diff options
Diffstat (limited to 'drivers/vulkan')
| -rw-r--r-- | drivers/vulkan/rendering_device_vulkan.cpp | 1943 | ||||
| -rw-r--r-- | drivers/vulkan/rendering_device_vulkan.h | 209 | ||||
| -rw-r--r-- | drivers/vulkan/vulkan_context.cpp | 977 | ||||
| -rw-r--r-- | drivers/vulkan/vulkan_context.h | 113 | ||||
| -rw-r--r-- | drivers/vulkan/vulkan_hooks.h | 58 |
5 files changed, 1594 insertions, 1706 deletions
diff --git a/drivers/vulkan/rendering_device_vulkan.cpp b/drivers/vulkan/rendering_device_vulkan.cpp index 0979ae9e16..6c833c8a26 100644 --- a/drivers/vulkan/rendering_device_vulkan.cpp +++ b/drivers/vulkan/rendering_device_vulkan.cpp @@ -1,32 +1,32 @@ -/*************************************************************************/ -/* rendering_device_vulkan.cpp */ -/*************************************************************************/ -/* This file is part of: */ -/* GODOT ENGINE */ -/* https://godotengine.org */ -/*************************************************************************/ -/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ -/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ -/* */ -/* Permission is hereby granted, free of charge, to any person obtaining */ -/* a copy of this software and associated documentation files (the */ -/* "Software"), to deal in the Software without restriction, including */ -/* without limitation the rights to use, copy, modify, merge, publish, */ -/* distribute, sublicense, and/or sell copies of the Software, and to */ -/* permit persons to whom the Software is furnished to do so, subject to */ -/* the following conditions: */ -/* */ -/* The above copyright notice and this permission notice shall be */ -/* included in all copies or substantial portions of the Software. */ -/* */ -/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ -/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ -/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ -/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ -/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ -/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ -/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/*************************************************************************/ +/**************************************************************************/ +/* rendering_device_vulkan.cpp */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ #include "rendering_device_vulkan.h" @@ -36,18 +36,16 @@ #include "core/io/marshalls.h" #include "core/os/os.h" #include "core/templates/hashfuncs.h" -#include "core/version.h" #include "drivers/vulkan/vulkan_context.h" #include "thirdparty/misc/smolv.h" -#include "thirdparty/spirv-reflect/spirv_reflect.h" //#define FORCE_FULL_BARRIER static const uint32_t SMALL_ALLOCATION_MAX_SIZE = 4096; -// Get the Vulkan object information and possible stage access types (bitwise OR'd with incoming values) -RenderingDeviceVulkan::Buffer *RenderingDeviceVulkan::_get_buffer_from_owner(RID p_buffer, VkPipelineStageFlags &r_stage_mask, VkAccessFlags &r_access_mask, uint32_t p_post_barrier) { +// Get the Vulkan object information and possible stage access types (bitwise OR'd with incoming values). +RenderingDeviceVulkan::Buffer *RenderingDeviceVulkan::_get_buffer_from_owner(RID p_buffer, VkPipelineStageFlags &r_stage_mask, VkAccessFlags &r_access_mask, BitField<BarrierMask> p_post_barrier) { Buffer *buffer = nullptr; if (vertex_buffer_owner.owns(p_buffer)) { buffer = vertex_buffer_owner.get_or_null(p_buffer); @@ -55,11 +53,11 @@ RenderingDeviceVulkan::Buffer *RenderingDeviceVulkan::_get_buffer_from_owner(RID r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT; r_access_mask |= VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT; if (buffer->usage & VK_BUFFER_USAGE_STORAGE_BUFFER_BIT) { - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; } - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; } @@ -69,20 +67,20 @@ RenderingDeviceVulkan::Buffer *RenderingDeviceVulkan::_get_buffer_from_owner(RID r_access_mask |= VK_ACCESS_INDEX_READ_BIT; buffer = index_buffer_owner.get_or_null(p_buffer); } else if (uniform_buffer_owner.owns(p_buffer)) { - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; } - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; } r_access_mask |= VK_ACCESS_UNIFORM_READ_BIT; buffer = uniform_buffer_owner.get_or_null(p_buffer); } else if (texture_buffer_owner.owns(p_buffer)) { - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; r_access_mask |= VK_ACCESS_SHADER_READ_BIT; } - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; r_access_mask |= VK_ACCESS_SHADER_READ_BIT; } @@ -90,11 +88,11 @@ RenderingDeviceVulkan::Buffer *RenderingDeviceVulkan::_get_buffer_from_owner(RID buffer = &texture_buffer_owner.get_or_null(p_buffer)->buffer; } else if (storage_buffer_owner.owns(p_buffer)) { buffer = storage_buffer_owner.get_or_null(p_buffer); - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } @@ -108,8 +106,8 @@ RenderingDeviceVulkan::Buffer *RenderingDeviceVulkan::_get_buffer_from_owner(RID } static void update_external_dependency_for_store(VkSubpassDependency2KHR &dependency, bool is_sampled, bool is_storage, bool is_depth) { - // Transitioning from write to read, protect the shaders that may use this next - // Allow for copies/image layout transitions + // Transitioning from write to read, protect the shaders that may use this next. + // Allow for copies/image layout transitions. dependency.dstStageMask |= VK_PIPELINE_STAGE_TRANSFER_BIT; dependency.dstAccessMask |= VK_ACCESS_TRANSFER_READ_BIT; @@ -125,7 +123,7 @@ static void update_external_dependency_for_store(VkSubpassDependency2KHR &depend } if (is_depth) { - // Depth resources have additional stages that may be interested in them + // Depth resources have additional stages that may be interested in them. dependency.dstStageMask |= VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT; dependency.dstAccessMask |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT; } @@ -146,7 +144,7 @@ void RenderingDeviceVulkan::_add_dependency(RID p_id, RID p_depends_on) { } void RenderingDeviceVulkan::_free_dependencies(RID p_id) { - //direct dependencies must be freed + // Direct dependencies must be freed. HashMap<RID, HashSet<RID>>::Iterator E = dependency_map.find(p_id); if (E) { @@ -156,7 +154,7 @@ void RenderingDeviceVulkan::_free_dependencies(RID p_id) { dependency_map.remove(E); } - //reverse dependencies must be unreferenced + // Reverse dependencies must be unreferenced. E = reverse_dependency_map.find(p_id); if (E) { @@ -860,7 +858,7 @@ uint32_t RenderingDeviceVulkan::get_image_format_pixel_size(DataFormat p_format) case DATA_FORMAT_D24_UNORM_S8_UINT: return 4; case DATA_FORMAT_D32_SFLOAT_S8_UINT: - return 5; //? + return 5; // ? case DATA_FORMAT_BC1_RGB_UNORM_BLOCK: case DATA_FORMAT_BC1_RGB_SRGB_BLOCK: case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK: @@ -995,9 +993,12 @@ void RenderingDeviceVulkan::get_compressed_image_format_block_dimensions(DataFor case DATA_FORMAT_EAC_R11_SNORM_BLOCK: case DATA_FORMAT_EAC_R11G11_UNORM_BLOCK: case DATA_FORMAT_EAC_R11G11_SNORM_BLOCK: - case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK: //again, not sure about astc - case DATA_FORMAT_ASTC_4x4_SRGB_BLOCK: - case DATA_FORMAT_ASTC_5x4_UNORM_BLOCK: + case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK: // Again, not sure about astc. + case DATA_FORMAT_ASTC_4x4_SRGB_BLOCK: { + r_w = 4; + r_h = 4; + } break; + case DATA_FORMAT_ASTC_5x4_UNORM_BLOCK: // Unsupported case DATA_FORMAT_ASTC_5x4_SRGB_BLOCK: case DATA_FORMAT_ASTC_5x5_UNORM_BLOCK: case DATA_FORMAT_ASTC_5x5_SRGB_BLOCK: @@ -1008,10 +1009,16 @@ void RenderingDeviceVulkan::get_compressed_image_format_block_dimensions(DataFor case DATA_FORMAT_ASTC_8x5_UNORM_BLOCK: case DATA_FORMAT_ASTC_8x5_SRGB_BLOCK: case DATA_FORMAT_ASTC_8x6_UNORM_BLOCK: - case DATA_FORMAT_ASTC_8x6_SRGB_BLOCK: + case DATA_FORMAT_ASTC_8x6_SRGB_BLOCK: { + r_w = 4; + r_h = 4; + } break; case DATA_FORMAT_ASTC_8x8_UNORM_BLOCK: - case DATA_FORMAT_ASTC_8x8_SRGB_BLOCK: - case DATA_FORMAT_ASTC_10x5_UNORM_BLOCK: + case DATA_FORMAT_ASTC_8x8_SRGB_BLOCK: { + r_w = 8; + r_h = 8; + } break; + case DATA_FORMAT_ASTC_10x5_UNORM_BLOCK: // Unsupported case DATA_FORMAT_ASTC_10x5_SRGB_BLOCK: case DATA_FORMAT_ASTC_10x6_UNORM_BLOCK: case DATA_FORMAT_ASTC_10x6_SRGB_BLOCK: @@ -1073,7 +1080,7 @@ uint32_t RenderingDeviceVulkan::get_compressed_image_format_block_byte_size(Data case DATA_FORMAT_EAC_R11G11_UNORM_BLOCK: case DATA_FORMAT_EAC_R11G11_SNORM_BLOCK: return 16; - case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK: //again, not sure about astc + case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK: // Again, not sure about astc. case DATA_FORMAT_ASTC_4x4_SRGB_BLOCK: case DATA_FORMAT_ASTC_5x4_UNORM_BLOCK: case DATA_FORMAT_ASTC_5x4_SRGB_BLOCK: @@ -1101,7 +1108,7 @@ uint32_t RenderingDeviceVulkan::get_compressed_image_format_block_byte_size(Data case DATA_FORMAT_ASTC_12x10_SRGB_BLOCK: case DATA_FORMAT_ASTC_12x12_UNORM_BLOCK: case DATA_FORMAT_ASTC_12x12_SRGB_BLOCK: - return 8; //wrong + return 16; default: { } } @@ -1110,7 +1117,7 @@ uint32_t RenderingDeviceVulkan::get_compressed_image_format_block_byte_size(Data uint32_t RenderingDeviceVulkan::get_compressed_image_format_pixel_rshift(DataFormat p_format) { switch (p_format) { - case DATA_FORMAT_BC1_RGB_UNORM_BLOCK: //these formats are half byte size, so rshift is 1 + case DATA_FORMAT_BC1_RGB_UNORM_BLOCK: // These formats are half byte size, so rshift is 1. case DATA_FORMAT_BC1_RGB_SRGB_BLOCK: case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK: case DATA_FORMAT_BC1_RGBA_SRGB_BLOCK: @@ -1123,6 +1130,10 @@ uint32_t RenderingDeviceVulkan::get_compressed_image_format_pixel_rshift(DataFor case DATA_FORMAT_EAC_R11_UNORM_BLOCK: case DATA_FORMAT_EAC_R11_SNORM_BLOCK: return 1; + case DATA_FORMAT_ASTC_8x8_SRGB_BLOCK: + case DATA_FORMAT_ASTC_8x8_UNORM_BLOCK: { + return 2; + } default: { } } @@ -1184,7 +1195,7 @@ uint32_t RenderingDeviceVulkan::get_image_format_required_size(DataFormat p_form } uint32_t RenderingDeviceVulkan::get_image_required_mipmaps(uint32_t p_width, uint32_t p_height, uint32_t p_depth) { - //formats and block size don't really matter here since they can all go down to 1px (even if block is larger) + // Formats and block size don't really matter here since they can all go down to 1px (even if block is larger). uint32_t w = p_width; uint32_t h = p_height; uint32_t d = p_depth; @@ -1402,16 +1413,16 @@ Error RenderingDeviceVulkan::_insert_staging_block() { } Error RenderingDeviceVulkan::_staging_buffer_allocate(uint32_t p_amount, uint32_t p_required_align, uint32_t &r_alloc_offset, uint32_t &r_alloc_size, bool p_can_segment) { - //determine a block to use + // Determine a block to use. r_alloc_size = p_amount; while (true) { r_alloc_offset = 0; - //see if we can use current block + // See if we can use current block. if (staging_buffer_blocks[staging_buffer_current].frame_used == frames_drawn) { - //we used this block this frame, let's see if there is still room + // We used this block this frame, let's see if there is still room. uint32_t write_from = staging_buffer_blocks[staging_buffer_current].fill_amount; @@ -1425,107 +1436,107 @@ Error RenderingDeviceVulkan::_staging_buffer_allocate(uint32_t p_amount, uint32_ int32_t available_bytes = int32_t(staging_buffer_block_size) - int32_t(write_from); if ((int32_t)p_amount < available_bytes) { - //all is good, we should be ok, all will fit + // All is good, we should be ok, all will fit. r_alloc_offset = write_from; } else if (p_can_segment && available_bytes >= (int32_t)p_required_align) { - //ok all won't fit but at least we can fit a chunkie - //all is good, update what needs to be written to + // Ok all won't fit but at least we can fit a chunkie. + // All is good, update what needs to be written to. r_alloc_offset = write_from; r_alloc_size = available_bytes - (available_bytes % p_required_align); } else { - //can't fit it into this buffer. - //will need to try next buffer + // Can't fit it into this buffer. + // Will need to try next buffer. staging_buffer_current = (staging_buffer_current + 1) % staging_buffer_blocks.size(); - // before doing anything, though, let's check that we didn't manage to fill all blocks - // possible in a single frame + // Before doing anything, though, let's check that we didn't manage to fill all blocks. + // Possible in a single frame. if (staging_buffer_blocks[staging_buffer_current].frame_used == frames_drawn) { - //guess we did.. ok, let's see if we can insert a new block.. + // Guess we did.. ok, let's see if we can insert a new block. if ((uint64_t)staging_buffer_blocks.size() * staging_buffer_block_size < staging_buffer_max_size) { - //we can, so we are safe + // We can, so we are safe. Error err = _insert_staging_block(); if (err) { return err; } - //claim for this frame + // Claim for this frame. staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn; } else { // Ok, worst case scenario, all the staging buffers belong to this frame // and this frame is not even done. - // If this is the main thread, it means the user is likely loading a lot of resources at once, - // otherwise, the thread should just be blocked until the next frame (currently unimplemented) + // If this is the main thread, it means the user is likely loading a lot of resources at once,. + // Otherwise, the thread should just be blocked until the next frame (currently unimplemented). - if (false) { //separate thread from render + if (false) { // Separate thread from render. //block_until_next_frame() continue; } else { - //flush EVERYTHING including setup commands. IF not immediate, also need to flush the draw commands + // Flush EVERYTHING including setup commands. IF not immediate, also need to flush the draw commands. _flush(true); - //clear the whole staging buffer + // Clear the whole staging buffer. for (int i = 0; i < staging_buffer_blocks.size(); i++) { staging_buffer_blocks.write[i].frame_used = 0; staging_buffer_blocks.write[i].fill_amount = 0; } - //claim current + // Claim current. staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn; } } } else { - //not from current frame, so continue and try again + // Not from current frame, so continue and try again. continue; } } } else if (staging_buffer_blocks[staging_buffer_current].frame_used <= frames_drawn - frame_count) { - //this is an old block, which was already processed, let's reuse + // This is an old block, which was already processed, let's reuse. staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn; staging_buffer_blocks.write[staging_buffer_current].fill_amount = 0; } else { - //this block may still be in use, let's not touch it unless we have to, so.. can we create a new one? + // This block may still be in use, let's not touch it unless we have to, so.. can we create a new one? if ((uint64_t)staging_buffer_blocks.size() * staging_buffer_block_size < staging_buffer_max_size) { - //we are still allowed to create a new block, so let's do that and insert it for current pos + // We are still allowed to create a new block, so let's do that and insert it for current pos. Error err = _insert_staging_block(); if (err) { return err; } - //claim for this frame + // Claim for this frame. staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn; } else { - // oops, we are out of room and we can't create more. - // let's flush older frames. + // Oops, we are out of room and we can't create more. + // Let's flush older frames. // The logic here is that if a game is loading a lot of data from the main thread, it will need to be stalled anyway. // If loading from a separate thread, we can block that thread until next frame when more room is made (not currently implemented, though). if (false) { - //separate thread from render + // Separate thread from render. //block_until_next_frame() - continue; //and try again + continue; // And try again. } else { _flush(false); for (int i = 0; i < staging_buffer_blocks.size(); i++) { - //clear all blocks but the ones from this frame + // Clear all blocks but the ones from this frame. int block_idx = (i + staging_buffer_current) % staging_buffer_blocks.size(); if (staging_buffer_blocks[block_idx].frame_used == frames_drawn) { - break; //ok, we reached something from this frame, abort + break; // Ok, we reached something from this frame, abort. } staging_buffer_blocks.write[block_idx].frame_used = 0; staging_buffer_blocks.write[block_idx].fill_amount = 0; } - //claim for current frame + // Claim for current frame. staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn; } } } - //all was good, break + // All was good, break. break; } @@ -1535,7 +1546,7 @@ Error RenderingDeviceVulkan::_staging_buffer_allocate(uint32_t p_amount, uint32_ } Error RenderingDeviceVulkan::_buffer_update(Buffer *p_buffer, size_t p_offset, const uint8_t *p_data, size_t p_data_size, bool p_use_draw_command_buffer, uint32_t p_required_align) { - //submitting may get chunked for various reasons, so convert this to a task + // Submitting may get chunked for various reasons, so convert this to a task. size_t to_submit = p_data_size; size_t submit_from = 0; @@ -1548,7 +1559,7 @@ Error RenderingDeviceVulkan::_buffer_update(Buffer *p_buffer, size_t p_offset, c return err; } - //map staging buffer (It's CPU and coherent) + // Map staging buffer (It's CPU and coherent). void *data_ptr = nullptr; { @@ -1556,12 +1567,12 @@ Error RenderingDeviceVulkan::_buffer_update(Buffer *p_buffer, size_t p_offset, c ERR_FAIL_COND_V_MSG(vkerr, ERR_CANT_CREATE, "vmaMapMemory failed with error " + itos(vkerr) + "."); } - //copy to staging buffer + // Copy to staging buffer. memcpy(((uint8_t *)data_ptr) + block_write_offset, p_data + submit_from, block_write_amount); - //unmap + // Unmap. vmaUnmapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation); - //insert a command to copy this + // Insert a command to copy this. VkBufferCopy region; region.srcOffset = block_write_offset; @@ -1579,21 +1590,21 @@ Error RenderingDeviceVulkan::_buffer_update(Buffer *p_buffer, size_t p_offset, c return OK; } -void RenderingDeviceVulkan::_memory_barrier(VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_sccess, bool p_sync_with_draw) { +void RenderingDeviceVulkan::_memory_barrier(VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_access, bool p_sync_with_draw) { VkMemoryBarrier mem_barrier; mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER; mem_barrier.pNext = nullptr; mem_barrier.srcAccessMask = p_src_access; - mem_barrier.dstAccessMask = p_dst_sccess; + mem_barrier.dstAccessMask = p_dst_access; if (p_src_stage_mask == 0 || p_dst_stage_mask == 0) { - return; //no barrier, since this is invalid + return; // No barrier, since this is invalid. } vkCmdPipelineBarrier(p_sync_with_draw ? frames[frame].draw_command_buffer : frames[frame].setup_command_buffer, p_src_stage_mask, p_dst_stage_mask, 0, 1, &mem_barrier, 0, nullptr, 0, nullptr); } void RenderingDeviceVulkan::_full_barrier(bool p_sync_with_draw) { - //used for debug + // Used for debug. _memory_barrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_ACCESS_INDIRECT_COMMAND_READ_BIT | VK_ACCESS_INDEX_READ_BIT | @@ -1628,14 +1639,14 @@ void RenderingDeviceVulkan::_full_barrier(bool p_sync_with_draw) { p_sync_with_draw); } -void RenderingDeviceVulkan::_buffer_memory_barrier(VkBuffer buffer, uint64_t p_from, uint64_t p_size, VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_sccess, bool p_sync_with_draw) { +void RenderingDeviceVulkan::_buffer_memory_barrier(VkBuffer buffer, uint64_t p_from, uint64_t p_size, VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_access, bool p_sync_with_draw) { VkBufferMemoryBarrier buffer_mem_barrier; buffer_mem_barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; buffer_mem_barrier.pNext = nullptr; buffer_mem_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; buffer_mem_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; buffer_mem_barrier.srcAccessMask = p_src_access; - buffer_mem_barrier.dstAccessMask = p_dst_sccess; + buffer_mem_barrier.dstAccessMask = p_dst_access; buffer_mem_barrier.buffer = buffer; buffer_mem_barrier.offset = p_from; buffer_mem_barrier.size = p_size; @@ -1655,15 +1666,13 @@ RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const T image_create_info.pNext = nullptr; image_create_info.flags = 0; -#ifndef _MSC_VER -#warning TODO check for support via RenderingDevice to enable on mobile when possible -#endif + // TODO: Check for support via RenderingDevice to enable on mobile when possible. #ifndef ANDROID_ENABLED // vkCreateImage fails with format list on Android (VK_ERROR_OUT_OF_HOST_MEMORY) - VkImageFormatListCreateInfoKHR format_list_create_info; //keep out of the if, needed for creation - Vector<VkFormat> allowed_formats; //keep out of the if, needed for creation + VkImageFormatListCreateInfoKHR format_list_create_info; // Keep out of the if, needed for creation. + Vector<VkFormat> allowed_formats; // Keep out of the if, needed for creation. #endif if (p_format.shareable_formats.size()) { image_create_info.flags |= VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT; @@ -1733,10 +1742,10 @@ RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const T ERR_FAIL_INDEX_V(p_format.samples, TEXTURE_SAMPLES_MAX, RID()); - image_create_info.samples = rasterization_sample_count[p_format.samples]; + image_create_info.samples = _ensure_supported_sample_count(p_format.samples); image_create_info.tiling = (p_format.usage_bits & TEXTURE_USAGE_CPU_READ_BIT) ? VK_IMAGE_TILING_LINEAR : VK_IMAGE_TILING_OPTIMAL; - //usage + // Usage. image_create_info.usage = 0; if (p_format.usage_bits & TEXTURE_USAGE_SAMPLING_BIT) { @@ -1800,7 +1809,7 @@ RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const T } { - //validate that this image is supported for the intended use + // Validate that this image is supported for the intended use. VkFormatProperties properties; vkGetPhysicalDeviceFormatProperties(context->get_physical_device(), image_create_info.format, &properties); VkFormatFeatureFlags flags; @@ -1841,7 +1850,7 @@ RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const T } } - //some view validation + // Some view validation. if (p_view.format_override != DATA_FORMAT_MAX) { ERR_FAIL_INDEX_V(p_view.format_override, DATA_FORMAT_MAX, RID()); @@ -1851,7 +1860,7 @@ RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const T ERR_FAIL_INDEX_V(p_view.swizzle_b, TEXTURE_SWIZZLE_MAX, RID()); ERR_FAIL_INDEX_V(p_view.swizzle_a, TEXTURE_SWIZZLE_MAX, RID()); - //allocate memory + // Allocate memory. uint32_t width, height; uint32_t image_size = get_image_format_required_size(p_format.format, p_format.width, p_format.height, p_format.depth, p_format.mipmaps, &width, &height); @@ -1884,23 +1893,24 @@ RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const T texture.mipmaps = image_create_info.mipLevels; texture.base_mipmap = 0; texture.base_layer = 0; + texture.is_resolve_buffer = p_format.is_resolve_buffer; texture.usage_flags = p_format.usage_bits; texture.samples = p_format.samples; texture.allowed_shared_formats = p_format.shareable_formats; - //set base layout based on usage priority + // Set base layout based on usage priority. if (p_format.usage_bits & TEXTURE_USAGE_SAMPLING_BIT) { - //first priority, readable + // First priority, readable. texture.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; } else if (p_format.usage_bits & TEXTURE_USAGE_STORAGE_BIT) { - //second priority, storage + // Second priority, storage. texture.layout = VK_IMAGE_LAYOUT_GENERAL; } else if (p_format.usage_bits & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) { - //third priority, color or depth + // Third priority, color or depth. texture.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; @@ -1925,7 +1935,7 @@ RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const T texture.bound = false; - //create view + // Create view. VkImageViewCreateInfo image_view_create_info; image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; @@ -1982,7 +1992,7 @@ RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const T ERR_FAIL_V_MSG(RID(), "vkCreateImageView failed with error " + itos(err) + "."); } - //barrier to set layout + // Barrier to set layout. { VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; @@ -2004,10 +2014,13 @@ RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const T } RID id = texture_owner.make_rid(texture); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif if (p_data.size()) { for (uint32_t i = 0; i < image_create_info.arrayLayers; i++) { - _texture_update(id, i, p_data[i], RD::BARRIER_MASK_ALL, true); + _texture_update(id, i, p_data[i], RD::BARRIER_MASK_ALL_BARRIERS, true); } } return id; @@ -2019,13 +2032,13 @@ RID RenderingDeviceVulkan::texture_create_shared(const TextureView &p_view, RID Texture *src_texture = texture_owner.get_or_null(p_with_texture); ERR_FAIL_COND_V(!src_texture, RID()); - if (src_texture->owner.is_valid()) { //ahh this is a share + if (src_texture->owner.is_valid()) { // Ahh this is a share. p_with_texture = src_texture->owner; src_texture = texture_owner.get_or_null(src_texture->owner); - ERR_FAIL_COND_V(!src_texture, RID()); //this is a bug + ERR_FAIL_COND_V(!src_texture, RID()); // This is a bug. } - //create view + // Create view. Texture texture = *src_texture; @@ -2086,7 +2099,7 @@ RID RenderingDeviceVulkan::texture_create_shared(const TextureView &p_view, RID usage_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_USAGE_CREATE_INFO; usage_info.pNext = nullptr; if (p_view.format_override != DATA_FORMAT_MAX) { - //need to validate usage with vulkan + // Need to validate usage with vulkan. usage_info.usage = 0; @@ -2133,6 +2146,9 @@ RID RenderingDeviceVulkan::texture_create_shared(const TextureView &p_view, RID texture.owner = p_with_texture; RID id = texture_owner.make_rid(texture); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif _add_dependency(id, p_with_texture); return id; @@ -2145,9 +2161,9 @@ RID RenderingDeviceVulkan::texture_create_from_extension(TextureType p_type, Dat Texture texture; texture.image = image; - // if we leave texture.allocation as a nullptr, would that be enough to detect we don't "own" the image? - // also leave texture.allocation_info alone - // we'll set texture.view later on + // If we leave texture.allocation as a nullptr, would that be enough to detect we don't "own" the image? + // Also leave texture.allocation_info alone. + // We'll set texture.view later on. texture.type = p_type; texture.format = p_format; texture.samples = p_samples; @@ -2155,14 +2171,35 @@ RID RenderingDeviceVulkan::texture_create_from_extension(TextureType p_type, Dat texture.height = p_height; texture.depth = p_depth; texture.layers = p_layers; - texture.mipmaps = 0; // maybe make this settable too? + texture.mipmaps = 1; texture.usage_flags = p_flags; texture.base_mipmap = 0; texture.base_layer = 0; texture.allowed_shared_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_UNORM); texture.allowed_shared_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_SRGB); - // Do we need to do something with texture.layout ? + // Set base layout based on usage priority. + + if (texture.usage_flags & TEXTURE_USAGE_SAMPLING_BIT) { + // First priority, readable. + texture.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; + + } else if (texture.usage_flags & TEXTURE_USAGE_STORAGE_BIT) { + // Second priority, storage. + + texture.layout = VK_IMAGE_LAYOUT_GENERAL; + + } else if (texture.usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) { + // Third priority, color or depth. + + texture.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; + + } else if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) { + texture.layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; + + } else { + texture.layout = VK_IMAGE_LAYOUT_GENERAL; + } if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) { texture.read_aspect_mask = VK_IMAGE_ASPECT_DEPTH_BIT; @@ -2176,7 +2213,7 @@ RID RenderingDeviceVulkan::texture_create_from_extension(TextureType p_type, Dat texture.barrier_aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT; } - // Create a view for us to use + // Create a view for us to use. VkImageViewCreateInfo image_view_create_info; image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; @@ -2207,7 +2244,7 @@ RID RenderingDeviceVulkan::texture_create_from_extension(TextureType p_type, Dat VK_COMPONENT_SWIZZLE_A }; - // hardcode for now, maybe make this settable from outside.. + // Hardcode for now, maybe make this settable from outside. image_view_create_info.components.r = component_swizzles[TEXTURE_SWIZZLE_R]; image_view_create_info.components.g = component_swizzles[TEXTURE_SWIZZLE_G]; image_view_create_info.components.b = component_swizzles[TEXTURE_SWIZZLE_B]; @@ -2230,7 +2267,7 @@ RID RenderingDeviceVulkan::texture_create_from_extension(TextureType p_type, Dat ERR_FAIL_V_MSG(RID(), "vkCreateImageView failed with error " + itos(err) + "."); } - //barrier to set layout + // Barrier to set layout. { VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; @@ -2252,6 +2289,9 @@ RID RenderingDeviceVulkan::texture_create_from_extension(TextureType p_type, Dat } RID id = texture_owner.make_rid(texture); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif return id; } @@ -2262,10 +2302,10 @@ RID RenderingDeviceVulkan::texture_create_shared_from_slice(const TextureView &p Texture *src_texture = texture_owner.get_or_null(p_with_texture); ERR_FAIL_COND_V(!src_texture, RID()); - if (src_texture->owner.is_valid()) { //ahh this is a share + if (src_texture->owner.is_valid()) { // Ahh this is a share. p_with_texture = src_texture->owner; src_texture = texture_owner.get_or_null(src_texture->owner); - ERR_FAIL_COND_V(!src_texture, RID()); //this is a bug + ERR_FAIL_COND_V(!src_texture, RID()); // This is a bug. } ERR_FAIL_COND_V_MSG(p_slice_type == TEXTURE_SLICE_CUBEMAP && (src_texture->type != TEXTURE_TYPE_CUBE && src_texture->type != TEXTURE_TYPE_CUBE_ARRAY), RID(), @@ -2277,7 +2317,7 @@ RID RenderingDeviceVulkan::texture_create_shared_from_slice(const TextureView &p ERR_FAIL_COND_V_MSG(p_slice_type == TEXTURE_SLICE_2D_ARRAY && (src_texture->type != TEXTURE_TYPE_2D_ARRAY), RID(), "Can only create an array slice from a 2D array mipmap"); - //create view + // Create view. ERR_FAIL_UNSIGNED_INDEX_V(p_mipmap, src_texture->mipmaps, RID()); ERR_FAIL_COND_V(p_mipmap + p_mipmaps > src_texture->mipmaps, RID()); @@ -2377,12 +2417,15 @@ RID RenderingDeviceVulkan::texture_create_shared_from_slice(const TextureView &p texture.owner = p_with_texture; RID id = texture_owner.make_rid(texture); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif _add_dependency(id, p_with_texture); return id; } -Error RenderingDeviceVulkan::texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, uint32_t p_post_barrier) { +Error RenderingDeviceVulkan::texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, BitField<BarrierMask> p_post_barrier) { return _texture_update(p_texture, p_layer, p_data, p_post_barrier, false); } @@ -2402,7 +2445,7 @@ static _ALWAYS_INLINE_ void _copy_region(uint8_t const *__restrict p_src, uint8_ } } -Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, uint32_t p_post_barrier, bool p_use_setup_queue) { +Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, BitField<BarrierMask> p_post_barrier, bool p_use_setup_queue) { _THREAD_SAFE_METHOD_ ERR_FAIL_COND_V_MSG((draw_list || compute_list) && !p_use_setup_queue, ERR_INVALID_PARAMETER, @@ -2414,7 +2457,7 @@ Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, co if (texture->owner != RID()) { p_texture = texture->owner; texture = texture_owner.get_or_null(texture->owner); - ERR_FAIL_COND_V(!texture, ERR_BUG); //this is a bug + ERR_FAIL_COND_V(!texture, ERR_BUG); // This is a bug. } ERR_FAIL_COND_V_MSG(texture->bound, ERR_CANT_ACQUIRE_RESOURCE, @@ -2436,7 +2479,7 @@ Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, co if (required_align == 1) { required_align = get_image_format_pixel_size(texture->format); } - if ((required_align % 4) != 0) { //alignment rules are really strange + if ((required_align % 4) != 0) { // Alignment rules are really strange. required_align *= 4; } @@ -2449,7 +2492,7 @@ Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, co VkCommandBuffer command_buffer = p_use_setup_queue ? frames[frame].setup_command_buffer : frames[frame].draw_command_buffer; - //barrier to transfer + // Barrier to transfer. { VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; @@ -2483,7 +2526,7 @@ Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, co const uint8_t *read_ptr_mipmap = r + mipmap_offset; image_size = image_total - mipmap_offset; - for (uint32_t z = 0; z < depth; z++) { //for 3D textures, depth may be > 0 + for (uint32_t z = 0; z < depth; z++) { // For 3D textures, depth may be > 0. const uint8_t *read_ptr = read_ptr_mipmap + image_size * z / depth; @@ -2505,7 +2548,7 @@ Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, co uint8_t *write_ptr; - { //map + { // Map. void *data_ptr = nullptr; VkResult vkerr = vmaMapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation, &data_ptr); ERR_FAIL_COND_V_MSG(vkerr, ERR_CANT_CREATE, "vmaMapMemory failed with error " + itos(vkerr) + "."); @@ -2520,11 +2563,11 @@ Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, co ERR_FAIL_COND_V(region_h % block_h, ERR_BUG); if (block_w != 1 || block_h != 1) { - //compressed image (blocks) - //must copy a block region + // Compressed image (blocks). + // Must copy a block region. uint32_t block_size = get_compressed_image_format_block_byte_size(texture->format); - //re-create current variables in blocky format + // Re-create current variables in blocky format. uint32_t xb = x / block_w; uint32_t yb = y / block_h; uint32_t wb = width / block_w; @@ -2533,19 +2576,19 @@ Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, co uint32_t region_hb = region_h / block_h; _copy_region(read_ptr, write_ptr, xb, yb, region_wb, region_hb, wb, block_size); } else { - //regular image (pixels) - //must copy a pixel region + // Regular image (pixels). + // Must copy a pixel region. _copy_region(read_ptr, write_ptr, x, y, region_w, region_h, width, pixel_size); } - { //unmap + { // Unmap. vmaUnmapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation); } VkBufferImageCopy buffer_image_copy; buffer_image_copy.bufferOffset = alloc_offset; - buffer_image_copy.bufferRowLength = 0; //tightly packed - buffer_image_copy.bufferImageHeight = 0; //tightly packed + buffer_image_copy.bufferRowLength = 0; // Tightly packed. + buffer_image_copy.bufferImageHeight = 0; // Tightly packed. buffer_image_copy.imageSubresource.aspectMask = texture->read_aspect_mask; buffer_image_copy.imageSubresource.mipLevel = mm_i; @@ -2572,19 +2615,19 @@ Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, co logic_height = MAX(1u, logic_height >> 1); } - //barrier to restore layout + // Barrier to restore layout. { uint32_t barrier_flags = 0; uint32_t access_flags = 0; - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_TRANSFER) { + if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) { barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; } @@ -2659,7 +2702,7 @@ Vector<uint8_t> RenderingDeviceVulkan::_texture_get_data_from_image(Texture *tex const uint8_t *slice_read_ptr = ((uint8_t *)img_mem) + layout.offset + z * layout.depthPitch; if (block_size > 1) { - //compressed + // Compressed. uint32_t line_width = (block_size * (width / blockw)); for (uint32_t y = 0; y < height / blockh; y++) { const uint8_t *rptr = slice_read_ptr + y * layout.rowPitch; @@ -2669,7 +2712,7 @@ Vector<uint8_t> RenderingDeviceVulkan::_texture_get_data_from_image(Texture *tex } } else { - //uncompressed + // Uncompressed. for (uint32_t y = 0; y < height; y++) { const uint8_t *rptr = slice_read_ptr + y * layout.rowPitch; uint8_t *wptr = write_ptr + y * pixel_size * width; @@ -2705,19 +2748,19 @@ Vector<uint8_t> RenderingDeviceVulkan::texture_get_data(RID p_texture, uint32_t ERR_FAIL_COND_V(p_layer >= layer_count, Vector<uint8_t>()); if (tex->usage_flags & TEXTURE_USAGE_CPU_READ_BIT) { - //does not need anything fancy, map and read. + // Does not need anything fancy, map and read. return _texture_get_data_from_image(tex, tex->image, tex->allocation, p_layer); } else { - //compute total image size + // Compute total image size. uint32_t width, height, depth; uint32_t buffer_size = get_image_format_required_size(tex->format, tex->width, tex->height, tex->depth, tex->mipmaps, &width, &height, &depth); - //allocate buffer - VkCommandBuffer command_buffer = frames[frame].draw_command_buffer; //makes more sense to retrieve + // Allocate buffer. + VkCommandBuffer command_buffer = frames[frame].draw_command_buffer; // Makes more sense to retrieve. Buffer tmp_buffer; _buffer_allocate(&tmp_buffer, buffer_size, VK_BUFFER_USAGE_TRANSFER_DST_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_HOST, VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT); - { //Source image barrier + { // Source image barrier. