i++) { Texture *p = texture_owner.get_or_null(t->proxies[i]); @@ -673,7 +655,7 @@ void TextureStorage::texture_2d_initialize(RID p_texture, const Ref<Image> &p_im texture.height = p_image->get_height(); texture.alloc_width = texture.width; texture.alloc_height = texture.height; - texture.mipmaps = p_image->get_mipmap_count(); + texture.mipmaps = p_image->get_mipmap_count() + 1; texture.format = p_image->get_format(); texture.type = Texture::TYPE_2D; texture.target = GL_TEXTURE_2D; @@ -708,9 +690,37 @@ void TextureStorage::texture_proxy_initialize(RID p_texture, RID p_base) { texture_owner.initialize_rid(p_texture, proxy_tex); } +RID TextureStorage::texture_create_external(Texture::Type p_type, Image::Format p_format, unsigned int p_image, int p_width, int p_height, int p_depth, int p_layers, RS::TextureLayeredType p_layered_type) { + Texture texture; + texture.active = true; + texture.is_external = true; + texture.type = p_type; + + switch (p_type) { + case Texture::TYPE_2D: { + texture.target = GL_TEXTURE_2D; + } break; + case Texture::TYPE_3D: { + texture.target = GL_TEXTURE_3D; + } break; + case Texture::TYPE_LAYERED: { + texture.target = GL_TEXTURE_2D_ARRAY; + } break; + } + + texture.real_format = texture.format = p_format; + texture.tex_id = p_image; + texture.alloc_width = texture.width = p_width; + texture.alloc_height = texture.height = p_height; + texture.depth = p_depth; + texture.layers = p_layers; + texture.layered_type = p_layered_type; + + return texture_owner.make_rid(texture); +} + void TextureStorage::texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer) { - // only 1 layer so far - texture_set_data(p_texture, p_image); + texture_set_data(p_texture, p_image, p_layer); #ifdef TOOLS_ENABLED Texture *tex = texture_owner.get_or_null(p_texture); @@ -719,14 +729,32 @@ void TextureStorage::texture_2d_update(RID p_texture, const Ref<Image> &p_image, } void TextureStorage::texture_proxy_update(RID p_texture, RID p_proxy_to) { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + ERR_FAIL_COND(!tex->is_proxy); + Texture *proxy_to = texture_owner.get_or_null(p_proxy_to); + ERR_FAIL_COND(!proxy_to); + ERR_FAIL_COND(proxy_to->is_proxy); + + if (tex->proxy_to.is_valid()) { + Texture *prev_tex = texture_owner.get_or_null(tex->proxy_to); + ERR_FAIL_COND(!prev_tex); + prev_tex->proxies.erase(p_texture); + } + + *tex = *proxy_to; + + tex->proxy_to = p_proxy_to; + tex->is_render_target = false; + tex->is_proxy = true; + tex->proxies.clear(); + proxy_to->proxies.push_back(p_texture); } void TextureStorage::texture_2d_placeholder_initialize(RID p_texture) { //this could be better optimized to reuse an existing image , done this way //for now to get it working - Ref<Image> image; - image.instantiate(); - image->create(4, 4, false, Image::FORMAT_RGBA8); + Ref<Image> image = Image::create_empty(4, 4, false, Image::FORMAT_RGBA8); image->fill(Color(1, 0, 1, 1)); texture_2d_initialize(p_texture, image); @@ -735,9 +763,7 @@ void TextureStorage::texture_2d_placeholder_initialize(RID p_texture) { void TextureStorage::texture_2d_layered_placeholder_initialize(RID p_texture, RenderingServer::TextureLayeredType p_layered_type) { //this could be better optimized to reuse an existing image , done this way //for now to get it working - Ref<Image> image; - image.instantiate(); - image->create(4, 4, false, Image::FORMAT_RGBA8); + Ref<Image> image = Image::create_empty(4, 4, false, Image::FORMAT_RGBA8); image->fill(Color(1, 0, 1, 1)); Vector<Ref<Image>> images; @@ -756,9 +782,7 @@ void TextureStorage::texture_2d_layered_placeholder_initialize(RID p_texture, Re void TextureStorage::texture_3d_placeholder_initialize(RID p_texture) { //this could be better optimized to reuse an existing image , done this way //for now to get it working - Ref<Image> image; - image.instantiate(); - image->create(4, 4, false, Image::FORMAT_RGBA8); + Ref<Image> image = Image::create_empty(4, 4, false, Image::FORMAT_RGBA8); image->fill(Color(1, 0, 1, 1)); Vector<Ref<Image>> images; @@ -782,6 +806,7 @@ Ref<Image> TextureStorage::texture_2d_get(RID p_texture) const { #ifdef GLES_OVER_GL // OpenGL 3.3 supports glGetTexImage which is faster and simpler than glReadPixels. + // It also allows for reading compressed textures, mipmaps, and more formats. Vector<uint8_t> data; int data_size = Image::get_image_data_size(texture->alloc_width, texture->alloc_height, texture->real_format, texture->mipmaps > 1); @@ -812,16 +837,71 @@ Ref<Image> TextureStorage::texture_2d_get(RID p_texture) const { data.resize(data_size); ERR_FAIL_COND_V(data.size() == 0, Ref<Image>()); - Ref<Image> image; - image.instantiate(); - image->create(texture->width, texture->height, texture->mipmaps > 1, texture->real_format, data); + Ref<Image> image = Image::create_from_data(texture->width, texture->height, texture->mipmaps > 1, texture->real_format, data); ERR_FAIL_COND_V(image->is_empty(), Ref<Image>()); if (texture->format != texture->real_format) { image->convert(texture->format); } #else - // Support for Web and Mobile will come later. - Ref<Image> image; + + Vector<uint8_t> data; + + // On web and mobile we always read an RGBA8 image with no mipmaps. + int data_size = Image::get_image_data_size(texture->alloc_width, texture->alloc_height, Image::FORMAT_RGBA8, false); + + data.resize(data_size * 2); //add some memory at the end, just in case for buggy drivers + uint8_t *w = data.ptrw(); + + GLuint temp_framebuffer; + glGenFramebuffers(1, &temp_framebuffer); + + GLuint temp_color_texture; + glGenTextures(1, &temp_color_texture); + + glBindFramebuffer(GL_FRAMEBUFFER, temp_framebuffer); + + glBindTexture(GL_TEXTURE_2D, temp_color_texture); + glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, texture->alloc_width, texture->alloc_height, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL); + + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, temp_color_texture, 0); + + glDepthMask(GL_FALSE); + glDisable(GL_DEPTH_TEST); + glDisable(GL_CULL_FACE); + glDisable(GL_BLEND); + glDepthFunc(GL_LEQUAL); + glColorMask(1, 1, 1, 1); + glActiveTexture(GL_TEXTURE0); + glBindTexture(GL_TEXTURE_2D, texture->tex_id); + + glViewport(0, 0, texture->alloc_width, texture->alloc_height); + glClearColor(0.0, 0.0, 0.0, 0.0); + glClear(GL_COLOR_BUFFER_BIT); + + CopyEffects::get_singleton()->copy_to_rect(Rect2i(0, 0, 1.0, 1.0)); + + glReadPixels(0, 0, texture->alloc_width, texture->alloc_height, GL_RGBA, GL_UNSIGNED_BYTE, &w[0]); + + glBindFramebuffer(GL_FRAMEBUFFER, 0); + glDeleteTextures(1, &temp_color_texture); + glDeleteFramebuffers(1, &temp_framebuffer); + + data.resize(data_size); + + ERR_FAIL_COND_V(data.size() == 0, Ref<Image>()); + Ref<Image> image = Image::create_from_data(texture->width, texture->height, false, Image::FORMAT_RGBA8, data); + ERR_FAIL_COND_V(image->is_empty(), Ref<Image>()); + + if (texture->format != Image::FORMAT_RGBA8) { + image->convert(texture->format); + } + + if (texture->mipmaps > 1) { + image->generate_mipmaps(); + } + #endif #ifdef TOOLS_ENABLED @@ -875,7 +955,7 @@ void TextureStorage::texture_replace(RID p_texture, RID p_by_texture) { //delete last, so proxies can be updated texture_owner.free(p_by_texture); - //decal_atlas_mark_dirty_on_texture(p_texture); + texture_atlas_mark_dirty_on_texture(p_texture); } void TextureStorage::texture_set_size_override(RID p_texture, int p_width, int p_height) { @@ -970,6 +1050,10 @@ Size2 