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -2773,7 +2816,7 @@ Vector<uint8_t> RenderingDeviceVulkan::texture_get_data(RID p_texture, uint32_t offset += size; } - { //restore src + { // Restore src. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -2837,7 +2880,7 @@ Size2i RenderingDeviceVulkan::texture_size(RID p_texture) { return Size2i(tex->width, tex->height); } -Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, const Vector3 &p_from, const Vector3 &p_to, const Vector3 &p_size, uint32_t p_src_mipmap, uint32_t p_dst_mipmap, uint32_t p_src_layer, uint32_t p_dst_layer, uint32_t p_post_barrier) { +Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, const Vector3 &p_from, const Vector3 &p_to, const Vector3 &p_size, uint32_t p_src_mipmap, uint32_t p_dst_mipmap, uint32_t p_src_layer, uint32_t p_dst_layer, BitField<BarrierMask> p_post_barrier) { _THREAD_SAFE_METHOD_ Texture *src_tex = texture_owner.get_or_null(p_from_texture); @@ -2888,9 +2931,9 @@ Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, VkCommandBuffer command_buffer = frames[frame].draw_command_buffer; { - //PRE Copy the image + // PRE Copy the image. - { //Source + { // Source. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -2910,7 +2953,7 @@ Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); } - { //Dest + { // Dest. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -2931,7 +2974,7 @@ Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); } - //COPY + // COPY. { VkImageCopy image_copy_region; @@ -2958,19 +3001,19 @@ Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, vkCmdCopyImage(command_buffer, src_tex->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_tex->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &image_copy_region); } - // RESTORE LAYOUT for SRC and DST + // RESTORE LAYOUT for SRC and DST. uint32_t barrier_flags = 0; uint32_t access_flags = 0; - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_TRANSFER) { + if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) { barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; } @@ -2979,7 +3022,7 @@ Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; } - { //restore src + { // Restore src. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -2999,7 +3042,7 @@ Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); } - { //make dst readable + { // Make dst readable. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; @@ -3032,7 +3075,7 @@ Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, return OK; } -Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID p_to_texture, uint32_t p_post_barrier) { +Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID p_to_texture, BitField<BarrierMask> p_post_barrier) { _THREAD_SAFE_METHOD_ Texture *src_tex = texture_owner.get_or_null(p_from_texture); @@ -3066,9 +3109,9 @@ Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID VkCommandBuffer command_buffer = frames[frame].draw_command_buffer; { - //PRE Copy the image + // PRE Copy the image. - { //Source + { // Source. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -3088,7 +3131,7 @@ Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); } - { //Dest + { // Dest. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -3109,7 +3152,7 @@ Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); } - //COPY + // COPY. { VkImageResolve image_copy_region; @@ -3136,19 +3179,19 @@ Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID vkCmdResolveImage(command_buffer, src_tex->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_tex->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &image_copy_region); } - // RESTORE LAYOUT for SRC and DST + // RESTORE LAYOUT for SRC and DST. uint32_t barrier_flags = 0; uint32_t access_flags = 0; - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_TRANSFER) { + if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) { barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; } @@ -3157,7 +3200,7 @@ Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; } - { //restore src + { // Restore src. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -3177,7 +3220,7 @@ Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID vkCmdPipelineBarrier(command_buffer, VK_ACCESS_TRANSFER_WRITE_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); } - { //make dst readable + { // Make dst readable. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; @@ -3203,7 +3246,7 @@ Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID return OK; } -Error RenderingDeviceVulkan::texture_clear(RID p_texture, const Color &p_color, uint32_t p_base_mipmap, uint32_t p_mipmaps, uint32_t p_base_layer, uint32_t p_layers, uint32_t p_post_barrier) { +Error RenderingDeviceVulkan::texture_clear(RID p_texture, const Color &p_color, uint32_t p_base_mipmap, uint32_t p_mipmaps, uint32_t p_base_layer, uint32_t p_layers, BitField<BarrierMask> p_post_barrier) { _THREAD_SAFE_METHOD_ Texture *src_tex = texture_owner.get_or_null(p_texture); @@ -3230,13 +3273,13 @@ Error RenderingDeviceVulkan::texture_clear(RID p_texture, const Color &p_color, VkImageLayout clear_layout = (src_tex->layout == VK_IMAGE_LAYOUT_GENERAL) ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL; - // NOTE: Perhaps the valid stages/accesses for a given owner should be a property of the owner. (Here and places like _get_buffer_from_owner) + // NOTE: Perhaps the valid stages/accesses for a given owner should be a property of the owner. (Here and places like _get_buffer_from_owner.) const VkPipelineStageFlags valid_texture_stages = VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; constexpr VkAccessFlags read_access = VK_ACCESS_SHADER_READ_BIT; constexpr VkAccessFlags read_write_access = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; const VkAccessFlags valid_texture_access = (src_tex->usage_flags & TEXTURE_USAGE_STORAGE_BIT) ? read_write_access : read_access; - { // Barrier from previous access with optional layout change (see clear_layout logic above) + { // Barrier from previous access with optional layout change (see clear_layout logic above). VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -3272,19 +3315,19 @@ Error RenderingDeviceVulkan::texture_clear(RID p_texture, const Color &p_color, vkCmdClearColorImage(command_buffer, src_tex->image, clear_layout, &clear_color, 1, &range); - { // Barrier to post clear accesses (changing back the layout if needed) + { // Barrier to post clear accesses (changing back the layout if needed). uint32_t barrier_flags = 0; uint32_t access_flags = 0; - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_TRANSFER) { + if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) { barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; } @@ -3323,44 +3366,44 @@ Error RenderingDeviceVulkan::texture_clear(RID p_texture, const Color &p_color, return OK; } -bool RenderingDeviceVulkan::texture_is_format_supported_for_usage(DataFormat p_format, uint32_t p_usage) const { +bool RenderingDeviceVulkan::texture_is_format_supported_for_usage(DataFormat p_format, BitField<RenderingDevice::TextureUsageBits> p_usage) const { ERR_FAIL_INDEX_V(p_format, DATA_FORMAT_MAX, false); _THREAD_SAFE_METHOD_ - //validate that this image is supported for the intended use + // Validate that this image is supported for the intended use. VkFormatProperties properties; vkGetPhysicalDeviceFormatProperties(context->get_physical_device(), vulkan_formats[p_format], &properties); VkFormatFeatureFlags flags; - if (p_usage & TEXTURE_USAGE_CPU_READ_BIT) { + if (p_usage.has_flag(TEXTURE_USAGE_CPU_READ_BIT)) { flags = properties.linearTilingFeatures; } else { flags = properties.optimalTilingFeatures; } - if (p_usage & TEXTURE_USAGE_SAMPLING_BIT && !(flags & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT)) { + if (p_usage.has_flag(TEXTURE_USAGE_SAMPLING_BIT) && !(flags & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT)) { return false; } - if (p_usage & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT)) { + if (p_usage.has_flag(TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) && !(flags & VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT)) { return false; } - if (p_usage & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT)) { + if (p_usage.has_flag(TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) && !(flags & VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT)) { return false; } - if (p_usage & TEXTURE_USAGE_STORAGE_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT)) { + if (p_usage.has_flag(TEXTURE_USAGE_STORAGE_BIT) && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT)) { return false; } - if (p_usage & TEXTURE_USAGE_STORAGE_ATOMIC_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_ATOMIC_BIT)) { + if (p_usage.has_flag(TEXTURE_USAGE_STORAGE_ATOMIC_BIT) && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_ATOMIC_BIT)) { return false; } // Validation via VK_FORMAT_FEATURE_FRAGMENT_SHADING_RATE_ATTACHMENT_BIT_KHR fails if VRS attachment is not supported. - if (p_usage & TEXTURE_USAGE_VRS_ATTACHMENT_BIT && p_format != DATA_FORMAT_R8_UINT) { + if (p_usage.has_flag(TEXTURE_USAGE_VRS_ATTACHMENT_BIT) && p_format != DATA_FORMAT_R8_UINT) { return false; } @@ -3372,12 +3415,12 @@ bool RenderingDeviceVulkan::texture_is_format_supported_for_usage(DataFormat p_f /********************/ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentFormat> &p_attachments, const Vector<FramebufferPass> &p_passes, InitialAction p_initial_action, FinalAction p_final_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, uint32_t p_view_count, Vector<TextureSamples> *r_samples) { - // Set up dependencies from/to external equivalent to the default (implicit) one, and then amend them + // Set up dependencies from/to external equivalent to the default (implicit) one, and then amend them. const VkPipelineStageFlags default_access_mask = VK_ACCESS_INPUT_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | - VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | // From Section 7.1 of Vulkan API Spec v1.1.148 + VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | // From Section 7.1 of Vulkan API Spec v1.1.148. VK_ACCESS_FRAGMENT_SHADING_RATE_ATTACHMENT_READ_BIT_KHR; VkPipelineStageFlags reading_stages = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_TRANSFER_BIT; @@ -3390,6 +3433,16 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF LocalVector<int32_t> attachment_last_pass; attachment_last_pass.resize(p_attachments.size()); + if (p_view_count > 1) { + const VulkanContext::MultiviewCapabilities capabilities = context->get_multiview_capabilities(); + + // This only works with multiview! + ERR_FAIL_COND_V_MSG(!capabilities.is_supported, VK_NULL_HANDLE, "Multiview not supported"); + + // Make sure we limit this to the number of views we support. + ERR_FAIL_COND_V_MSG(p_view_count > capabilities.max_view_count, VK_NULL_HANDLE, "Hardware does not support requested number of views for Multiview render pass"); + } + // These are only used if we use multiview but we need to define them in scope. const uint32_t view_mask = (1 << p_view_count) - 1; const uint32_t correlation_mask = (1 << p_view_count) - 1; @@ -3413,27 +3466,27 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF description.pNext = nullptr; description.flags = 0; description.format = vulkan_formats[p_attachments[i].format]; - description.samples = rasterization_sample_count[p_attachments[i].samples]; + description.samples = _ensure_supported_sample_count(p_attachments[i].samples); bool is_sampled = p_attachments[i].usage_flags & TEXTURE_USAGE_SAMPLING_BIT; bool is_storage = p_attachments[i].usage_flags & TEXTURE_USAGE_STORAGE_BIT; bool is_depth = p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; // We can setup a framebuffer where we write to our VRS texture to set it up. - // We make the assumption here that if our texture is actually used as our VRS attachment, - // it is used as such for each subpass. This is fairly certain seeing the restrictions on subpasses. + // We make the assumption here that if our texture is actually used as our VRS attachment. + // It is used as such for each subpass. This is fairly certain seeing the restrictions on subpasses. bool is_vrs = p_attachments[i].usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT && i == p_passes[0].vrs_attachment; if (is_vrs) { - // For VRS we only read, there is no writing to this texture + // For VRS we only read, there is no writing to this texture. description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; description.initialLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD; } else { - // For each UNDEFINED, assume the prior use was a *read*, as we'd be discarding the output of a write - // Also, each UNDEFINED will do an immediate layout transition (write), s.t. we must ensure execution synchronization vs. + // For each UNDEFINED, assume the prior use was a *read*, as we'd be discarding the output of a write. + // Also, each UNDEFINED will do an immediate layout transition (write), s.t. we must ensure execution synchronization vs // the read. If this is a performance issue, one could track the actual last accessor of each resource, adding only that - // stage + // stage. switch (is_depth ? p_initial_depth_action : p_initial_action) { case INITIAL_ACTION_CLEAR_REGION: @@ -3450,7 +3503,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF } else { description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; - description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there + description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there. dependency_from_external.srcStageMask |= reading_stages; } } break; @@ -3467,7 +3520,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF } else { description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; - description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there + description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there. dependency_from_external.srcStageMask |= reading_stages; } } break; @@ -3478,13 +3531,13 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; } else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) { description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; - description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there + description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there. description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; dependency_from_external.srcStageMask |= reading_stages; } else { description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; - description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there + description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there. dependency_from_external.srcStageMask |= reading_stages; } } break; @@ -3501,12 +3554,12 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF } else { description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; - description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there + description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there. dependency_from_external.srcStageMask |= reading_stages; } } break; default: { - ERR_FAIL_V(VK_NULL_HANDLE); //should never reach here + ERR_FAIL_V(VK_NULL_HANDLE); // Should never reach here. } } } @@ -3517,7 +3570,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF int last_pass = p_passes.size() - 1; if (is_depth) { - //likely missing depth resolve? + // Likely missing depth resolve? if (p_passes[last_pass].depth_attachment == i) { used_last = true; } @@ -3527,14 +3580,15 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF } } else { if (p_passes[last_pass].resolve_attachments.size()) { - //if using resolve attachments, check resolve attachments + // If using resolve attachments, check resolve attachments. for (int j = 0; j < p_passes[last_pass].resolve_attachments.size(); j++) { if (p_passes[last_pass].resolve_attachments[j] == i) { used_last = true; break; } } - } else { + } + if (!used_last) { for (int j = 0; j < p_passes[last_pass].color_attachments.size(); j++) { if (p_passes[last_pass].color_attachments[j] == i) { used_last = true; @@ -3567,13 +3621,13 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF } if (is_vrs) { - // We don't change our VRS texture during this process + // We don't change our VRS texture during this process. description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; description.finalLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; - // TODO do we need to update our external dependency ? + // TODO: Do we need to update our external dependency? // update_external_dependency_for_store(dependency_to_external, is_sampled, is_storage, false); } else { switch (is_depth ? final_depth_action : final_action) { @@ -3591,8 +3645,8 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF } else { description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; - description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there - // TODO: What does this mean about the next usage (and thus appropriate dependency masks + description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there. + // TODO: What does this mean about the next usage (and thus appropriate dependency masks. } } break; case FINAL_ACTION_DISCARD: { @@ -3607,7 +3661,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF } else { description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; - description.finalLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there + description.finalLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there. } } break; case FINAL_ACTION_CONTINUE: { @@ -3622,12 +3676,12 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF } else { description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; - description.finalLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there + description.finalLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there. } } break; default: { - ERR_FAIL_V(VK_NULL_HANDLE); //should never reach here + ERR_FAIL_V(VK_NULL_HANDLE); // Should never reach here. } } } @@ -3689,7 +3743,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF reference.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; attachment_last_pass[attachment] = i; } - reference.aspectMask = 0; + reference.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; color_references.push_back(reference); } @@ -3711,7 +3765,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF reference.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; attachment_last_pass[attachment] = i; } - reference.aspectMask = 0; // TODO we need to set this here, possibly VK_IMAGE_ASPECT_COLOR_BIT ?? + reference.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; input_references.push_back(reference); } @@ -3737,10 +3791,10 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF bool multisample = p_attachments[attachment].samples > TEXTURE_SAMPLES_1; ERR_FAIL_COND_V_MSG(multisample, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), resolve attachments can't be multisample."); reference.attachment = attachment_remap[attachment]; - reference.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; // VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; + reference.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; // VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL attachment_last_pass[attachment] = i; } - reference.aspectMask = 0; + reference.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; resolve_references.push_back(reference); } @@ -3755,7 +3809,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer depth format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass."); depth_stencil_reference.attachment = attachment_remap[attachment]; depth_stencil_reference.layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; - depth_stencil_reference.aspectMask = 0; + depth_stencil_reference.aspectMask = VK_IMAGE_ASPECT_NONE; attachment_last_pass[attachment] = i; if (is_multisample_first) { @@ -3781,9 +3835,9 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF vrs_reference.pNext = nullptr; vrs_reference.attachment = attachment_remap[attachment]; vrs_reference.layout = VK_IMAGE_LAYOUT_FRAGMENT_SHADING_RATE_ATTACHMENT_OPTIMAL_KHR; - vrs_reference.aspectMask = 0; + vrs_reference.aspectMask = VK_IMAGE_ASPECT_NONE; - Size2i texel_size = context->get_vrs_capabilities().max_texel_size; + Size2i texel_size = context->get_vrs_capabilities().texel_size; VkFragmentShadingRateAttachmentInfoKHR &vrs_attachment_info = vrs_attachment_info_array[i]; vrs_attachment_info.sType = VK_STRUCTURE_TYPE_FRAGMENT_SHADING_RATE_ATTACHMENT_INFO_KHR; @@ -3806,7 +3860,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), preserve attachment (" + itos(j) + ")."); if (attachment_last_pass[attachment] != i) { - //preserve can still be used to keep depth or color from being discarded after use + // Preserve can still be used to keep depth or color from being discarded after use. attachment_last_pass[attachment] = i; preserve_references.push_back(attachment); } @@ -3875,14 +3929,14 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF subpass_dependencies.push_back(dependency); } /* - // NOTE: Big Mallet Approach -- any layout transition causes a full barrier + // NOTE: Big Mallet Approach -- any layout transition causes a full barrier. if (reference.layout != description.initialLayout) { - // NOTE: this should be smarter based on the texture's knowledge of its previous role + // NOTE: This should be smarter based on the texture's knowledge of its previous role. dependency_from_external.srcStageMask |= VK_PIPELINE_STAGE_ALL_COMMANDS_BIT; dependency_from_external.srcAccessMask |= VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT; } if (reference.layout != description.finalLayout) { - // NOTE: this should be smarter based on the texture's knowledge of its subsequent role + // NOTE: This should be smarter based on the texture's knowledge of its subsequent role. dependency_to_external.dstStageMask |= VK_PIPELINE_STAGE_ALL_COMMANDS_BIT; dependency_to_external.dstAccessMask |= VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT; } @@ -3922,18 +3976,11 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF Vector<uint32_t> view_masks; VkRenderPassMultiviewCreateInfo render_pass_multiview_create_info; - if (p_view_count > 1) { - // this may no longer be needed with the new settings already including this + if ((p_view_count > 1) && !context->supports_renderpass2()) { + // This is only required when using vkCreateRenderPass, we add it if vkCreateRenderPass2KHR is not supported + // resulting this in being passed to our vkCreateRenderPass fallback. - const VulkanContext::MultiviewCapabilities capabilities = context->get_multiview_capabilities(); - - // For now this only works with multiview! - ERR_FAIL_COND_V_MSG(!capabilities.is_supported, VK_NULL_HANDLE, "Multiview not supported"); - - // Make sure we limit this to the number of views we support. - ERR_FAIL_COND_V_MSG(p_view_count > capabilities.max_view_count, VK_NULL_HANDLE, "Hardware does not support requested number of views for Multiview render pass"); - - // Set view masks for each subpass + // Set view masks for each subpass. for (uint32_t i = 0; i < subpasses.size(); i++) { view_masks.push_back(view_mask); } @@ -3971,7 +4018,7 @@ RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_c passes.push_back(pass); return framebuffer_format_create_multipass(p_format, passes, p_view_count); } -RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_create_multipass(const Vector<AttachmentFormat> &p_attachments, Vector<FramebufferPass> &p_passes, uint32_t p_view_count) { +RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_create_multipass(const Vector<AttachmentFormat> &p_attachments, const Vector<FramebufferPass> &p_passes, uint32_t p_view_count) { _THREAD_SAFE_METHOD_ FramebufferFormatKey key; @@ -3981,14 +4028,14 @@ RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_c const RBMap<FramebufferFormatKey, FramebufferFormatID>::Element *E = framebuffer_format_cache.find(key); if (E) { - //exists, return + // Exists, return. return E->get(); } Vector<TextureSamples> samples; - VkRenderPass render_pass = _render_pass_create(p_attachments, p_passes, INITIAL_ACTION_CLEAR, FINAL_ACTION_READ, INITIAL_ACTION_CLEAR, FINAL_ACTION_READ, p_view_count, &samples); //actions don't matter for this use case + VkRenderPass render_pass = _render_pass_create(p_attachments, p_passes, INITIAL_ACTION_CLEAR, FINAL_ACTION_READ, INITIAL_ACTION_CLEAR, FINAL_ACTION_READ, p_view_count, &samples); // Actions don't matter for this use case. - if (render_pass == VK_NULL_HANDLE) { //was likely invalid + if (render_pass == VK_NULL_HANDLE) { // Was likely invalid. return INVALID_ID; } FramebufferFormatID id = FramebufferFormatID(framebuffer_format_cache.size()) | (FramebufferFormatID(ID_TYPE_FRAMEBUFFER_FORMAT) << FramebufferFormatID(ID_BASE_SHIFT)); @@ -4009,7 +4056,7 @@ RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_c const RBMap<FramebufferFormatKey, FramebufferFormatID>::Element *E = framebuffer_format_cache.find(key); if (E) { - //exists, return + // Exists, return. return E->get(); } @@ -4019,7 +4066,7 @@ RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_c subpass.flags = 0; subpass.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; subpass.viewMask = 0; - subpass.inputAttachmentCount = 0; //unsupported for now + subpass.inputAttachmentCount = 0; // Unsupported for now. subpass.pInputAttachments = nullptr; subpass.colorAttachmentCount = 0; subpass.pColorAttachments = nullptr; @@ -4046,7 +4093,7 @@ RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_c ERR_FAIL_COND_V_MSG(res, 0, "vkCreateRenderPass2KHR for empty fb failed with error " + itos(res) + "."); - if (render_pass == VK_NULL_HANDLE) { //was likely invalid + if (render_pass == VK_NULL_HANDLE) { // Was likely invalid. return INVALID_ID; } @@ -4082,7 +4129,11 @@ RID RenderingDeviceVulkan::framebuffer_create_empty(const Size2i &p_size, Textur framebuffer.size = p_size; framebuffer.view_count = 1; - return framebuffer_owner.make_rid(framebuffer); + RID id = framebuffer_owner.make_rid(framebuffer); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + return id; } RID RenderingDeviceVulkan::framebuffer_create(const Vector<RID> &p_texture_attachments, FramebufferFormatID p_format_check, uint32_t p_view_count) { @@ -4100,7 +4151,11 @@ RID RenderingDeviceVulkan::framebuffer_create(const Vector<RID> &p_texture_attac } else if (texture && texture->usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT) { pass.vrs_attachment = i; } else { - pass.color_attachments.push_back(texture ? i : FramebufferPass::ATTACHMENT_UNUSED); + if (texture && texture->is_resolve_buffer) { + pass.resolve_attachments.push_back(i); + } else { + pass.color_attachments.push_back(texture ? i : FramebufferPass::ATTACHMENT_UNUSED); + } } } @@ -4110,7 +4165,7 @@ RID RenderingDeviceVulkan::framebuffer_create(const Vector<RID> &p_texture_attac return framebuffer_create_multipass(p_texture_attachments, passes, p_format_check, p_view_count); } -RID RenderingDeviceVulkan::framebuffer_create_multipass(const Vector<RID> &p_texture_attachments, Vector<FramebufferPass> &p_passes, FramebufferFormatID p_format_check, uint32_t p_view_count) { +RID RenderingDeviceVulkan::framebuffer_create_multipass(const Vector<RID> &p_texture_attachments, const Vector<FramebufferPass> &p_passes, FramebufferFormatID p_format_check, uint32_t p_view_count) { _THREAD_SAFE_METHOD_ Vector<AttachmentFormat> attachments; @@ -4130,9 +4185,9 @@ RID RenderingDeviceVulkan::framebuffer_create_multipass(const Vector<RID> &p_tex size.height = texture->height; size_set = true; } else if (texture->usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT) { - // If this is not the first attachement we assume this is used as the VRS attachment - // in this case this texture will be 1/16th the size of the color attachement. - // So we skip the size check + // If this is not the first attachment we assume this is used as the VRS attachment. + // In this case this texture will be 1/16th the size of the color attachment. + // So we skip the size check. } else { ERR_FAIL_COND_V_MSG((uint32_t)size.width != texture->width || (uint32_t)size.height != texture->height, RID(), "All textures in a framebuffer should be the same size."); @@ -4162,6 +4217,9 @@ RID RenderingDeviceVulkan::framebuffer_create_multipass(const Vector<RID> &p_tex framebuffer.view_count = p_view_count; RID id = framebuffer_owner.make_rid(framebuffer); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif for (int i = 0; i < p_texture_attachments.size(); i++) { if (p_texture_attachments[i].is_valid()) { @@ -4181,6 +4239,22 @@ RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_get_form return framebuffer->format_id; } +bool RenderingDeviceVulkan::framebuffer_is_valid(RID p_framebuffer) const { + _THREAD_SAFE_METHOD_ + + return framebuffer_owner.owns(p_framebuffer); +} + +void RenderingDeviceVulkan::framebuffer_set_invalidation_callback(RID p_framebuffer, InvalidationCallback p_callback, void *p_userdata) { + _THREAD_SAFE_METHOD_ + + Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_framebuffer); + ERR_FAIL_COND(!framebuffer); + + framebuffer->invalidated_callback = p_callback; + framebuffer->invalidated_callback_userdata = p_userdata; +} + /*****************/ /**** SAMPLER ****/ /*****************/ @@ -4223,7 +4297,11 @@ RID RenderingDeviceVulkan::sampler_create(const SamplerState &p_state) { VkResult res = vkCreateSampler(device, &sampler_create_info, nullptr, &sampler); ERR_FAIL_COND_V_MSG(res, RID(), "vkCreateSampler failed with error " + itos(res) + "."); - return sampler_owner.make_rid(sampler); + RID id = sampler_owner.make_rid(sampler); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + return id; } /**********************/ @@ -4252,10 +4330,14 @@ RID RenderingDeviceVulkan::vertex_buffer_create(uint32_t p_size_bytes, const Vec _buffer_memory_barrier(buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT, false); } - return vertex_buffer_owner.make_rid(buffer); + RID id = vertex_buffer_owner.make_rid(buffer); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + return id; } -// Internally reference counted, this ID is warranted to be unique for the same description, but needs to be freed as many times as it was allocated +// Internally reference counted, this ID is warranted to be unique for the same description, but needs to be freed as many times as it was allocated. RenderingDevice::VertexFormatID RenderingDeviceVulkan::vertex_format_create(const Vector<VertexAttribute> &p_vertex_formats) { _THREAD_SAFE_METHOD_ @@ -4267,7 +4349,7 @@ RenderingDevice::VertexFormatID RenderingDeviceVulkan::vertex_format_create(cons return *idptr; } - //does not exist, create one and cache it + // Does not exist, create one and cache it. VertexDescriptionCache vdcache; vdcache.bindings = memnew_arr(VkVertexInputBindingDescription, p_vertex_formats.size()); vdcache.attributes = memnew_arr(VkVertexInputAttributeDescription, p_vertex_formats.size()); @@ -4307,7 +4389,7 @@ RenderingDevice::VertexFormatID RenderingDeviceVulkan::vertex_format_create(cons return id; } -RID RenderingDeviceVulkan::vertex_array_create(uint32_t p_vertex_count, VertexFormatID p_vertex_format, const Vector<RID> &p_src_buffers) { +RID RenderingDeviceVulkan::vertex_array_create(uint32_t p_vertex_count, VertexFormatID p_vertex_format, const Vector<RID> &p_src_buffers, const Vector<uint64_t> &p_offsets) { _THREAD_SAFE_METHOD_ ERR_FAIL_COND_V(!vertex_formats.has(p_vertex_format), RID()); @@ -4321,27 +4403,34 @@ RID RenderingDeviceVulkan::vertex_array_create(uint32_t p_vertex_count, VertexFo VertexArray vertex_array; + if (p_offsets.is_empty()) { + vertex_array.offsets.resize_zeroed(p_src_buffers.size()); + } else { + ERR_FAIL_COND_V(p_offsets.size() != p_src_buffers.size(), RID()); + vertex_array.offsets = p_offsets; + } + vertex_array.vertex_count = p_vertex_count; vertex_array.description = p_vertex_format; - vertex_array.max_instances_allowed = 0xFFFFFFFF; //by default as many as you want + vertex_array.max_instances_allowed = 0xFFFFFFFF; // By default as many as you want. for (int i = 0; i < p_src_buffers.size(); i++) { Buffer *buffer = vertex_buffer_owner.get_or_null(p_src_buffers[i]); - //validate with buffer + // Validate with buffer. { const VertexAttribute &atf = vd.vertex_formats[i]; uint32_t element_size = get_format_vertex_size(atf.format); - ERR_FAIL_COND_V(element_size == 0, RID()); //should never happens since this was prevalidated + ERR_FAIL_COND_V(element_size == 0, RID()); // Should never happens since this was prevalidated. if (atf.frequency == VERTEX_FREQUENCY_VERTEX) { - //validate size for regular drawing + // Validate size for regular drawing. uint64_t total_size = uint64_t(atf.stride) * (p_vertex_count - 1) + atf.offset + element_size; ERR_FAIL_COND_V_MSG(total_size > buffer->size, RID(), "Attachment (" + itos(i) + ") will read past the end of the buffer."); } else { - //validate size for instances drawing + // Validate size for instances drawing. uint64_t available = buffer->size - atf.offset; ERR_FAIL_COND_V_MSG(available < element_size, RID(), "Attachment (" + itos(i) + ") uses instancing, but it's just too small."); @@ -4352,7 +4441,6 @@ RID RenderingDeviceVulkan::vertex_array_create(uint32_t p_vertex_count, VertexFo } vertex_array.buffers.push_back(buffer->buffer); - vertex_array.offsets.push_back(0); //offset unused, but passing anyway } RID id = vertex_array_owner.make_rid(vertex_array); @@ -4387,7 +4475,7 @@ RID RenderingDeviceVulkan::index_buffer_create(uint32_t p_index_count, IndexBuff const uint16_t *index16 = (const uint16_t *)r; for (uint32_t i = 0; i < p_index_count; i++) { if (p_use_restart_indices && index16[i] == 0xFFFF) { - continue; //restart index, ignore + continue; // Restart index, ignore. } index_buffer.max_index = MAX(index16[i], index_buffer.max_index); } @@ -4395,7 +4483,7 @@ RID RenderingDeviceVulkan::index_buffer_create(uint32_t p_index_count, IndexBuff const uint32_t *index32 = (const uint32_t *)r; for (uint32_t i = 0; i < p_index_count; i++) { if (p_use_restart_indices && index32[i] == 0xFFFFFFFF) { - continue; //restart index, ignore + continue; // Restart index, ignore. } index_buffer.max_index = MAX(index32[i], index_buffer.max_index); } @@ -4413,7 +4501,11 @@ RID RenderingDeviceVulkan::index_buffer_create(uint32_t p_index_count, IndexBuff _buffer_update(&index_buffer, 0, r, data_size); _buffer_memory_barrier(index_buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_INDEX_READ_BIT, false); } - return index_buffer_owner.make_rid(index_buffer); + RID id = index_buffer_owner.make_rid(index_buffer); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + return id; } RID RenderingDeviceVulkan::index_array_create(RID p_index_buffer, uint32_t p_index_offset, uint32_t p_index_count) { @@ -4443,14 +4535,6 @@ RID RenderingDeviceVulkan::index_array_create(RID p_index_buffer, uint32_t p_ind /**** SHADER ****/ /****************/ -static const char *shader_stage_names[RenderingDevice::SHADER_STAGE_MAX] = { - "Vertex", - "Fragment", - "TesselationControl", - "TesselationEvaluation", - "Compute" -}; - static const char *shader_uniform_names[RenderingDevice::UNIFORM_TYPE_MAX] = { "Sampler", "CombinedSampler", "Texture", "Image", "TextureBuffer", "SamplerTextureBuffer", "ImageBuffer", "UniformBuffer", "StorageBuffer", "InputAttachment" }; @@ -4481,214 +4565,22 @@ String RenderingDeviceVulkan::_shader_uniform_debug(RID p_shader, int p_set) { } return ret; } -#if 0 -bool RenderingDeviceVulkan::_uniform_add_binding(Vector<Vector<VkDescriptorSetLayoutBinding> > &bindings, Vector<Vector<UniformInfo> > &uniform_infos, const glslang::TObjectReflection &reflection, RenderingDevice::ShaderStage p_stage, Shader::PushConstant &push_constant, String *r_error) { - VkDescriptorSetLayoutBinding layout_binding; - UniformInfo info; - - switch (reflection.getType()->getBasicType()) { - case glslang::EbtSampler: { - //print_line("DEBUG: IsSampler"); - if (reflection.getType()->getSampler().dim == glslang::EsdBuffer) { - //texture buffers - if (reflection.getType()->getSampler().isCombined()) { - layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER; - info.type = UNIFORM_TYPE_SAMPLER_WITH_TEXTURE_BUFFER; - //print_line("DEBUG: SAMPLER: texel combined"); - } else if (reflection.getType()->getSampler().isTexture()) { - layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER; - info.type = UNIFORM_TYPE_TEXTURE_BUFFER; - //print_line("DEBUG: SAMPLER: texel alone"); - } else if (reflection.getType()->getSampler().isImage()) { - layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER; - info.type = UNIFORM_TYPE_IMAGE_BUFFER; - //print_line("DEBUG: SAMPLER: texel buffer"); - } else { - if (r_error) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' is of unsupported buffer type."; - } - return false; - } - } else if (reflection.getType()->getSampler().isCombined()) { - layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; - info.type = UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; - //print_line("DEBUG: SAMPLER: combined"); - } else if (reflection.getType()->getSampler().isPureSampler()) { - layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLER; - info.type = UNIFORM_TYPE_SAMPLER; - //print_line("DEBUG: SAMPLER: sampler"); - } else if (reflection.getType()->getSampler().isTexture()) { - layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE; - info.type = UNIFORM_TYPE_TEXTURE; - //print_line("DEBUG: SAMPLER: image"); - } else if (reflection.getType()->getSampler().isImage()) { - layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE; - info.type = UNIFORM_TYPE_IMAGE; - //print_line("DEBUG: SAMPLER: storage image"); - } else { - //print_line("DEBUG: sampler unknown"); - if (r_error) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' is of unsupported sampler type."; - } - return false; - } - - if (reflection.getType()->isArray()) { - layout_binding.descriptorCount = reflection.getType()->getArraySizes()->getCumulativeSize(); - //print_line("DEBUG: array of size: " + itos(layout_binding.descriptorCount)); - } else { - layout_binding.descriptorCount = 1; - } - - info.length = layout_binding.descriptorCount; - - } break; - /*case glslang::EbtStruct: { - print_line("DEBUG: Struct"); - - } break;*/ - case glslang::EbtBlock: { - //print_line("DEBUG: Block"); - if (reflection.getType()->getQualifier().storage == glslang::EvqUniform) { - if (reflection.getType()->getQualifier().layoutPushConstant) { - uint32_t len = reflection.size; - if (push_constant.push_constant_size != 0 && push_constant.push_constant_size != len) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' push constants for different stages should all be the same size."; - return false; - } - push_constant.push_constant_size = len; - push_constant.push_constants_vk_stage |= shader_stage_masks[p_stage]; - return true; - } - //print_line("DEBUG: Uniform buffer"); - layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; - info.type = UNIFORM_TYPE_UNIFORM_BUFFER; - } else if (reflection.getType()->getQualifier().storage == glslang::EvqBuffer) { - layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; - info.type = UNIFORM_TYPE_STORAGE_BUFFER; - //print_line("DEBUG: Storage buffer"); - } else { - if (r_error) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' is of unsupported block type: (" + itos(reflection.getType()->getQualifier().storage) + ")."; - } - return false; - } - - if (reflection.getType()->isArray()) { - layout_binding.descriptorCount = reflection.getType()->getArraySizes()->getCumulativeSize(); - //print_line("DEBUG: array of size: " + itos(layout_binding.descriptorCount)); - } else { - layout_binding.descriptorCount = 1; - } - - info.length = reflection.size; - - } break; - /*case glslang::EbtReference: { - } break;*/ - /*case glslang::EbtAtomicUint: { - } break;*/ - default: { - if (reflection.getType()->getQualifier().hasOffset() || reflection.name.find(".") != std::string::npos) { - //member of uniform block? - return true; - } - - if (r_error) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' unsupported uniform type."; - } - return false; - } - } - - if (!reflection.getType()->getQualifier().hasBinding()) { - if (r_error) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' lacks a binding number."; - } - return false; - } - - uint32_t set = reflection.getType()->getQualifier().hasSet() ? reflection.getType()->getQualifier().layoutSet : 0; - - if (set >= MAX_UNIFORM_SETS) { - if (r_error) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' uses a set (" + itos(set) + ") index larger than what is supported (" + itos(MAX_UNIFORM_SETS) + ")."; - } - return false; - } - - if (set >= limits.maxBoundDescriptorSets) { - if (r_error) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' uses a set (" + itos(set) + ") index larger than what is supported by the hardware (" + itos(limits.maxBoundDescriptorSets) + ")."; - } - return false; - } - - uint32_t binding = reflection.getType()->getQualifier().layoutBinding; - - if (set < (uint32_t)bindings.size()) { - //check if this already exists - for (int i = 0; i < bindings[set].size(); i++) { - if (bindings[set][i].binding == binding) { - //already exists, verify that it's the same type - if (bindings[set][i].descriptorType != layout_binding.descriptorType) { - if (r_error) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(binding) + " with different uniform type."; - } - return false; - } - - //also, verify that it's the same size - if (bindings[set][i].descriptorCount != layout_binding.descriptorCount || uniform_infos[set][i].length != info.length) { - if (r_error) { - *r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(binding) + " with different uniform size."; - } - return false; - } - - //just append stage mask and return - bindings.write[set].write[i].stageFlags |= shader_stage_masks[p_stage]; - uniform_infos.write[set].write[i].stages |= 1 << p_stage; - return true; - } - } - } - layout_binding.binding = binding; - layout_binding.stageFlags = shader_stage_masks[p_stage]; - layout_binding.pImmutableSamplers = nullptr; //no support for this yet - - info.stages = 1 << p_stage; - info.binding = binding; - - if (set >= (uint32_t)bindings.size()) { - bindings.resize(set + 1); - uniform_infos.resize(set + 1); - } -#if 0 - print_line("stage: " + String(shader_stage_names[p_stage]) + " set: " + itos(set) + " binding: " + itos(info.binding) + " type:" + shader_uniform_names[info.type] + " length: " + itos(info.length)); -#endif - bindings.write[set].push_back(layout_binding); - uniform_infos.write[set].push_back(info); - - return true; -} -#endif -//version 1: initial -//version 2: Added shader name -//version 3: Added writable +// Version 1: initial. +// Version 2: Added shader name. +// Version 3: Added writable. #define SHADER_BINARY_VERSION 3 String RenderingDeviceVulkan::shader_get_binary_cache_key() const { - return "Vulkan-SV" + itos(SHADER_BINARY_VERSION) + "-" + String(VERSION_NUMBER) + "-" + String(VERSION_HASH); + return "Vulkan-SV" + itos(SHADER_BINARY_VERSION); } struct RenderingDeviceVulkanShaderBinaryDataBinding { uint32_t type; uint32_t binding; uint32_t stages; - uint32_t length; //size of arrays (in total elements), or ubos (in bytes * total elements) + uint32_t length; // Size of arrays (in total elements), or ubos (in bytes * total elements). uint32_t writable; }; @@ -4705,351 +4597,73 @@ struct RenderingDeviceVulkanShaderBinarySpecializationConstant { struct RenderingDeviceVulkanShaderBinaryData { uint32_t vertex_input_mask; - uint32_t fragment_outputs; - uint32_t specialization_constant_count; + uint32_t fragment_output_mask; + uint32_t specialization_constants_count; uint32_t is_compute; uint32_t compute_local_size[3]; uint32_t set_count; uint32_t push_constant_size; - uint32_t push_constants_vk_stage; + uint32_t push_constant_vk_stages_mask; uint32_t stage_count; uint32_t shader_name_len; }; Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Vector<ShaderStageSPIRVData> &p_spirv, const String &p_shader_name) { - RenderingDeviceVulkanShaderBinaryData binary_data; - binary_data.vertex_input_mask = 0; - binary_data.fragment_outputs = 0; - binary_data.specialization_constant_count = 0; - binary_data.is_compute = 0; - binary_data.compute_local_size[0] = 0; - binary_data.compute_local_size[1] = 0; - binary_data.compute_local_size[2] = 0; - binary_data.set_count = 0; - binary_data.push_constant_size = 0; - binary_data.push_constants_vk_stage = 0; - - Vector<Vector<RenderingDeviceVulkanShaderBinaryDataBinding>> uniform_info; //set bindings - Vector<RenderingDeviceVulkanShaderBinarySpecializationConstant> specialization_constants; - - uint32_t stages_processed = 0; - - for (int i = 0; i < p_spirv.size(); i++) { - if (p_spirv[i].shader_stage == SHADER_STAGE_COMPUTE) { - binary_data.is_compute = true; - ERR_FAIL_COND_V_MSG(p_spirv.size() != 1, Vector<uint8_t>(), - "Compute shaders can only receive one stage, dedicated to compute."); - } - ERR_FAIL_COND_V_MSG(stages_processed & (1 << p_spirv[i].shader_stage), Vector<uint8_t>(), - "Stage " + String(shader_stage_names[p_spirv[i].shader_stage]) + " submitted more than once."); - - { - SpvReflectShaderModule module; - const uint8_t *spirv = p_spirv[i].spir_v.ptr(); - SpvReflectResult result = spvReflectCreateShaderModule(p_spirv[i].spir_v.size(), spirv, &module); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed parsing shader."); - - if (binary_data.is_compute) { - binary_data.compute_local_size[0] = module.entry_points->local_size.x; - binary_data.compute_local_size[1] = module.entry_points->local_size.y; - binary_data.compute_local_size[2] = module.entry_points->local_size.z; - } - uint32_t binding_count = 0; - result = spvReflectEnumerateDescriptorBindings(&module, &binding_count, nullptr); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating descriptor bindings."); - - uint32_t stage = p_spirv[i].shader_stage; - - if (binding_count > 0) { - //Parse bindings - - Vector<SpvReflectDescriptorBinding *> bindings; - bindings.resize(binding_count); - result = spvReflectEnumerateDescriptorBindings(&module, &binding_count, bindings.ptrw()); - - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed getting descriptor bindings."); - - for (uint32_t j = 0; j < binding_count; j++) { - const SpvReflectDescriptorBinding &binding = *bindings[j]; - - RenderingDeviceVulkanShaderBinaryDataBinding info; - - bool need_array_dimensions = false; - bool need_block_size = false; - bool may_be_writable = false; - - switch (binding.descriptor_type) { - case SPV_REFLECT_DESCRIPTOR_TYPE_SAMPLER: { - info.type = UNIFORM_TYPE_SAMPLER; - need_array_dimensions = true; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER: { - info.type = UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; - need_array_dimensions = true; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_SAMPLED_IMAGE: { - info.type = UNIFORM_TYPE_TEXTURE; - need_array_dimensions = true; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_IMAGE: { - info.type = UNIFORM_TYPE_IMAGE; - need_array_dimensions = true; - may_be_writable = true; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER: { - info.type = UNIFORM_TYPE_TEXTURE_BUFFER; - need_array_dimensions = true; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER: { - info.type = UNIFORM_TYPE_IMAGE_BUFFER; - need_array_dimensions = true; - may_be_writable = true; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_UNIFORM_BUFFER: { - info.type = UNIFORM_TYPE_UNIFORM_BUFFER; - need_block_size = true; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_BUFFER: { - info.type = UNIFORM_TYPE_STORAGE_BUFFER; - need_block_size = true; - may_be_writable = true; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC: { - ERR_PRINT("Dynamic uniform buffer not supported."); - continue; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC: { - ERR_PRINT("Dynamic storage buffer not supported."); - continue; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_INPUT_ATTACHMENT: { - info.type = UNIFORM_TYPE_INPUT_ATTACHMENT; - need_array_dimensions = true; - } break; - case SPV_REFLECT_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR: { - ERR_PRINT("Acceleration structure not supported."); - continue; - } break; - } - - if (need_array_dimensions) { - if (binding.array.dims_count == 0) { - info.length = 1; - } else { - for (uint32_t k = 0; k < binding.array.dims_count; k++) { - if (k == 0) { - info.length = binding.array.dims[0]; - } else { - info.length *= binding.array.dims[k]; - } - } - } - - } else if (need_block_size) { - info.length = binding.block.size; - } else { - info.length = 0; - } - - if (may_be_writable) { - info.writable = !(binding.type_description->decoration_flags & SPV_REFLECT_DECORATION_NON_WRITABLE) && !(binding.block.decoration_flags & SPV_REFLECT_DECORATION_NON_WRITABLE); - } else { - info.writable = false; - } - - info.binding = binding.binding; - uint32_t set = binding.set; - - ERR_FAIL_COND_V_MSG(set >= MAX_UNIFORM_SETS, Vector<uint8_t>(), - "On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' uses a set (" + itos(set) + ") index larger than what is supported (" + itos(MAX_UNIFORM_SETS) + ")."); - - ERR_FAIL_COND_V_MSG(set >= limits.maxBoundDescriptorSets, Vector<uint8_t>(), - "On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' uses a set (" + itos(set) + ") index larger than what is supported by the hardware (" + itos(limits.maxBoundDescriptorSets) + ")."); - - if (set < (uint32_t)uniform_info.size()) { - //check if this already exists - bool exists = false; - for (int k = 0; k < uniform_info[set].size(); k++) { - if (uniform_info[set][k].binding == (uint32_t)info.binding) { - //already exists, verify that it's the same type - ERR_FAIL_COND_V_MSG(uniform_info[set][k].type != info.type, Vector<uint8_t>(), - "On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(info.binding) + " with different uniform type."); - - //also, verify that it's the same size - ERR_FAIL_COND_V_MSG(uniform_info[set][k].length != info.length, Vector<uint8_t>(), - "On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(info.binding) + " with different uniform size."); - - //also, verify that it has the same writability - ERR_FAIL_COND_V_MSG(uniform_info[set][k].writable != info.writable, Vector<uint8_t>(), - "On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(info.binding) + " with different writability."); - - //just append stage mask and return - uniform_info.write[set].write[k].stages |= 1 << stage; - exists = true; - break; - } - } - - if (exists) { - continue; //merged - } - } - - info.stages = 1 << stage; - - if (set >= (uint32_t)uniform_info.size()) { - uniform_info.resize(set + 1); - } - - uniform_info.write[set].push_back(info); - } - } - - { - //specialization constants - - uint32_t sc_count = 0; - result = spvReflectEnumerateSpecializationConstants(&module, &sc_count, nullptr); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating specialization constants."); - - if (sc_count) { - Vector<SpvReflectSpecializationConstant *> spec_constants; - spec_constants.resize(sc_count); - - result = spvReflectEnumerateSpecializationConstants(&module, &sc_count, spec_constants.ptrw()); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining specialization constants."); - - for (uint32_t j = 0; j < sc_count; j++) { - int32_t existing = -1; - RenderingDeviceVulkanShaderBinarySpecializationConstant sconst; - SpvReflectSpecializationConstant *spc = spec_constants[j]; - - sconst.constant_id = spc->constant_id; - sconst.int_value = 0.0; // clear previous value JIC - switch (spc->constant_type) { - case SPV_REFLECT_SPECIALIZATION_CONSTANT_BOOL: { - sconst.type = PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; - sconst.bool_value = spc->default_value.int_bool_value != 0; - } break; - case SPV_REFLECT_SPECIALIZATION_CONSTANT_INT: { - sconst.type = PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT; - sconst.int_value = spc->default_value.int_bool_value; - } break; - case SPV_REFLECT_SPECIALIZATION_CONSTANT_FLOAT: { - sconst.type = PIPELINE_SPECIALIZATION_CONSTANT_TYPE_FLOAT; - sconst.float_value = spc->default_value.float_value; - } break; - } - sconst.stage_flags = 1 << p_spirv[i].shader_stage; - - for (int k = 0; k < specialization_constants.size(); k++) { - if (specialization_constants[k].constant_id == sconst.constant_id) { - ERR_FAIL_COND_V_MSG(specialization_constants[k].type != sconst.type, Vector<uint8_t>(), "More than one specialization constant used for id (" + itos(sconst.constant_id) + "), but their types differ."); - ERR_FAIL_COND_V_MSG(specialization_constants[k].int_value != sconst.int_value, Vector<uint8_t>(), "More than one specialization constant used for id (" + itos(sconst.constant_id) + "), but their default values differ."); - existing = k; - break; - } - } - - if (existing > 0) { - specialization_constants.write[existing].stage_flags |= sconst.stage_flags; - } else { - specialization_constants.push_back(sconst); - } - } - } - } - - if (stage == SHADER_STAGE_VERTEX) { - uint32_t iv_count = 0; - result = spvReflectEnumerateInputVariables(&module, &iv_count, nullptr); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating input variables."); - - if (iv_count) { - Vector<SpvReflectInterfaceVariable *> input_vars; - input_vars.resize(iv_count); - - result = spvReflectEnumerateInputVariables(&module, &iv_count, input_vars.ptrw()); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining input variables."); - - for (uint32_t j = 0; j < iv_count; j++) { - if (input_vars[j] && input_vars[j]->decoration_flags == 0) { //regular input - binary_data.vertex_input_mask |= (1 << uint32_t(input_vars[j]->location)); - } - } - } - } - - if (stage == SHADER_STAGE_FRAGMENT) { - uint32_t ov_count = 0; - result = spvReflectEnumerateOutputVariables(&module, &ov_count, nullptr); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating output variables."); - - if (ov_count) { - Vector<SpvReflectInterfaceVariable *> output_vars; - output_vars.resize(ov_count); + SpirvReflectionData spirv_data; + if (_reflect_spirv(p_spirv, spirv_data) != OK) { + return Vector<uint8_t>(); + } - result = spvReflectEnumerateOutputVariables(&module, &ov_count, output_vars.ptrw()); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining output variables."); + ERR_FAIL_COND_V_MSG((uint32_t)spirv_data.uniforms.size() > limits.maxBoundDescriptorSets, Vector<uint8_t>(), + "Number of uniform sets is larger than what is supported by the hardware (" + itos(limits.maxBoundDescriptorSets) + ")."); - for (uint32_t j = 0; j < ov_count; j++) { - const SpvReflectInterfaceVariable *refvar = output_vars[j]; - if (refvar != nullptr && refvar->built_in != SpvBuiltInFragDepth) { - binary_data.fragment_outputs |= 1 << refvar->location; - } - } - } + // Collect reflection data into binary data. + RenderingDeviceVulkanShaderBinaryData binary_data; + Vector<Vector<RenderingDeviceVulkanShaderBinaryDataBinding>> uniform_info; // Set bindings. + Vector<RenderingDeviceVulkanShaderBinarySpecializationConstant> specialization_constants; + { + binary_data.vertex_input_mask = spirv_data.vertex_input_mask; + binary_data.fragment_output_mask = spirv_data.fragment_output_mask; + binary_data.specialization_constants_count = spirv_data.specialization_constants.size(); + binary_data.is_compute = spirv_data.is_compute; + binary_data.compute_local_size[0] = spirv_data.compute_local_size[0]; + binary_data.compute_local_size[1] = spirv_data.compute_local_size[1]; + binary_data.compute_local_size[2] = spirv_data.compute_local_size[2]; + binary_data.set_count = spirv_data.uniforms.size(); + binary_data.push_constant_size = spirv_data.push_constant_size; + for (uint32_t i = 0; i < SHADER_STAGE_MAX; i++) { + if (spirv_data.push_constant_stages_mask.has_flag((ShaderStage)(1 << i))) { + binary_data.push_constant_vk_stages_mask |= shader_stage_masks[i]; } + } - uint32_t pc_count = 0; - result = spvReflectEnumeratePushConstantBlocks(&module, &pc_count, nullptr); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating push constants."); - - if (pc_count) { - ERR_FAIL_COND_V_MSG(pc_count > 1, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "': Only one push constant is supported, which should be the same across shader stages."); - - Vector<SpvReflectBlockVariable *> pconstants; - pconstants.resize(pc_count); - result = spvReflectEnumeratePushConstantBlocks(&module, &pc_count, pconstants.ptrw()); - ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining push constants."); -#if 0 - if (pconstants[0] == nullptr) { - Ref<FileAccess> f = FileAccess::open("res://popo.spv", FileAccess::WRITE); - f->store_buffer((const uint8_t *)&SpirV[0], SpirV.size() * sizeof(uint32_t)); - } -#endif - - ERR_FAIL_COND_V_MSG(binary_data.push_constant_size && binary_data.push_constant_size != pconstants[0]->size, Vector<uint8_t>(), - "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "': Push constant block must be the same across shader stages."); - - binary_data.push_constant_size = pconstants[0]->size; - binary_data.push_constants_vk_stage |= shader_stage_masks[stage]; - - //print_line("Stage: " + String(shader_stage_names[stage]) + " push constant of size=" + itos(push_constant.push_constant_size)); + for (const Vector<SpirvReflectionData::Uniform> &spirv_set : spirv_data.uniforms) { + Vector<RenderingDeviceVulkanShaderBinaryDataBinding> set_bindings; + for (const SpirvReflectionData::Uniform &spirv_uniform : spirv_set) { + RenderingDeviceVulkanShaderBinaryDataBinding binding{}; + binding.type = (uint32_t)spirv_uniform.type; + binding.binding = spirv_uniform.binding; + binding.stages = (uint32_t)spirv_uniform.stages_mask; + binding.length = spirv_uniform.length; + binding.writable = (uint32_t)spirv_uniform.writable; + set_bindings.push_back(binding); } - - // Destroy the reflection data when no longer required. - spvReflectDestroyShaderModule(&module); + uniform_info.push_back(set_bindings); } - stages_processed |= (1 << p_spirv[i].shader_stage); + for (const SpirvReflectionData::SpecializationConstant &spirv_sc : spirv_data.specialization_constants) { + RenderingDeviceVulkanShaderBinarySpecializationConstant spec_constant{}; + spec_constant.type = (uint32_t)spirv_sc.type; + spec_constant.constant_id = spirv_sc.constant_id; + spec_constant.int_value = spirv_sc.int_value; + spec_constant.stage_flags = (uint32_t)spirv_sc.stages_mask; + specialization_constants.push_back(spec_constant); + } } Vector<Vector<uint8_t>> compressed_stages; Vector<uint32_t> smolv_size; - Vector<uint32_t> zstd_size; //if 0, stdno t used + Vector<uint32_t> zstd_size; // If 0, zstd not used. uint32_t stages_binary_size = 0; @@ -5061,7 +4675,7 @@ Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Ve ERR_FAIL_V_MSG(Vector<uint8_t>(), "Error compressing shader stage :" + String(shader_stage_names[p_spirv[i].shader_stage])); } else { smolv_size.push_back(smolv.size()); - { //zstd + { // zstd. Vector<uint8_t> zstd; zstd.resize(Compression::get_max_compressed_buffer_size(smolv.size(), Compression::MODE_ZSTD)); int dst_size = Compression::compress(zstd.ptrw(), &smolv[0], smolv.size(), Compression::MODE_ZSTD); @@ -5074,7 +4688,7 @@ Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Ve Vector<uint8_t> smv; smv.resize(smolv.size()); memcpy(smv.ptrw(), &smolv[0], smolv.size()); - zstd_size.push_back(0); //not using zstd + zstd_size.push_back(0); // Not using zstd. compressed_stages.push_back(smv); } } @@ -5086,7 +4700,7 @@ Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Ve stages_binary_size += s; } - binary_data.specialization_constant_count = specialization_constants.size(); + binary_data.specialization_constants_count = specialization_constants.size(); binary_data.set_count = uniform_info.size(); binary_data.stage_count = p_spirv.size(); @@ -5094,12 +4708,12 @@ Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Ve binary_data.shader_name_len = shader_name_utf.length(); - uint32_t total_size = sizeof(uint32_t) * 3; //header + version + main datasize; + uint32_t total_size = sizeof(uint32_t) * 3; // Header + version + main datasize;. total_size += sizeof(RenderingDeviceVulkanShaderBinaryData); total_size += binary_data.shader_name_len; - if ((binary_data.shader_name_len % 4) != 0) { //alignment rules are really strange + if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange. total_size += 4 - (binary_data.shader_name_len % 4); } @@ -5110,7 +4724,7 @@ Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Ve total_size += sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) * specialization_constants.size(); - total_size += compressed_stages.size() * sizeof(uint32_t) * 3; //sizes + total_size += compressed_stages.size() * sizeof(uint32_t) * 3; // Sizes. total_size += stages_binary_size; Vector<uint8_t> ret; @@ -5119,9 +4733,9 @@ Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Ve uint32_t offset = 0; uint8_t *binptr = ret.ptrw(); binptr[0] = 'G'; - binptr[1] = 'V'; + binptr[1] = 'S'; binptr[2] = 'B'; - binptr[3] = 'D'; //godot vulkan binary data + binptr[3] = 'D'; // Godot Shader Binary Data. offset += 4; encode_uint32(SHADER_BINARY_VERSION, binptr + offset); offset += sizeof(uint32_t); @@ -5132,7 +4746,7 @@ Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Ve memcpy(binptr + offset, shader_name_utf.ptr(), binary_data.shader_name_len); offset += binary_data.shader_name_len; - if ((binary_data.shader_name_len % 4) != 0) { //alignment rules are really strange + if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange. offset += 4 - (binary_data.shader_name_len % 4); } @@ -5180,9 +4794,9 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ uint32_t binsize = p_shader_binary.size(); uint32_t read_offset = 0; - //consistency check + // Consistency check. ERR_FAIL_COND_V(binsize < sizeof(uint32_t) * 3 + sizeof(RenderingDeviceVulkanShaderBinaryData), RID()); - ERR_FAIL_COND_V(binptr[0] != 'G' || binptr[1] != 'V' || binptr[2] != 'B' || binptr[3] != 'D', RID()); + ERR_FAIL_COND_V(binptr[0] != 'G' || binptr[1] != 'S' || binptr[2] != 'B' || binptr[3] != 'D', RID()); uint32_t bin_version = decode_uint32(binptr + 4); ERR_FAIL_COND_V(bin_version != SHADER_BINARY_VERSION, RID()); @@ -5192,12 +4806,12 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ const RenderingDeviceVulkanShaderBinaryData &binary_data = *(reinterpret_cast<const RenderingDeviceVulkanShaderBinaryData *>(binptr + 12)); Shader::PushConstant push_constant; - push_constant.push_constant_size = binary_data.push_constant_size; - push_constant.push_constants_vk_stage = binary_data.push_constants_vk_stage; + push_constant.size = binary_data.push_constant_size; + push_constant.vk_stages_mask = binary_data.push_constant_vk_stages_mask; uint32_t vertex_input_mask = binary_data.vertex_input_mask; - uint32_t fragment_outputs = binary_data.fragment_outputs; + uint32_t fragment_output_mask = binary_data.fragment_output_mask; bool is_compute = binary_data.is_compute; @@ -5210,7 +4824,7 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ if (binary_data.shader_name_len) { name.parse_utf8((const char *)(binptr + read_offset), binary_data.shader_name_len); read_offset += binary_data.shader_name_len; - if ((binary_data.shader_name_len % 4) != 0) { //alignment rules are really strange + if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange. read_offset += 4 - (binary_data.shader_name_len % 4); } } @@ -5293,11 +4907,11 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ read_offset += set_size; } - ERR_FAIL_COND_V(read_offset + binary_data.specialization_constant_count * sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) >= binsize, RID()); + ERR_FAIL_COND_V(read_offset + binary_data.specialization_constants_count * sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) >= binsize, RID()); Vector<Shader::SpecializationConstant> specialization_constants; - for (uint32_t i = 0; i < binary_data.specialization_constant_count; i++) { + for (uint32_t i = 0; i < binary_data.specialization_constants_count; i++) { const RenderingDeviceVulkanShaderBinarySpecializationConstant &src_sc = *(reinterpret_cast<const RenderingDeviceVulkanShaderBinarySpecializationConstant *>(binptr + read_offset)); Shader::SpecializationConstant sc; sc.constant.int_value = src_sc.int_value; @@ -5327,7 +4941,7 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ const uint8_t *src_smolv = nullptr; if (zstd_size > 0) { - //decompress to smolv + // Decompress to smolv. smolv.resize(smolv_size); int dec_smolv_size = Compression::decompress(smolv.ptrw(), smolv.size(), binptr + read_offset, zstd_size, Compression::MODE_ZSTD); ERR_FAIL_COND_V(dec_smolv_size != (int32_t)smolv_size, RID()); @@ -5356,14 +4970,14 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ ERR_FAIL_COND_V(read_offset != binsize, RID()); - //all good, let's create modules + // All good, let's create modules. _THREAD_SAFE_METHOD_ Shader shader; shader.vertex_input_mask = vertex_input_mask; - shader.fragment_output_mask = fragment_outputs; + shader.fragment_output_mask = fragment_output_mask; shader.push_constant = push_constant; shader.is_compute = is_compute; shader.compute_local_size[0] = compute_local_size[0]; @@ -5393,30 +5007,22 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ break; } - const VkShaderStageFlagBits shader_stage_bits[SHADER_STAGE_MAX] = { - VK_SHADER_STAGE_VERTEX_BIT, - VK_SHADER_STAGE_FRAGMENT_BIT, - VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT, - VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT, - VK_SHADER_STAGE_COMPUTE_BIT, - }; - VkPipelineShaderStageCreateInfo shader_stage; shader_stage.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; shader_stage.pNext = nullptr; shader_stage.flags = 0; - shader_stage.stage = shader_stage_bits[stage_type[i]]; + shader_stage.stage = shader_stage_masks[stage_type[i]]; shader_stage.module = module; shader_stage.pName = "main"; shader_stage.pSpecializationInfo = nullptr; shader.pipeline_stages.push_back(shader_stage); } - //proceed to create descriptor sets + // Proceed to create descriptor sets. if (success) { for (int i = 0; i < set_bindings.size(); i++) { - //empty ones are fine if they were not used according to spec (binding count will be 0) + // Empty ones are fine if they were not used according to spec (binding count will be 0). VkDescriptorSetLayoutCreateInfo layout_create_info; layout_create_info.