TextureStorage::texture_size_with_proxy(RID p_texture) { } } +RID TextureStorage::texture_get_rd_texture_rid(RID p_texture, bool p_srgb) const { + return RID(); +} + void TextureStorage::texture_set_data(RID p_texture, const Ref<Image> &p_image, int p_layer) { Texture *texture = texture_owner.get_or_null(p_texture); @@ -1009,7 +1093,7 @@ void TextureStorage::texture_set_data(RID p_texture, const Ref<Image> &p_image, img->resize_to_po2(false); } - GLenum blit_target = (texture->target == GL_TEXTURE_CUBE_MAP) ? _cube_side_enum[p_layer] : GL_TEXTURE_2D; + GLenum blit_target = (texture->target == GL_TEXTURE_CUBE_MAP) ? _cube_side_enum[p_layer] : texture->target; Vector<uint8_t> read = img->get_data(); @@ -1066,7 +1150,11 @@ void TextureStorage::texture_set_data(RID p_texture, const Ref<Image> &p_image, glCompressedTexImage2D(blit_target, i, internal_format, bw, bh, 0, size, &read[ofs]); } else { glPixelStorei(GL_UNPACK_ALIGNMENT, 1); - glTexImage2D(blit_target, i, internal_format, w, h, 0, format, type, &read[ofs]); + if (texture->target == GL_TEXTURE_2D_ARRAY) { + glTexSubImage3D(GL_TEXTURE_2D_ARRAY, i, 0, 0, p_layer, w, h, 0, format, type, &read[ofs]); + } else { + glTexImage2D(blit_target, i, internal_format, w, h, 0, format, type, &read[ofs]); + } } tsize += size; @@ -1143,6 +1231,217 @@ RID TextureStorage::texture_create_radiance_cubemap(RID p_source, int p_resoluti return RID(); } +/* TEXTURE ATLAS API */ + +void TextureStorage::texture_add_to_texture_atlas(RID p_texture) { + if (!texture_atlas.textures.has(p_texture)) { + TextureAtlas::Texture t; + t.users = 1; + texture_atlas.textures[p_texture] = t; + texture_atlas.dirty = true; + } else { + TextureAtlas::Texture *t = texture_atlas.textures.getptr(p_texture); + t->users++; + } +} + +void TextureStorage::texture_remove_from_texture_atlas(RID p_texture) { + TextureAtlas::Texture *t = texture_atlas.textures.getptr(p_texture); + ERR_FAIL_COND(!t); + t->users--; + if (t->users == 0) { + texture_atlas.textures.erase(p_texture); + // Do not mark it dirty, there is no need to since it remains working. + } +} + +void TextureStorage::texture_atlas_mark_dirty_on_texture(RID p_texture) { + if (texture_atlas.textures.has(p_texture)) { + texture_atlas.dirty = true; // Mark it dirty since it was most likely modified. + } +} + +void TextureStorage::texture_atlas_remove_texture(RID p_texture) { + if (texture_atlas.textures.has(p_texture)) { + texture_atlas.textures.erase(p_texture); + // There is not much a point of making it dirty, texture can be removed next time the atlas is updated. + } +} + +GLuint TextureStorage::texture_atlas_get_texture() const { + return texture_atlas.texture; +} + +void TextureStorage::update_texture_atlas() { + CopyEffects *copy_effects = CopyEffects::get_singleton(); + ERR_FAIL_NULL(copy_effects); + + if (!texture_atlas.dirty) { + return; //nothing to do + } + + texture_atlas.dirty = false; + + if (texture_atlas.texture != 0) { + glDeleteTextures(1, &texture_atlas.texture); + texture_atlas.texture = 0; + glDeleteFramebuffers(1, &texture_atlas.framebuffer); + texture_atlas.framebuffer = 0; + } + + const int border = 2; + + if (texture_atlas.textures.size()) { + //generate atlas + Vector<TextureAtlas::SortItem> itemsv; + itemsv.resize(texture_atlas.textures.size()); + int base_size = 8; + + int idx = 0; + + for (const KeyValue<RID, TextureAtlas::Texture> &E : texture_atlas.textures) { + TextureAtlas::SortItem &si = itemsv.write[idx]; + + Texture *src_tex = get_texture(E.key); + + si.size.width = (src_tex->width / border) + 1; + si.size.height = (src_tex->height / border) + 1; + si.pixel_size = Size2i(src_tex->width, src_tex->height); + + if (base_size < si.size.width) { + base_size = nearest_power_of_2_templated(si.size.width); + } + + si.texture = E.key; + idx++; + } + + //sort items by size + itemsv.sort(); + + //attempt to create atlas + int item_count = itemsv.size(); + TextureAtlas::SortItem *items = itemsv.ptrw(); + + int atlas_height = 0; + + while (true) { + Vector<int> v_offsetsv; + v_offsetsv.resize(base_size); + + int *v_offsets = v_offsetsv.ptrw(); + memset(v_offsets, 0, sizeof(int) * base_size); + + int max_height = 0; + + for (int i = 0; i < item_count; i++) { + //best fit + TextureAtlas::SortItem &si = items[i]; + int best_idx = -1; + int best_height = 0x7FFFFFFF; + for (int j = 0; j <= base_size - si.size.width; j++) { + int height = 0; + for (int k = 0; k < si.size.width; k++) { + int h = v_offsets[k + j]; + if (h > height) { + height = h; + if (height > best_height) { + break; //already bad + } + } + } + + if (height < best_height) { + best_height = height; + best_idx = j; + } + } + + //update + for (int k = 0; k < si.size.width; k++) { + v_offsets[k + best_idx] = best_height + si.size.height; + } + + si.pos.x = best_idx; + si.pos.y = best_height; + + if (si.pos.y + si.size.height > max_height) { + max_height = si.pos.y + si.size.height; + } + } + + if (max_height <= base_size * 2) { + atlas_height = max_height; + break; //good ratio, break; + } + + base_size *= 2; + } + + texture_atlas.size.width = base_size * border; + texture_atlas.size.height = nearest_power_of_2_templated(atlas_height * border); + + for (int i = 0; i < item_count; i++) { + TextureAtlas::Texture *t = texture_atlas.textures.getptr(items[i].texture); + t->uv_rect.position = items[i].pos * border + Vector2i(border / 2, border / 2); + t->uv_rect.size = items[i].pixel_size; + + t->uv_rect.position /= Size2(texture_atlas.size); + t->uv_rect.size /= Size2(texture_atlas.size); + } + } else { + texture_atlas.size.width = 4; + texture_atlas.size.height = 4; + } + + { // Atlas Texture initialize. + // TODO validate texture atlas size with maximum texture size + glGenTextures(1, &texture_atlas.texture); + glActiveTexture(GL_TEXTURE0); + glBindTexture(GL_TEXTURE_2D, texture_atlas.texture); + glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, texture_atlas.size.width, texture_atlas.size.height, 0, GL_RGBA, GL_UNSIGNED_BYTE, nullptr); + + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, 0); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 1); + + glGenFramebuffers(1, &texture_atlas.framebuffer); + glBindFramebuffer(GL_FRAMEBUFFER, texture_atlas.framebuffer); + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, texture_atlas.texture, 0); + + GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER); + + if (status != GL_FRAMEBUFFER_COMPLETE) { + glDeleteFramebuffers(1, &texture_atlas.framebuffer); + texture_atlas.framebuffer = 0; + glDeleteTextures(1, &texture_atlas.texture); + texture_atlas.texture = 0; + WARN_PRINT("Could not create texture atlas, status: " + get_framebuffer_error(status)); + return; + } + glViewport(0, 0, texture_atlas.size.width, texture_atlas.size.height); + glClearColor(0.0, 0.0, 0.0, 0.0); + glClear(GL_COLOR_BUFFER_BIT); + glBindTexture(GL_TEXTURE_2D, 0); + } + + glDisable(GL_BLEND); + + if (texture_atlas.textures.size()) { + for (const KeyValue<RID, TextureAtlas::Texture> &E : texture_atlas.textures) { + TextureAtlas::Texture *t = texture_atlas.textures.getptr(E.key); + Texture *src_tex = get_texture(E.key); + glActiveTexture(GL_TEXTURE0); + glBindTexture(GL_TEXTURE_2D, src_tex->tex_id); + copy_effects->copy_to_rect(t->uv_rect); + } + } + glBindFramebuffer(GL_FRAMEBUFFER, 0); +} + /* DECAL API */ RID TextureStorage::decal_allocate() { @@ -1211,6 +1510,8 @@ void TextureStorage::_update_render_target(RenderTarget *rt) { return; } + Config *config = Config::get_singleton(); + rt->color_internal_format = rt->is_transparent ? GL_RGBA8 : GL_RGB10_A2; rt->color_format = GL_RGBA; rt->color_type = rt->is_transparent ? GL_UNSIGNED_BYTE : GL_UNSIGNED_INT_2_10_10_10_REV; @@ -1221,31 +1522,74 @@ void TextureStorage::_update_render_target(RenderTarget *rt) { glDepthMask(GL_FALSE); { - /* Front FBO */ + Texture *texture; + bool use_multiview = rt->view_count > 1 && config->multiview_supported; + GLenum texture_target = use_multiview ? GL_TEXTURE_2D_ARRAY : GL_TEXTURE_2D; - Texture *texture = get_texture(rt->texture); - ERR_FAIL_COND(!texture); + /* Front FBO */ - // framebuffer glGenFramebuffers(1, &rt->fbo); glBindFramebuffer(GL_FRAMEBUFFER, rt->fbo); // color - glGenTextures(1, &rt->color); - glBindTexture(GL_TEXTURE_2D, rt->color); + if (rt->overridden.color.is_valid()) { + texture = get_texture(rt->overridden.color); + ERR_FAIL_COND(!texture); - glTexImage2D(GL_TEXTURE_2D, 0, rt->color_internal_format, rt->size.x, rt->size.y, 0, rt->color_format, rt->color_type, nullptr); + rt->color = texture->tex_id; + rt->size = Size2i(texture->width, texture->height); + } else { + texture = get_texture(rt->texture); + ERR_FAIL_COND(!texture); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + glGenTextures(1, &rt->color); + glBindTexture(texture_target, rt->color); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + if (use_multiview) { + glTexImage3D(texture_target, 0, rt->color_internal_format, rt->size.x, rt->size.y, rt->view_count, 0, rt->color_format, rt->color_type, nullptr); + } else { + glTexImage2D(texture_target, 0, rt->color_internal_format, rt->size.x, rt->size.y, 0, rt->color_format, rt->color_type, nullptr); + } + + glTexParameteri(texture_target, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + glTexParameteri(texture_target, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + glTexParameteri(texture_target, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTexParameteri(texture_target, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + } + if (use_multiview) { + glFramebufferTextureMultiviewOVR(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, rt->color, 0, 0, rt->view_count); + } else { + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, rt->color, 0); + } - glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, rt->color, 0); + // depth + if (rt->overridden.depth.is_valid()) { + texture = get_texture(rt->overridden.depth); + ERR_FAIL_COND(!texture); - GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER); + rt->depth = texture->tex_id; + } else { + glGenTextures(1, &rt->depth); + glBindTexture(texture_target, rt->depth); + + if (use_multiview) { + glTexImage3D(texture_target, 0, GL_DEPTH_COMPONENT24, rt->size.x, rt->size.y, rt->view_count, 0, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, nullptr); + } else { + glTexImage2D(texture_target, 0, GL_DEPTH_COMPONENT24, rt->size.x, rt->size.y, 0, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, nullptr); + } + + glTexParameteri(texture_target, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + glTexParameteri(texture_target, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + glTexParameteri(texture_target, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTexParameteri(texture_target, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + } + if (use_multiview) { + glFramebufferTextureMultiviewOVR(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, rt->depth, 0, 0, rt->view_count); + } else { + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, rt->depth, 0); + } + GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER); if (status != GL_FRAMEBUFFER_COMPLETE) { glDeleteFramebuffers(1, &rt->fbo); glDeleteTextures(1, &rt->color); @@ -1253,25 +1597,38 @@ void TextureStorage::_update_render_target(RenderTarget *rt) { rt->size.x = 0; rt->size.y = 0; rt->color = 0; - texture->tex_id = 0; - texture->active = false; + rt->depth = 0; + if (rt->overridden.color.is_null()) { + texture->tex_id = 0; + texture->active = false; + } WARN_PRINT("Could not create render target, status: " + get_framebuffer_error(status)); return; } - texture->format = rt->image_format; - texture->real_format = rt->image_format; - texture->type = Texture::TYPE_2D; - texture->target = GL_TEXTURE_2D; - texture->gl_format_cache = rt->color_format; - texture->gl_type_cache = GL_UNSIGNED_BYTE; - texture->gl_internal_format_cache = rt->color_internal_format; - texture->tex_id = rt->color; - texture->width = rt->size.x; - texture->alloc_width = rt->size.x; - texture->height = rt->size.y; - texture->alloc_height = rt->size.y; - texture->active = true; + if (rt->overridden.color.is_valid()) { + texture->is_render_target = true; + } else { + texture->format = rt->image_format; + texture->real_format = rt->image_format; + texture->target = texture_target; + if (rt->view_count > 1 && config->multiview_supported) { + texture->type = Texture::TYPE_LAYERED; + texture->layers = rt->view_count; + } else { + texture->type = Texture::TYPE_2D; + texture->layers = 1; + } + texture->gl_format_cache = rt->color_format; + texture->gl_type_cache = GL_UNSIGNED_BYTE; + texture->gl_internal_format_cache = rt->color_internal_format; + texture->tex_id = rt->color; + texture->width = rt->size.x; + texture->alloc_width = rt->size.x; + texture->height = rt->size.y; + texture->alloc_height = rt->size.y; + texture->active = true; + } } glClearColor(0, 0, 0, 0); @@ -1337,19 +1694,51 @@ void TextureStorage::_clear_render_target(RenderTarget *rt) { return; } + // Dispose of the cached fbo's and the allocated textures + for (KeyValue<uint32_t, RenderTarget::RTOverridden::FBOCacheEntry> &E : rt->overridden.fbo_cache) { + glDeleteTextures(E.value.allocated_textures.size(), E.value.allocated_textures.ptr()); + // Don't delete the current FBO, we'll do that a couple lines down. + if (E.value.fbo != rt->fbo) { + glDeleteFramebuffers(1, &E.value.fbo); + } + } + rt->overridden.fbo_cache.clear(); + if (rt->fbo) { glDeleteFramebuffers(1, &rt->fbo); - glDeleteTextures(1, &rt->color); rt->fbo = 0; - rt->color = 0; } - Texture *tex = get_texture(rt->texture); - tex->alloc_height = 0; - tex->alloc_width = 0; - tex->width = 0; - tex->height = 0; - tex->active = false; + if (rt->overridden.color.is_null()) { + if (rt->texture.is_valid()) { + Texture *tex = get_texture(rt->texture); + tex->alloc_height = 0; + tex->alloc_width = 0; + tex->width = 0; + tex->height = 0; + tex->active = false; + } + } else { + Texture *tex = get_texture(rt->overridden.color); + tex->is_render_target = false; + } + + if (rt->overridden.color.is_valid()) { + rt->overridden.color = RID(); + } else if (rt->color) { + glDeleteTextures(1, &rt->color); + } + rt->color = 0; + + if (rt->overridden.depth.is_valid()) { + rt->overridden.depth = RID(); + } else if (rt->depth) { + glDeleteTextures(1, &rt->depth); + } + rt->depth = 0; + + rt->overridden.velocity = RID(); + rt->overridden.is_overridden = false; if (rt->backbuffer_fbo != 0) { glDeleteFramebuffers(1, &rt->backbuffer_fbo); @@ -1357,6 +1746,7 @@ void TextureStorage::_clear_render_target(RenderTarget *rt) { rt->backbuffer = 0; rt->backbuffer_fbo = 0; } + _render_target_clear_sdf(rt); } RID TextureStorage::render_target_create() { @@ -1381,7 +1771,9 @@ void TextureStorage::render_target_free(RID p_rid) { Texture *t = get_texture(rt->texture); if (t) { t->is_render_target = false; - texture_free(rt->texture); + if (rt->overridden.color.is_null()) { + texture_free(rt->texture); + } //memdelete(t); } render_target_owner.free(p_rid); @@ -1405,13 +1797,17 @@ void TextureStorage::render_target_set_size(RID p_render_target, int p_width, in RenderTarget *rt = render_target_owner.get_or_null(p_render_target); ERR_FAIL_COND(!rt); - if (p_width == rt->size.x && p_height == rt->size.y) { + if (p_width == rt->size.x && p_height == rt->size.y && p_view_count == rt->view_count) { + return; + } + if (rt->overridden.color.is_valid()) { return; } _clear_render_target(rt); rt->size = Size2i(p_width, p_height); + rt->view_count = p_view_count; _update_render_target(rt); } @@ -1424,10 +1820,91 @@ Size2i TextureStorage::render_target_get_size(RID p_render_target) const { return rt->size; } +void TextureStorage::render_target_set_override(RID p_render_target, RID p_color_texture, RID p_depth_texture, RID p_velocity_texture) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + ERR_FAIL_COND(rt->direct_to_screen); + + rt->overridden.velocity = p_velocity_texture; + + if (rt->overridden.color == p_color_texture && rt->overridden.depth == p_depth_texture) { + return; + } + + if (p_color_texture.is_null() && p_depth_texture.is_null()) { + _clear_render_target(rt); + _update_render_target(rt); + return; + } + + if (!rt->overridden.is_overridden) { + _clear_render_target(rt); + } + + rt->overridden.color = p_color_texture; + rt->overridden.depth = p_depth_texture; + rt->overridden.is_overridden = true; + + uint32_t hash_key = hash_murmur3_one_64(p_color_texture.get_id()); + hash_key = hash_murmur3_one_64(p_depth_texture.get_id(), hash_key); + hash_key = hash_fmix32(hash_key); + + RBMap<uint32_t, RenderTarget::RTOverridden::FBOCacheEntry>::Element *cache; + if ((cache = rt->overridden.fbo_cache.find(hash_key)) != nullptr) { + rt->fbo = cache->get().fbo; + rt->color = cache->get().color; + rt->depth = cache->get().depth; + rt->size = cache->get().size; + rt->texture = p_color_texture; + return; + } + + _update_render_target(rt); + + RenderTarget::RTOverridden::FBOCacheEntry new_entry; + new_entry.fbo = rt->fbo; + new_entry.color = rt->color; + new_entry.depth = rt->depth; + new_entry.size = rt->size; + // Keep track of any textures we had to allocate because they weren't overridden. + if (p_color_texture.is_null()) { + new_entry.allocated_textures.push_back(rt->color); + } + if (p_depth_texture.is_null()) { + new_entry.allocated_textures.push_back(rt->depth); + } + rt->overridden.fbo_cache.insert(hash_key, new_entry); +} + +RID TextureStorage::render_target_get_override_color(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + return rt->overridden.color; +} + +RID TextureStorage::render_target_get_override_depth(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + return rt->overridden.depth; +} + +RID TextureStorage::render_target_get_override_velocity(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + return rt->overridden.velocity; +} + RID TextureStorage::render_target_get_texture(RID p_render_target) { RenderTarget *rt = render_target_owner.get_or_null(p_render_target); ERR_FAIL_COND_V(!rt, RID()); + if (rt->overridden.color.is_valid()) { + return rt->overridden.color; + } + return rt->texture; } @@ -1437,8 +1914,10 @@ void TextureStorage::render_target_set_transparent(RID p_render_target, bool p_t rt->is_transparent = p_transparent; - _clear_render_target(rt); - _update_render_target(rt); + if (rt->overridden.color.is_null()) { + _clear_render_target(rt); + _update_render_target(rt); + } } bool TextureStorage::render_target_get_transparent(RID p_render_target) const { @@ -1459,6 +1938,11 @@ void TextureStorage::render_target_set_direct_to_screen(RID p_render_target, boo // those functions change how they operate depending on the value of DIRECT_TO_SCREEN _clear_render_target(rt); rt->direct_to_screen = p_direct_to_screen; + if (rt->direct_to_screen) { + rt->overridden.color = RID(); + rt->overridden.depth = RID(); + rt->overridden.velocity = RID(); + } _update_render_target(rt); } @@ -1491,6 +1975,7 @@ void TextureStorage::render_target_set_msaa(RID p_render_target, RS::ViewportMSA } WARN_PRINT("2D MSAA is not yet supported for GLES3."); + _clear_render_target(rt); rt->msaa = p_msaa; _update_render_target(rt); @@ -1541,13 +2026,283 @@ void TextureStorage::render_target_do_clear_request(RID p_render_target) { } void TextureStorage::render_target_set_sdf_size_and_scale(RID p_render_target, RS::ViewportSDFOversize p_size, RS::ViewportSDFScale p_scale) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + if (rt->sdf_oversize == p_size && rt->sdf_scale == p_scale) { + return; + } + + rt->sdf_oversize = p_size; + rt->sdf_scale = p_scale; + + _render_target_clear_sdf(rt); +} + +Rect2i TextureStorage::_render_target_get_sdf_rect(const RenderTarget *rt) const { + Size2i margin; + int scale; + switch (rt->sdf_oversize) { + case RS::VIEWPORT_SDF_OVERSIZE_100_PERCENT: { + scale = 100; + } break; + case RS::VIEWPORT_SDF_OVERSIZE_120_PERCENT: { + scale = 120; + } break; + case RS::VIEWPORT_SDF_OVERSIZE_150_PERCENT: { + scale = 150; + } break; + case RS::VIEWPORT_SDF_OVERSIZE_200_PERCENT: { + scale = 200; + } break; + default: { + } + } + + margin = (rt->size * scale / 100) - rt->size; + + Rect2i r(Vector2i(), rt->size); + r.position -= margin; + r.size += margin * 2; + + return r; } Rect2i TextureStorage::render_target_get_sdf_rect(RID p_render_target) const { - return Rect2i(); + const RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, Rect2i()); + + return _render_target_get_sdf_rect(rt); } void TextureStorage::render_target_mark_sdf_enabled(RID p_render_target, bool p_enabled) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + + rt->sdf_enabled = p_enabled; +} + +bool TextureStorage::render_target_is_sdf_enabled(RID p_render_target) const { + const RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, false); + + return rt->sdf_enabled; +} + +GLuint TextureStorage::render_target_get_sdf_texture(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, 0); + if (rt->sdf_texture_read == 0) { + Texture *texture = texture_owner.get_or_null(default_gl_textures[DEFAULT_GL_TEXTURE_BLACK]); + return texture->tex_id; + } + + return rt->sdf_texture_read; +} + +void TextureStorage::_render_target_allocate_sdf(RenderTarget *rt) { + ERR_FAIL_COND(rt->sdf_texture_write_fb != 0); + + Size2i size = _render_target_get_sdf_rect(rt).size; + + glGenTextures(1, &rt->sdf_texture_write); + glActiveTexture(GL_TEXTURE0); + glBindTexture(GL_TEXTURE_2D, rt->sdf_texture_write); + glTexImage2D(GL_TEXTURE_2D, 0, GL_R8, size.width, size.height, 0, GL_RED, GL_UNSIGNED_BYTE, nullptr); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, 0); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 1); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + + glGenFramebuffers(1, &rt->sdf_texture_write_fb); + glBindFramebuffer(GL_FRAMEBUFFER, rt->sdf_texture_write_fb); + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, rt->sdf_texture_write, 0); + + int scale; + switch (rt->sdf_scale) { + case RS::VIEWPORT_SDF_SCALE_100_PERCENT: { + scale = 100; + } break; + case RS::VIEWPORT_SDF_SCALE_50_PERCENT: { + scale = 50; + } break; + case RS::VIEWPORT_SDF_SCALE_25_PERCENT: { + scale = 25; + } break; + default: { + scale = 100; + } break; + } + + rt->process_size = size * scale / 100; + rt->process_size.x = MAX(rt->process_size.x, 1); + rt->process_size.y = MAX(rt->process_size.y, 1); + + glGenTextures(2, rt->sdf_texture_process); + glBindTexture(GL_TEXTURE_2D, rt->sdf_texture_process[0]); + glTexImage2D(GL_TEXTURE_2D, 0, GL_RG16I, rt->process_size.width, rt->process_size.height, 0, GL_RG_INTEGER, GL_SHORT, nullptr); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, 0); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 1); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + + glBindTexture(GL_TEXTURE_2D, rt->sdf_texture_process[1]); + glTexImage2D(GL_TEXTURE_2D, 0, GL_RG16I, rt->process_size.width, rt->process_size.height, 0, GL_RG_INTEGER, GL_SHORT, nullptr); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, 0); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 1); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + + glGenTextures(1, &rt->sdf_texture_read); + glBindTexture(GL_TEXTURE_2D, rt->sdf_texture_read); + glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, rt->process_size.width, rt->process_size.height, 0, GL_RGBA, GL_UNSIGNED_BYTE, nullptr); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, 0); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 1); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); +} + +void TextureStorage::_render_target_clear_sdf(RenderTarget *rt) { + if (rt->sdf_texture_write_fb != 0) { + glDeleteTextures(1, &rt->sdf_texture_read); + glDeleteTextures(1, &rt->sdf_texture_write); + glDeleteTextures(2, rt->sdf_texture_process); + glDeleteFramebuffers(1, &rt->sdf_texture_write_fb); + rt->sdf_texture_read = 0; + rt->sdf_texture_write = 0; + rt->sdf_texture_process[0] = 0; + rt->sdf_texture_process[1] = 0; + rt->sdf_texture_write_fb = 0; + } +} + +GLuint TextureStorage::render_target_get_sdf_framebuffer(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, 0); + + if (rt->sdf_texture_write_fb == 0) { + _render_target_allocate_sdf(rt); + } + + return rt->sdf_texture_write_fb; +} +void TextureStorage::render_target_sdf_process(RID p_render_target) { + CopyEffects *copy_effects = CopyEffects::get_singleton(); + + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + ERR_FAIL_COND(rt->sdf_texture_write_fb == 0); + + Rect2i r = _render_target_get_sdf_rect(rt); + + Size2i size = r.size; + int32_t shift = 0; + + bool shrink = false; + + switch (rt->sdf_scale) { + case RS::VIEWPORT_SDF_SCALE_50_PERCENT: { + size[0] >>= 1; + size[1] >>= 1; + shift = 1; + shrink = true; + } break; + case RS::VIEWPORT_SDF_SCALE_25_PERCENT: { + size[0] >>= 2; + size[1] >>= 2; + shift = 2; + shrink = true; + } break; + default: { + }; + } + + GLuint temp_fb; + glGenFramebuffers(1, &temp_fb); + glBindFramebuffer(GL_FRAMEBUFFER, temp_fb); + + // Load + CanvasSdfShaderGLES3::ShaderVariant variant = shrink ? CanvasSdfShaderGLES3::MODE_LOAD_SHRINK : CanvasSdfShaderGLES3::MODE_LOAD; + bool success = sdf_shader.shader.version_bind_shader(sdf_shader.shader_version, variant); + if (!success) { + return; + } + + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::BASE_SIZE, r.size, sdf_shader.shader_version, variant); + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::SIZE, size, sdf_shader.shader_version, variant); + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::STRIDE, 0, sdf_shader.shader_version, variant); + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::SHIFT, shift, sdf_shader.shader_version, variant); + + glActiveTexture(GL_TEXTURE0); + glBindTexture(GL_TEXTURE_2D, rt->sdf_texture_write); + + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, rt->sdf_texture_process[0], 0); + glViewport(0, 0, size.width, size.height); + glEnable(GL_SCISSOR_TEST); + glScissor(0, 0, size.width, size.height); + + copy_effects->draw_screen_triangle(); + + // Process + + int stride = nearest_power_of_2_templated(MAX(size.width, size.height) / 2); + + variant = CanvasSdfShaderGLES3::MODE_PROCESS; + success = sdf_shader.shader.version_bind_shader(sdf_shader.shader_version, variant); + if (!success) { + return; + } + + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::BASE_SIZE, r.size, sdf_shader.shader_version, variant); + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::SIZE, size, sdf_shader.shader_version, variant); + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::STRIDE, stride, sdf_shader.shader_version, variant); + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::SHIFT, shift, sdf_shader.shader_version, variant); + + bool swap = false; + + //jumpflood + while (stride > 0) { + glBindTexture(GL_TEXTURE_2D, 0); + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, rt->sdf_texture_process[swap ? 0 : 1], 0); + glBindTexture(GL_TEXTURE_2D, rt->sdf_texture_process[swap ? 1 : 0]); + + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::STRIDE, stride, sdf_shader.shader_version, variant); + + copy_effects->draw_screen_triangle(); + + stride /= 2; + swap = !swap; + } + + // Store + variant = shrink ? CanvasSdfShaderGLES3::MODE_STORE_SHRINK : CanvasSdfShaderGLES3::MODE_STORE; + success = sdf_shader.shader.version_bind_shader(sdf_shader.shader_version, variant); + if (!success) { + return; + } + + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::BASE_SIZE, r.size, sdf_shader.shader_version, variant); + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::SIZE, size, sdf_shader.shader_version, variant); + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::STRIDE, stride, sdf_shader.shader_version, variant); + sdf_shader.shader.version_set_uniform(CanvasSdfShaderGLES3::SHIFT, shift, sdf_shader.shader_version, variant); + + glBindTexture(GL_TEXTURE_2D, 0); + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, rt->sdf_texture_read, 0); + glBindTexture(GL_TEXTURE_2D, rt->sdf_texture_process[swap ? 1 : 0]); + + copy_effects->draw_screen_triangle(); + + glBindTexture(GL_TEXTURE_2D, 0); + glBindFramebuffer(GL_FRAMEBUFFER, system_fbo); + glDeleteFramebuffers(1, &temp_fb); + glDisable(GL_SCISSOR_TEST); } void TextureStorage::render_target_copy_to_back_buffer(RID p_render_target, const Rect2i &p_region, bool p_gen_mipmaps) { diff --git a/drivers/gles3/storage/texture_storage.h b/drivers/gles3/storage/texture_storage.h index 7e083e48e8..169c50638d 100644 --- a/drivers/gles3/storage/texture_storage.h +++ b/drivers/gles3/storage/texture_storage.h @@ -39,6 +39,8 @@ #include "servers/rendering/renderer_compositor.h" #include "servers/rendering/storage/texture_storage.h" +#include "../shaders/canvas_sdf.glsl.gen.h" + // This must come first to avoid windows.h mess #include "platform_config.h" #ifndef OPENGL_INCLUDE_H @@ -84,18 +86,8 @@ namespace GLES3 { #define _GL_TEXTURE_EXTERNAL_OES 0x8D65 -#ifdef GLES_OVER_GL -#define _GL_HALF_FLOAT_OES 0x140B -#else -#define _GL_HALF_FLOAT_OES 0x8D61 -#endif - #define _EXT_TEXTURE_CUBE_MAP_SEAMLESS 0x884F -#define _RED_OES 0x1903 - -#define _DEPTH_COMPONENT24_OES 0x81A6 - #ifndef GLES_OVER_GL #define glClearDepth glClearDepthf #endif //!GLES_OVER_GL @@ -128,26 +120,16 @@ struct CanvasTexture { RS::CanvasItemTextureRepeat texture_repeat = RS::CANVAS_ITEM_TEXTURE_REPEAT_DEFAULT; }; -/* CANVAS SHADOW */ - -struct CanvasLightShadow { - RID self; - int size; - int height; - GLuint fbo; - GLuint depth; - GLuint distance; //for older devices -}; - struct RenderTarget; struct Texture { RID self; bool is_proxy = false; + bool is_external = false; bool is_render_target = false; - RID proxy_to = RID(); + RID proxy_to; Vector<RID> proxies; String path; @@ -206,6 +188,7 @@ struct Texture { void copy_from(const Texture &o) { proxy_to = o.proxy_to; is_proxy = o.is_proxy; + is_external = o.is_external; width = o.width; height = o.height; alloc_width = o.alloc_width; @@ -324,10 +307,12 @@ private: struct RenderTarget { Point2i position = Point2i(0, 0); Size2i size = Size2i(0, 0); + uint32_t view_count = 1; int mipmap_count = 1; RID self; GLuint fbo = 0; GLuint color = 0; + GLuint depth = 0; GLuint backbuffer_fbo = 0; GLuint backbuffer = 0; @@ -336,12 +321,37 @@ struct RenderTarget { GLuint color_type = GL_UNSIGNED_BYTE; Image::Format image_format = Image::FORMAT_RGBA8; + GLuint sdf_texture_write = 0; + GLuint sdf_texture_write_fb = 0; + GLuint sdf_texture_process[2] = { 0, 0 }; + GLuint sdf_texture_read = 0; + RS::ViewportSDFOversize sdf_oversize = RS::VIEWPORT_SDF_OVERSIZE_120_PERCENT; + RS::ViewportSDFScale sdf_scale = RS::VIEWPORT_SDF_SCALE_50_PERCENT; + Size2i process_size; + bool sdf_enabled = false; + bool is_transparent = false; bool direct_to_screen = false; bool used_in_frame = false; RS::ViewportMSAA msaa = RS::VIEWPORT_MSAA_DISABLED; + struct RTOverridden { + bool is_overridden = false; + RID color; + RID depth; + RID velocity; + + struct FBOCacheEntry { + GLuint fbo; + GLuint color; + GLuint depth; + Size2i size; + Vector<GLuint> allocated_textures; + }; + RBMap<uint32_t, FBOCacheEntry> fbo_cache; + } overridden; + RID texture; Color clear_color = Color(1, 1, 1, 1); @@ -361,16 +371,44 @@ private: RID_Owner<CanvasTexture, true> canvas_texture_owner; - /* CANVAS SHADOW */ - - RID_PtrOwner<CanvasLightShadow> canvas_light_shadow_owner; - /* Texture API */ mutable RID_Owner<Texture> texture_owner; Ref<Image> _get_gl_image_and_format(const Ref<Image> &p_image, Image::Format p_format, Image::Format &r_real_format, GLenum &r_gl_format, GLenum &r_gl_internal_format, GLenum &r_gl_type, bool &r_compressed, bool p_force_decompress) const; + /* TEXTURE ATLAS API */ + + struct TextureAtlas { + struct Texture { + int users; + Rect2 uv_rect; + }; + + struct SortItem { + RID texture; + Size2i pixel_size; + Size2i size; + Point2i pos; + + bool operator<(const SortItem &p_item) const { + //sort larger to smaller + if (size.height == p_item.size.height) { + return size.width > p_item.size.width; + } else { + return size.height > p_item.size.height; + } + } + }; + + HashMap<RID, Texture> textures; + bool dirty = true; + + GLuint texture = 0; + GLuint framebuffer = 0; + Size2i size; + } texture_atlas; + /* Render Target API */ mutable RID_Owner<RenderTarget> render_target_owner; @@ -378,6 +416,14 @@ private: void _clear_render_target(RenderTarget *rt); void _update_render_target(RenderTarget *rt); void _create_render_target_backbuffer(RenderTarget *rt); + void _render_target_allocate_sdf(RenderTarget *rt); + void _render_target_clear_sdf(RenderTarget *rt); + Rect2i _render_target_get_sdf_rect(const RenderTarget *rt) const; + + struct RenderTargetSDF { + CanvasSdfShaderGLES3 shader; + RID shader_version; + } sdf_shader; public: static TextureStorage *get_singleton(); @@ -404,10 +450,6 @@ public: virtual void canvas_texture_set_texture_filter(RID p_item, RS::CanvasItemTextureFilter p_filter) override; virtual void canvas_texture_set_texture_repeat(RID p_item, RS::CanvasItemTextureRepeat p_repeat) override; - /* CANVAS SHADOW */ - - RID canvas_light_shadow_buffer_create(int p_width); - /* Texture API */ Texture *get_texture(RID p_rid) { @@ -431,6 +473,8 @@ public: virtual void texture_3d_initialize(RID p_texture, Image::Format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data) override; virtual void texture_proxy_initialize(RID p_texture, RID p_base) override; //all slices, then all the mipmaps, must be coherent + RID texture_create_external(Texture::Type p_type, Image::Format p_format, unsigned int p_image, int p_width, int p_height, int p_depth, int p_layers, RS::TextureLayeredType p_layered_type = RS::TEXTURE_LAYERED_2D_ARRAY); + virtual void texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer = 0) override; virtual void texture_3d_update(RID p_texture, const Vector<Ref<Image>> &p_data) override{}; virtual void texture_proxy_update(RID p_proxy, RID p_base) override; @@ -461,6 +505,8 @@ public: virtual Size2 texture_size_with_proxy(RID p_proxy) override; + virtual RID texture_get_rd_texture_rid(RID p_texture, bool p_srgb = false) const override; + void texture_set_data(RID p_texture, const Ref<Image> &p_image, int p_layer = 0); void texture_set_data_partial(RID p_texture, const Ref<Image> &p_image, int src_x, int src_y, int src_w, int src_h, int dst_x, int dst_y, int p_dst_mip, int p_layer = 0); //Ref<Image> texture_get_data(RID p_texture, int p_layer = 0) const; @@ -473,6 +519,25 @@ public: void texture_bind(RID p_texture, uint32_t p_texture_no); RID texture_create_radiance_cubemap(RID p_source, int p_resolution = -1) const; + /* TEXTURE ATLAS API */ + + void update_texture_atlas(); + + GLuint texture_atlas_get_texture() const; + _FORCE_INLINE_ Rect2 texture_atlas_get_texture_rect(RID p_texture) { + TextureAtlas::Texture *t = texture_atlas.textures.getptr(p_texture); + if (!t) { + return Rect2(); + } + + return t->uv_rect; + } + + void texture_add_to_texture_atlas(RID p_texture); + void texture_remove_from_texture_atlas(RID p_texture); + void texture_atlas_mark_dirty_on_texture(RID p_texture); + void texture_atlas_remove_texture(RID p_texture); + /* DECAL API */ virtual RID decal_allocate() override; @@ -534,9 +599,13 @@ public: void render_target_disable_clear_request(RID p_render_target) override; void render_target_do_clear_request(RID p_render_target) override; - void render_target_set_sdf_size_and_scale(RID p_render_target, RS::ViewportSDFOversize p_size, RS::ViewportSDFScale p_scale) override; - Rect2i render_target_get_sdf_rect(RID p_render_target) const override; - void render_target_mark_sdf_enabled(RID p_render_target, bool p_enabled) override; + virtual void render_target_set_sdf_size_and_scale(RID p_render_target, RS::ViewportSDFOversize p_size, RS::ViewportSDFScale p_scale) override; + virtual Rect2i render_target_get_sdf_rect(RID p_render_target) const override; + GLuint render_target_get_sdf_texture(RID p_render_target); + GLuint render_target_get_sdf_framebuffer(RID p_render_target); + void render_target_sdf_process(RID p_render_target); + virtual void render_target_mark_sdf_enabled(RID p_render_target, bool p_enabled) override; + bool render_target_is_sdf_enabled(RID p_render_target) const; void render_target_copy_to_back_buffer(RID p_render_target, const Rect2i &p_region, bool p_gen_mipmaps); void render_target_clear_back_buffer(RID p_render_target, const Rect2i &p_region, const Color &p_color); @@ -547,12 +616,10 @@ public: virtual void render_target_set_vrs_texture(RID p_render_target, RID p_texture) override {} virtual RID render_target_get_vrs_texture(RID p_render_target) const override { return RID(); } - virtual void render_target_set_override_color(RID p_render_target, RID p_texture) override {} - virtual RID render_target_get_override_color(RID p_render_target) const override { return RID(); } - virtual void render_target_set_override_depth(RID p_render_target, RID p_texture) override {} - virtual RID render_target_get_override_depth(RID p_render_target) const override { return RID(); } - virtual void