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; layout_create_info.pNext = nullptr; @@ -5435,13 +5041,13 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ Shader::Set set; set.descriptor_set_layout = layout; set.uniform_info = uniform_info[i]; - //sort and hash + // Sort and hash. set.uniform_info.sort(); - uint32_t format = 0; //no format, default + uint32_t format = 0; // No format, default. if (set.uniform_info.size()) { - //has data, needs an actual format; + // Has data, needs an actual format. UniformSetFormat usformat; usformat.uniform_info = set.uniform_info; RBMap<UniformSetFormat, uint32_t>::Element *E = uniform_set_format_cache.find(usformat); @@ -5459,7 +5065,7 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ } if (success) { - //create pipeline layout + // Create pipeline layout. VkPipelineLayoutCreateInfo pipeline_layout_create_info; pipeline_layout_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; pipeline_layout_create_info.pNext = nullptr; @@ -5477,10 +5083,10 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ // Needs to be declared in this outer scope, otherwise it may not outlive its assignment // to pipeline_layout_create_info. VkPushConstantRange push_constant_range; - if (push_constant.push_constant_size) { - push_constant_range.stageFlags = push_constant.push_constants_vk_stage; + if (push_constant.size) { + push_constant_range.stageFlags = push_constant.vk_stages_mask; push_constant_range.offset = 0; - push_constant_range.size = push_constant.push_constant_size; + push_constant_range.size = push_constant.size; pipeline_layout_create_info.pushConstantRangeCount = 1; pipeline_layout_create_info.pPushConstantRanges = &push_constant_range; @@ -5498,7 +5104,7 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ } if (!success) { - //clean up if failed + // Clean up if failed. for (int i = 0; i < shader.pipeline_stages.size(); i++) { vkDestroyShaderModule(device, shader.pipeline_stages[i].module, nullptr); } @@ -5510,7 +5116,11 @@ RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_ ERR_FAIL_V_MSG(RID(), error_text); } - return shader_owner.make_rid(shader); + RID id = shader_owner.make_rid(shader); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + return id; } uint32_t RenderingDeviceVulkan::shader_get_vertex_input_attribute_mask(RID p_shader) { @@ -5543,10 +5153,14 @@ RID RenderingDeviceVulkan::uniform_buffer_create(uint32_t p_size_bytes, const Ve _buffer_update(&buffer, 0, r, data_size); _buffer_memory_barrier(buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_UNIFORM_READ_BIT, false); } - return uniform_buffer_owner.make_rid(buffer); + RID id = uniform_buffer_owner.make_rid(buffer); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + return id; } -RID RenderingDeviceVulkan::storage_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data, uint32_t p_usage) { +RID RenderingDeviceVulkan::storage_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data, BitField<StorageBufferUsage> p_usage) { _THREAD_SAFE_METHOD_ ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(), "Creating buffers with data is forbidden during creation of a draw list"); @@ -5557,7 +5171,7 @@ RID RenderingDeviceVulkan::storage_buffer_create(uint32_t p_size_bytes, const Ve Buffer buffer; uint32_t flags = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT; - if (p_usage & STORAGE_BUFFER_USAGE_DISPATCH_INDIRECT) { + if (p_usage.has_flag(STORAGE_BUFFER_USAGE_DISPATCH_INDIRECT)) { flags |= VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT; } Error err = _buffer_allocate(&buffer, p_size_bytes, flags, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0); @@ -5613,8 +5227,12 @@ RID RenderingDeviceVulkan::texture_buffer_create(uint32_t p_size_elements, DataF ERR_FAIL_V_MSG(RID(), "Unable to create buffer view, error " + itos(res) + "."); } - //allocate the view - return texture_buffer_owner.make_rid(texture_buffer); + // Allocate the view. + RID id = texture_buffer_owner.make_rid(texture_buffer); +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + return id; } RenderingDeviceVulkan::DescriptorPool *RenderingDeviceVulkan::_descriptor_pool_allocate(const DescriptorPoolKey &p_key) { @@ -5632,17 +5250,17 @@ RenderingDeviceVulkan::DescriptorPool *RenderingDeviceVulkan::_descriptor_pool_a } if (!pool) { - //create a new one + // Create a new one. pool = memnew(DescriptorPool); pool->usage = 0; VkDescriptorPoolCreateInfo descriptor_pool_create_info; descriptor_pool_create_info.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO; descriptor_pool_create_info.pNext = nullptr; - descriptor_pool_create_info.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT; // can't think how somebody may NOT need this flag.. + descriptor_pool_create_info.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT; // Can't think how somebody may NOT need this flag. descriptor_pool_create_info.maxSets = max_descriptors_per_pool; Vector<VkDescriptorPoolSize> sizes; - //here comes more vulkan API strangeness + // Here comes more vulkan API strangeness. if (p_key.uniform_type[UNIFORM_TYPE_SAMPLER]) { VkDescriptorPoolSize s; @@ -5742,7 +5360,7 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, ERR_FAIL_COND_V_MSG(p_shader_set >= (uint32_t)shader->sets.size() || shader->sets[p_shader_set].uniform_info.size() == 0, RID(), "Desired set (" + itos(p_shader_set) + ") not used by shader."); - //see that all sets in shader are satisfied + // See that all sets in shader are satisfied. const Shader::Set &set = shader->sets[p_shader_set]; @@ -5755,11 +5373,11 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, Vector<VkWriteDescriptorSet> writes; DescriptorPoolKey pool_key; - //to keep them alive until update call + // To keep them alive until update call. List<Vector<VkDescriptorBufferInfo>> buffer_infos; List<Vector<VkBufferView>> buffer_views; List<Vector<VkDescriptorImageInfo>> image_infos; - //used for verification to make sure a uniform set does not use a framebuffer bound texture + // Used for verification to make sure a uniform set does not use a framebuffer bound texture. LocalVector<UniformSet::AttachableTexture> attachable_textures; Vector<Texture *> mutable_sampled_textures; Vector<Texture *> mutable_storage_textures; @@ -5780,14 +5398,14 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, ERR_FAIL_COND_V_MSG(uniform.uniform_type != set_uniform.type, RID(), "Mismatch uniform type for binding (" + itos(set_uniform.binding) + "), set (" + itos(p_shader_set) + "). Expected '" + shader_uniform_names[set_uniform.type] + "', supplied: '" + shader_uniform_names[uniform.uniform_type] + "'."); - VkWriteDescriptorSet write; //common header + VkWriteDescriptorSet write; // Common header. write.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET; write.pNext = nullptr; - write.dstSet = VK_NULL_HANDLE; //will assign afterwards when everything is valid + write.dstSet = VK_NULL_HANDLE; // Will assign afterwards when everything is valid. write.dstBinding = set_uniform.binding; write.dstArrayElement = 0; write.descriptorCount = 0; - write.descriptorType = VK_DESCRIPTOR_TYPE_MAX_ENUM; //Invalid value. + write.descriptorType = VK_DESCRIPTOR_TYPE_MAX_ENUM; // Invalid value. write.pImageInfo = nullptr; write.pBufferInfo = nullptr; write.pTexelBufferView = nullptr; @@ -5860,13 +5478,11 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, } if (texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT) { - //can also be used as storage, add to mutable sampled + // Can also be used as storage, add to mutable sampled. mutable_sampled_textures.push_back(texture); } - if (texture->owner.is_valid()) { - texture = texture_owner.get_or_null(texture->owner); - ERR_FAIL_COND_V(!texture, RID()); //bug, should never happen - } + + DEV_ASSERT(!texture->owner.is_valid() || texture_owner.get_or_null(texture->owner)); img_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; @@ -5913,14 +5529,11 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, } if (texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT) { - //can also be used as storage, add to mutable sampled + // Can also be used as storage, add to mutable sampled. mutable_sampled_textures.push_back(texture); } - if (texture->owner.is_valid()) { - texture = texture_owner.get_or_null(texture->owner); - ERR_FAIL_COND_V(!texture, RID()); //bug, should never happen - } + DEV_ASSERT(!texture->owner.is_valid() || texture_owner.get_or_null(texture->owner)); img_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; @@ -5961,14 +5574,11 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, img_info.imageView = texture->view; if (texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT) { - //can also be used as storage, add to mutable sampled + // Can also be used as storage, add to mutable sampled. mutable_storage_textures.push_back(texture); } - if (texture->owner.is_valid()) { - texture = texture_owner.get_or_null(texture->owner); - ERR_FAIL_COND_V(!texture, RID()); //bug, should never happen - } + DEV_ASSERT(!texture->owner.is_valid() || texture_owner.get_or_null(texture->owner)); img_info.imageLayout = VK_IMAGE_LAYOUT_GENERAL; @@ -6057,7 +5667,7 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, type_size = uniform.get_id_count() / 2; } break; case UNIFORM_TYPE_IMAGE_BUFFER: { - //todo + // Todo. } break; case UNIFORM_TYPE_UNIFORM_BUFFER: { @@ -6093,7 +5703,7 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, } ERR_FAIL_COND_V_MSG(!buffer, RID(), "Storage buffer supplied (binding: " + itos(uniform.binding) + ") is invalid."); - //if 0, then it's sized on link time + // If 0, then it's sized on link time. ERR_FAIL_COND_V_MSG(set_uniform.length > 0 && buffer->size != (uint32_t)set_uniform.length, RID(), "Storage buffer supplied (binding: " + itos(uniform.binding) + ") size (" + itos(buffer->size) + " does not match size of shader uniform: (" + itos(set_uniform.length) + ")."); @@ -6130,10 +5740,7 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, img_info.sampler = VK_NULL_HANDLE; img_info.imageView = texture->view; - if (texture->owner.is_valid()) { - texture = texture_owner.get_or_null(texture->owner); - ERR_FAIL_COND_V(!texture, RID()); //bug, should never happen - } + DEV_ASSERT(!texture->owner.is_valid() || texture_owner.get_or_null(texture->owner)); img_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; @@ -6160,7 +5767,7 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, pool_key.uniform_type[set_uniform.type] += type_size; } - //need a descriptor pool + // Need a descriptor pool. DescriptorPool *pool = _descriptor_pool_allocate(pool_key); ERR_FAIL_COND_V(!pool, RID()); @@ -6177,7 +5784,7 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, VkResult res = vkAllocateDescriptorSets(device, &descriptor_set_allocate_info, &descriptor_set); if (res) { - _descriptor_pool_free(pool_key, pool); // meh + _descriptor_pool_free(pool_key, pool); // Meh. ERR_FAIL_V_MSG(RID(), "Cannot allocate descriptor sets, error " + itos(res) + "."); } @@ -6193,7 +5800,10 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, uniform_set.shader_id = p_shader; RID id = uniform_set_owner.make_rid(uniform_set); - //add dependencies +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + // Add dependencies. _add_dependency(id, p_shader); for (uint32_t i = 0; i < uniform_count; i++) { const Uniform &uniform = uniforms[i]; @@ -6203,7 +5813,7 @@ RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, } } - //write the contents + // Write the contents. if (writes.size()) { for (int i = 0; i < writes.size(); i++) { writes.write[i].dstSet = descriptor_set; @@ -6218,14 +5828,14 @@ bool RenderingDeviceVulkan::uniform_set_is_valid(RID p_uniform_set) { return uniform_set_owner.owns(p_uniform_set); } -void RenderingDeviceVulkan::uniform_set_set_invalidation_callback(RID p_uniform_set, UniformSetInvalidatedCallback p_callback, void *p_userdata) { +void RenderingDeviceVulkan::uniform_set_set_invalidation_callback(RID p_uniform_set, InvalidationCallback p_callback, void *p_userdata) { UniformSet *us = uniform_set_owner.get_or_null(p_uniform_set); ERR_FAIL_COND(!us); us->invalidated_callback = p_callback; us->invalidated_callback_userdata = p_userdata; } -Error RenderingDeviceVulkan::buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const void *p_data, uint32_t p_post_barrier) { +Error RenderingDeviceVulkan::buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const void *p_data, BitField<BarrierMask> p_post_barrier) { _THREAD_SAFE_METHOD_ ERR_FAIL_COND_V_MSG(draw_list, ERR_INVALID_PARAMETER, @@ -6235,8 +5845,8 @@ Error RenderingDeviceVulkan::buffer_update(RID p_buffer, uint32_t p_offset, uint VkPipelineStageFlags dst_stage_mask = 0; VkAccessFlags dst_access = 0; - if (p_post_barrier & BARRIER_MASK_TRANSFER) { - // Protect subsequent updates... + if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) { + // Protect subsequent updates. dst_stage_mask = VK_PIPELINE_STAGE_TRANSFER_BIT; dst_access = VK_ACCESS_TRANSFER_WRITE_BIT; } @@ -6248,7 +5858,7 @@ Error RenderingDeviceVulkan::buffer_update(RID p_buffer, uint32_t p_offset, uint ERR_FAIL_COND_V_MSG(p_offset + p_size > buffer->size, ERR_INVALID_PARAMETER, "Attempted to write buffer (" + itos((p_offset + p_size) - buffer->size) + " bytes) past the end."); - // no barrier should be needed here + // No barrier should be needed here. // _buffer_memory_barrier(buffer->buffer, p_offset, p_size, dst_stage_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_access, VK_ACCESS_TRANSFER_WRITE_BIT, true); Error err = _buffer_update(buffer, p_offset, (uint8_t *)p_data, p_size, p_post_barrier); @@ -6271,7 +5881,7 @@ Error RenderingDeviceVulkan::buffer_update(RID p_buffer, uint32_t p_offset, uint return err; } -Error RenderingDeviceVulkan::buffer_clear(RID p_buffer, uint32_t p_offset, uint32_t p_size, uint32_t p_post_barrier) { +Error RenderingDeviceVulkan::buffer_clear(RID p_buffer, uint32_t p_offset, uint32_t p_size, BitField<BarrierMask> p_post_barrier) { _THREAD_SAFE_METHOD_ ERR_FAIL_COND_V_MSG((p_size % 4) != 0, ERR_INVALID_PARAMETER, @@ -6283,8 +5893,8 @@ Error RenderingDeviceVulkan::buffer_clear(RID p_buffer, uint32_t p_offset, uint3 VkPipelineStageFlags dst_stage_mask = 0; VkAccessFlags dst_access = 0; - if (p_post_barrier & BARRIER_MASK_TRANSFER) { - // Protect subsequent updates... + if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) { + // Protect subsequent updates. dst_stage_mask = VK_PIPELINE_STAGE_TRANSFER_BIT; dst_access = VK_ACCESS_TRANSFER_WRITE_BIT; } @@ -6297,7 +5907,7 @@ Error RenderingDeviceVulkan::buffer_clear(RID p_buffer, uint32_t p_offset, uint3 ERR_FAIL_COND_V_MSG(p_offset + p_size > buffer->size, ERR_INVALID_PARAMETER, "Attempted to write buffer (" + itos((p_offset + p_size) - buffer->size) + " bytes) past the end."); - // should not be needed + // Should not be needed. // _buffer_memory_barrier(buffer->buffer, p_offset, p_size, dst_stage_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_access, VK_ACCESS_TRANSFER_WRITE_BIT, p_post_barrier); vkCmdFillBuffer(frames[frame].draw_command_buffer, buffer->buffer, p_offset, p_size, 0); @@ -6318,17 +5928,17 @@ Error RenderingDeviceVulkan::buffer_clear(RID p_buffer, uint32_t p_offset, uint3 Vector<uint8_t> RenderingDeviceVulkan::buffer_get_data(RID p_buffer) { _THREAD_SAFE_METHOD_ - // It could be this buffer was just created + // It could be this buffer was just created. VkPipelineShaderStageCreateFlags src_stage_mask = VK_PIPELINE_STAGE_TRANSFER_BIT; VkAccessFlags src_access_mask = VK_ACCESS_TRANSFER_WRITE_BIT; - // Get the vulkan buffer and the potential stage/access possible - Buffer *buffer = _get_buffer_from_owner(p_buffer, src_stage_mask, src_access_mask, BARRIER_MASK_ALL); + // Get the vulkan buffer and the potential stage/access possible. + Buffer *buffer = _get_buffer_from_owner(p_buffer, src_stage_mask, src_access_mask, BARRIER_MASK_ALL_BARRIERS); if (!buffer) { ERR_FAIL_V_MSG(Vector<uint8_t>(), "Buffer is either invalid or this type of buffer can't be retrieved. Only Index and Vertex buffers allow retrieving."); } // Make sure no one is using the buffer -- the "false" gets us to the same command buffer as below. - _buffer_memory_barrier(buffer->buffer, 0, buffer->size, src_stage_mask, src_access_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_READ_BIT, false); + _buffer_memory_barrier(buffer->buffer, 0, buffer->size, src_stage_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, src_access_mask, VK_ACCESS_TRANSFER_READ_BIT, false); VkCommandBuffer command_buffer = frames[frame].setup_command_buffer; @@ -6338,8 +5948,8 @@ Vector<uint8_t> RenderingDeviceVulkan::buffer_get_data(RID p_buffer) { region.srcOffset = 0; region.dstOffset = 0; region.size = buffer->size; - vkCmdCopyBuffer(command_buffer, buffer->buffer, tmp_buffer.buffer, 1, ®ion); //dst buffer is in CPU, but I wonder if src buffer needs a barrier for this.. - //flush everything so memory can be safely mapped + vkCmdCopyBuffer(command_buffer, buffer->buffer, tmp_buffer.buffer, 1, ®ion); // Dst buffer is in CPU, but I wonder if src buffer needs a barrier for this. + // Flush everything so memory can be safely mapped. _flush(true); void *buffer_mem; @@ -6364,10 +5974,10 @@ Vector<uint8_t> RenderingDeviceVulkan::buffer_get_data(RID p_buffer) { /**** RENDER PIPELINE ****/ /*************************/ -RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferFormatID p_framebuffer_format, VertexFormatID p_vertex_format, RenderPrimitive p_render_primitive, const PipelineRasterizationState &p_rasterization_state, const PipelineMultisampleState &p_multisample_state, const PipelineDepthStencilState &p_depth_stencil_state, const PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags, uint32_t p_for_render_pass, const Vector<PipelineSpecializationConstant> &p_specialization_constants) { +RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferFormatID p_framebuffer_format, VertexFormatID p_vertex_format, RenderPrimitive p_render_primitive, const PipelineRasterizationState &p_rasterization_state, const PipelineMultisampleState &p_multisample_state, const PipelineDepthStencilState &p_depth_stencil_state, const PipelineColorBlendState &p_blend_state, BitField<PipelineDynamicStateFlags> p_dynamic_state_flags, uint32_t p_for_render_pass, const Vector<PipelineSpecializationConstant> &p_specialization_constants) { _THREAD_SAFE_METHOD_ - //needs a shader + // Needs a shader. Shader *shader = shader_owner.get_or_null(p_shader); ERR_FAIL_COND_V(!shader, RID()); @@ -6375,13 +5985,13 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma "Compute shaders can't be used in render pipelines"); if (p_framebuffer_format == INVALID_ID) { - //if nothing provided, use an empty one (no attachments) + // If nothing provided, use an empty one (no attachments). p_framebuffer_format = framebuffer_format_create(Vector<AttachmentFormat>()); } ERR_FAIL_COND_V(!framebuffer_formats.has(p_framebuffer_format), RID()); const FramebufferFormat &fb_format = framebuffer_formats[p_framebuffer_format]; - { //validate shader vs framebuffer + { // Validate shader vs framebuffer. ERR_FAIL_COND_V_MSG(p_for_render_pass >= uint32_t(fb_format.E->key().passes.size()), RID(), "Render pass requested for pipeline creation (" + itos(p_for_render_pass) + ") is out of bounds"); const FramebufferPass &pass = fb_format.E->key().passes[p_for_render_pass]; @@ -6394,17 +6004,17 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma ERR_FAIL_COND_V_MSG(shader->fragment_output_mask != output_mask, RID(), "Mismatch fragment shader output mask (" + itos(shader->fragment_output_mask) + ") and framebuffer color output mask (" + itos(output_mask) + ") when binding both in render pipeline."); } - //vertex + // Vertex. VkPipelineVertexInputStateCreateInfo pipeline_vertex_input_state_create_info; if (p_vertex_format != INVALID_ID) { - //uses vertices, else it does not + // Uses vertices, else it does not. ERR_FAIL_COND_V(!vertex_formats.has(p_vertex_format), RID()); const VertexDescriptionCache &vd = vertex_formats[p_vertex_format]; pipeline_vertex_input_state_create_info = vd.create_info; - //validate with inputs + // Validate with inputs. for (uint32_t i = 0; i < 32; i++) { if (!(shader->vertex_input_mask & (1UL << i))) { continue; @@ -6421,7 +6031,7 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma } } else { - //does not use vertices + // Does not use vertices. pipeline_vertex_input_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO; pipeline_vertex_input_state_create_info.pNext = nullptr; pipeline_vertex_input_state_create_info.flags = 0; @@ -6433,7 +6043,7 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma ERR_FAIL_COND_V_MSG(shader->vertex_input_mask != 0, RID(), "Shader contains vertex inputs, but no vertex input description was provided for pipeline creation."); } - //input assembly + // Input assembly. ERR_FAIL_INDEX_V(p_render_primitive, RENDER_PRIMITIVE_MAX, RID()); @@ -6459,7 +6069,7 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma input_assembly_create_info.topology = topology_list[p_render_primitive]; input_assembly_create_info.primitiveRestartEnable = (p_render_primitive == RENDER_PRIMITIVE_TRIANGLE_STRIPS_WITH_RESTART_INDEX); - //tessellation + // Tessellation. VkPipelineTessellationStateCreateInfo tessellation_create_info; tessellation_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO; tessellation_create_info.pNext = nullptr; @@ -6471,12 +6081,12 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma viewport_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO; viewport_state_create_info.pNext = nullptr; viewport_state_create_info.flags = 0; - viewport_state_create_info.viewportCount = 1; //if VR extensions are supported at some point, this will have to be customizable in the framebuffer format + viewport_state_create_info.viewportCount = 1; // If VR extensions are supported at some point, this will have to be customizable in the framebuffer format. viewport_state_create_info.pViewports = nullptr; viewport_state_create_info.scissorCount = 1; viewport_state_create_info.pScissors = nullptr; - //rasterization + // Rasterization. VkPipelineRasterizationStateCreateInfo rasterization_state_create_info; rasterization_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO; rasterization_state_create_info.pNext = nullptr; @@ -6493,24 +6103,24 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma ERR_FAIL_INDEX_V(p_rasterization_state.cull_mode, 3, RID()); rasterization_state_create_info.cullMode = cull_mode[p_rasterization_state.cull_mode]; rasterization_state_create_info.frontFace = (p_rasterization_state.front_face == POLYGON_FRONT_FACE_CLOCKWISE ? VK_FRONT_FACE_CLOCKWISE : VK_FRONT_FACE_COUNTER_CLOCKWISE); - rasterization_state_create_info.depthBiasEnable = p_rasterization_state.depth_bias_enable; + rasterization_state_create_info.depthBiasEnable = p_rasterization_state.depth_bias_enabled; rasterization_state_create_info.depthBiasConstantFactor = p_rasterization_state.depth_bias_constant_factor; rasterization_state_create_info.depthBiasClamp = p_rasterization_state.depth_bias_clamp; rasterization_state_create_info.depthBiasSlopeFactor = p_rasterization_state.depth_bias_slope_factor; rasterization_state_create_info.lineWidth = p_rasterization_state.line_width; - //multisample + // Multisample. VkPipelineMultisampleStateCreateInfo multisample_state_create_info; multisample_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO; multisample_state_create_info.pNext = nullptr; multisample_state_create_info.flags = 0; - multisample_state_create_info.rasterizationSamples = rasterization_sample_count[p_multisample_state.sample_count]; + multisample_state_create_info.rasterizationSamples = _ensure_supported_sample_count(p_multisample_state.sample_count); multisample_state_create_info.sampleShadingEnable = p_multisample_state.enable_sample_shading; multisample_state_create_info.minSampleShading = p_multisample_state.min_sample_shading; Vector<VkSampleMask> sample_mask; if (p_multisample_state.sample_mask.size()) { - //use sample mask + // Use sample mask. const int rasterization_sample_mask_expected_size[TEXTURE_SAMPLES_MAX] = { 1, 2, 4, 8, 16, 32, 64 }; @@ -6528,7 +6138,7 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma multisample_state_create_info.alphaToCoverageEnable = p_multisample_state.enable_alpha_to_coverage; multisample_state_create_info.alphaToOneEnable = p_multisample_state.enable_alpha_to_one; - //depth stencil + // Depth stencil. VkPipelineDepthStencilStateCreateInfo depth_stencil_state_create_info; depth_stencil_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO; @@ -6568,7 +6178,7 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma depth_stencil_state_create_info.minDepthBounds = p_depth_stencil_state.depth_range_min; depth_stencil_state_create_info.maxDepthBounds = p_depth_stencil_state.depth_range_max; - //blend state + // Blend state. VkPipelineColorBlendStateCreateInfo color_blend_state_create_info; color_blend_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO; color_blend_state_create_info.pNext = nullptr; @@ -6639,42 +6249,42 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma color_blend_state_create_info.blendConstants[2] = p_blend_state.blend_constant.b; color_blend_state_create_info.blendConstants[3] = p_blend_state.blend_constant.a; - //dynamic state + // Dynamic state. VkPipelineDynamicStateCreateInfo dynamic_state_create_info; dynamic_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO; dynamic_state_create_info.pNext = nullptr; dynamic_state_create_info.flags = 0; - Vector<VkDynamicState> dynamic_states; //vulkan is weird.. + Vector<VkDynamicState> dynamic_states; // Vulkan is weird. - dynamic_states.push_back(VK_DYNAMIC_STATE_VIEWPORT); //viewport and scissor are always dynamic + dynamic_states.push_back(VK_DYNAMIC_STATE_VIEWPORT); // Viewport and scissor are always dynamic. dynamic_states.push_back(VK_DYNAMIC_STATE_SCISSOR); - if (p_dynamic_state_flags & DYNAMIC_STATE_LINE_WIDTH) { + if (p_dynamic_state_flags.has_flag(DYNAMIC_STATE_LINE_WIDTH)) { dynamic_states.push_back(VK_DYNAMIC_STATE_LINE_WIDTH); } - if (p_dynamic_state_flags & DYNAMIC_STATE_DEPTH_BIAS) { + if (p_dynamic_state_flags.has_flag(DYNAMIC_STATE_DEPTH_BIAS)) { dynamic_states.push_back(VK_DYNAMIC_STATE_DEPTH_BIAS); } - if (p_dynamic_state_flags & DYNAMIC_STATE_BLEND_CONSTANTS) { + if (p_dynamic_state_flags.has_flag(DYNAMIC_STATE_BLEND_CONSTANTS)) { dynamic_states.push_back(VK_DYNAMIC_STATE_BLEND_CONSTANTS); } - if (p_dynamic_state_flags & DYNAMIC_STATE_DEPTH_BOUNDS) { + if (p_dynamic_state_flags.has_flag(DYNAMIC_STATE_DEPTH_BOUNDS)) { dynamic_states.push_back(VK_DYNAMIC_STATE_DEPTH_BOUNDS); } - if (p_dynamic_state_flags & DYNAMIC_STATE_STENCIL_COMPARE_MASK) { + if (p_dynamic_state_flags.has_flag(DYNAMIC_STATE_STENCIL_COMPARE_MASK)) { dynamic_states.push_back(VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK); } - if (p_dynamic_state_flags & DYNAMIC_STATE_STENCIL_WRITE_MASK) { + if (p_dynamic_state_flags.has_flag(DYNAMIC_STATE_STENCIL_WRITE_MASK)) { dynamic_states.push_back(VK_DYNAMIC_STATE_STENCIL_WRITE_MASK); } - if (p_dynamic_state_flags & DYNAMIC_STATE_STENCIL_REFERENCE) { + if (p_dynamic_state_flags.has_flag(DYNAMIC_STATE_STENCIL_REFERENCE)) { dynamic_states.push_back(VK_DYNAMIC_STATE_STENCIL_REFERENCE); } @@ -6686,19 +6296,19 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma VkPipelineFragmentShadingRateStateCreateInfoKHR vrs_create_info; if (context->get_vrs_capabilities().attachment_vrs_supported) { // If VRS is used, this defines how the different VRS types are combined. - // combinerOps[0] decides how we use the output of pipeline and primitive (drawcall) VRS - // combinerOps[1] decides how we use the output of combinerOps[0] and our attachment VRS + // combinerOps[0] decides how we use the output of pipeline and primitive (drawcall) VRS. + // combinerOps[1] decides how we use the output of combinerOps[0] and our attachment VRS. vrs_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_FRAGMENT_SHADING_RATE_STATE_CREATE_INFO_KHR; vrs_create_info.pNext = nullptr; vrs_create_info.fragmentSize = { 4, 4 }; - vrs_create_info.combinerOps[0] = VK_FRAGMENT_SHADING_RATE_COMBINER_OP_KEEP_KHR; // We don't use pipeline/primitive VRS so this really doesn't matter - vrs_create_info.combinerOps[1] = VK_FRAGMENT_SHADING_RATE_COMBINER_OP_REPLACE_KHR; // always use the outcome of attachment VRS if enabled + vrs_create_info.combinerOps[0] = VK_FRAGMENT_SHADING_RATE_COMBINER_OP_KEEP_KHR; // We don't use pipeline/primitive VRS so this really doesn't matter. + vrs_create_info.combinerOps[1] = VK_FRAGMENT_SHADING_RATE_COMBINER_OP_REPLACE_KHR; // Always use the outcome of attachment VRS if enabled. graphics_pipeline_nextptr = &vrs_create_info; } - //finally, pipeline create info + // Finally, pipeline create info. VkGraphicsPipelineCreateInfo graphics_pipeline_create_info; graphics_pipeline_create_info.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO; @@ -6716,9 +6326,9 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma specialization_info.resize(pipeline_stages.size()); specialization_map_entries.resize(pipeline_stages.size()); for (int i = 0; i < shader->specialization_constants.size(); i++) { - //see if overridden + // See if overridden. const Shader::SpecializationConstant &sc = shader->specialization_constants[i]; - data_ptr[i] = sc.constant.int_value; //just copy the 32 bits + data_ptr[i] = sc.constant.int_value; // Just copy the 32 bits. for (int j = 0; j < p_specialization_constants.size(); j++) { const PipelineSpecializationConstant &psc = p_specialization_constants[j]; @@ -6782,10 +6392,10 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateGraphicsPipelines failed with error " + itos(err) + " for shader '" + shader->name + "'."); pipeline.set_formats = shader->set_formats; - pipeline.push_constant_stages = shader->push_constant.push_constants_vk_stage; + pipeline.push_constant_stages_mask = shader->push_constant.vk_stages_mask; pipeline.pipeline_layout = shader->pipeline_layout; pipeline.shader = p_shader; - pipeline.push_constant_size = shader->push_constant.push_constant_size; + pipeline.push_constant_size = shader->push_constant.size; #ifdef DEBUG_ENABLED pipeline.validation.dynamic_state = p_dynamic_state_flags; @@ -6813,9 +6423,12 @@ RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferForma }; pipeline.validation.primitive_minimum = primitive_minimum[p_render_primitive]; #endif - //create ID to associate with this pipeline + // Create ID to associate with this pipeline. RID id = render_pipeline_owner.make_rid(pipeline); - //now add all the dependencies +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + // Now add all the dependencies. _add_dependency(id, p_shader); return id; } @@ -6832,14 +6445,14 @@ bool RenderingDeviceVulkan::render_pipeline_is_valid(RID p_pipeline) { RID RenderingDeviceVulkan::compute_pipeline_create(RID p_shader, const Vector<PipelineSpecializationConstant> &p_specialization_constants) { _THREAD_SAFE_METHOD_ - //needs a shader + // Needs a shader. Shader *shader = shader_owner.get_or_null(p_shader); ERR_FAIL_COND_V(!shader, RID()); ERR_FAIL_COND_V_MSG(!shader->is_compute, RID(), "Non-compute shaders can't be used in compute pipelines"); - //finally, pipeline create info + // Finally, pipeline create info. VkComputePipelineCreateInfo compute_pipeline_create_info; compute_pipeline_create_info.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO; @@ -6859,9 +6472,9 @@ RID RenderingDeviceVulkan::compute_pipeline_create(RID p_shader, const Vector<Pi specialization_constant_data.resize(shader->specialization_constants.size()); uint32_t *data_ptr = specialization_constant_data.ptrw(); for (int i = 0; i < shader->specialization_constants.size(); i++) { - //see if overridden + // See if overridden. const Shader::SpecializationConstant &sc = shader->specialization_constants[i]; - data_ptr[i] = sc.constant.int_value; //just copy the 32 bits + data_ptr[i] = sc.constant.int_value; // Just copy the 32 bits. for (int j = 0; j < p_specialization_constants.size(); j++) { const PipelineSpecializationConstant &psc = p_specialization_constants[j]; @@ -6894,17 +6507,20 @@ RID RenderingDeviceVulkan::compute_pipeline_create(RID p_shader, const Vector<Pi ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateComputePipelines failed with error " + itos(err) + "."); pipeline.set_formats = shader->set_formats; - pipeline.push_constant_stages = shader->push_constant.push_constants_vk_stage; + pipeline.push_constant_stages_mask = shader->push_constant.vk_stages_mask; pipeline.pipeline_layout = shader->pipeline_layout; pipeline.shader = p_shader; - pipeline.push_constant_size = shader->push_constant.push_constant_size; + pipeline.push_constant_size = shader->push_constant.size; pipeline.local_group_size[0] = shader->compute_local_size[0]; pipeline.local_group_size[1] = shader->compute_local_size[1]; pipeline.local_group_size[2] = shader->compute_local_size[2]; - //create ID to associate with this pipeline + // Create ID to associate with this pipeline. RID id = compute_pipeline_owner.make_rid(pipeline); - //now add all the dependencies +#ifdef DEV_ENABLED + set_resource_name(id, "RID:" + itos(id.get_id())); +#endif + // Now add all the dependencies. _add_dependency(id, p_shader); return id; } @@ -6934,7 +6550,7 @@ RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::screen_get_framebuff _THREAD_SAFE_METHOD_ ERR_FAIL_COND_V_MSG(local_device.is_valid(), INVALID_ID, "Local devices have no screen"); - //very hacky, but not used often per frame so I guess ok + // Very hacky, but not used often per frame so I guess ok. VkFormat vkformat = context->get_screen_format(); DataFormat format = DATA_FORMAT_MAX; for (int i = 0; i < DATA_FORMAT_MAX; i++) { @@ -7036,7 +6652,7 @@ Error RenderingDeviceVulkan::_draw_list_setup_framebuffer(Framebuffer *p_framebu vk.view_count = p_framebuffer->view_count; if (!p_framebuffer->framebuffers.has(vk)) { - //need to create this version + // Need to create this version. Framebuffer::Version version; version.render_pass = _render_pass_create(framebuffer_formats[p_framebuffer->format_id].E->key().attachments, framebuffer_formats[p_framebuffer->format_id].E->key().passes, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_framebuffer->view_count); @@ -7112,7 +6728,7 @@ Error RenderingDeviceVulkan::_draw_list_render_pass_begin(Framebuffer *framebuff } if (color_index < p_clear_colors.size() && texture->usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) { - ERR_FAIL_INDEX_V(color_index, p_clear_colors.size(), ERR_BUG); //a bug + ERR_FAIL_INDEX_V(color_index, p_clear_colors.size(), ERR_BUG); // A bug. Color clear_color = p_clear_colors[color_index]; clear_value.color.float32[0] = clear_color.r; clear_value.color.float32[1] = clear_color.g; @@ -7143,7 +6759,7 @@ Error RenderingDeviceVulkan::_draw_list_render_pass_begin(Framebuffer *framebuff ERR_CONTINUE_MSG(!(texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT), "Supplied storage texture " + itos(i) + " for draw list is not set to be used for storage."); if (texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT) { - //must change layout to general + // Must change layout to general. VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -7171,7 +6787,7 @@ Error RenderingDeviceVulkan::_draw_list_render_pass_begin(Framebuffer *framebuff vkCmdBeginRenderPass(command_buffer, &render_pass_begin, subpass_contents); - //mark textures as bound + // Mark textures as bound. draw_list_bound_textures.clear(); draw_list_unbind_color_textures = p_final_color_action != FINAL_ACTION_CONTINUE; draw_list_unbind_depth_textures = p_final_depth_action != FINAL_ACTION_CONTINUE; @@ -7188,12 +6804,12 @@ Error RenderingDeviceVulkan::_draw_list_render_pass_begin(Framebuffer *framebuff return OK; } -void RenderingDeviceVulkan::_draw_list_insert_clear_region(DrawList *draw_list, Framebuffer *framebuffer, Point2i viewport_offset, Point2i viewport_size, bool p_clear_color, const Vector<Color> &p_clear_colors, bool p_clear_depth, float p_depth, uint32_t p_stencil) { +void RenderingDeviceVulkan::_draw_list_insert_clear_region(DrawList *p_draw_list, Framebuffer *p_framebuffer, Point2i p_viewport_offset, Point2i p_viewport_size, bool p_clear_color, const Vector<Color> &p_clear_colors, bool p_clear_depth, float p_depth, uint32_t p_stencil) { Vector<VkClearAttachment> clear_attachments; int color_index = 0; int texture_index = 0; - for (int i = 0; i < framebuffer->texture_ids.size(); i++) { - Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]); + for (int i = 0; i < p_framebuffer->texture_ids.size(); i++) { + Texture *texture = texture_owner.get_or_null(p_framebuffer->texture_ids[i]); if (!texture) { texture_index++; @@ -7226,12 +6842,12 @@ void RenderingDeviceVulkan::_draw_list_insert_clear_region(DrawList *draw_list, VkClearRect cr; cr.baseArrayLayer = 0; cr.layerCount = 1; - cr.rect.offset.x = viewport_offset.x; - cr.rect.offset.y = viewport_offset.y; - cr.rect.extent.width = viewport_size.width; - cr.rect.extent.height = viewport_size.height; + cr.rect.offset.x = p_viewport_offset.x; + cr.rect.offset.y = p_viewport_offset.y; + cr.rect.extent.width = p_viewport_size.width; + cr.rect.extent.height = p_viewport_size.height; - vkCmdClearAttachments(draw_list->command_buffer, clear_attachments.size(), clear_attachments.ptr(), 1, &cr); + vkCmdClearAttachments(p_draw_list->command_buffer, clear_attachments.size(), clear_attachments.ptr(), 1, &cr); } RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_begin(RID p_framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values, float p_clear_depth, uint32_t p_clear_stencil, const Rect2 &p_region, const Vector<RID> &p_storage_textures) { @@ -7248,7 +6864,7 @@ RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_begin(RID p_framebu bool needs_clear_color = false; bool needs_clear_depth = false; - if (p_region != Rect2() && p_region != Rect2(Vector2(), viewport_size)) { //check custom region + if (p_region != Rect2() && p_region != Rect2(Vector2(), viewport_size)) { // Check custom region. Rect2i viewport(viewport_offset, viewport_size); Rect2i regioni = p_region; if (!(regioni.position.x >= viewport.position.x) && (regioni.position.y >= viewport.position.y) && @@ -7277,15 +6893,17 @@ RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_begin(RID p_framebu } } - if (p_initial_color_action == INITIAL_ACTION_CLEAR || needs_clear_color) { //check clear values + if (p_initial_color_action == INITIAL_ACTION_CLEAR || needs_clear_color) { // Check clear values. int color_count = 0; for (int i = 0; i < framebuffer->texture_ids.size(); i++) { Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]); // We only check for our VRS usage bit if this is not the first texture id. // If it is the first we're likely populating our VRS texture. - // Bit dirty but.. + // Bit dirty but... if (!texture || (!(texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) && !(i != 0 && texture->usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT))) { - color_count++; + if (!texture || !texture->is_resolve_buffer) { + color_count++; + } } } ERR_FAIL_COND_V_MSG(p_clear_color_values.size() != color_count, INVALID_ID, "Clear color values supplied (" + itos(p_clear_color_values.size()) + ") differ from the amount required for framebuffer color attachments (" + itos(color_count) + ")."); @@ -7355,7 +6973,7 @@ Error RenderingDeviceVulkan::draw_list_begin_split(RID p_framebuffer, uint32_t p bool needs_clear_color = false; bool needs_clear_depth = false; - if (p_region != Rect2() && p_region != Rect2(Vector2(), viewport_size)) { //check custom region + if (p_region != Rect2() && p_region != Rect2(Vector2(), viewport_size)) { // Check custom region. Rect2i viewport(viewport_offset, viewport_size); Rect2i regioni = p_region; if (!(regioni.position.x >= viewport.position.x) && (regioni.position.y >= viewport.position.y) && @@ -7377,7 +6995,7 @@ Error RenderingDeviceVulkan::draw_list_begin_split(RID p_framebuffer, uint32_t p } } - if (p_initial_color_action == INITIAL_ACTION_CLEAR || needs_clear_color) { //check clear values + if (p_initial_color_action == INITIAL_ACTION_CLEAR || needs_clear_color) { // Check clear values. int color_count = 0; for (int i = 0; i < framebuffer->texture_ids.size(); i++) { @@ -7463,7 +7081,7 @@ RenderingDeviceVulkan::DrawList *RenderingDeviceVulkan::_get_draw_list_ptr(DrawL return nullptr; } - uint64_t index = p_id & ((DrawListID(1) << DrawListID(ID_BASE_SHIFT)) - 1); //mask + uint64_t index = p_id & ((DrawListID(1) << DrawListID(ID_BASE_SHIFT)) - 1); // Mask. if (index >= draw_list_count) { return nullptr; @@ -7475,6 +7093,16 @@ RenderingDeviceVulkan::DrawList *RenderingDeviceVulkan::_get_draw_list_ptr(DrawL } } +void RenderingDeviceVulkan::draw_list_set_blend_constants(DrawListID p_list, const Color &p_color) { + DrawList *dl = _get_draw_list_ptr(p_list); + ERR_FAIL_COND(!dl); +#ifdef DEBUG_ENABLED + ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified."); +#endif + + vkCmdSetBlendConstants(dl->command_buffer, p_color.components); +} + void RenderingDeviceVulkan::draw_list_bind_render_pipeline(DrawListID p_list, RID p_render_pipeline) { DrawList *dl = _get_draw_list_ptr(p_list); ERR_FAIL_COND(!dl); @@ -7489,7 +7117,7 @@ void RenderingDeviceVulkan::draw_list_bind_render_pipeline(DrawListID p_list, RI #endif if (p_render_pipeline == dl->state.pipeline) { - return; //redundant state, return. + return; // Redundant state, return. } dl->state.pipeline = p_render_pipeline; @@ -7498,17 +7126,17 @@ void RenderingDeviceVulkan::draw_list_bind_render_pipeline(DrawListID p_list, RI vkCmdBindPipeline(dl->command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline->pipeline); if (dl->state.pipeline_shader != pipeline->shader) { - // shader changed, so descriptor sets may become incompatible. + // Shader changed, so descriptor sets may become incompatible. - //go through ALL sets, and unbind them (and all those above) if the format is different + // Go through ALL sets, and unbind them (and all those above) if the format is different. - uint32_t pcount = pipeline->set_formats.size(); //formats count in this pipeline + uint32_t pcount = pipeline->set_formats.size(); // Formats count in this pipeline. dl->state.set_count = MAX(dl->state.set_count, pcount); - const uint32_t *pformats = pipeline->set_formats.ptr(); //pipeline set formats + const uint32_t *pformats = pipeline->set_formats.ptr(); // Pipeline set formats. - bool sets_valid = true; //once invalid, all above become invalid + bool sets_valid = true; // Once invalid, all above become invalid. for (uint32_t i = 0; i < pcount; i++) { - //if a part of the format is different, invalidate it (and the rest) + // If a part of the format is different, invalidate it (and the rest). if (!sets_valid || dl->state.sets[i].pipeline_expected_format != pformats[i]) { dl->state.sets[i].bound = false; dl->state.sets[i].pipeline_expected_format = pformats[i]; @@ -7517,14 +7145,14 @@ void RenderingDeviceVulkan::draw_list_bind_render_pipeline(DrawListID p_list, RI } for (uint32_t i = pcount; i < dl->state.set_count; i++) { - //unbind the ones above (not used) if exist + // Unbind the ones above (not used) if exist. dl->state.sets[i].bound = false; } - dl->state.set_count = pcount; //update set count + dl->state.set_count = pcount; // Update set count. if (pipeline->push_constant_size) { - dl->state.pipeline_push_constant_stages = pipeline->push_constant_stages; + dl->state.pipeline_push_constant_stages = pipeline->push_constant_stages_mask; #ifdef DEBUG_ENABLED dl->validation.pipeline_push_constant_supplied = false; #endif @@ -7534,7 +7162,7 @@ void RenderingDeviceVulkan::draw_list_bind_render_pipeline(DrawListID p_list, RI } #ifdef DEBUG_ENABLED - //update render pass pipeline info + // Update render pass pipeline info. dl->validation.pipeline_active = true; dl->validation.pipeline_dynamic_state = pipeline->validation.dynamic_state; dl->validation.pipeline_vertex_format = pipeline->validation.vertex_format; @@ -7564,8 +7192,8 @@ void RenderingDeviceVulkan::draw_list_bind_uniform_set(DrawListID p_list, RID p_ dl->state.set_count = p_index; } - dl->state.sets[p_index].descriptor_set = uniform_set->descriptor_set; //update set pointer - dl->state.sets[p_index].bound = false; //needs rebind + dl->state.sets[p_index].descriptor_set = uniform_set->descriptor_set; // Update set pointer. + dl->state.sets[p_index].bound = false; // Needs rebind. dl->state.sets[p_index].uniform_set_format = uniform_set->format; dl->state.sets[p_index].uniform_set = p_uniform_set; @@ -7583,7 +7211,7 @@ void RenderingDeviceVulkan::draw_list_bind_uniform_set(DrawListID p_list, RID p_ } #ifdef DEBUG_ENABLED - { //validate that textures bound are not attached as framebuffer bindings + { // Validate that textures bound are not attached as framebuffer bindings. uint32_t attachable_count = uniform_set->attachable_textures.size(); const UniformSet::AttachableTexture *attachable_ptr = uniform_set->attachable_textures.ptr(); uint32_t bound_count = draw_list_bound_textures.size(); @@ -7609,7 +7237,7 @@ void RenderingDeviceVulkan::draw_list_bind_vertex_array(DrawListID p_list, RID p ERR_FAIL_COND(!vertex_array); if (dl->state.vertex_array == p_vertex_array) { - return; //already set + return; // Already set. } dl->state.vertex_array = p_vertex_array; @@ -7633,7 +7261,7 @@ void RenderingDeviceVulkan::draw_list_bind_index_array(DrawListID p_list, RID p_ ERR_FAIL_COND(!index_array); if (dl->state.index_array == p_index_array) { - return; //already set + return; // Already set. } dl->state.index_array = p_index_array; @@ -7643,7 +7271,7 @@ void RenderingDeviceVulkan::draw_list_bind_index_array(DrawListID p_list, RID p_ dl->validation.index_array_size = index_array->indices; dl->validation.index_array_offset = index_array->offset; - vkCmdBindIndexBuffer(dl->command_buffer, index_array->buffer, index_array->offset, index_array->index_type); + vkCmdBindIndexBuffer(dl->command_buffer, index_array->buffer, 0, index_array->index_type); } void RenderingDeviceVulkan::draw_list_set_line_width(DrawListID p_list, float p_width) { @@ -7685,30 +7313,30 @@ void RenderingDeviceVulkan::draw_list_draw(DrawListID p_list, bool p_use_indices ERR_FAIL_COND_MSG(!dl->validation.pipeline_active, "No render pipeline was set before attempting to draw."); if (dl->validation.pipeline_vertex_format != INVALID_ID) { - //pipeline uses vertices, validate format + // Pipeline uses vertices, validate format. ERR_FAIL_COND_MSG(dl->validation.vertex_format == INVALID_ID, "No vertex array was bound, and render pipeline expects vertices."); - //make sure format is right + // Make sure format is right. ERR_FAIL_COND_MSG(dl->validation.pipeline_vertex_format != dl->validation.vertex_format, "The vertex format used to create the pipeline does not match the vertex format bound."); - //make sure number of instances is valid + // Make sure number of instances is valid. ERR_FAIL_COND_MSG(p_instances > dl->validation.vertex_max_instances_allowed, "Number of instances requested (" + itos(p_instances) + " is larger than the maximum number supported by the bound vertex array (" + itos(dl->validation.vertex_max_instances_allowed) + ")."); } if (dl->validation.pipeline_push_constant_size > 0) { - //using push constants, check that they were supplied + // Using push constants, check that they were supplied. ERR_FAIL_COND_MSG(!dl->validation.pipeline_push_constant_supplied, "The shader in this pipeline requires a push constant to be set before drawing, but it's not present."); } #endif - //Bind descriptor sets + // Bind descriptor sets. for (uint32_t i = 0; i < dl->state.set_count; i++) { if (dl->state.sets[i].pipeline_expected_format == 0) { - continue; //nothing expected by this pipeline + continue; // Nothing expected by this pipeline. } #ifdef DEBUG_ENABLED if (dl->state.sets[i].pipeline_expected_format != dl->state.sets[i].uniform_set_format) { @@ -7723,7 +7351,7 @@ void RenderingDeviceVulkan::draw_list_draw(DrawListID p_list, bool p_use_indices } #endif if (!dl->state.sets[i].bound) { - //All good, see if this requires re-binding + // All good, see if this requires re-binding. vkCmdBindDescriptorSets(dl->command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS, dl->state.pipeline_layout, i, 1, &dl->state.sets[i].descriptor_set, 0, nullptr); dl->state.sets[i].bound = true; } @@ -7737,12 +7365,6 @@ void RenderingDeviceVulkan::draw_list_draw(DrawListID p_list, bool p_use_indices ERR_FAIL_COND_MSG(!dl->validation.index_array_size, "Draw command requested indices, but no index buffer was set."); - if (dl->validation.pipeline_vertex_format != INVALID_ID) { - //uses vertices, do some vertex validations - ERR_FAIL_COND_MSG(dl->validation.vertex_array_size < dl->validation.index_array_max_index, - "Index array references (max index: " + itos(dl->validation.index_array_max_index) + ") indices beyond the vertex array size (" + itos(dl->validation.vertex_array_size) + ")."); - } - ERR_FAIL_COND_MSG(dl->validation.pipeline_uses_restart_indices != dl->validation.index_buffer_uses_restart_indices, "The usage of restart indices in index buffer does not match the render primitive in the pipeline."); #endif @@ -7864,7 +7486,7 @@ Error RenderingDeviceVulkan::draw_list_switch_to_next_pass_split(uint32_t p_spli } Error RenderingDeviceVulkan::_draw_list_allocate(const Rect2i &p_viewport, uint32_t p_splits, uint32_t p_subpass) { - // Lock while draw_list is active + // Lock while draw_list is active. _THREAD_SAFE_LOCK_ if (p_splits == 0) { @@ -7891,7 +7513,7 @@ Error RenderingDeviceVulkan::_draw_list_allocate(const Rect2i &p_viewport, uint3 VkCommandBuffer command_buffer; VkCommandBufferAllocateInfo cmdbuf; - //no command buffer exists, create it. + // No command buffer exists, create it. cmdbuf.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO; cmdbuf.pNext = nullptr; cmdbuf.commandPool = split_draw_list_allocators[i].command_pool; @@ -7910,7 +7532,7 @@ Error RenderingDeviceVulkan::_draw_list_allocate(const Rect2i &p_viewport, uint3 draw_list_split = true; for (uint32_t i = 0; i < p_splits; i++) { - //take a command buffer and initialize it + // Take a command buffer and initialize it. VkCommandBuffer command_buffer = split_draw_list_allocators[i].command_buffers[frame]; VkCommandBufferInheritanceInfo inheritance_info; @@ -7920,7 +7542,7 @@ Error RenderingDeviceVulkan::_draw_list_allocate(const Rect2i &p_viewport, uint3 inheritance_info.subpass = p_subpass; inheritance_info.framebuffer = draw_list_vkframebuffer; inheritance_info.occlusionQueryEnable = false; - inheritance_info.queryFlags = 0; //? + inheritance_info.queryFlags = 0; // ? inheritance_info.pipelineStatistics = 0; VkCommandBufferBeginInfo cmdbuf_begin; @@ -7953,7 +7575,7 @@ Error RenderingDeviceVulkan::_draw_list_allocate(const Rect2i &p_viewport, uint3 void RenderingDeviceVulkan::_draw_list_free(Rect2i *r_last_viewport) { if (draw_list_split) { - //send all command buffers + // Send all command buffers. VkCommandBuffer *command_buffers = (VkCommandBuffer *)alloca(sizeof(VkCommandBuffer) * draw_list_count); for (uint32_t i = 0; i < draw_list_count; i++) { vkEndCommandBuffer(draw_list[i].command_buffer); @@ -7973,16 +7595,16 @@ void RenderingDeviceVulkan::_draw_list_free(Rect2i *r_last_viewport) { if (r_last_viewport) { *r_last_viewport = draw_list->viewport; } - //just end the list + // Just end the list. memdelete(draw_list); draw_list = nullptr; } - // draw_list is no longer active + // Draw_list is no longer active. _THREAD_SAFE_UNLOCK_ } -void RenderingDeviceVulkan::draw_list_end(uint32_t p_post_barrier) { +void RenderingDeviceVulkan::draw_list_end(BitField<BarrierMask> p_post_barrier) { _THREAD_SAFE_METHOD_ ERR_FAIL_COND_MSG(!draw_list, "Immediate draw list is already inactive."); @@ -7993,7 +7615,7 @@ void RenderingDeviceVulkan::draw_list_end(uint32_t p_post_barrier) { for (int i = 0; i < draw_list_bound_textures.size(); i++) { Texture *texture = texture_owner.get_or_null(draw_list_bound_textures[i]); - ERR_CONTINUE(!texture); //wtf + ERR_CONTINUE(!texture); // Wtf. if (draw_list_unbind_color_textures && (texture->usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT)) { texture->bound = false; } @@ -8004,15 +7626,15 @@ void RenderingDeviceVulkan::draw_list_end(uint32_t p_post_barrier) { uint32_t barrier_flags = 0; uint32_t access_flags = 0; - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT /*| VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT*/; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT /*| VK_ACCESS_INDIRECT_COMMAND_READ_BIT*/; } - if (p_post_barrier & BARRIER_MASK_TRANSFER) { + if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) { barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT; } @@ -8065,8 +7687,8 @@ void RenderingDeviceVulkan::draw_list_end(uint32_t p_post_barrier) { draw_list_storage_textures.clear(); // To ensure proper synchronization, we must make sure rendering is done before: - // * Some buffer is copied - // * Another render pass happens (since we may be done) + // * Some buffer is copied. + // * Another render pass happens (since we may be done). VkMemoryBarrier mem_barrier; mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER; @@ -8091,7 +7713,7 @@ RenderingDevice::ComputeListID RenderingDeviceVulkan::compute_list_begin(bool p_ ERR_FAIL_COND_V_MSG(!p_allow_draw_overlap && draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time."); ERR_FAIL_COND_V_MSG(compute_list != nullptr, INVALID_ID, "Only one draw/compute list can be active at the same time."); - // Lock while compute_list is active + // Lock while compute_list is active. _THREAD_SAFE_LOCK_ compute_list = memnew(ComputeList); @@ -8111,7 +7733,7 @@ void RenderingDeviceVulkan::compute_list_bind_compute_pipeline(ComputeListID p_l ERR_FAIL_COND(!pipeline); if (p_compute_pipeline == cl->state.pipeline) { - return; //redundant state, return. + return; // Redundant state, return. } cl->state.pipeline = p_compute_pipeline; @@ -8120,17 +7742,17 @@ void RenderingDeviceVulkan::compute_list_bind_compute_pipeline(ComputeListID p_l vkCmdBindPipeline(cl->command_buffer, VK_PIPELINE_BIND_POINT_COMPUTE, pipeline->pipeline); if (cl->state.pipeline_shader != pipeline->shader) { - // shader changed, so descriptor sets may become incompatible. + // Shader changed, so descriptor sets may become incompatible. - //go through ALL sets, and unbind them (and all those above) if the format is different + // Go through ALL sets, and unbind them (and all those above) if the format is different. - uint32_t pcount = pipeline->set_formats.size(); //formats count in this pipeline + uint32_t pcount = pipeline->set_formats.size(); // Formats count in this pipeline. cl->state.set_count = MAX(cl->state.set_count, pcount); - const uint32_t *pformats = pipeline->set_formats.ptr(); //pipeline set formats + const uint32_t *pformats = pipeline->set_formats.ptr(); // Pipeline set formats. - bool sets_valid = true; //once invalid, all above become invalid + bool sets_valid = true; // Once invalid, all above become invalid. for (uint32_t i = 0; i < pcount; i++) { - //if a part of the format is different, invalidate it (and the rest) + // If a part of the format is different, invalidate it (and the rest). if (!sets_valid || cl->state.sets[i].pipeline_expected_format != pformats[i]) { cl->state.sets[i].bound = false; cl->state.sets[i].pipeline_expected_format = pformats[i]; @@ -8139,14 +7761,14 @@ void RenderingDeviceVulkan::compute_list_bind_compute_pipeline(ComputeListID p_l } for (uint32_t i = pcount; i < cl->state.set_count; i++) { - //unbind the ones above (not used) if exist + // Unbind the ones above (not used) if exist. cl->state.sets[i].bound = false; } - cl->state.set_count = pcount; //update set count + cl->state.set_count = pcount; // Update set count. if (pipeline->push_constant_size) { - cl->state.pipeline_push_constant_stages = pipeline->push_constant_stages; + cl->state.pipeline_push_constant_stages = pipeline->push_constant_stages_mask; #ifdef DEBUG_ENABLED cl->validation.pipeline_push_constant_supplied = false; #endif @@ -8159,7 +7781,7 @@ void RenderingDeviceVulkan::compute_list_bind_compute_pipeline(ComputeListID p_l } #ifdef DEBUG_ENABLED - //update compute pass pipeline info + // Update compute pass pipeline info. cl->validation.pipeline_active = true; cl->validation.pipeline_push_constant_size = pipeline->push_constant_size; #endif @@ -8187,8 +7809,8 @@ void RenderingDeviceVulkan::compute_list_bind_uniform_set(ComputeListID p_list, cl->state.set_count = p_index; } - cl->state.sets[p_index].descriptor_set = uniform_set->descriptor_set; //update set pointer - cl->state.sets[p_index].bound = false; //needs rebind + cl->state.sets[p_index].descriptor_set = uniform_set->descriptor_set; // Update set pointer. + cl->state.sets[p_index].bound = false; // Needs rebind. cl->state.sets[p_index].uniform_set_format = uniform_set->format; cl->state.sets[p_index].uniform_set = p_uniform_set; @@ -8291,7 +7913,7 @@ void RenderingDeviceVulkan::compute_list_bind_uniform_set(ComputeListID p_list, textures_to_storage[i]->layout = VK_IMAGE_LAYOUT_GENERAL; - cl->state.textures_to_sampled_layout.insert(textures_to_storage[i]); //needs to go back to sampled layout afterwards + cl->state.textures_to_sampled_layout.insert(textures_to_storage[i]); // Needs to go back to sampled layout afterwards. } } @@ -8304,7 +7926,7 @@ void RenderingDeviceVulkan::compute_list_bind_uniform_set(ComputeListID p_list, } #if 0 - { //validate that textures bound are not attached as framebuffer bindings + { // Validate that textures bound are not attached as framebuffer bindings. uint32_t attachable_count = uniform_set->attachable_textures.size(); const RID *attachable_ptr = uniform_set->attachable_textures.ptr(); uint32_t bound_count = draw_list_bound_textures.size(); @@ -8364,18 +7986,18 @@ void RenderingDeviceVulkan::compute_list_dispatch(ComputeListID p_list, uint32_t ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw."); if (cl->validation.pipeline_push_constant_size > 0) { - //using push constants, check that they were supplied + // Using push constants, check that they were supplied. ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied, "The shader in this pipeline requires a push constant to be set before drawing, but it's not present."); } #endif - //Bind descriptor sets + // Bind descriptor sets. for (uint32_t i = 0; i < cl->state.set_count; i++) { if (cl->state.sets[i].pipeline_expected_format == 0) { - continue; //nothing expected by this pipeline + continue; // Nothing expected by this pipeline. } #ifdef DEBUG_ENABLED if (cl->state.sets[i].pipeline_expected_format != cl->state.sets[i].uniform_set_format) { @@ -8390,7 +8012,7 @@ void RenderingDeviceVulkan::compute_list_dispatch(ComputeListID p_list, uint32_t } #endif if (!cl->state.sets[i].bound) { - //All good, see if this requires re-binding + // All good, see if this requires re-binding. vkCmdBindDescriptorSets(cl->command_buffer, VK_PIPELINE_BIND_POINT_COMPUTE, cl->state.pipeline_layout, i, 1, &cl->state.sets[i].descriptor_set, 0, nullptr); cl->state.sets[i].bound = true; } @@ -8416,7 +8038,7 @@ void RenderingDeviceVulkan::compute_list_dispatch_threads(ComputeListID p_list, ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw."); if (cl->validation.pipeline_push_constant_size > 0) { - //using push constants, check that they were supplied + // Using push constants, check that they were supplied. ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied, "The shader in this pipeline requires a push constant to be set before drawing, but it's not present."); } @@ -8447,18 +8069,18 @@ void RenderingDeviceVulkan::compute_list_dispatch_indirect(ComputeListID p_list, ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw."); if (cl->validation.pipeline_push_constant_size > 0) { - //using push constants, check that they were supplied + // Using push constants, check that they were supplied. ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied, "The shader in this pipeline requires a push constant to be set before drawing, but it's not present."); } #endif - //Bind descriptor sets + // Bind descriptor sets. for (uint32_t i = 0; i < cl->state.set_count; i++) { if (cl->state.sets[i].pipeline_expected_format == 0) { - continue; //nothing expected by this pipeline + continue; // Nothing expected by this pipeline. } #ifdef DEBUG_ENABLED if (cl->state.sets[i].pipeline_expected_format != cl->state.sets[i].uniform_set_format) { @@ -8473,7 +8095,7 @@ void RenderingDeviceVulkan::compute_list_dispatch_indirect(ComputeListID p_list, } #endif if (!cl->state.sets[i].bound) { - //All good, see if this requires re-binding + // All good, see if this requires re-binding. vkCmdBindDescriptorSets(cl->command_buffer, VK_PIPELINE_BIND_POINT_COMPUTE, cl->state.pipeline_layout, i, 1, &cl->state.sets[i].descriptor_set, 0, nullptr); cl->state.sets[i].bound = true; } @@ -8490,20 +8112,20 @@ void RenderingDeviceVulkan::compute_list_add_barrier(ComputeListID p_list) { #endif } -void RenderingDeviceVulkan::compute_list_end(uint32_t p_post_barrier) { +void RenderingDeviceVulkan::compute_list_end(BitField<BarrierMask> p_post_barrier) { ERR_FAIL_COND(!compute_list); uint32_t barrier_flags = 0; uint32_t access_flags = 0; - if (p_post_barrier & BARRIER_MASK_COMPUTE) { + if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) { barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } - if (p_post_barrier & BARRIER_MASK_RASTER) { + if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) { barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT; access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT; } - if (p_post_barrier & BARRIER_MASK_TRANSFER) { + if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) { barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT; } @@ -8567,47 +8189,49 @@ void RenderingDeviceVulkan::compute_list_end(uint32_t p_post_barrier) { memdelete(compute_list); compute_list = nullptr; - // compute_list is no longer active + // Compute_list is no longer active. _THREAD_SAFE_UNLOCK_ } -void RenderingDeviceVulkan::barrier(uint32_t p_from, uint32_t p_to) { +void RenderingDeviceVulkan::barrier(BitField<BarrierMask> p_from, BitField<BarrierMask> p_to) { uint32_t src_barrier_flags = 0; uint32_t src_access_flags = 0; - if (p_from & BARRIER_MASK_COMPUTE) { - src_barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; - src_access_flags |= VK_ACCESS_SHADER_WRITE_BIT; - } - if (p_from & BARRIER_MASK_RASTER) { - src_barrier_flags |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT; - src_access_flags |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT; - } - if (p_from & BARRIER_MASK_TRANSFER) { - src_barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; - src_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; - } if (p_from == 0) { src_barrier_flags = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT; + } else { + if (p_from.has_flag(BARRIER_MASK_COMPUTE)) { + src_barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; + src_access_flags |= VK_ACCESS_SHADER_WRITE_BIT; + } + if (p_from.has_flag(BARRIER_MASK_RASTER)) { + src_barrier_flags |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT; + src_access_flags |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT; + } + if (p_from.has_flag(BARRIER_MASK_TRANSFER)) { + src_barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; + src_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; + } } uint32_t dst_barrier_flags = 0; uint32_t dst_access_flags = 0; - if (p_to & BARRIER_MASK_COMPUTE) { - dst_barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; - dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; - } - if (p_to & BARRIER_MASK_RASTER) { - dst_barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT; - dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT; - } - if (p_to & BARRIER_MASK_TRANSFER) { - dst_barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; - dst_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT; - } if (p_to == 0) { dst_barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } else { + if (p_to.has_flag(BARRIER_MASK_COMPUTE)) { + dst_barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; + dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; + } + if (p_to.has_flag(BARRIER_MASK_RASTER)) { + dst_barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT; + dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT; + } + if (p_to.has_flag(BARRIER_MASK_TRANSFER)) { + dst_barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; + dst_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT; + } } _memory_barrier(src_barrier_flags, dst_barrier_flags, src_access_flags, dst_access_flags, true); @@ -8660,7 +8284,7 @@ void RenderingDeviceVulkan::draw_list_render_secondary_to_framebuffer(ID p_frame "Draw list index (" + itos(i) + ") is created with a framebuffer format incompatible with this render pass."); if (dl->validation.active) { - //needs to be closed, so close it. + // Needs to be closed, so close it. vkEndCommandBuffer(dl->command_buffer); dl->validation.active = false; } @@ -8677,7 +8301,15 @@ void RenderingDeviceVulkan::draw_list_render_secondary_to_framebuffer(ID p_frame #endif void RenderingDeviceVulkan::_free_internal(RID p_id) { - //push everything so it's disposed of next time this frame index is processed (means, it's safe to do it) +#ifdef DEV_ENABLED + String resource_name; + if (resource_names.has(p_id)) { + resource_name = resource_names[p_id]; + resource_names.erase(p_id); + } +#endif + + // Push everything so it's disposed of next time this frame index is processed (means, it's safe to do it). if (texture_owner.owns(p_id)) { Texture *texture = texture_owner.get_or_null(p_id); frames[frame].textures_to_dispose_of.push_back(*texture); @@ -8685,6 +8317,11 @@ void RenderingDeviceVulkan::_free_internal(RID p_id) { } else if (framebuffer_owner.owns(p_id)) { Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_id); frames[frame].framebuffers_to_dispose_of.push_back(*framebuffer); + + if (framebuffer->invalidated_callback != nullptr) { + framebuffer->invalidated_callback(framebuffer->invalidated_callback_userdata); + } + framebuffer_owner.free(p_id); } else if (sampler_owner.owns(p_id)) { VkSampler *sampler = sampler_owner.get_or_null(p_id); @@ -8741,30 +8378,34 @@ void RenderingDeviceVulkan::_free_internal(RID p_id) { frames[frame].compute_pipelines_to_dispose_of.push_back(*pipeline); compute_pipeline_owner.free(p_id); } else { +#ifdef DEV_ENABLED + ERR_PRINT("Attempted to free invalid ID: " + itos(p_id.get_id()) + " " + resource_name); +#else ERR_PRINT("Attempted to free invalid ID: " + itos(p_id.get_id())); +#endif } } void RenderingDeviceVulkan::free(RID p_id) { _THREAD_SAFE_METHOD_ - _free_dependencies(p_id); //recursively erase dependencies first, to avoid potential API problems + _free_dependencies(p_id); // Recursively erase dependencies first, to avoid potential API problems. _free_internal(p_id); } -// The full list of resources that can be named is in the VkObjectType enum +// The full list of resources that can be named is in the VkObjectType enum. // We just expose the resources that are owned and can be accessed easily. void RenderingDeviceVulkan::set_resource_name(RID p_id, const String p_name) { if (texture_owner.owns(p_id)) { Texture *texture = texture_owner.get_or_null(p_id); if (texture->owner.is_null()) { - // Don't set the source texture's name when calling on a texture view + // Don't set the source texture's name when calling on a texture view. context->set_object_name(VK_OBJECT_TYPE_IMAGE, uint64_t(texture->image), p_name); } context->set_object_name(VK_OBJECT_TYPE_IMAGE_VIEW, uint64_t(texture->view), p_name + " View"); } else if (framebuffer_owner.owns(p_id)) { //Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_id); - // Not implemented for now as the relationship between Framebuffer and RenderPass is very complex + // Not implemented for now as the relationship between Framebuffer and RenderPass is very complex. } else if (sampler_owner.owns(p_id)) { VkSampler *sampler = sampler_owner.get_or_null(p_id); context->set_object_name(VK_OBJECT_TYPE_SAMPLER, uint64_t(*sampler), p_name); @@ -8803,7 +8444,11 @@ void RenderingDeviceVulkan::set_resource_name(RID p_id, const String p_name) { context->set_object_name(VK_OBJECT_TYPE_PIPELINE_LAYOUT, uint64_t(pipeline->pipeline_layout), p_name + " Layout"); } else { ERR_PRINT("Attempted to name invalid ID: " + itos(p_id.get_id())); + return; } +#ifdef DEV_ENABLED + resource_names[p_id] = p_name; +#endif } void RenderingDeviceVulkan::draw_command_begin_label(String p_label_name, const Color p_color) { @@ -8847,17 +8492,17 @@ void RenderingDeviceVulkan::_finalize_command_bufers() { ERR_PRINT("Found open compute list at the end of the frame, this should never happen (further compute will likely not work)."); } - { //complete the setup buffer (that needs to be processed before anything else) + { // Complete the setup buffer (that needs to be processed before anything else). vkEndCommandBuffer(frames[frame].setup_command_buffer); vkEndCommandBuffer(frames[frame].draw_command_buffer); } } void RenderingDeviceVulkan::_begin_frame() { - //erase pending resources + // Erase pending resources. _free_pending_resources(frame); - //create setup command buffer and set as the setup buffer + // Create setup command buffer and set as the setup buffer. { VkCommandBufferBeginInfo cmdbuf_begin; @@ -8876,13 +8521,13 @@ void RenderingDeviceVulkan::_begin_frame() { if (local_device.is_null()) { context->append_command_buffer(frames[frame].draw_command_buffer); - context->set_setup_buffer(frames[frame].setup_command_buffer); //append now so it's added before everything else + context->set_setup_buffer(frames[frame].setup_command_buffer); // Append now so it's added before everything else. } } - //advance current frame + // Advance current frame. frames_drawn++; - //advance staging buffer if used + // Advance staging buffer if used. if (staging_buffer_used) { staging_buffer_current = (staging_buffer_current + 1) % staging_buffer_blocks.size(); staging_buffer_used = false; @@ -8900,6 +8545,25 @@ void RenderingDeviceVulkan::_begin_frame() { frames[frame].index = Engine::get_singleton()->get_frames_drawn(); } +VkSampleCountFlagBits RenderingDeviceVulkan::_ensure_supported_sample_count(TextureSamples p_requested_sample_count) const { + VkSampleCountFlags sample_count_flags = limits.framebufferColorSampleCounts & limits.framebufferDepthSampleCounts; + + if (sample_count_flags & rasterization_sample_count[p_requested_sample_count]) { + // The requested sample count is supported. + return rasterization_sample_count[p_requested_sample_count]; + } else { + // Find the closest lower supported sample count. + VkSampleCountFlagBits sample_count = rasterization_sample_count[p_requested_sample_count]; + while (sample_count > VK_SAMPLE_COUNT_1_BIT) { + if (sample_count_flags & sample_count) { + return sample_count; + } + sample_count = (VkSampleCountFlagBits)(sample_count >> 1); + } + } + return VK_SAMPLE_COUNT_1_BIT; +} + void RenderingDeviceVulkan::swap_buffers() { ERR_FAIL_COND_MSG(local_device.is_valid(), "Local devices can't swap buffers."); _THREAD_SAFE_METHOD_ @@ -8907,7 +8571,7 @@ void RenderingDeviceVulkan::swap_buffers() { _finalize_command_bufers(); screen_prepared = false; - //swap buffers + // Swap buffers. context->swap_buffers(); frame = (frame + 1) % frame_count; @@ -8953,15 +8617,15 @@ VmaPool RenderingDeviceVulkan::_find_or_create_small_allocs_pool(uint32_t p_mem_ pci.pMemoryAllocateNext = nullptr; VmaPool pool = VK_NULL_HANDLE; VkResult res = vmaCreatePool(allocator, &pci, &pool); - small_allocs_pools[p_mem_type_index] = pool; // Don't try to create it again if failed the first time + small_allocs_pools[p_mem_type_index] = pool; // Don't try to create it again if failed the first time. ERR_FAIL_COND_V_MSG(res, pool, "vmaCreatePool failed with error " + itos(res) + "."); return pool; } void RenderingDeviceVulkan::_free_pending_resources(int p_frame) { - //free in dependency usage order, so nothing weird happens - //pipelines + // Free in dependency usage order, so nothing weird happens. + // Pipelines. while (frames[p_frame].render_pipelines_to_dispose_of.front()) { RenderPipeline *pipeline = &frames[p_frame].render_pipelines_to_dispose_of.front()->get(); @@ -8978,7 +8642,7 @@ void RenderingDeviceVulkan::_free_pending_resources(int p_frame) { frames[p_frame].compute_pipelines_to_dispose_of.pop_front(); } - //uniform sets + // Uniform sets. while (frames[p_frame].uniform_sets_to_dispose_of.front()) { UniformSet *uniform_set = &frames[p_frame].uniform_sets_to_dispose_of.front()->get(); @@ -8988,7 +8652,7 @@ void RenderingDeviceVulkan::_free_pending_resources(int p_frame) { frames[p_frame].uniform_sets_to_dispose_of.pop_front(); } - //buffer views + // Buffer views. while (frames[p_frame].buffer_views_to_dispose_of.front()) { VkBufferView buffer_view = frames[p_frame].buffer_views_to_dispose_of.front()->get(); @@ -8997,19 +8661,19 @@ void RenderingDeviceVulkan::_free_pending_resources(int p_frame) { frames[p_frame].buffer_views_to_dispose_of.pop_front(); } - //shaders + // Shaders. while (frames[p_frame].shaders_to_dispose_of.front()) { Shader *shader = &frames[p_frame].shaders_to_dispose_of.front()->get(); - //descriptor set layout for each set + // Descriptor set layout for each set. for (int i = 0; i < shader->sets.size(); i++) { vkDestroyDescriptorSetLayout(device, shader->sets[i].descriptor_set_layout, nullptr); } - //pipeline layout + // Pipeline layout. vkDestroyPipelineLayout(device, shader->pipeline_layout, nullptr); - //shaders themselves + // Shaders themselves. for (int i = 0; i < shader->pipeline_stages.size(); i++) { vkDestroyShaderModule(device, shader->pipeline_stages[i].module, nullptr); } @@ -9017,7 +8681,7 @@ void RenderingDeviceVulkan::_free_pending_resources(int p_frame) { frames[p_frame].shaders_to_dispose_of.pop_front(); } - //samplers + // Samplers. while (frames[p_frame].samplers_to_dispose_of.front()) { VkSampler sampler = frames[p_frame].samplers_to_dispose_of.front()->get(); @@ -9026,12 +8690,12 @@ void RenderingDeviceVulkan::_free_pending_resources(int p_frame) { frames[p_frame].samplers_to_dispose_of.pop_front(); } - //framebuffers + // Framebuffers. while (frames[p_frame].framebuffers_to_dispose_of.front()) { Framebuffer *framebuffer = &frames[p_frame].framebuffers_to_dispose_of.front()->get(); for (const KeyValue<Framebuffer::VersionKey, Framebuffer::Version> &E : framebuffer->framebuffers) { - //first framebuffer, then render pass because it depends on it + // First framebuffer, then render pass because it depends on it. vkDestroyFramebuffer(device, E.value.framebuffer, nullptr); vkDestroyRenderPass(device, E.value.render_pass, nullptr); } @@ -9039,23 +8703,23 @@ void RenderingDeviceVulkan::_free_pending_resources(int p_frame) { frames[p_frame].framebuffers_to_dispose_of.pop_front(); } - //textures + // Textures. while (frames[p_frame].textures_to_dispose_of.front()) { Texture *texture = &frames[p_frame].textures_to_dispose_of.front()->get(); if (texture->bound) { - WARN_PRINT("Deleted a texture while it was bound.."); + WARN_PRINT("Deleted a texture while it was bound."); } vkDestroyImageView(device, texture->view, nullptr); if (texture->owner.is_null()) { - //actually owns the image and the allocation too + // Actually owns the image and the allocation too. image_memory -= texture->allocation_info.size; vmaDestroyImage(allocator, texture->image, texture->allocation); } frames[p_frame].textures_to_dispose_of.pop_front(); } - //buffers + // Buffers. while (frames[p_frame].buffers_to_dispose_of.front()) { _buffer_free(&frames[p_frame].buffers_to_dispose_of.front()->get()); @@ -9087,9 +8751,9 @@ uint64_t RenderingDeviceVulkan::get_memory_usage(MemoryType p_type) const { void RenderingDeviceVulkan::_flush(bool p_current_frame) { if (local_device.is_valid() && !p_current_frame) { - return; //flushing previous frames has no effect with local device + return; // Flushing previous frames has no effect with local device. } - //not doing this crashes RADV (undefined behavior) + // Not doing this crashes RADV (undefined behavior). if (p_current_frame) { vkEndCommandBuffer(frames[frame].setup_command_buffer); vkEndCommandBuffer(frames[frame].draw_command_buffer); @@ -9113,7 +8777,7 @@ void RenderingDeviceVulkan::_flush(bool p_current_frame) { } else { context->flush(p_current_frame, p_current_frame); - //re-create the setup command + // Re-create the setup command. if (p_current_frame) { VkCommandBufferBeginInfo cmdbuf_begin; cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO; @@ -9123,7 +8787,7 @@ void RenderingDeviceVulkan::_flush(bool p_current_frame) { VkResult err = vkBeginCommandBuffer(frames[frame].setup_command_buffer, &cmdbuf_begin); ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + "."); - context->set_setup_buffer(frames[frame].setup_command_buffer); //append now so it's added before everything else + context->set_setup_buffer(frames[frame].setup_command_buffer); // Append now so it's added before everything else. } if (p_current_frame) { @@ -9141,7 +8805,7 @@ void RenderingDeviceVulkan::_flush(bool p_current_frame) { } void RenderingDeviceVulkan::initialize(VulkanContext *p_context, bool p_local_device) { - // get our device capabilities + // Get our device capabilities. { device_capabilities.version_major = p_context->get_vulkan_major(); device_capabilities.version_minor = p_context->get_vulkan_minor(); @@ -9154,12 +8818,12 @@ void RenderingDeviceVulkan::initialize(VulkanContext *p_context, bool p_local_de local_device = p_context->local_device_create(); device = p_context->local_device_get_vk_device(local_device); } else { - frame_count = p_context->get_swapchain_image_count() + 1; //always need one extra to ensure it's unused at any time, without having to use a fence for this. + frame_count = p_context->get_swapchain_image_count() + 1; // Always need one extra to ensure it's unused at any time, without having to use a fence for this. } limits = p_context->get_device_limits(); max_timestamp_query_elements = 256; - { //initialize allocator + { // Initialize allocator. VmaAllocatorCreateInfo allocatorInfo; memset(&allocatorInfo, 0, sizeof(VmaAllocatorCreateInfo)); @@ -9171,11 +8835,11 @@ void RenderingDeviceVulkan::initialize(VulkanContext *p_context, bool p_local_de frames.resize(frame_count); frame = 0; - //create setup and frame buffers + // Create setup and frame buffers. for (int i = 0; i < frame_count; i++) { frames[i].index = 0; - { //create command pool, one per frame is recommended + { // Create command pool, one per frame is recommended. VkCommandPoolCreateInfo cmd_pool_info; cmd_pool_info.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO; cmd_pool_info.pNext = nullptr; @@ -9186,10 +8850,10 @@ void RenderingDeviceVulkan::initialize(VulkanContext *p_context, bool p_local_de ERR_FAIL_COND_MSG(res, "vkCreateCommandPool failed with error " + itos(res) + "."); } - { //create command buffers + { // Create command buffers. VkCommandBufferAllocateInfo cmdbuf; - //no command buffer exists, create it. + // No command buffer exists, create it. cmdbuf.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO; cmdbuf.pNext = nullptr; cmdbuf.commandPool = frames[i].command_pool; @@ -9204,7 +8868,7 @@ void RenderingDeviceVulkan::initialize(VulkanContext *p_context, bool p_local_de } { - //create query pool + // Create query pool. VkQueryPoolCreateInfo query_pool_create_info; query_pool_create_info.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO; query_pool_create_info.flags = 0; @@ -9226,8 +8890,8 @@ void RenderingDeviceVulkan::initialize(VulkanContext *p_context, bool p_local_de } { - //begin the first command buffer for the first frame, so - //setting up things can be done in the meantime until swap_buffers(), which is called before advance. + // Begin the first command buffer for the first frame, so + // setting up things can be done in the meantime until swap_buffers(), which is called before advance. VkCommandBufferBeginInfo cmdbuf_begin; cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO; cmdbuf_begin.pNext = nullptr; @@ -9240,42 +8904,40 @@ void RenderingDeviceVulkan::initialize(VulkanContext *p_context, bool p_local_de err = vkBeginCommandBuffer(frames[0].draw_command_buffer, &cmdbuf_begin); ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + "."); if (local_device.is_null()) { - context->set_setup_buffer(frames[0].setup_command_buffer); //append now so it's added before everything else + context->set_setup_buffer(frames[0].setup_command_buffer); // Append now so it's added before everything else. context->append_command_buffer(frames[0].draw_command_buffer); } } - // Note: If adding new project settings here, also duplicate their definition in - // rendering_server.cpp for headless doctool. - staging_buffer_block_size = GLOBAL_DEF("rendering/vulkan/staging_buffer/block_size_kb", 256); + staging_buffer_block_size = GLOBAL_GET("rendering/rendering_device/staging_buffer/block_size_kb"); staging_buffer_block_size = MAX(4u, staging_buffer_block_size); - staging_buffer_block_size *= 1024; //kb -> bytes - staging_buffer_max_size = GLOBAL_DEF("rendering/vulkan/staging_buffer/max_size_mb", 128); + staging_buffer_block_size *= 1024; // Kb -> bytes. + staging_buffer_max_size = GLOBAL_GET("rendering/rendering_device/staging_buffer/max_size_mb"); staging_buffer_max_size = MAX(1u, staging_buffer_max_size); staging_buffer_max_size *= 1024 * 1024; if (staging_buffer_max_size < staging_buffer_block_size * 4) { - //validate enough blocks + // Validate enough blocks. staging_buffer_max_size = staging_buffer_block_size * 4; } - texture_upload_region_size_px = GLOBAL_DEF("rendering/vulkan/staging_buffer/texture_upload_region_size_px", 64); + texture_upload_region_size_px = GLOBAL_GET("rendering/rendering_device/staging_buffer/texture_upload_region_size_px"); texture_upload_region_size_px = nearest_power_of_2_templated(texture_upload_region_size_px); - frames_drawn = frame_count; //start from frame count, so everything else is immediately old + frames_drawn = frame_count; // Start from frame count, so everything else is immediately old. - //ensure current staging block is valid and at least one per frame exists + // Ensure current staging block is valid and at least one per frame exists. staging_buffer_current = 0; staging_buffer_used = false; for (int i = 0; i < frame_count; i++) { - //staging was never used, create a block + // Staging was never used, create a block. Error err = _insert_staging_block(); ERR_CONTINUE(err != OK); } - max_descriptors_per_pool = GLOBAL_DEF("rendering/vulkan/descriptor_pools/max_descriptors_per_pool", 64); + max_descriptors_per_pool = GLOBAL_GET("rendering/rendering_device/vulkan/max_descriptors_per_pool"); - //check to make sure DescriptorPoolKey is good + // Check to make sure DescriptorPoolKey is good. static_assert(sizeof(uint64_t) * 3 >= UNIFORM_TYPE_MAX * sizeof(uint16_t)); draw_list = nullptr; @@ -9296,6 +8958,11 @@ void RenderingDeviceVulkan::_free_rids(T &p_owner, const char *p_type) { WARN_PRINT(vformat("%d RIDs of type \"%s\" were leaked.", owned.size(), p_type)); } for (const RID &E : owned) { +#ifdef DEV_ENABLED + if (resource_names.has(E)) { + print_line(String(" - ") + resource_names[E]); + } +#endif free(E); } } @@ -9305,7 +8972,7 @@ void RenderingDeviceVulkan::capture_timestamp(const String &p_name) { ERR_FAIL_COND_MSG(draw_list != nullptr, "Capturing timestamps during draw list creation is not allowed. Offending timestamp was: " + p_name); ERR_FAIL_COND(frames[frame].timestamp_count >= max_timestamp_query_elements); - //this should be optional for profiling, else it will slow things down + // This should be optional for profiling, else it will slow things down. { VkMemoryBarrier memoryBarrier; @@ -9431,7 +9098,7 @@ uint64_t RenderingDeviceVulkan::get_driver_resource(DriverResource p_resource, R return uint64_t(render_pipeline->pipeline); } break; default: { - // not supported for this driver + // Not supported for this driver. return 0; } break; } @@ -9468,9 +9135,9 @@ static void mult64to128(uint64_t u, uint64_t v, uint64_t &h, uint64_t &l) { uint64_t RenderingDeviceVulkan::get_captured_timestamp_gpu_time(uint32_t p_index) const { ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, 0); - // this sucks because timestampPeriod multiplier is a float, while the timestamp is 64 bits nanosecs. - // so, in cases like nvidia which give you enormous numbers and 1 as multiplier, multiplying is next to impossible - // need to do 128 bits fixed point multiplication to get the right value + // This sucks because timestampPeriod multiplier is a float, while the timestamp is 64 bits nanosecs. + // So, in cases like nvidia which give you enormous numbers and 1 as multiplier, multiplying is next to impossible. + // Need to do 128 bits fixed point multiplication to get the right value. uint64_t shift_bits = 16; @@ -9563,6 +9230,10 @@ uint64_t RenderingDeviceVulkan::limit_get(Limit p_limit) const { return limits.maxComputeWorkGroupSize[1]; case LIMIT_MAX_COMPUTE_WORKGROUP_SIZE_Z: return limits.maxComputeWorkGroupSize[2]; + case LIMIT_MAX_VIEWPORT_DIMENSIONS_X: + return limits.maxViewportDimensions[0]; + case LIMIT_MAX_VIEWPORT_DIMENSIONS_Y: + return limits.maxViewportDimensions[1]; case LIMIT_SUBGROUP_SIZE: { VulkanContext::SubgroupCapabilities subgroup_capabilities = context->get_subgroup_capabilities(); return subgroup_capabilities.size; @@ -9575,6 +9246,12 @@ uint64_t RenderingDeviceVulkan::limit_get(Limit p_limit) const { VulkanContext::SubgroupCapabilities subgroup_capabilities = context->get_subgroup_capabilities(); return subgroup_capabilities.supported_operations_flags_rd(); } + case LIMIT_VRS_TEXEL_WIDTH: { + return context->get_vrs_capabilities().texel_size.x; + } + case LIMIT_VRS_TEXEL_HEIGHT: { + return context->get_vrs_capabilities().texel_size.y; + } default: ERR_FAIL_V(0); } @@ -9583,7 +9260,7 @@ uint64_t RenderingDeviceVulkan::limit_get(Limit p_limit) const { } void RenderingDeviceVulkan::finalize() { - //free all resources + // Free all resources. _flush(false); @@ -9601,7 +9278,7 @@ void RenderingDeviceVulkan::finalize() { _free_rids(framebuffer_owner, "Framebuffer"); _free_rids(sampler_owner, "Sampler"); { - //for textures it's a bit more difficult because they may be shared + // For textures it's a bit more difficult because they may be shared. List<RID> owned; texture_owner.get_owned_list(&owned); if (owned.size()) { @@ -9610,23 +9287,33 @@ void RenderingDeviceVulkan::finalize() { } else { WARN_PRINT(vformat("%d RIDs of type \"Texture\" were leaked.", owned.size())); } - //free shared first + // Free shared first. for (List<RID>::Element *E = owned.front(); E;) { List<RID>::Element *N = E->next(); if (texture_is_shared(E->get())) { +#ifdef DEV_ENABLED + if (resource_names.has(E->get())) { + print_line(String(" - ") + resource_names[E->get()]); + } +#endif free(E->get()); owned.erase(E); } E = N; } - //free non shared second, this will avoid an error trying to free unexisting textures due to dependencies. + // Free non shared second, this will avoid an error trying to free unexisting textures due to dependencies. for (const RID &E : owned) { +#ifdef DEV_ENABLED + if (resource_names.has(E)) { + print_line(String(" - ") + resource_names[E]); + } +#endif free(E); } } } - //free everything pending + // Free everything pending. for (int i = 0; i < frame_count; i++) { int f = (frame + i) % frame_count; _free_pending_resources(f); @@ -9662,7 +9349,7 @@ void RenderingDeviceVulkan::finalize() { } framebuffer_formats.clear(); - //all these should be clear at this point + // All these should be clear at this point. ERR_FAIL_COND(descriptor_pools.size()); ERR_FAIL_COND(dependency_map.size()); ERR_FAIL_COND(reverse_dependency_map.size()); diff --git a/drivers/vulkan/rendering_device_vulkan.h b/drivers/vulkan/rendering_device_vulkan.h index 7c8021251f..b2c2a03a87 100644 --- a/drivers/vulkan/rendering_device_vulkan.h +++ b/drivers/vulkan/rendering_device_vulkan.h @@ -1,32 +1,32 @@ -/*************************************************************************/ -/* rendering_device_vulkan.h */ -/*************************************************************************/ -/* This file is part of: */ -/* GODOT ENGINE */ -/* https://godotengine.org */ -/*************************************************************************/ -/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ -/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ -/* */ -/* Permission is hereby granted, free of charge, to any person obtaining */ -/* a copy of this software and associated documentation files (the */ -/* "Software"), to deal in the Software without restriction, including */ -/* without limitation the rights to use, copy, modify, merge, publish, */ -/* distribute, sublicense, and/or sell copies of the Software, and to */ -/* permit persons to whom the Software is furnished to do so, subject to */ -/* the following conditions: */ -/* */ -/* The above copyright notice and this permission notice shall be */ -/* included in all copies or substantial portions of the Software. */ -/* */ -/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ -/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ -/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ -/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ -/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ -/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ -/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/*************************************************************************/ +/**************************************************************************/ +/* rendering_device_vulkan.h */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ #ifndef RENDERING_DEVICE_VULKAN_H #define RENDERING_DEVICE_VULKAN_H @@ -96,13 +96,13 @@ class RenderingDeviceVulkan : public RenderingDevice { ID_TYPE_SPLIT_DRAW_LIST, ID_TYPE_COMPUTE_LIST, ID_TYPE_MAX, - ID_BASE_SHIFT = 58 //5 bits for ID types + ID_BASE_SHIFT = 58 // 5 bits for ID types. }; VkDevice device = VK_NULL_HANDLE; - HashMap<RID, HashSet<RID>> dependency_map; //IDs to IDs that depend on it - HashMap<RID, HashSet<RID>> reverse_dependency_map; //same as above, but in reverse + HashMap<RID, HashSet<RID>> dependency_map; // IDs to IDs that depend on it. + HashMap<RID, HashSet<RID>> reverse_dependency_map; // Same as above, but in reverse. void _add_dependency(RID p_id, RID p_depends_on); void _free_dependencies(RID p_id); @@ -150,9 +150,11 @@ class RenderingDeviceVulkan : public RenderingDevice { bool used_in_raster = false; bool used_in_compute = false; + bool is_resolve_buffer = false; + uint32_t read_aspect_mask = 0; uint32_t barrier_aspect_mask = 0; - bool bound = false; //bound to framebffer + bool bound = false; // Bound to framebffer. RID owner; }; @@ -160,7 +162,7 @@ class RenderingDeviceVulkan : public RenderingDevice { uint32_t texture_upload_region_size_px = 0; Vector<uint8_t> _texture_get_data_from_image(Texture *tex, VkImage p_image, VmaAllocation p_allocation, uint32_t p_layer, bool p_2d = false); - Error _texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, uint32_t p_post_barrier, bool p_use_setup_queue); + Error _texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, BitField<BarrierMask> p_post_barrier, bool p_use_setup_queue); /*****************/ /**** SAMPLER ****/ @@ -214,7 +216,7 @@ class RenderingDeviceVulkan : public RenderingDevice { uint32_t usage = 0; VkBuffer buffer = VK_NULL_HANDLE; VmaAllocation allocation = nullptr; - VkDescriptorBufferInfo buffer_info; //used for binding + VkDescriptorBufferInfo buffer_info; // Used for binding. Buffer() { } }; @@ -224,8 +226,8 @@ class RenderingDeviceVulkan : public RenderingDevice { Error _buffer_update(Buffer *p_buffer, size_t p_offset, const uint8_t *p_data, size_t p_data_size, bool p_use_draw_command_buffer = false, uint32_t p_required_align = 32); void _full_barrier(bool p_sync_with_draw); - void _memory_barrier(VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_sccess, bool p_sync_with_draw); - void _buffer_memory_barrier(VkBuffer buffer, uint64_t p_from, uint64_t p_size, VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_sccess, bool p_sync_with_draw); + void _memory_barrier(VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_access, bool p_sync_with_draw); + void _buffer_memory_barrier(VkBuffer buffer, uint64_t p_from, uint64_t p_size, VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_access, bool p_sync_with_draw); /*********************/ /**** FRAMEBUFFER ****/ @@ -256,7 +258,7 @@ class RenderingDeviceVulkan : public RenderingDevice { const FramebufferPass *key_pass_ptr = p_key.passes.ptr(); for (uint32_t i = 0; i < pass_size; i++) { - { //compare color attachments + { // Compare color attachments. uint32_t attachment_size = pass_ptr[i].color_attachments.size(); uint32_t key_attachment_size = key_pass_ptr[i].color_attachments.size(); if (attachment_size != key_attachment_size) { @@ -271,7 +273,7 @@ class RenderingDeviceVulkan : public RenderingDevice { } } } - { //compare input attachments + { // Compare input attachments. uint32_t attachment_size = pass_ptr[i].input_attachments.size(); uint32_t key_attachment_size = key_pass_ptr[i].input_attachments.size(); if (attachment_size != key_attachment_size) { @@ -286,7 +288,7 @@ class RenderingDeviceVulkan : public RenderingDevice { } } } - { //compare resolve attachments + { // Compare resolve attachments. uint32_t attachment_size = pass_ptr[i].resolve_attachments.size(); uint32_t key_attachment_size = key_pass_ptr[i].resolve_attachments.size(); if (attachment_size != key_attachment_size) { @@ -301,7 +303,7 @@ class RenderingDeviceVulkan : public RenderingDevice { } } } - { //compare preserve attachments + { // Compare preserve attachments. uint32_t attachment_size = pass_ptr[i].preserve_attachments.size(); uint32_t key_attachment_size = key_pass_ptr[i].preserve_attachments.size(); if (attachment_size != key_attachment_size) { @@ -343,7 +345,7 @@ class RenderingDeviceVulkan : public RenderingDevice { } } - return false; //equal + return false; // Equal. } }; @@ -353,9 +355,9 @@ class RenderingDeviceVulkan : public RenderingDevice { RBMap<FramebufferFormatKey, FramebufferFormatID> framebuffer_format_cache; struct FramebufferFormat { const RBMap<FramebufferFormatKey, FramebufferFormatID>::Element *E; - VkRenderPass render_pass = VK_NULL_HANDLE; //here for constructing shaders, never used, see section (7.2. Render Pass Compatibility from Vulkan spec) + VkRenderPass render_pass = VK_NULL_HANDLE; // Here for constructing shaders, never used, see section (7.2. Render Pass Compatibility from Vulkan spec). Vector<TextureSamples> pass_samples; - uint32_t view_count = 1; // number of views + uint32_t view_count = 1; // Number of views. }; HashMap<FramebufferFormatID, FramebufferFormat> framebuffer_formats; @@ -392,10 +394,12 @@ class RenderingDeviceVulkan : public RenderingDevice { uint32_t storage_mask = 0; Vector<RID> texture_ids; + InvalidationCallback invalidated_callback = nullptr; + void *invalidated_callback_userdata = nullptr; struct Version { VkFramebuffer framebuffer = VK_NULL_HANDLE; - VkRenderPass render_pass = VK_NULL_HANDLE; //this one is owned + VkRenderPass render_pass = VK_NULL_HANDLE; // This one is owned. uint32_t subpass_count = 1; }; @@ -452,7 +456,7 @@ class RenderingDeviceVulkan : public RenderingDevice { return false; } } - return true; //they are equal + return true; // They are equal. } } @@ -497,14 +501,14 @@ class RenderingDeviceVulkan : public RenderingDevice { int vertex_count = 0; uint32_t max_instances_allowed = 0; - Vector<VkBuffer> buffers; //not owned, just referenced + Vector<VkBuffer> buffers; // Not owned, just referenced. Vector<VkDeviceSize> offsets; }; RID_Owner<VertexArray, true> vertex_array_owner; struct IndexBuffer : public Buffer { - uint32_t max_index = 0; //used for validation + uint32_t max_index = 0; // Used for validation. uint32_t index_count = 0; VkIndexType index_type = VK_INDEX_TYPE_NONE_NV; bool supports_restart_indices = false; @@ -513,8 +517,8 @@ class RenderingDeviceVulkan : public RenderingDevice { RID_Owner<IndexBuffer, true> index_buffer_owner; struct IndexArray { - uint32_t max_index = 0; //remember the maximum index here too, for validation - VkBuffer buffer; //not owned, inherited from index buffer + uint32_t max_index = 0; // Remember the maximum index here too, for validation. + VkBuffer buffer; // Not owned, inherited from index buffer. uint32_t offset = 0; uint32_t indices = 0; VkIndexType index_type = VK_INDEX_TYPE_NONE_NV; @@ -539,16 +543,12 @@ class RenderingDeviceVulkan : public RenderingDevice { // As a result, we need to figure out quickly when something is no longer "compatible". // in order to avoid costly rebinds. - enum { - MAX_UNIFORM_SETS = 16 - }; - struct UniformInfo { UniformType type = UniformType::UNIFORM_TYPE_MAX; bool writable = false; int binding = 0; uint32_t stages = 0; - int length = 0; //size of arrays (in total elements), or ubos (in bytes * total elements) + int length = 0; // Size of arrays (in total elements), or ubos (in bytes * total elements). bool operator!=(const UniformInfo &p_info) const { return (binding != p_info.binding || type != p_info.type || writable != p_info.writable || stages != p_info.stages || length != p_info.length); @@ -620,12 +620,12 @@ class RenderingDeviceVulkan : public RenderingDevice { VkDescriptorSetLayout descriptor_set_layout = VK_NULL_HANDLE; }; - uint32_t vertex_input_mask = 0; //inputs used, this is mostly for validation + uint32_t vertex_input_mask = 0; // Inputs used, this is mostly for validation. uint32_t fragment_output_mask = 0; struct PushConstant { - uint32_t push_constant_size = 0; - uint32_t push_constants_vk_stage = 0; + uint32_t size = 0; + uint32_t vk_stages_mask = 0; }; PushConstant push_constant; @@ -643,7 +643,7 @@ class RenderingDeviceVulkan : public RenderingDevice { Vector<VkPipelineShaderStageCreateInfo> pipeline_stages; Vector<SpecializationConstant> specialization_constants; VkPipelineLayout pipeline_layout = VK_NULL_HANDLE; - String name; //used for debug + String name; // Used for debug. }; String _shader_uniform_debug(RID p_shader, int p_set = -1); @@ -715,7 +715,7 @@ class RenderingDeviceVulkan : public RenderingDevice { RID_Owner<Buffer, true> uniform_buffer_owner; RID_Owner<Buffer, true> storage_buffer_owner; - //texture buffer needs a view + // Texture buffer needs a view. struct TextureBuffer { Buffer buffer; VkBufferView view = VK_NULL_HANDLE; @@ -738,16 +738,16 @@ class RenderingDeviceVulkan : public RenderingDevice { DescriptorPool *pool = nullptr; DescriptorPoolKey pool_key; VkDescriptorSet descriptor_set = VK_NULL_HANDLE; - //VkPipelineLayout pipeline_layout; //not owned, inherited from shader + //VkPipelineLayout pipeline_layout; // Not owned, inherited from shader. struct AttachableTexture { uint32_t bind; RID texture; }; - LocalVector<AttachableTexture> attachable_textures; //used for validation - Vector<Texture *> mutable_sampled_textures; //used for layout change - Vector<Texture *> mutable_storage_textures; //used for layout change - UniformSetInvalidatedCallback invalidated_callback = nullptr; + LocalVector<AttachableTexture> attachable_textures; // Used for validation. + Vector<Texture *> mutable_sampled_textures; // Used for layout change. + Vector<Texture *> mutable_storage_textures; // Used for layout change. + InvalidationCallback invalidated_callback = nullptr; void *invalidated_callback_userdata = nullptr; }; @@ -769,7 +769,7 @@ class RenderingDeviceVulkan : public RenderingDevice { // was not supplied as intended. struct RenderPipeline { - //Cached values for validation + // Cached values for validation. #ifdef DEBUG_ENABLED struct Validation { FramebufferFormatID framebuffer_format = 0; @@ -781,13 +781,13 @@ class RenderingDeviceVulkan : public RenderingDevice { uint32_t primitive_divisor = 0; } validation; #endif - //Actual pipeline + // Actual pipeline. RID shader; Vector<uint32_t> set_formats; - VkPipelineLayout pipeline_layout = VK_NULL_HANDLE; // not owned, needed for push constants + VkPipelineLayout pipeline_layout = VK_NULL_HANDLE; // Not owned, needed for push constants. VkPipeline pipeline = VK_NULL_HANDLE; uint32_t push_constant_size = 0; - uint32_t push_constant_stages = 0; + uint32_t push_constant_stages_mask = 0; }; RID_Owner<RenderPipeline, true> render_pipeline_owner; @@ -795,10 +795,10 @@ class RenderingDeviceVulkan : public RenderingDevice { struct ComputePipeline { RID shader; Vector<uint32_t> set_formats; - VkPipelineLayout pipeline_layout = VK_NULL_HANDLE; // not owned, needed for push constants + VkPipelineLayout pipeline_layout = VK_NULL_HANDLE; // Not owned, needed for push constants. VkPipeline pipeline = VK_NULL_HANDLE; uint32_t push_constant_size = 0; - uint32_t push_constant_stages = 0; + uint32_t push_constant_stages_mask = 0; uint32_t local_group_size[3] = { 0, 0, 0 }; }; @@ -821,7 +821,7 @@ class RenderingDeviceVulkan : public RenderingDevice { struct SplitDrawListAllocator { VkCommandPool command_pool = VK_NULL_HANDLE; - Vector<VkCommandBuffer> command_buffers; //one for each frame + Vector<VkCommandBuffer> command_buffers; // One for each frame. }; Vector<SplitDrawListAllocator> split_draw_list_allocators; @@ -901,11 +901,11 @@ class RenderingDeviceVulkan : public RenderingDevice { bool draw_list_unbind_color_textures = false; bool draw_list_unbind_depth_textures = false; - void _draw_list_insert_clear_region(DrawList *draw_list, Framebuffer *framebuffer, Point2i viewport_offset, Point2i viewport_size, bool p_clear_color, const Vector<Color> &p_clear_colors, bool p_clear_depth, float p_depth, uint32_t p_stencil); + void _draw_list_insert_clear_region(DrawList *p_draw_list, Framebuffer *p_framebuffer, Point2i p_viewport_offset, Point2i p_viewport_size, bool p_clear_color, const Vector<Color> &p_clear_colors, bool p_clear_depth, float p_depth, uint32_t p_stencil); Error _draw_list_setup_framebuffer(Framebuffer *p_framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, VkFramebuffer *r_framebuffer, VkRenderPass *r_render_pass, uint32_t *r_subpass_count); Error _draw_list_render_pass_begin(Framebuffer *framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_colors, float p_clear_depth, uint32_t p_clear_stencil, Point2i viewport_offset, Point2i viewport_size, VkFramebuffer vkframebuffer, VkRenderPass render_pass, VkCommandBuffer command_buffer, VkSubpassContents subpass_contents, const Vector<RID> &p_storage_textures); _FORCE_INLINE_ DrawList *_get_draw_list_ptr(DrawListID p_id); - Buffer *_get_buffer_from_owner(RID p_buffer, VkPipelineStageFlags &dst_stage_mask, VkAccessFlags &dst_access, uint32_t p_post_barrier); + Buffer *_get_buffer_from_owner(RID p_buffer, VkPipelineStageFlags &dst_stage_mask, VkAccessFlags &dst_access, BitField<BarrierMask> p_post_barrier); Error _draw_list_allocate(const Rect2i &p_viewport, uint32_t p_splits, uint32_t p_subpass); void _draw_list_free(Rect2i *r_last_viewport = nullptr); @@ -973,7 +973,7 @@ class RenderingDeviceVulkan : public RenderingDevice { // when the frame is cycled. struct Frame { - //list in usage order, from last to free to first to free + // List in usage order, from last to free to first to free. List<Buffer> buffers_to_dispose_of; List<Texture> textures_to_dispose_of; List<Framebuffer> framebuffers_to_dispose_of; @@ -985,8 +985,8 @@ class RenderingDeviceVulkan : public RenderingDevice { List<ComputePipeline> compute_pipelines_to_dispose_of; VkCommandPool command_pool = VK_NULL_HANDLE; - VkCommandBuffer setup_command_buffer = VK_NULL_HANDLE; //used at the beginning of every frame for set-up - VkCommandBuffer draw_command_buffer = VK_NULL_HANDLE; //used at the beginning of every frame for set-up + VkCommandBuffer setup_command_buffer = VK_NULL_HANDLE; // Used at the beginning of every frame for set-up. + VkCommandBuffer draw_command_buffer = VK_NULL_HANDLE; // Used at the beginning of every frame for set-up. struct Timestamp { String description; @@ -1007,9 +1007,9 @@ class RenderingDeviceVulkan : public RenderingDevice { uint32_t max_timestamp_query_elements = 0; - TightLocalVector<Frame> frames; //frames available, for main device they are cycled (usually 3), for local devices only 1 - int frame = 0; //current frame - int frame_count = 0; //total amount of frames + TightLocalVector<Frame> frames; // Frames available, for main device they are cycled (usually 3), for local devices only 1. + int frame = 0; // Current frame. + int frame_count = 0; // Total amount of frames. uint64_t frames_drawn = 0; RID local_device; bool local_device_processing = false; @@ -1036,36 +1036,44 @@ class RenderingDeviceVulkan : public RenderingDevice { void _finalize_command_bufers(); void _begin_frame(); +#ifdef DEV_ENABLED + HashMap<RID, String> resource_names; +#endif + + VkSampleCountFlagBits _ensure_supported_sample_count(TextureSamples p_requested_sample_count) const; + public: virtual RID texture_create(const TextureFormat &p_format, const TextureView &p_view, const Vector<Vector<uint8_t>> &p_data = Vector<Vector<uint8_t>>()); virtual RID texture_create_shared(const TextureView &p_view, RID p_with_texture); virtual RID texture_create_from_extension(TextureType p_type, DataFormat p_format, TextureSamples p_samples, uint64_t p_flags, uint64_t p_image, uint64_t p_width, uint64_t p_height, uint64_t p_depth, uint64_t p_layers); virtual RID texture_create_shared_from_slice(const TextureView &p_view, RID p_with_texture, uint32_t p_layer, uint32_t p_mipmap, uint32_t p_mipmaps = 1, TextureSliceType p_slice_type = TEXTURE_SLICE_2D); - virtual Error texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, uint32_t p_post_barrier = BARRIER_MASK_ALL); + virtual Error texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, BitField<BarrierMask> p_post_barrier = BARRIER_MASK_ALL_BARRIERS); virtual Vector<uint8_t> texture_get_data(RID p_texture, uint32_t p_layer); - virtual bool texture_is_format_supported_for_usage(DataFormat p_format, uint32_t p_usage) const; + virtual bool texture_is_format_supported_for_usage(DataFormat p_format, BitField<RenderingDevice::TextureUsageBits> p_usage) const; virtual bool texture_is_shared(RID p_texture); virtual bool texture_is_valid(RID p_texture); virtual Size2i texture_size(RID p_texture); - virtual Error texture_copy(RID p_from_texture, RID p_to_texture, const Vector3 &p_from, const Vector3 &p_to, const Vector3 &p_size, uint32_t p_src_mipmap, uint32_t p_dst_mipmap, uint32_t p_src_layer, uint32_t p_dst_layer, uint32_t p_post_barrier = BARRIER_MASK_ALL); - virtual Error texture_clear(RID p_texture, const Color &p_color, uint32_t p_base_mipmap, uint32_t p_mipmaps, uint32_t p_base_layer, uint32_t p_layers, uint32_t p_post_barrier = BARRIER_MASK_ALL); - virtual Error texture_resolve_multisample(RID p_from_texture, RID p_to_texture, uint32_t p_post_barrier = BARRIER_MASK_ALL); + virtual Error texture_copy(RID p_from_texture, RID p_to_texture, const Vector3 &p_from, const Vector3 &p_to, const Vector3 &p_size, uint32_t p_src_mipmap, uint32_t p_dst_mipmap, uint32_t p_src_layer, uint32_t p_dst_layer, BitField<BarrierMask> p_post_barrier = BARRIER_MASK_ALL_BARRIERS); + virtual Error texture_clear(RID p_texture, const Color &p_color, uint32_t p_base_mipmap, uint32_t p_mipmaps, uint32_t p_base_layer, uint32_t p_layers, BitField<BarrierMask> p_post_barrier = BARRIER_MASK_ALL_BARRIERS); + virtual Error texture_resolve_multisample(RID p_from_texture, RID p_to_texture, BitField<BarrierMask> p_post_barrier = BARRIER_MASK_ALL_BARRIERS); /*********************/ /**** FRAMEBUFFER ****/ /*********************/ virtual FramebufferFormatID framebuffer_format_create(const Vector<AttachmentFormat> &p_format, uint32_t p_view_count = 1); - virtual FramebufferFormatID framebuffer_format_create_multipass(const Vector<AttachmentFormat> &p_attachments, Vector<FramebufferPass> &p_passes, uint32_t p_view_count = 1); + virtual FramebufferFormatID framebuffer_format_create_multipass(const Vector<AttachmentFormat> &p_attachments, const Vector<FramebufferPass> &p_passes, uint32_t p_view_count = 1); virtual FramebufferFormatID framebuffer_format_create_empty(TextureSamples p_samples = TEXTURE_SAMPLES_1); virtual TextureSamples framebuffer_format_get_texture_samples(FramebufferFormatID p_format, uint32_t p_pass = 0); virtual RID framebuffer_create(const Vector<RID> &p_texture_attachments, FramebufferFormatID p_format_check = INVALID_ID, uint32_t p_view_count = 1); - virtual RID framebuffer_create_multipass(const Vector<RID> &p_texture_attachments, Vector<FramebufferPass> &p_passes, FramebufferFormatID p_format_check = INVALID_ID, uint32_t p_view_count = 1); + virtual RID framebuffer_create_multipass(const Vector<RID> &p_texture_attachments, const Vector<FramebufferPass> &p_passes, FramebufferFormatID p_format_check = INVALID_ID, uint32_t p_view_count = 1); virtual RID framebuffer_create_empty(const Size2i &p_size, TextureSamples p_samples = TEXTURE_SAMPLES_1, FramebufferFormatID p_format_check = INVALID_ID); + virtual bool framebuffer_is_valid(RID p_framebuffer) const; + virtual void framebuffer_set_invalidation_callback(RID p_framebuffer, InvalidationCallback p_callback, void *p_userdata); virtual FramebufferFormatID framebuffer_get_format(RID p_framebuffer); @@ -1081,9 +1089,9 @@ public: virtual RID vertex_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data = Vector<uint8_t>(), bool p_use_as_storage = false); - // Internally reference counted, this ID is warranted to be unique for the same description, but needs to be freed as many times as it was allocated + // Internally reference counted, this ID is warranted to be unique for the same description, but needs to be freed as many times as it was allocated. virtual VertexFormatID vertex_format_create(const Vector<VertexAttribute> &p_vertex_formats); - virtual RID vertex_array_create(uint32_t p_vertex_count, VertexFormatID p_vertex_format, const Vector<RID> &p_src_buffers); + virtual RID vertex_array_create(uint32_t p_vertex_count, VertexFormatID p_vertex_format, const Vector<RID> &p_src_buffers, const Vector<uint64_t> &p_offsets = Vector<uint64_t>()); virtual RID index_buffer_create(uint32_t p_size_indices, IndexBufferFormat p_format, const Vector<uint8_t> &p_data = Vector<uint8_t>(), bool p_use_restart_indices = false); @@ -1105,22 +1113,22 @@ public: /*****************/ virtual RID uniform_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data = Vector<uint8_t>()); - virtual RID storage_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data = Vector<uint8_t>(), uint32_t p_usage = 0); + virtual RID storage_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data = Vector<uint8_t>(), BitField<StorageBufferUsage> p_usage = 0); virtual RID texture_buffer_create(uint32_t p_size_elements, DataFormat p_format, const Vector<uint8_t> &p_data = Vector<uint8_t>()); virtual RID uniform_set_create(const Vector<Uniform> &p_uniforms, RID p_shader, uint32_t p_shader_set); virtual bool uniform_set_is_valid(RID p_uniform_set); - virtual void uniform_set_set_invalidation_callback(RID p_uniform_set, UniformSetInvalidatedCallback p_callback, void *p_userdata); + virtual void uniform_set_set_invalidation_callback(RID p_uniform_set, InvalidationCallback p_callback, void *p_userdata); - virtual Error buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const void *p_data, uint32_t p_post_barrier = BARRIER_MASK_ALL); //works for any buffer - virtual Error buffer_clear(RID p_buffer, uint32_t p_offset, uint32_t p_size, uint32_t p_post_barrier = BARRIER_MASK_ALL); + virtual Error buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const void *p_data, BitField<BarrierMask> p_post_barrier = BARRIER_MASK_ALL_BARRIERS); // Works for any buffer. + virtual Error buffer_clear(RID p_buffer, uint32_t p_offset, uint32_t p_size, BitField<BarrierMask> p_post_barrier = BARRIER_MASK_ALL_BARRIERS); virtual Vector<uint8_t> buffer_get_data(RID p_buffer); /*************************/ /**** RENDER PIPELINE ****/ /*************************/ - virtual RID render_pipeline_create(RID p_shader, FramebufferFormatID p_framebuffer_format, VertexFormatID p_vertex_format, RenderPrimitive p_render_primitive, const PipelineRasterizationState &p_rasterization_state, const PipelineMultisampleState &p_multisample_state, const PipelineDepthStencilState &p_depth_stencil_state, const PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags = 0, uint32_t p_for_render_pass = 0, const Vector<PipelineSpecializationConstant> &p_specialization_constants = Vector<PipelineSpecializationConstant>()); + virtual RID render_pipeline_create(RID p_shader, FramebufferFormatID p_framebuffer_format, VertexFormatID p_vertex_format, RenderPrimitive p_render_primitive, const PipelineRasterizationState &p_rasterization_state, const PipelineMultisampleState &p_multisample_state, const PipelineDepthStencilState &p_depth_stencil_state, const PipelineColorBlendState &p_blend_state, BitField<PipelineDynamicStateFlags> p_dynamic_state_flags = 0, uint32_t p_for_render_pass = 0, const Vector<PipelineSpecializationConstant> &p_specialization_constants = Vector<PipelineSpecializationConstant>()); virtual bool render_pipeline_is_valid(RID p_pipeline); /**************************/ @@ -1147,6 +1155,7 @@ public: virtual DrawListID draw_list_begin(RID p_framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values = Vector<Color>(), float p_clear_depth = 1.0, uint32_t p_clear_stencil = 0, const Rect2 &p_region = Rect2(), const Vector<RID> &p_storage_textures = Vector<RID>()); virtual Error draw_list_begin_split(RID p_framebuffer, uint32_t p_splits, DrawListID *r_split_ids, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values = Vector<Color>(), float p_clear_depth = 1.0, uint32_t p_clear_stencil = 0, const Rect2 &p_region = Rect2(), const Vector<RID> &p_storage_textures = Vector<RID>()); + virtual void draw_list_set_blend_constants(DrawListID p_list, const Color &p_color); virtual void draw_list_bind_render_pipeline(DrawListID p_list, RID p_render_pipeline); virtual void draw_list_bind_uniform_set(DrawListID p_list, RID p_uniform_set, uint32_t p_index); virtual void draw_list_bind_vertex_array(DrawListID p_list, RID p_vertex_array); @@ -1163,7 +1172,7 @@ public: virtual DrawListID draw_list_switch_to_next_pass(); virtual Error draw_list_switch_to_next_pass_split(uint32_t p_splits, DrawListID *r_split_ids); - virtual void draw_list_end(uint32_t p_post_barrier = BARRIER_MASK_ALL); + virtual void draw_list_end(BitField<BarrierMask> p_post_barrier = BARRIER_MASK_ALL_BARRIERS); /***********************/ /**** COMPUTE LISTS ****/ @@ -1178,9 +1187,9 @@ public: virtual void compute_list_dispatch(ComputeListID p_list, uint32_t p_x_groups, uint32_t p_y_groups, uint32_t p_z_groups); virtual void compute_list_dispatch_threads(ComputeListID p_list, uint32_t p_x_threads, uint32_t p_y_threads, uint32_t p_z_threads); virtual void compute_list_dispatch_indirect(ComputeListID p_list, RID p_buffer, uint32_t p_offset); - virtual void compute_list_end(uint32_t p_post_barrier = BARRIER_MASK_ALL); + virtual void compute_list_end(BitField<BarrierMask> p_post_barrier = BARRIER_MASK_ALL_BARRIERS); - virtual void barrier(uint32_t p_from = BARRIER_MASK_ALL, uint32_t p_to = BARRIER_MASK_ALL); + virtual void barrier(BitField<BarrierMask> p_from = BARRIER_MASK_ALL_BARRIERS, BitField<BarrierMask> p_to = BARRIER_MASK_ALL_BARRIERS); virtual void full_barrier(); /**************/ @@ -1210,10 +1219,10 @@ public: void initialize(VulkanContext *p_context, bool p_local_device = false); void finalize(); - virtual void swap_buffers(); //for main device + virtual void swap_buffers(); // For main device. - virtual void submit(); //for local device - virtual void sync(); //for local device + virtual void submit(); // For local device. + virtual void sync(); // For local device. virtual uint32_t get_frame_delay() const; diff --git a/drivers/vulkan/vulkan_context.cpp b/drivers/vulkan/vulkan_context.cpp index 0d8a3310fd..0bccefede8 100644 --- a/drivers/vulkan/vulkan_context.cpp +++ b/drivers/vulkan/vulkan_context.cpp @@ -1,32 +1,32 @@ -/*************************************************************************/ -/* vulkan_context.cpp */ -/*************************************************************************/ -/* This file is part of: */ -/* GODOT ENGINE */ -/* https://godotengine.org */ -/*************************************************************************/ -/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ -/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ -/* */ -/* Permission is hereby granted, free of charge, to any person obtaining */ -/* a copy of this software and associated documentation files (the */ -/* "Software"), to deal in the Software without restriction, including */ -/* without limitation the rights to use, copy, modify, merge, publish, */ -/* distribute, sublicense, and/or sell copies of the Software, and to */ -/* permit persons to whom the Software is furnished to do so, subject to */ -/* the following conditions: */ -/* */ -/* The above copyright notice and this permission notice shall be */ -/* included in all copies or substantial portions of the Software. */ -/* */ -/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ -/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ -/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ -/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ -/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ -/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ -/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/*************************************************************************/ +/**************************************************************************/ +/* vulkan_context.cpp */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ #include "vulkan_context.h" @@ -48,15 +48,119 @@ VulkanHooks *VulkanContext::vulkan_hooks = nullptr; -VkResult VulkanContext::vkCreateRenderPass2KHR(VkDevice device, const VkRenderPassCreateInfo2 *pCreateInfo, const VkAllocationCallbacks *pAllocator, VkRenderPass *pRenderPass) { - if (fpCreateRenderPass2KHR == nullptr) { - fpCreateRenderPass2KHR = (PFN_vkCreateRenderPass2KHR)vkGetInstanceProcAddr(inst, "vkCreateRenderPass2KHR"); +Vector<VkAttachmentReference> VulkanContext::_convert_VkAttachmentReference2(uint32_t p_count, const VkAttachmentReference2 *p_refs) { + Vector<VkAttachmentReference> att_refs; + + if (p_refs != nullptr) { + for (uint32_t i = 0; i < p_count; i++) { + // We lose aspectMask in this conversion but we don't use it currently. + + VkAttachmentReference ref = { + p_refs[i].attachment, /* attachment */ + p_refs[i].layout /* layout */ + }; + + att_refs.push_back(ref); + } } - if (fpCreateRenderPass2KHR == nullptr) { - return VK_ERROR_EXTENSION_NOT_PRESENT; + return att_refs; +} + +VkResult VulkanContext::vkCreateRenderPass2KHR(VkDevice p_device, const VkRenderPassCreateInfo2 *p_create_info, const VkAllocationCallbacks *p_allocator, VkRenderPass *p_render_pass) { + if (is_device_extension_enabled(VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME)) { + if (fpCreateRenderPass2KHR == nullptr) { + fpCreateRenderPass2KHR = (PFN_vkCreateRenderPass2KHR)vkGetDeviceProcAddr(p_device, "vkCreateRenderPass2KHR"); + } + + if (fpCreateRenderPass2KHR == nullptr) { + return VK_ERROR_EXTENSION_NOT_PRESENT; + } else { + return (fpCreateRenderPass2KHR)(p_device, p_create_info, p_allocator, p_render_pass); + } } else { - return (fpCreateRenderPass2KHR)(device, pCreateInfo, pAllocator, pRenderPass); + // need to fall back on vkCreateRenderPass + + const void *next = p_create_info->pNext; // ATM we only support multiview which should work if supported. + + Vector<VkAttachmentDescription> attachments; + for (uint32_t i = 0; i < p_create_info->attachmentCount; i++) { + // Basically the old layout just misses type and next. + VkAttachmentDescription att = { + p_create_info->pAttachments[i].flags, /* flags */ + p_create_info->pAttachments[i].format, /* format */ + p_create_info->pAttachments[i].samples, /* samples */ + p_create_info->pAttachments[i].loadOp, /* loadOp */ + p_create_info->pAttachments[i].storeOp, /* storeOp */ + p_create_info->pAttachments[i].stencilLoadOp, /* stencilLoadOp */ + p_create_info->pAttachments[i].stencilStoreOp, /* stencilStoreOp */ + p_create_info->pAttachments[i].initialLayout, /* initialLayout */ + p_create_info->pAttachments[i].finalLayout /* finalLayout */ + }; + + attachments.push_back(att); + } + + Vector<VkSubpassDescription> subpasses; + for (uint32_t i = 0; i < p_create_info->subpassCount; i++) { + // Here we need to do more, again it's just stripping out type and next + // but we have VkAttachmentReference2 to convert to VkAttachmentReference. + // Also viewmask is not supported but we don't use it outside of multiview. + + Vector<VkAttachmentReference> input_attachments = _convert_VkAttachmentReference2(p_create_info->pSubpasses[i].inputAttachmentCount, p_create_info->pSubpasses[i].pInputAttachments); + Vector<VkAttachmentReference> color_attachments = _convert_VkAttachmentReference2(p_create_info->pSubpasses[i].colorAttachmentCount, p_create_info->pSubpasses[i].pColorAttachments); + Vector<VkAttachmentReference> resolve_attachments = _convert_VkAttachmentReference2(p_create_info->pSubpasses[i].colorAttachmentCount, p_create_info->pSubpasses[i].pResolveAttachments); + Vector<VkAttachmentReference> depth_attachments = _convert_VkAttachmentReference2(p_create_info->pSubpasses[i].colorAttachmentCount, p_create_info->pSubpasses[i].pDepthStencilAttachment); + + VkSubpassDescription subpass = { + p_create_info->pSubpasses[i].flags, /* flags */ + p_create_info->pSubpasses[i].pipelineBindPoint, /* pipelineBindPoint */ + p_create_info->pSubpasses[i].inputAttachmentCount, /* inputAttachmentCount */ + input_attachments.size() == 0 ? nullptr : input_attachments.ptr(), /* pInputAttachments */ + p_create_info->pSubpasses[i].colorAttachmentCount, /* colorAttachmentCount */ + color_attachments.size() == 0 ? nullptr : color_attachments.ptr(), /* pColorAttachments */ + resolve_attachments.size() == 0 ? nullptr : resolve_attachments.ptr(), /* pResolveAttachments */ + depth_attachments.size() == 0 ? nullptr : depth_attachments.ptr(), /* pDepthStencilAttachment */ + p_create_info->pSubpasses[i].preserveAttachmentCount, /* preserveAttachmentCount */ + p_create_info->pSubpasses[i].pPreserveAttachments /* pPreserveAttachments */ + }; + + subpasses.push_back(subpass); + } + + Vector<VkSubpassDependency> dependencies; + for (uint32_t i = 0; i < p_create_info->dependencyCount; i++) { + // We lose viewOffset here but again I don't believe we use this anywhere. + VkSubpassDependency dep = { + p_create_info->pDependencies[i].srcSubpass, /* srcSubpass */ + p_create_info->pDependencies[i].dstSubpass, /* dstSubpass */ + p_create_info->pDependencies[i].srcStageMask, /* srcStageMask */ + p_create_info->pDependencies[i].dstStageMask, /* dstStageMask */ + p_create_info->pDependencies[i].srcAccessMask, /* srcAccessMask */ + p_create_info->pDependencies[i].dstAccessMask, /* dstAccessMask */ + p_create_info->pDependencies[i].dependencyFlags, /* dependencyFlags */ + }; + + dependencies.push_back(dep); + } + + // CorrelatedViewMask is not supported in vkCreateRenderPass but we + // currently only use this for multiview. + // We'll need to look into this. + + VkRenderPassCreateInfo create_info = { + VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO, /* sType */ + next, /* pNext*/ + p_create_info->flags, /* flags */ + (uint32_t)attachments.size(), /* attachmentCount */ + attachments.ptr(), /* pAttachments */ + (uint32_t)subpasses.size(), /* subpassCount */ + subpasses.ptr(), /* pSubpasses */ + (uint32_t)dependencies.size(), /* */ + dependencies.ptr(), /* */ + }; + + return vkCreateRenderPass(device, &create_info, p_allocator, p_render_pass); } } @@ -85,19 +189,6 @@ VKAPI_ATTR VkBool32 VKAPI_CALL VulkanContext::_debug_messenger_callback( return VK_FALSE; } - // Workaround for Vulkan-Loader usability bug: https://github.com/KhronosGroup/Vulkan-Loader/issues/262. - if (strstr(pCallbackData->pMessage, "wrong ELF class: ELFCLASS32") != nullptr) { - return VK_FALSE; - } - -#ifdef WINDOWS_ENABLED - // Some software installs Vulkan overlays in Windows registry and never cleans them up on uninstall. - // So we get spammy error level messages from the loader about those - make them verbose instead. - if (strstr(pCallbackData->pMessage, "loader_get_json: Failed to open JSON file") != nullptr) { - messageSeverity = VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT; - } -#endif - if (pCallbackData->pMessageIdName && strstr(pCallbackData->pMessageIdName, "UNASSIGNED-CoreValidation-DrawState-ClearCmdBeforeDraw") != nullptr) { return VK_FALSE; } @@ -237,7 +328,7 @@ Error VulkanContext::_get_preferred_validation_layers(uint32_t *count, const cha { "VK_LAYER_GOOGLE_threading", "VK_LAYER_LUNARG_parameter_validation", "VK_LAYER_LUNARG_object_tracker", "VK_LAYER_LUNARG_core_validation", "VK_LAYER_GOOGLE_unique_objects" } }; - // Clear out-arguments + // Clear out-arguments. *count = 0; if (names != nullptr) { *names = nullptr; @@ -287,9 +378,7 @@ Error VulkanContext::_obtain_vulkan_version() { uint32_t api_version; VkResult res = func(&api_version); if (res == VK_SUCCESS) { - vulkan_major = VK_API_VERSION_MAJOR(api_version); - vulkan_minor = VK_API_VERSION_MINOR(api_version); - vulkan_patch = VK_API_VERSION_PATCH(api_version); + instance_api_version = api_version; } else { // According to the documentation this shouldn't fail with anything except a memory allocation error // in which case we're in deep trouble anyway. @@ -297,74 +386,179 @@ Error VulkanContext::_obtain_vulkan_version() { } } else { print_line("vkEnumerateInstanceVersion not available, assuming Vulkan 1.0."); - } - - // We don't go above 1.2. - if ((vulkan_major > 1) || (vulkan_major == 1 && vulkan_minor > 2)) { - vulkan_major = 1; - vulkan_minor = 2; - vulkan_patch = 0; + instance_api_version = VK_API_VERSION_1_0; } return OK; } -Error VulkanContext::_initialize_extensions() { - uint32_t instance_extension_count = 0; +bool VulkanContext::instance_extensions_initialized = false; +HashMap<CharString, bool> VulkanContext::requested_instance_extensions; + +void VulkanContext::register_requested_instance_extension(const CharString &extension_name, bool p_required) { + ERR_FAIL_COND_MSG(instance_extensions_initialized, "You can only registered extensions before the Vulkan instance is created"); + ERR_FAIL_COND(requested_instance_extensions.has(extension_name)); + + requested_instance_extensions[extension_name] = p_required; +} + +Error VulkanContext::_initialize_instance_extensions() { + enabled_instance_extension_names.clear(); + + // Make sure our core extensions are here + register_requested_instance_extension(VK_KHR_SURFACE_EXTENSION_NAME, true); + register_requested_instance_extension(_get_platform_surface_extension(), true); + + if (_use_validation_layers()) { + register_requested_instance_extension(VK_EXT_DEBUG_REPORT_EXTENSION_NAME, false); + } - enabled_extension_count = 0; - enabled_debug_utils = false; - enabled_debug_report = false; - // Look for instance extensions. - VkBool32 surfaceExtFound = 0; - VkBool32 platformSurfaceExtFound = 0; - memset(extension_names, 0, sizeof(extension_names)); + register_requested_instance_extension(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME, false); + // Only enable debug utils in verbose mode or DEV_ENABLED. + // End users would get spammed with messages of varying verbosity due to the + // mess that thirdparty layers/extensions and drivers seem to leave in their + // wake, making the Windows registry a bottomless pit of broken layer JSON. +#ifdef DEV_ENABLED + bool want_debug_utils = true; +#else + bool want_debug_utils = OS::get_singleton()->is_stdout_verbose(); +#endif + if (want_debug_utils) { + register_requested_instance_extension(VK_EXT_DEBUG_UTILS_EXTENSION_NAME, false); + } + + // Load instance extensions that are available... + uint32_t instance_extension_count = 0; VkResult err = vkEnumerateInstanceExtensionProperties(nullptr, &instance_extension_count, nullptr); ERR_FAIL_COND_V(err != VK_SUCCESS && err != VK_INCOMPLETE, ERR_CANT_CREATE); + ERR_FAIL_COND_V_MSG(instance_extension_count == 0, ERR_CANT_CREATE, "No instance extensions found, is a driver installed?"); - if (instance_extension_count > 0) { - VkExtensionProperties *instance_extensions = (VkExtensionProperties *)malloc(sizeof(VkExtensionProperties) * instance_extension_count); - err = vkEnumerateInstanceExtensionProperties(nullptr, &instance_extension_count, instance_extensions); - if (err != VK_SUCCESS && err != VK_INCOMPLETE) { - free(instance_extensions); - ERR_FAIL_V(ERR_CANT_CREATE); + VkExtensionProperties *instance_extensions = (VkExtensionProperties *)malloc(sizeof(VkExtensionProperties) * instance_extension_count); + err = vkEnumerateInstanceExtensionProperties(nullptr, &instance_extension_count, instance_extensions); + if (err != VK_SUCCESS && err != VK_INCOMPLETE) { + free(instance_extensions); + ERR_FAIL_V(ERR_CANT_CREATE); + } +#ifdef DEV_ENABLED + for (uint32_t i = 0; i < instance_extension_count; i++) { + print_verbose(String("VULKAN: Found instance extension ") + String(instance_extensions[i].extensionName)); + } +#endif + + // Enable all extensions that are supported and requested + for (uint32_t i = 0; i < instance_extension_count; i++) { + CharString extension_name(instance_extensions[i].extensionName); + if (requested_instance_extensions.has(extension_name)) { + enabled_instance_extension_names.insert(extension_name); } - for (uint32_t i = 0; i < instance_extension_count; i++) { - if (!strcmp(VK_KHR_SURFACE_EXTENSION_NAME, instance_extensions[i].extensionName)) { - surfaceExtFound = 1; - extension_names[enabled_extension_count++] = VK_KHR_SURFACE_EXTENSION_NAME; - } + } - if (!strcmp(_get_platform_surface_extension(), instance_extensions[i].extensionName)) { - platformSurfaceExtFound = 1; - extension_names[enabled_extension_count++] = _get_platform_surface_extension(); - } - if (!strcmp(VK_EXT_DEBUG_REPORT_EXTENSION_NAME, instance_extensions[i].extensionName)) { - if (_use_validation_layers()) { - extension_names[enabled_extension_count++] = VK_EXT_DEBUG_REPORT_EXTENSION_NAME; - enabled_debug_report = true; - } - } - if (!strcmp(VK_EXT_DEBUG_UTILS_EXTENSION_NAME, instance_extensions[i].extensionName)) { - extension_names[enabled_extension_count++] = VK_EXT_DEBUG_UTILS_EXTENSION_NAME; - enabled_debug_utils = true; - } - if (!strcmp(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME, instance_extensions[i].extensionName)) { - extension_names[enabled_extension_count++] = VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME; - } - if (enabled_extension_count >= MAX_EXTENSIONS) { + // Now check our requested extensions + for (KeyValue<CharString, bool> &requested_extension : requested_instance_extensions) { + if (!enabled_instance_extension_names.has(requested_extension.key)) { + if (requested_extension.value) { free(instance_extensions); - ERR_FAIL_V_MSG(ERR_BUG, "Enabled extension count reaches MAX_EXTENSIONS, BUG"); + ERR_FAIL_V_MSG(ERR_BUG, String("Required extension ") + String(requested_extension.key) + String(" not found, is a driver installed?")); + } else { + print_verbose(String("Optional extension ") + String(requested_extension.key) + String(" not found")); } } + } - free(instance_extensions); + free(instance_extensions); + + instance_extensions_initialized = true; + return OK; +} + +bool VulkanContext::device_extensions_initialized = false; +HashMap<CharString, bool> VulkanContext::requested_device_extensions; + +void VulkanContext::register_requested_device_extension(const CharString &extension_name, bool p_required) { + ERR_FAIL_COND_MSG(device_extensions_initialized, "You can only registered extensions before the Vulkan instance is created"); + ERR_FAIL_COND(requested_device_extensions.has(extension_name)); + + requested_device_extensions[extension_name] = p_required; +} + +Error VulkanContext::_initialize_device_extensions() { + // Look for device extensions. + enabled_device_extension_names.clear(); + + // Make sure our core extensions are here + register_requested_device_extension(VK_KHR_SWAPCHAIN_EXTENSION_NAME, true); + + register_requested_device_extension(VK_KHR_MULTIVIEW_EXTENSION_NAME, false); + register_requested_device_extension(VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME, false); + register_requested_device_extension(VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME, false); + register_requested_device_extension(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME, false); + register_requested_device_extension(VK_KHR_16BIT_STORAGE_EXTENSION_NAME, false); + + // TODO consider the following extensions: + // - VK_KHR_spirv_1_4 + // - VK_KHR_swapchain_mutable_format + // - VK_EXT_full_screen_exclusive + // - VK_EXT_hdr_metadata + // - VK_KHR_depth_stencil_resolve + + // Even though the user "enabled" the extension via the command + // line, we must make sure that it's enumerated for use with the + // device. Therefore, disable it here, and re-enable it again if + // enumerated. + if (VK_KHR_incremental_present_enabled) { + register_requested_device_extension(VK_KHR_INCREMENTAL_PRESENT_EXTENSION_NAME, false); + } + if (VK_GOOGLE_display_timing_enabled) { + register_requested_device_extension(VK_GOOGLE_DISPLAY_TIMING_EXTENSION_NAME, false); } - ERR_FAIL_COND_V_MSG(!surfaceExtFound, ERR_CANT_CREATE, "No surface extension found, is a driver installed?"); - ERR_FAIL_COND_V_MSG(!platformSurfaceExtFound, ERR_CANT_CREATE, "No platform surface extension found, is a driver installed?"); + // obtain available device extensions + uint32_t device_extension_count = 0; + VkResult err = vkEnumerateDeviceExtensionProperties(gpu, nullptr, &device_extension_count, nullptr); + ERR_FAIL_COND_V(err, ERR_CANT_CREATE); + ERR_FAIL_COND_V_MSG(device_extension_count == 0, ERR_CANT_CREATE, + "vkEnumerateDeviceExtensionProperties failed to find any extensions\n\n" + "Do you have a compatible Vulkan installable client driver (ICD) installed?\n" + "vkCreateInstance Failure"); + VkExtensionProperties *device_extensions = (VkExtensionProperties *)malloc(sizeof(VkExtensionProperties) * device_extension_count); + err = vkEnumerateDeviceExtensionProperties(gpu, nullptr, &device_extension_count, device_extensions); + if (err) { + free(device_extensions); + ERR_FAIL_V(ERR_CANT_CREATE); + } + +#ifdef DEV_ENABLED + for (uint32_t i = 0; i < device_extension_count; i++) { + print_verbose(String("VULKAN: Found device extension ") + String(device_extensions[i].extensionName)); + } +#endif + + // Enable all extensions that are supported and requested + for (uint32_t i = 0; i < device_extension_count; i++) { + CharString extension_name(device_extensions[i].extensionName); + if (requested_device_extensions.has(extension_name)) { + enabled_device_extension_names.insert(extension_name); + } + } + + // Now check our requested extensions + for (KeyValue<CharString, bool> &requested_extension : requested_device_extensions) { + if (!enabled_device_extension_names.has(requested_extension.key)) { + if (requested_extension.value) { + free(device_extensions); + ERR_FAIL_V_MSG(ERR_BUG, + String("vkEnumerateDeviceExtensionProperties failed to find the ") + String(requested_extension.key) + String(" extension.\n\nDo you have a compatible Vulkan installable client driver (ICD) installed?