render_target_set_override_velocity(RID p_render_target, RID p_texture) override {} - virtual RID render_target_get_override_velocity(RID p_render_target) const override { return RID(); } + virtual void render_target_set_override(RID p_render_target, RID p_color_texture, RID p_depth_texture, RID p_velocity_texture) override; + virtual RID render_target_get_override_color(RID p_render_target) const override; + virtual RID render_target_get_override_depth(RID p_render_target) const override; + virtual RID render_target_get_override_velocity(RID p_render_target) const override; virtual RID render_target_get_texture(RID p_render_target) override; diff --git a/drivers/gles3/storage/utilities.cpp b/drivers/gles3/storage/utilities.cpp index 16bacf1829..fe900c7cfb 100644 --- a/drivers/gles3/storage/utilities.cpp +++ b/drivers/gles3/storage/utilities.cpp @@ -38,20 +38,44 @@ #include "particles_storage.h" #include "texture_storage.h" +#include "servers/rendering/rendering_server_globals.h" + using namespace GLES3; Utilities *Utilities::singleton = nullptr; Utilities::Utilities() { singleton = this; + frame = 0; + for (int i = 0; i < FRAME_COUNT; i++) { + frames[i].index = 0; + glGenQueries(max_timestamp_query_elements, frames[i].queries); + + frames[i].timestamp_names.resize(max_timestamp_query_elements); + frames[i].timestamp_cpu_values.resize(max_timestamp_query_elements); + frames[i].timestamp_count = 0; + + frames[i].timestamp_result_names.resize(max_timestamp_query_elements); + frames[i].timestamp_cpu_result_values.resize(max_timestamp_query_elements); + frames[i].timestamp_result_values.resize(max_timestamp_query_elements); + frames[i].timestamp_result_count = 0; + } } Utilities::~Utilities() { singleton = nullptr; + for (int i = 0; i < FRAME_COUNT; i++) { + glDeleteQueries(max_timestamp_query_elements, frames[i].queries); + } } Vector<uint8_t> Utilities::buffer_get_data(GLenum p_target, GLuint p_buffer, uint32_t p_buffer_size) { Vector<uint8_t> ret; + + if (p_buffer_size == 0) { + return ret; + } + ret.resize(p_buffer_size); glBindBuffer(p_target, p_buffer); @@ -84,6 +108,10 @@ RS::InstanceType Utilities::get_base_type(RID p_rid) const { return RS::INSTANCE_LIGHT; } else if (GLES3::LightStorage::get_singleton()->owns_lightmap(p_rid)) { return RS::INSTANCE_LIGHTMAP; + } else if (GLES3::ParticlesStorage::get_singleton()->owns_particles(p_rid)) { + return RS::INSTANCE_PARTICLES; + } else if (GLES3::ParticlesStorage::get_singleton()->owns_particles_collision(p_rid)) { + return RS::INSTANCE_PARTICLES_COLLISION; } return RS::INSTANCE_NONE; } @@ -119,53 +147,18 @@ bool Utilities::free(RID p_rid) { } else if (GLES3::LightStorage::get_singleton()->owns_lightmap(p_rid)) { GLES3::LightStorage::get_singleton()->lightmap_free(p_rid); return true; - } else { - return false; - } - /* - else if (reflection_probe_owner.owns(p_rid)) { - // delete the texture - ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_rid); - reflection_probe->instance_remove_deps(); - - reflection_probe_owner.free(p_rid); - memdelete(reflection_probe); - - return true; - } else if (lightmap_capture_data_owner.owns(p_rid)) { - // delete the texture - LightmapCapture *lightmap_capture = lightmap_capture_data_owner.get_or_null(p_rid); - lightmap_capture->instance_remove_deps(); - - lightmap_capture_data_owner.free(p_rid); - memdelete(lightmap_capture); + } else if (GLES3::ParticlesStorage::get_singleton()->owns_particles(p_rid)) { + GLES3::ParticlesStorage::get_singleton()->particles_free(p_rid); return true; - - } else if (canvas_occluder_owner.owns(p_rid)) { - CanvasOccluder *co = canvas_occluder_owner.get_or_null(p_rid); - if (co->index_id) { - glDeleteBuffers(1, &co->index_id); - } - if (co->vertex_id) { - glDeleteBuffers(1, &co->vertex_id); - } - - canvas_occluder_owner.free(p_rid); - memdelete(co); - + } else if (GLES3::ParticlesStorage::get_singleton()->owns_particles_collision(p_rid)) { + GLES3::ParticlesStorage::get_singleton()->particles_collision_free(p_rid); return true; - - } else if (canvas_light_shadow_owner.owns(p_rid)) { - CanvasLightShadow *cls = canvas_light_shadow_owner.get_or_null(p_rid); - glDeleteFramebuffers(1, &cls->fbo); - glDeleteRenderbuffers(1, &cls->depth); - glDeleteTextures(1, &cls->distance); - canvas_light_shadow_owner.free(p_rid); - memdelete(cls); - + } else if (GLES3::ParticlesStorage::get_singleton()->owns_particles_collision_instance(p_rid)) { + GLES3::ParticlesStorage::get_singleton()->particles_collision_instance_free(p_rid); return true; + } else { + return false; } - */ } /* DEPENDENCIES */ @@ -183,6 +176,12 @@ void Utilities::base_update_dependency(RID p_base, DependencyTracker *p_instance } else if (LightStorage::get_singleton()->owns_light(p_base)) { Light *l = LightStorage::get_singleton()->get_light(p_base); p_instance->update_dependency(&l->dependency); + } else if (ParticlesStorage::get_singleton()->owns_particles(p_base)) { + Dependency *dependency = ParticlesStorage::get_singleton()->particles_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (ParticlesStorage::get_singleton()->owns_particles_collision(p_base)) { + Dependency *dependency = ParticlesStorage::get_singleton()->particles_collision_get_dependency(p_base); + p_instance->update_dependency(dependency); } } @@ -213,95 +212,78 @@ void Utilities::visibility_notifier_call(RID p_notifier, bool p_enter, bool p_de /* TIMING */ -//void Utilities::render_info_begin_capture() { -// info.snap = info.render; -//} - -//void Utilities::render_info_end_capture() { -// info.snap.object_count = info.render.object_count - info.snap.object_count; -// info.snap.draw_call_count = info.render.draw_call_count - info.snap.draw_call_count; -// info.snap.material_switch_count = info.render.material_switch_count - info.snap.material_switch_count; -// info.snap.surface_switch_count = info.render.surface_switch_count - info.snap.surface_switch_count; -// info.snap.shader_rebind_count = info.render.shader_rebind_count - info.snap.shader_rebind_count; -// info.snap.vertices_count = info.render.vertices_count - info.snap.vertices_count; -// info.snap._2d_item_count = info.render._2d_item_count - info.snap._2d_item_count; -// info.snap._2d_draw_call_count = info.render._2d_draw_call_count - info.snap._2d_draw_call_count; -//} - -//int Utilities::get_captured_render_info(RS::RenderInfo p_info) { -// switch (p_info) { -// case RS::INFO_OBJECTS_IN_FRAME: { -// return info.snap.object_count; -// } break; -// case RS::INFO_VERTICES_IN_FRAME: { -// return info.snap.vertices_count; -// } break; -// case RS::INFO_MATERIAL_CHANGES_IN_FRAME: { -// return info.snap.material_switch_count; -// } break; -// case RS::INFO_SHADER_CHANGES_IN_FRAME: { -// return info.snap.shader_rebind_count; -// } break; -// case RS::INFO_SURFACE_CHANGES_IN_FRAME: { -// return info.snap.surface_switch_count; -// } break; -// case RS::INFO_DRAW_CALLS_IN_FRAME: { -// return info.snap.draw_call_count; -// } break; -// /* -// case RS::INFO_2D_ITEMS_IN_FRAME: { -// return info.snap._2d_item_count; -// } break; -// case RS::INFO_2D_DRAW_CALLS_IN_FRAME: { -// return info.snap._2d_draw_call_count; -// } break; -// */ -// default: { -// return get_render_info(p_info); -// } -// } -//} - -//int Utilities::get_render_info(RS::RenderInfo p_info) { -// switch (p_info) { -// case RS::INFO_OBJECTS_IN_FRAME: -// return info.render_final.object_count; -// case RS::INFO_VERTICES_IN_FRAME: -// return info.render_final.vertices_count; -// case RS::INFO_MATERIAL_CHANGES_IN_FRAME: -// return info.render_final.material_switch_count; -// case RS::INFO_SHADER_CHANGES_IN_FRAME: -// return info.render_final.shader_rebind_count; -// case RS::INFO_SURFACE_CHANGES_IN_FRAME: -// return info.render_final.surface_switch_count; -// case RS::INFO_DRAW_CALLS_IN_FRAME: -// return info.render_final.draw_call_count; -// /* -// case RS::INFO_2D_ITEMS_IN_FRAME: -// return info.render_final._2d_item_count; -// case RS::INFO_2D_DRAW_CALLS_IN_FRAME: -// return info.render_final._2d_draw_call_count; -//*/ -// case RS::INFO_USAGE_VIDEO_MEM_TOTAL: -// return 0; //no idea -// case RS::INFO_VIDEO_MEM_USED: -// return info.vertex_mem + info.texture_mem; -// case RS::INFO_TEXTURE_MEM_USED: -// return info.texture_mem; -// case RS::INFO_VERTEX_MEM_USED: -// return info.vertex_mem; -// default: -// return 0; //no idea either -// } -//} +void Utilities::capture_timestamps_begin() { + capture_timestamp("Frame Begin"); +} + +void Utilities::capture_timestamp(const String &p_name) { + ERR_FAIL_COND(frames[frame].timestamp_count >= max_timestamp_query_elements); + +#ifdef GLES_OVER_GL + glQueryCounter(frames[frame].queries[frames[frame].timestamp_count], GL_TIMESTAMP); +#endif + + frames[frame].timestamp_names[frames[frame].timestamp_count] = p_name; + frames[frame].timestamp_cpu_values[frames[frame].timestamp_count] = OS::get_singleton()->get_ticks_usec(); + frames[frame].timestamp_count++; +} + +void Utilities::_capture_timestamps_begin() { + // frame is incremented at the end of the frame so this gives us the queries for frame - 2. By then they should be ready. + if (frames[frame].timestamp_count) { +#ifdef GLES_OVER_GL + for (uint32_t i = 0; i < frames[frame].timestamp_count; i++) { + uint64_t temp = 0; + glGetQueryObjectui64v(frames[frame].queries[i], GL_QUERY_RESULT, &temp); + frames[frame].timestamp_result_values[i] = temp; + } +#endif + SWAP(frames[frame].timestamp_names, frames[frame].timestamp_result_names); + SWAP(frames[frame].timestamp_cpu_values, frames[frame].timestamp_cpu_result_values); + } + + frames[frame].timestamp_result_count = frames[frame].timestamp_count; + frames[frame].timestamp_count = 0; + frames[frame].index = Engine::get_singleton()->get_frames_drawn(); + capture_timestamp("Internal Begin"); +} + +void Utilities::capture_timestamps_end() { + capture_timestamp("Internal End"); + frame = (frame + 1) % FRAME_COUNT; +} + +uint32_t Utilities::get_captured_timestamps_count() const { + return frames[frame].timestamp_result_count; +} + +uint64_t Utilities::get_captured_timestamps_frame() const { + return frames[frame].index; +} + +uint64_t Utilities::get_captured_timestamp_gpu_time(uint32_t p_index) const { + ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, 0); + return frames[frame].timestamp_result_values[p_index]; +} + +uint64_t Utilities::get_captured_timestamp_cpu_time(uint32_t p_index) const { + ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, 0); + return frames[frame].timestamp_cpu_result_values[p_index]; +} + +String Utilities::get_captured_timestamp_name(uint32_t p_index) const { + ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, String()); + return frames[frame].timestamp_result_names[p_index]; +} /* MISC */ void Utilities::update_dirty_resources() { MaterialStorage::get_singleton()->_update_global_shader_uniforms(); MaterialStorage::get_singleton()->_update_queued_materials(); - //MeshStorage::get_singleton()->_update_dirty_skeletons(); + MeshStorage::get_singleton()->_update_dirty_skeletons(); MeshStorage::get_singleton()->_update_dirty_multimeshes(); + TextureStorage::get_singleton()->update_texture_atlas(); } void Utilities::set_debug_generate_wireframes(bool p_generate) { @@ -355,4 +337,13 @@ String Utilities::get_video_adapter_api_version() const { return (const char *)glGetString(GL_VERSION); } +Size2i Utilities::get_maximum_viewport_size() const { + Config *config = Config::get_singleton(); + if (!config) { + return Size2i(); + } + + return Size2i(config->max_viewport_size[0], config->max_viewport_size[1]); +} + #endif // GLES3_ENABLED diff --git a/drivers/gles3/storage/utilities.h b/drivers/gles3/storage/utilities.h index e054f2f816..55a875958e 100644 --- a/drivers/gles3/storage/utilities.h +++ b/drivers/gles3/storage/utilities.h @@ -79,62 +79,35 @@ public: /* TIMING */ - struct Info { - uint64_t texture_mem = 0; - uint64_t vertex_mem = 0; - - struct Render { - uint32_t object_count; - uint32_t draw_call_count; - uint32_t material_switch_count; - uint32_t surface_switch_count; - uint32_t shader_rebind_count; - uint32_t vertices_count; - uint32_t _2d_item_count; - uint32_t _2d_draw_call_count; - - void reset() { - object_count = 0; - draw_call_count = 0; - material_switch_count = 0; - surface_switch_count = 0; - shader_rebind_count = 0; - vertices_count = 0; - _2d_item_count = 0; - _2d_draw_call_count = 0; - } - } render, render_final, snap; - - Info() { - render.reset(); - render_final.reset(); - } - - } info; - - virtual void capture_timestamps_begin() override {} - virtual void capture_timestamp(const String &p_name) override {} - virtual uint32_t get_captured_timestamps_count() const override { - return 0; - } - virtual uint64_t get_captured_timestamps_frame() const override { - return 0; - } - virtual uint64_t get_captured_timestamp_gpu_time(uint32_t p_index) const override { - return 0; - } - virtual uint64_t get_captured_timestamp_cpu_time(uint32_t p_index) const override { - return 0; - } - virtual String get_captured_timestamp_name(uint32_t p_index) const override { - return String(); - } - - // void render_info_begin_capture() override; - // void render_info_end_capture() override; - // int get_captured_render_info(RS::RenderInfo p_info) override; - - // int get_render_info(RS::RenderInfo p_info) override; +#define MAX_QUERIES 256 +#define FRAME_COUNT 3 + + struct Frame { + GLuint queries[MAX_QUERIES]; + TightLocalVector<String> timestamp_names; + TightLocalVector<uint64_t> timestamp_cpu_values; + uint32_t timestamp_count = 0; + TightLocalVector<String> timestamp_result_names; + TightLocalVector<uint64_t> timestamp_cpu_result_values; + TightLocalVector<uint64_t> timestamp_result_values; + uint32_t timestamp_result_count = 0; + uint64_t index = 0; + }; + + const uint32_t max_timestamp_query_elements = MAX_QUERIES; + + Frame frames[FRAME_COUNT]; // Frames for capturing timestamps. We use 3 so we don't need to wait for commands to complete + uint32_t frame = 0; + + virtual void capture_timestamps_begin() override; + virtual void capture_timestamp(const String &p_name) override; + virtual uint32_t get_captured_timestamps_count() const override; + virtual uint64_t get_captured_timestamps_frame() const override; + virtual uint64_t get_captured_timestamp_gpu_time(uint32_t p_index) const override; + virtual uint64_t get_captured_timestamp_cpu_time(uint32_t p_index) const override; + virtual String get_captured_timestamp_name(uint32_t p_index) const override; + void _capture_timestamps_begin(); + void capture_timestamps_end(); /* MISC */ @@ -150,6 +123,8 @@ public: virtual String get_video_adapter_vendor() const override; virtual RenderingDevice::DeviceType get_video_adapter_type() const override; virtual String get_video_adapter_api_version() const override; + + virtual Size2i get_maximum_viewport_size() const override; }; } // namespace GLES3 |