\nvkCreateInstance Failure")); + } else { + print_verbose(String("Optional extension ") + String(requested_extension.key) + String(" not found")); + } + } + } + + free(device_extensions); + + device_extensions_initialized = true; return OK; } @@ -441,7 +635,7 @@ String VulkanContext::SubgroupCapabilities::supported_stages_desc() const { res += ", STAGE_MESH_NV"; } - return res.substr(2); // Remove first ", " + return res.substr(2); // Remove first ", ". } uint32_t VulkanContext::SubgroupCapabilities::supported_operations_flags_rd() const { @@ -506,7 +700,7 @@ String VulkanContext::SubgroupCapabilities::supported_operations_desc() const { res += ", FEATURE_PARTITIONED_NV"; } - return res.substr(2); // Remove first ", " + return res.substr(2); // Remove first ", ". } Error VulkanContext::_check_capabilities() { @@ -522,6 +716,9 @@ Error VulkanContext::_check_capabilities() { vrs_capabilities.pipeline_vrs_supported = false; vrs_capabilities.primitive_vrs_supported = false; vrs_capabilities.attachment_vrs_supported = false; + vrs_capabilities.min_texel_size = Size2i(); + vrs_capabilities.max_texel_size = Size2i(); + vrs_capabilities.texel_size = Size2i(); multiview_capabilities.is_supported = false; multiview_capabilities.geometry_shader_is_supported = false; multiview_capabilities.tessellation_shader_is_supported = false; @@ -538,155 +735,176 @@ Error VulkanContext::_check_capabilities() { storage_buffer_capabilities.storage_push_constant_16_is_supported = false; storage_buffer_capabilities.storage_input_output_16 = false; - // Check for extended features. - PFN_vkGetPhysicalDeviceFeatures2 vkGetPhysicalDeviceFeatures2_func = (PFN_vkGetPhysicalDeviceFeatures2)vkGetInstanceProcAddr(inst, "vkGetPhysicalDeviceFeatures2"); - if (vkGetPhysicalDeviceFeatures2_func == nullptr) { - // In Vulkan 1.0 might be accessible under its original extension name. - vkGetPhysicalDeviceFeatures2_func = (PFN_vkGetPhysicalDeviceFeatures2)vkGetInstanceProcAddr(inst, "vkGetPhysicalDeviceFeatures2KHR"); - } - if (vkGetPhysicalDeviceFeatures2_func != nullptr) { - // Check our extended features. - VkPhysicalDeviceFragmentShadingRateFeaturesKHR vrs_features = { - /*sType*/ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FRAGMENT_SHADING_RATE_FEATURES_KHR, - /*pNext*/ nullptr, - /*pipelineFragmentShadingRate*/ false, - /*primitiveFragmentShadingRate*/ false, - /*attachmentFragmentShadingRate*/ false, - }; - - VkPhysicalDeviceShaderFloat16Int8FeaturesKHR shader_features = { - /*sType*/ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_FLOAT16_INT8_FEATURES_KHR, - /*pNext*/ &vrs_features, - /*shaderFloat16*/ false, - /*shaderInt8*/ false, - }; + if (is_instance_extension_enabled(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME)) { + // Check for extended features. + PFN_vkGetPhysicalDeviceFeatures2 vkGetPhysicalDeviceFeatures2_func = (PFN_vkGetPhysicalDeviceFeatures2)vkGetInstanceProcAddr(inst, "vkGetPhysicalDeviceFeatures2"); + if (vkGetPhysicalDeviceFeatures2_func == nullptr) { + // In Vulkan 1.0 might be accessible under its original extension name. + vkGetPhysicalDeviceFeatures2_func = (PFN_vkGetPhysicalDeviceFeatures2)vkGetInstanceProcAddr(inst, "vkGetPhysicalDeviceFeatures2KHR"); + } + if (vkGetPhysicalDeviceFeatures2_func != nullptr) { + // Check our extended features. + VkPhysicalDeviceFragmentShadingRateFeaturesKHR vrs_features = { + /*sType*/ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FRAGMENT_SHADING_RATE_FEATURES_KHR, + /*pNext*/ nullptr, + /*pipelineFragmentShadingRate*/ false, + /*primitiveFragmentShadingRate*/ false, + /*attachmentFragmentShadingRate*/ false, + }; - VkPhysicalDevice16BitStorageFeaturesKHR storage_feature = { - /*sType*/ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES_KHR, - /*pNext*/ &shader_features, - /*storageBuffer16BitAccess*/ false, - /*uniformAndStorageBuffer16BitAccess*/ false, - /*storagePushConstant16*/ false, - /*storageInputOutput16*/ false, - }; + VkPhysicalDeviceShaderFloat16Int8FeaturesKHR shader_features = { + /*sType*/ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_FLOAT16_INT8_FEATURES_KHR, + /*pNext*/ &vrs_features, + /*shaderFloat16*/ false, + /*shaderInt8*/ false, + }; - VkPhysicalDeviceMultiviewFeatures multiview_features = { - /*sType*/ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES, - /*pNext*/ &storage_feature, - /*multiview*/ false, - /*multiviewGeometryShader*/ false, - /*multiviewTessellationShader*/ false, - }; + VkPhysicalDevice16BitStorageFeaturesKHR storage_feature = { + /*sType*/ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES_KHR, + /*pNext*/ &shader_features, + /*storageBuffer16BitAccess*/ false, + /*uniformAndStorageBuffer16BitAccess*/ false, + /*storagePushConstant16*/ false, + /*storageInputOutput16*/ false, + }; - VkPhysicalDeviceFeatures2 device_features; - device_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; - device_features.pNext = &multiview_features; + VkPhysicalDeviceMultiviewFeatures multiview_features = { + /*sType*/ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES, + /*pNext*/ &storage_feature, + /*multiview*/ false, + /*multiviewGeometryShader*/ false, + /*multiviewTessellationShader*/ false, + }; - vkGetPhysicalDeviceFeatures2_func(gpu, &device_features); + VkPhysicalDeviceFeatures2 device_features; + device_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; + device_features.pNext = &multiview_features; - vrs_capabilities.pipeline_vrs_supported = vrs_features.pipelineFragmentShadingRate; - vrs_capabilities.primitive_vrs_supported = vrs_features.primitiveFragmentShadingRate; - vrs_capabilities.attachment_vrs_supported = vrs_features.attachmentFragmentShadingRate; + vkGetPhysicalDeviceFeatures2_func(gpu, &device_features); - multiview_capabilities.is_supported = multiview_features.multiview; - multiview_capabilities.geometry_shader_is_supported = multiview_features.multiviewGeometryShader; - multiview_capabilities.tessellation_shader_is_supported = multiview_features.multiviewTessellationShader; + // We must check that the relative extension is present before assuming a + // feature as enabled. Actually, according to the spec we shouldn't add the + // structs in pNext at all, but this works fine. + // See also: https://github.com/godotengine/godot/issues/65409 + if (is_device_extension_enabled(VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME)) { + vrs_capabilities.pipeline_vrs_supported = vrs_features.pipelineFragmentShadingRate; + vrs_capabilities.primitive_vrs_supported = vrs_features.primitiveFragmentShadingRate; + vrs_capabilities.attachment_vrs_supported = vrs_features.attachmentFragmentShadingRate; + } - shader_capabilities.shader_float16_is_supported = shader_features.shaderFloat16; - shader_capabilities.shader_int8_is_supported = shader_features.shaderInt8; + if (is_device_extension_enabled(VK_KHR_MULTIVIEW_EXTENSION_NAME)) { + multiview_capabilities.is_supported = multiview_features.multiview; + multiview_capabilities.geometry_shader_is_supported = multiview_features.multiviewGeometryShader; + multiview_capabilities.tessellation_shader_is_supported = multiview_features.multiviewTessellationShader; + } - storage_buffer_capabilities.storage_buffer_16_bit_access_is_supported = storage_feature.storageBuffer16BitAccess; - storage_buffer_capabilities.uniform_and_storage_buffer_16_bit_access_is_supported = storage_feature.uniformAndStorageBuffer16BitAccess; - storage_buffer_capabilities.storage_push_constant_16_is_supported = storage_feature.storagePushConstant16; - storage_buffer_capabilities.storage_input_output_16 = storage_feature.storageInputOutput16; - } + if (is_device_extension_enabled(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME)) { + shader_capabilities.shader_float16_is_supported = shader_features.shaderFloat16; + shader_capabilities.shader_int8_is_supported = shader_features.shaderInt8; + } - // Check extended properties. - PFN_vkGetPhysicalDeviceProperties2 device_properties_func = (PFN_vkGetPhysicalDeviceProperties2)vkGetInstanceProcAddr(inst, "vkGetPhysicalDeviceProperties2"); - if (device_properties_func == nullptr) { - // In Vulkan 1.0 might be accessible under its original extension name. - device_properties_func = (PFN_vkGetPhysicalDeviceProperties2)vkGetInstanceProcAddr(inst, "vkGetPhysicalDeviceProperties2KHR"); - } - if (device_properties_func != nullptr) { - VkPhysicalDeviceFragmentShadingRatePropertiesKHR vrsProperties; - VkPhysicalDeviceMultiviewProperties multiviewProperties; - VkPhysicalDeviceSubgroupProperties subgroupProperties; - VkPhysicalDeviceProperties2 physicalDeviceProperties; - void *nextptr = nullptr; + if (is_device_extension_enabled(VK_KHR_16BIT_STORAGE_EXTENSION_NAME)) { + storage_buffer_capabilities.storage_buffer_16_bit_access_is_supported = storage_feature.storageBuffer16BitAccess; + storage_buffer_capabilities.uniform_and_storage_buffer_16_bit_access_is_supported = storage_feature.uniformAndStorageBuffer16BitAccess; + storage_buffer_capabilities.storage_push_constant_16_is_supported = storage_feature.storagePushConstant16; + storage_buffer_capabilities.storage_input_output_16 = storage_feature.storageInputOutput16; + } + } - subgroupProperties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_PROPERTIES; - subgroupProperties.pNext = nextptr; - nextptr = &subgroupProperties; + // Check extended properties. + PFN_vkGetPhysicalDeviceProperties2 device_properties_func = (PFN_vkGetPhysicalDeviceProperties2)vkGetInstanceProcAddr(inst, "vkGetPhysicalDeviceProperties2"); + if (device_properties_func == nullptr) { + // In Vulkan 1.0 might be accessible under its original extension name. + device_properties_func = (PFN_vkGetPhysicalDeviceProperties2)vkGetInstanceProcAddr(inst, "vkGetPhysicalDeviceProperties2KHR"); + } + if (device_properties_func != nullptr) { + VkPhysicalDeviceFragmentShadingRatePropertiesKHR vrsProperties{}; + VkPhysicalDeviceMultiviewProperties multiviewProperties{}; + VkPhysicalDeviceSubgroupProperties subgroupProperties{}; + VkPhysicalDeviceProperties2 physicalDeviceProperties{}; + void *nextptr = nullptr; + + if (device_api_version >= VK_API_VERSION_1_1) { // Vulkan 1.1 or higher + subgroupProperties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_PROPERTIES; + subgroupProperties.pNext = nextptr; + + nextptr = &subgroupProperties; + } - if (multiview_capabilities.is_supported) { - multiviewProperties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_PROPERTIES; - multiviewProperties.pNext = nextptr; + if (multiview_capabilities.is_supported) { + multiviewProperties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_PROPERTIES; + multiviewProperties.pNext = nextptr; - nextptr = &multiviewProperties; - } + nextptr = &multiviewProperties; + } - if (vrs_capabilities.attachment_vrs_supported) { - vrsProperties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FRAGMENT_SHADING_RATE_PROPERTIES_KHR; - vrsProperties.pNext = nextptr; + if (vrs_capabilities.attachment_vrs_supported) { + vrsProperties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FRAGMENT_SHADING_RATE_PROPERTIES_KHR; + vrsProperties.pNext = nextptr; - nextptr = &vrsProperties; - } + nextptr = &vrsProperties; + } - physicalDeviceProperties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2; - physicalDeviceProperties.pNext = nextptr; + physicalDeviceProperties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2; + physicalDeviceProperties.pNext = nextptr; - device_properties_func(gpu, &physicalDeviceProperties); + device_properties_func(gpu, &physicalDeviceProperties); - subgroup_capabilities.size = subgroupProperties.subgroupSize; - subgroup_capabilities.supportedStages = subgroupProperties.supportedStages; - subgroup_capabilities.supportedOperations = subgroupProperties.supportedOperations; - // Note: quadOperationsInAllStages will be true if: - // - supportedStages has VK_SHADER_STAGE_ALL_GRAPHICS + VK_SHADER_STAGE_COMPUTE_BIT - // - supportedOperations has VK_SUBGROUP_FEATURE_QUAD_BIT - subgroup_capabilities.quadOperationsInAllStages = subgroupProperties.quadOperationsInAllStages; + subgroup_capabilities.size = subgroupProperties.subgroupSize; + subgroup_capabilities.supportedStages = subgroupProperties.supportedStages; + subgroup_capabilities.supportedOperations = subgroupProperties.supportedOperations; + // Note: quadOperationsInAllStages will be true if: + // - supportedStages has VK_SHADER_STAGE_ALL_GRAPHICS + VK_SHADER_STAGE_COMPUTE_BIT. + // - supportedOperations has VK_SUBGROUP_FEATURE_QUAD_BIT. + subgroup_capabilities.quadOperationsInAllStages = subgroupProperties.quadOperationsInAllStages; - if (vrs_capabilities.pipeline_vrs_supported || vrs_capabilities.primitive_vrs_supported || vrs_capabilities.attachment_vrs_supported) { - print_verbose("- Vulkan Variable Rate Shading supported:"); - if (vrs_capabilities.pipeline_vrs_supported) { - print_verbose(" Pipeline fragment shading rate"); - } - if (vrs_capabilities.primitive_vrs_supported) { - print_verbose(" Primitive fragment shading rate"); - } - if (vrs_capabilities.attachment_vrs_supported) { - // TODO expose these somehow to the end user - vrs_capabilities.min_texel_size.x = vrsProperties.minFragmentShadingRateAttachmentTexelSize.width; - vrs_capabilities.min_texel_size.y = vrsProperties.minFragmentShadingRateAttachmentTexelSize.height; - vrs_capabilities.max_texel_size.x = vrsProperties.maxFragmentShadingRateAttachmentTexelSize.width; - vrs_capabilities.max_texel_size.y = vrsProperties.maxFragmentShadingRateAttachmentTexelSize.height; + if (vrs_capabilities.pipeline_vrs_supported || vrs_capabilities.primitive_vrs_supported || vrs_capabilities.attachment_vrs_supported) { + print_verbose("- Vulkan Variable Rate Shading supported:"); + if (vrs_capabilities.pipeline_vrs_supported) { + print_verbose(" Pipeline fragment shading rate"); + } + if (vrs_capabilities.primitive_vrs_supported) { + print_verbose(" Primitive fragment shading rate"); + } + if (vrs_capabilities.attachment_vrs_supported) { + // TODO expose these somehow to the end user. + vrs_capabilities.min_texel_size.x = vrsProperties.minFragmentShadingRateAttachmentTexelSize.width; + vrs_capabilities.min_texel_size.y = vrsProperties.minFragmentShadingRateAttachmentTexelSize.height; + vrs_capabilities.max_texel_size.x = vrsProperties.maxFragmentShadingRateAttachmentTexelSize.width; + vrs_capabilities.max_texel_size.y = vrsProperties.maxFragmentShadingRateAttachmentTexelSize.height; + + // We'll attempt to default to a texel size of 16x16 + vrs_capabilities.texel_size.x = CLAMP(16, vrs_capabilities.min_texel_size.x, vrs_capabilities.max_texel_size.x); + vrs_capabilities.texel_size.y = CLAMP(16, vrs_capabilities.min_texel_size.y, vrs_capabilities.max_texel_size.y); + + print_verbose(String(" Attachment fragment shading rate") + String(", min texel size: (") + itos(vrs_capabilities.min_texel_size.x) + String(", ") + itos(vrs_capabilities.min_texel_size.y) + String(")") + String(", max texel size: (") + itos(vrs_capabilities.max_texel_size.x) + String(", ") + itos(vrs_capabilities.max_texel_size.y) + String(")")); + } - print_verbose(String(" Attachment fragment shading rate") + String(", min texel size: (") + itos(vrs_capabilities.min_texel_size.x) + String(", ") + itos(vrs_capabilities.min_texel_size.y) + String(")") + String(", max texel size: (") + itos(vrs_capabilities.max_texel_size.x) + String(", ") + itos(vrs_capabilities.max_texel_size.y) + String(")")); + } else { + print_verbose("- Vulkan Variable Rate Shading not supported"); } - } else { - print_verbose("- Vulkan Variable Rate Shading not supported"); - } + if (multiview_capabilities.is_supported) { + multiview_capabilities.max_view_count = multiviewProperties.maxMultiviewViewCount; + multiview_capabilities.max_instance_count = multiviewProperties.maxMultiviewInstanceIndex; - if (multiview_capabilities.is_supported) { - multiview_capabilities.max_view_count = multiviewProperties.maxMultiviewViewCount; - multiview_capabilities.max_instance_count = multiviewProperties.maxMultiviewInstanceIndex; + print_verbose("- Vulkan multiview supported:"); + print_verbose(" max view count: " + itos(multiview_capabilities.max_view_count)); + print_verbose(" max instances: " + itos(multiview_capabilities.max_instance_count)); + } else { + print_verbose("- Vulkan multiview not supported"); + } - print_verbose("- Vulkan multiview supported:"); - print_verbose(" max view count: " + itos(multiview_capabilities.max_view_count)); - print_verbose(" max instances: " + itos(multiview_capabilities.max_instance_count)); + print_verbose("- Vulkan subgroup:"); + print_verbose(" size: " + itos(subgroup_capabilities.size)); + print_verbose(" stages: " + subgroup_capabilities.supported_stages_desc()); + print_verbose(" supported ops: " + subgroup_capabilities.supported_operations_desc()); + if (subgroup_capabilities.quadOperationsInAllStages) { + print_verbose(" quad operations in all stages"); + } } else { - print_verbose("- Vulkan multiview not supported"); + print_verbose("- Couldn't call vkGetPhysicalDeviceProperties2"); } - - print_verbose("- Vulkan subgroup:"); - print_verbose(" size: " + itos(subgroup_capabilities.size)); - print_verbose(" stages: " + subgroup_capabilities.supported_stages_desc()); - print_verbose(" supported ops: " + subgroup_capabilities.supported_operations_desc()); - if (subgroup_capabilities.quadOperationsInAllStages) { - print_verbose(" quad operations in all stages"); - } - } else { - print_verbose("- Couldn't call vkGetPhysicalDeviceProperties2"); } return OK; @@ -698,27 +916,40 @@ Error VulkanContext::_create_instance() { // Initialize extensions. { - Error err = _initialize_extensions(); + Error err = _initialize_instance_extensions(); if (err != OK) { return err; } } - CharString cs = ProjectSettings::get_singleton()->get("application/config/name").operator String().utf8(); + int enabled_extension_count = 0; + const char *enabled_extension_names[MAX_EXTENSIONS]; + ERR_FAIL_COND_V(enabled_instance_extension_names.size() > MAX_EXTENSIONS, ERR_CANT_CREATE); + for (const CharString &extension_name : enabled_instance_extension_names) { + enabled_extension_names[enabled_extension_count++] = extension_name.ptr(); + } + + // We'll set application version to the Vulkan version we're developing against, even if our instance is based on + // an older Vulkan version, devices can still support newer versions of Vulkan. + // The exception is when we're on Vulkan 1.0, we should not set this to anything but 1.0. + // Note that this value is only used by validation layers to warn us about version issues. + uint32_t application_api_version = instance_api_version == VK_API_VERSION_1_0 ? VK_API_VERSION_1_0 : VK_API_VERSION_1_2; + + CharString cs = GLOBAL_GET("application/config/name").operator String().utf8(); const VkApplicationInfo app = { /*sType*/ VK_STRUCTURE_TYPE_APPLICATION_INFO, /*pNext*/ nullptr, /*pApplicationName*/ cs.get_data(), - /*applicationVersion*/ 0, + /*applicationVersion*/ 0, // It would be really nice if we store a version number in project settings, say "application/config/version" /*pEngineName*/ VERSION_NAME, /*engineVersion*/ VK_MAKE_VERSION(VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH), - /*apiVersion*/ VK_MAKE_VERSION(vulkan_major, vulkan_minor, 0) + /*apiVersion*/ application_api_version }; VkInstanceCreateInfo inst_info{}; inst_info.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO; inst_info.pApplicationInfo = &app; inst_info.enabledExtensionCount = enabled_extension_count; - inst_info.ppEnabledExtensionNames = (const char *const *)extension_names; + inst_info.ppEnabledExtensionNames = (const char *const *)enabled_extension_names; if (_use_validation_layers()) { _get_preferred_validation_layers(&inst_info.enabledLayerCount, &inst_info.ppEnabledLayerNames); } @@ -728,10 +959,10 @@ Error VulkanContext::_create_instance() { * After the instance is created, we use the instance-based * function to register the final callback. */ - VkDebugUtilsMessengerCreateInfoEXT dbg_messenger_create_info; - VkDebugReportCallbackCreateInfoEXT dbg_report_callback_create_info{}; - if (enabled_debug_utils) { - // VK_EXT_debug_utils style + VkDebugUtilsMessengerCreateInfoEXT dbg_messenger_create_info = {}; + VkDebugReportCallbackCreateInfoEXT dbg_report_callback_create_info = {}; + if (is_instance_extension_enabled(VK_EXT_DEBUG_UTILS_EXTENSION_NAME)) { + // VK_EXT_debug_utils style. dbg_messenger_create_info.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT; dbg_messenger_create_info.pNext = nullptr; dbg_messenger_create_info.flags = 0; @@ -743,7 +974,7 @@ Error VulkanContext::_create_instance() { dbg_messenger_create_info.pfnUserCallback = _debug_messenger_callback; dbg_messenger_create_info.pUserData = this; inst_info.pNext = &dbg_messenger_create_info; - } else if (enabled_debug_report) { + } else if (is_instance_extension_enabled(VK_EXT_DEBUG_REPORT_EXTENSION_NAME)) { dbg_report_callback_create_info.sType = VK_STRUCTURE_TYPE_DEBUG_REPORT_CALLBACK_CREATE_INFO_EXT; dbg_report_callback_create_info.flags = VK_DEBUG_REPORT_INFORMATION_BIT_EXT | VK_DEBUG_REPORT_WARNING_BIT_EXT | @@ -783,7 +1014,7 @@ Error VulkanContext::_create_instance() { volkLoadInstance(inst); #endif - if (enabled_debug_utils) { + if (is_instance_extension_enabled(VK_EXT_DEBUG_UTILS_EXTENSION_NAME)) { // Setup VK_EXT_debug_utils function pointers always (we use them for debug labels and names). CreateDebugUtilsMessengerEXT = (PFN_vkCreateDebugUtilsMessengerEXT)vkGetInstanceProcAddr(inst, "vkCreateDebugUtilsMessengerEXT"); @@ -824,7 +1055,7 @@ Error VulkanContext::_create_instance() { ERR_FAIL_V(ERR_CANT_CREATE); break; } - } else if (enabled_debug_report) { + } else if (is_instance_extension_enabled(VK_EXT_DEBUG_REPORT_EXTENSION_NAME)) { CreateDebugReportCallbackEXT = (PFN_vkCreateDebugReportCallbackEXT)vkGetInstanceProcAddr(inst, "vkCreateDebugReportCallbackEXT"); DebugReportMessageEXT = (PFN_vkDebugReportMessageEXT)vkGetInstanceProcAddr(inst, "vkDebugReportMessageEXT"); DestroyDebugReportCallbackEXT = (PFN_vkDestroyDebugReportCallbackEXT)vkGetInstanceProcAddr(inst, "vkDestroyDebugReportCallbackEXT"); @@ -902,8 +1133,8 @@ Error VulkanContext::_create_physical_device(VkSurfaceKHR p_surface) { } } else { // TODO: At least on Linux Laptops integrated GPUs fail with Vulkan in many instances. - // The device should really be a preference, but for now choosing a discrete GPU over the - // integrated one is better than the default. + // The device should really be a preference, but for now choosing a discrete GPU over the + // integrated one is better than the default. int type_selected = -1; print_verbose("Vulkan devices:"); @@ -1005,12 +1236,6 @@ Error VulkanContext::_create_physical_device(VkSurfaceKHR p_surface) { free(physical_devices); - // Look for device extensions. - uint32_t device_extension_count = 0; - VkBool32 swapchainExtFound = 0; - enabled_extension_count = 0; - memset(extension_names, 0, sizeof(extension_names)); - // Get identifier properties. vkGetPhysicalDeviceProperties(gpu, &gpu_props); @@ -1030,89 +1255,19 @@ Error VulkanContext::_create_physical_device(VkSurfaceKHR p_surface) { } } - print_line( - "Vulkan API " + itos(vulkan_major) + "." + itos(vulkan_minor) + "." + itos(vulkan_patch) + - " - " + "Using Vulkan Device #" + itos(device_index) + ": " + device_vendor + " - " + device_name); - + // Get device version device_api_version = gpu_props.apiVersion; - err = vkEnumerateDeviceExtensionProperties(gpu, nullptr, &device_extension_count, nullptr); - ERR_FAIL_COND_V(err, ERR_CANT_CREATE); - - if (device_extension_count > 0) { - VkExtensionProperties *device_extensions = (VkExtensionProperties *)malloc(sizeof(VkExtensionProperties) * device_extension_count); - err = vkEnumerateDeviceExtensionProperties(gpu, nullptr, &device_extension_count, device_extensions); - if (err) { - free(device_extensions); - ERR_FAIL_V(ERR_CANT_CREATE); - } - - for (uint32_t i = 0; i < device_extension_count; i++) { - if (!strcmp(VK_KHR_SWAPCHAIN_EXTENSION_NAME, device_extensions[i].extensionName)) { - swapchainExtFound = 1; - extension_names[enabled_extension_count++] = VK_KHR_SWAPCHAIN_EXTENSION_NAME; - } - if (!strcmp(VK_KHR_MULTIVIEW_EXTENSION_NAME, device_extensions[i].extensionName)) { - // If multiview is supported, enable it. - extension_names[enabled_extension_count++] = VK_KHR_MULTIVIEW_EXTENSION_NAME; - } - if (!strcmp(VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME, device_extensions[i].extensionName)) { - // if shading rate image is supported, enable it - extension_names[enabled_extension_count++] = VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME; - } - if (!strcmp(VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME, device_extensions[i].extensionName)) { - extension_names[enabled_extension_count++] = VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME; - } - if (enabled_extension_count >= MAX_EXTENSIONS) { - free(device_extensions); - ERR_FAIL_V_MSG(ERR_BUG, "Enabled extension count reaches MAX_EXTENSIONS, BUG"); - } - } - - if (VK_KHR_incremental_present_enabled) { - // Even though the user "enabled" the extension via the command - // line, we must make sure that it's enumerated for use with the - // device. Therefore, disable it here, and re-enable it again if - // enumerated. - VK_KHR_incremental_present_enabled = false; - for (uint32_t i = 0; i < device_extension_count; i++) { - if (!strcmp(VK_KHR_INCREMENTAL_PRESENT_EXTENSION_NAME, device_extensions[i].extensionName)) { - extension_names[enabled_extension_count++] = VK_KHR_INCREMENTAL_PRESENT_EXTENSION_NAME; - VK_KHR_incremental_present_enabled = true; - } - if (enabled_extension_count >= MAX_EXTENSIONS) { - free(device_extensions); - ERR_FAIL_V_MSG(ERR_BUG, "Enabled extension count reaches MAX_EXTENSIONS, BUG"); - } - } - } + // Output our device version + print_line("Vulkan API " + get_device_api_version() + " - " + "Using Vulkan Device #" + itos(device_index) + ": " + device_vendor + " - " + device_name); - if (VK_GOOGLE_display_timing_enabled) { - // Even though the user "enabled" the extension via the command - // line, we must make sure that it's enumerated for use with the - // device. Therefore, disable it here, and re-enable it again if - // enumerated. - VK_GOOGLE_display_timing_enabled = false; - for (uint32_t i = 0; i < device_extension_count; i++) { - if (!strcmp(VK_GOOGLE_DISPLAY_TIMING_EXTENSION_NAME, device_extensions[i].extensionName)) { - extension_names[enabled_extension_count++] = VK_GOOGLE_DISPLAY_TIMING_EXTENSION_NAME; - VK_GOOGLE_display_timing_enabled = true; - } - if (enabled_extension_count >= MAX_EXTENSIONS) { - free(device_extensions); - ERR_FAIL_V_MSG(ERR_BUG, "Enabled extension count reaches MAX_EXTENSIONS, BUG"); - } - } + { + Error _err = _initialize_device_extensions(); + if (_err != OK) { + return _err; } - - free(device_extensions); } - ERR_FAIL_COND_V_MSG(!swapchainExtFound, ERR_CANT_CREATE, - "vkEnumerateDeviceExtensionProperties failed to find the " VK_KHR_SWAPCHAIN_EXTENSION_NAME - " extension.\n\nDo you have a compatible Vulkan installable client driver (ICD) installed?\n" - "vkCreateInstance Failure"); - // Call with nullptr data to get count. vkGetPhysicalDeviceQueueFamilyProperties(gpu, &queue_family_count, nullptr); ERR_FAIL_COND_V(queue_family_count == 0, ERR_CANT_CREATE); @@ -1173,9 +1328,9 @@ Error VulkanContext::_create_device() { }; nextptr = &shader_features; - VkPhysicalDeviceFragmentShadingRateFeaturesKHR vrs_features; + VkPhysicalDeviceFragmentShadingRateFeaturesKHR vrs_features = {}; if (vrs_capabilities.pipeline_vrs_supported || vrs_capabilities.primitive_vrs_supported || vrs_capabilities.attachment_vrs_supported) { - // insert into our chain to enable these features if they are available + // Insert into our chain to enable these features if they are available. vrs_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FRAGMENT_SHADING_RATE_FEATURES_KHR; vrs_features.pNext = nextptr; vrs_features.pipelineFragmentShadingRate = vrs_capabilities.pipeline_vrs_supported; @@ -1185,10 +1340,10 @@ Error VulkanContext::_create_device() { nextptr = &vrs_features; } - VkPhysicalDeviceVulkan11Features vulkan11features; - VkPhysicalDevice16BitStorageFeaturesKHR storage_feature; - VkPhysicalDeviceMultiviewFeatures multiview_features; - if (vulkan_major > 1 || vulkan_minor >= 2) { + VkPhysicalDeviceVulkan11Features vulkan11features = {}; + VkPhysicalDevice16BitStorageFeaturesKHR storage_feature = {}; + VkPhysicalDeviceMultiviewFeatures multiview_features = {}; + if (device_api_version >= VK_API_VERSION_1_2) { // In Vulkan 1.2 and newer we use a newer struct to enable various features. vulkan11features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_1_FEATURES; @@ -1207,7 +1362,7 @@ Error VulkanContext::_create_device() { vulkan11features.shaderDrawParameters = 0; nextptr = &vulkan11features; } else { - // On Vulkan 1.0 and 1.1 we use our older structs to initialise these features. + // On Vulkan 1.0 and 1.1 we use our older structs to initialize these features. storage_feature.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES_KHR; storage_feature.pNext = nextptr; storage_feature.storageBuffer16BitAccess = storage_buffer_capabilities.storage_buffer_16_bit_access_is_supported; @@ -1216,7 +1371,7 @@ Error VulkanContext::_create_device() { storage_feature.storageInputOutput16 = storage_buffer_capabilities.storage_input_output_16; nextptr = &storage_feature; - if (vulkan_major == 1 && vulkan_minor == 1) { + if (device_api_version >= VK_API_VERSION_1_1) { // any Vulkan 1.1.x version multiview_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES; multiview_features.pNext = nextptr; multiview_features.multiview = multiview_capabilities.is_supported; @@ -1226,6 +1381,13 @@ Error VulkanContext::_create_device() { } } + uint32_t enabled_extension_count = 0; + const char *enabled_extension_names[MAX_EXTENSIONS]; + ERR_FAIL_COND_V(enabled_device_extension_names.size() > MAX_EXTENSIONS, ERR_CANT_CREATE); + for (const CharString &extension_name : enabled_device_extension_names) { + enabled_extension_names[enabled_extension_count++] = extension_name.ptr(); + } + VkDeviceCreateInfo sdevice = { /*sType*/ VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO, /*pNext*/ nextptr, @@ -1235,7 +1397,7 @@ Error VulkanContext::_create_device() { /*enabledLayerCount*/ 0, /*ppEnabledLayerNames*/ nullptr, /*enabledExtensionCount*/ enabled_extension_count, - /*ppEnabledExtensionNames*/ (const char *const *)extension_names, + /*ppEnabledExtensionNames*/ (const char *const *)enabled_extension_names, /*pEnabledFeatures*/ &physical_device_features, // If specific features are required, pass them in here. }; if (separate_present_queue) { @@ -1323,7 +1485,7 @@ Error VulkanContext::_initialize_queues(VkSurfaceKHR p_surface) { GET_DEVICE_PROC_ADDR(device, GetSwapchainImagesKHR); GET_DEVICE_PROC_ADDR(device, AcquireNextImageKHR); GET_DEVICE_PROC_ADDR(device, QueuePresentKHR); - if (VK_GOOGLE_display_timing_enabled) { + if (is_device_extension_enabled(VK_GOOGLE_DISPLAY_TIMING_EXTENSION_NAME)) { GET_DEVICE_PROC_ADDR(device, GetRefreshCycleDurationGOOGLE); GET_DEVICE_PROC_ADDR(device, GetPastPresentationTimingGOOGLE); } @@ -1354,7 +1516,7 @@ Error VulkanContext::_initialize_queues(VkSurfaceKHR p_surface) { color_space = surfFormats[0].colorSpace; } else { // These should be ordered with the ones we want to use on top and fallback modes further down - // we want a 32bit RGBA unsigned normalised buffer or similar. + // we want a 32bit RGBA unsigned normalized buffer or similar. const VkFormat allowed_formats[] = { VK_FORMAT_B8G8R8A8_UNORM, VK_FORMAT_R8G8B8A8_UNORM @@ -1436,6 +1598,24 @@ bool VulkanContext::_use_validation_layers() { return Engine::get_singleton()->is_validation_layers_enabled(); } +VkExtent2D VulkanContext::_compute_swapchain_extent(const VkSurfaceCapabilitiesKHR &p_surf_capabilities, int *p_window_width, int *p_window_height) const { + // Width and height are either both 0xFFFFFFFF, or both not 0xFFFFFFFF. + if (p_surf_capabilities.currentExtent.width == 0xFFFFFFFF) { + // If the surface size is undefined, the size is set to the size + // of the images requested, which must fit within the minimum and + // maximum values. + VkExtent2D extent = {}; + extent.width = CLAMP((uint32_t)(*p_window_width), p_surf_capabilities.minImageExtent.width, p_surf_capabilities.maxImageExtent.width); + extent.height = CLAMP((uint32_t)(*p_window_height), p_surf_capabilities.minImageExtent.height, p_surf_capabilities.maxImageExtent.height); + return extent; + } else { + // If the surface size is defined, the swap chain size must match. + *p_window_width = p_surf_capabilities.currentExtent.width; + *p_window_height = p_surf_capabilities.currentExtent.height; + return p_surf_capabilities.currentExtent; + } +} + Error VulkanContext::_window_create(DisplayServer::WindowID p_window_id, DisplayServer::VSyncMode p_vsync_mode, VkSurfaceKHR p_surface, int p_width, int p_height) { ERR_FAIL_COND_V(windows.has(p_window_id), ERR_INVALID_PARAMETER); @@ -1576,32 +1756,7 @@ Error VulkanContext::_update_swap_chain(Window *window) { ERR_FAIL_V(ERR_CANT_CREATE); } - VkExtent2D swapchainExtent; - // Width and height are either both 0xFFFFFFFF, or both not 0xFFFFFFFF. - if (surfCapabilities.currentExtent.width == 0xFFFFFFFF) { - // If the surface size is undefined, the size is set to the size - // of the images requested, which must fit within the minimum and - // maximum values. - swapchainExtent.width = window->width; - swapchainExtent.height = window->height; - - if (swapchainExtent.width < surfCapabilities.minImageExtent.width) { - swapchainExtent.width = surfCapabilities.minImageExtent.width; - } else if (swapchainExtent.width > surfCapabilities.maxImageExtent.width) { - swapchainExtent.width = surfCapabilities.maxImageExtent.width; - } - - if (swapchainExtent.height < surfCapabilities.minImageExtent.height) { - swapchainExtent.height = surfCapabilities.minImageExtent.height; - } else if (swapchainExtent.height > surfCapabilities.maxImageExtent.height) { - swapchainExtent.height = surfCapabilities.maxImageExtent.height; - } - } else { - // If the surface size is defined, the swap chain size must match. - swapchainExtent = surfCapabilities.currentExtent; - window->width = surfCapabilities.currentExtent.width; - window->height = surfCapabilities.currentExtent.height; - } + VkExtent2D swapchainExtent = _compute_swapchain_extent(surfCapabilities, &window->width, &window->height); if (window->width == 0 || window->height == 0) { free(presentModes); @@ -1611,17 +1766,17 @@ Error VulkanContext::_update_swap_chain(Window *window) { // The FIFO present mode is guaranteed by the spec to be supported // and to have no tearing. It's a great default present mode to use. - // There are times when you may wish to use another present mode. The - // following code shows how to select them, and the comments provide some - // reasons you may wish to use them. + // There are times when you may wish to use another present mode. The + // following code shows how to select them, and the comments provide some + // reasons you may wish to use them. // // It should be noted that Vulkan 1.0 doesn't provide a method for - // synchronizing rendering with the presentation engine's display. There + // synchronizing rendering with the presentation engine's display. There // is a method provided for throttling rendering with the display, but // there are some presentation engines for which this method will not work. // If an application doesn't throttle its rendering, and if it renders much // faster than the refresh rate of the display, this can waste power on - // mobile devices. That is because power is being spent rendering images + // mobile devices. That is because power is being spent rendering images // that may never be seen. // VK_PRESENT_MODE_IMMEDIATE_KHR is for applications that don't care about @@ -1666,7 +1821,22 @@ Error VulkanContext::_update_swap_chain(Window *window) { if (present_mode_available) { window->presentMode = requested_present_mode; } else { - WARN_PRINT("Requested VSync mode is not available!"); + String present_mode_string; + switch (window->vsync_mode) { + case DisplayServer::VSYNC_MAILBOX: + present_mode_string = "Mailbox"; + break; + case DisplayServer::VSYNC_ADAPTIVE: + present_mode_string = "Adaptive"; + break; + case DisplayServer::VSYNC_ENABLED: + present_mode_string = "Enabled"; + break; + case DisplayServer::VSYNC_DISABLED: + present_mode_string = "Disabled"; + break; + } + WARN_PRINT(vformat("The requested V-Sync mode %s is not available. Falling back to V-Sync mode Enabled.", present_mode_string)); window->vsync_mode = DisplayServer::VSYNC_ENABLED; // Set to default. } @@ -1684,7 +1854,7 @@ Error VulkanContext::_update_swap_chain(Window *window) { // If maxImageCount is 0, we can ask for as many images as we want; // otherwise we're limited to maxImageCount. if ((surfCapabilities.maxImageCount > 0) && (desiredNumOfSwapchainImages > surfCapabilities.maxImageCount)) { - // Application must settle for fewer images than desired: + // Application must settle for fewer images than desired. desiredNumOfSwapchainImages = surfCapabilities.maxImageCount; } @@ -1695,18 +1865,22 @@ Error VulkanContext::_update_swap_chain(Window *window) { preTransform = surfCapabilities.currentTransform; } - // Find a supported composite alpha mode - one of these is guaranteed to be set. VkCompositeAlphaFlagBitsKHR compositeAlpha = VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR; - VkCompositeAlphaFlagBitsKHR compositeAlphaFlags[4] = { - VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR, - VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR, - VK_COMPOSITE_ALPHA_POST_MULTIPLIED_BIT_KHR, - VK_COMPOSITE_ALPHA_INHERIT_BIT_KHR, - }; - for (uint32_t i = 0; i < ARRAY_SIZE(compositeAlphaFlags); i++) { - if (surfCapabilities.supportedCompositeAlpha & compositeAlphaFlags[i]) { - compositeAlpha = compositeAlphaFlags[i]; - break; + + if (OS::get_singleton()->is_layered_allowed() || !(surfCapabilities.supportedCompositeAlpha & compositeAlpha)) { + // Find a supported composite alpha mode - one of these is guaranteed to be set. + VkCompositeAlphaFlagBitsKHR compositeAlphaFlags[4] = { + VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR, + VK_COMPOSITE_ALPHA_POST_MULTIPLIED_BIT_KHR, + VK_COMPOSITE_ALPHA_INHERIT_BIT_KHR, + VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR, + }; + + for (uint32_t i = 0; i < ARRAY_SIZE(compositeAlphaFlags); i++) { + if (surfCapabilities.supportedCompositeAlpha & compositeAlphaFlags[i]) { + compositeAlpha = compositeAlphaFlags[i]; + break; + } } } @@ -2028,14 +2202,14 @@ Error VulkanContext::prepare_buffers() { } do { - // Get the index of the next available swapchain image: + // Get the index of the next available swapchain image. err = fpAcquireNextImageKHR(device, w->swapchain, UINT64_MAX, w->image_acquired_semaphores[frame_index], VK_NULL_HANDLE, &w->current_buffer); if (err == VK_ERROR_OUT_OF_DATE_KHR) { // Swapchain is out of date (e.g. the window was resized) and - // must be recreated: + // must be recreated. print_verbose("Vulkan: Early out of date swapchain, recreating."); // resize_notify(); _update_swap_chain(w); @@ -2066,9 +2240,9 @@ Error VulkanContext::swap_buffers() { VkResult err; #if 0 - if (VK_GOOGLE_display_timing_enabled) { + if (is_device_extension_enabled(VK_GOOGLE_DISPLAY_TIMING_EXTENSION_NAME)) { // Look at what happened to previous presents, and make appropriate - // adjustments in timing: + // adjustments in timing. DemoUpdateTargetIPD(demo); // Note: a real application would position its geometry to that it's in @@ -2187,7 +2361,7 @@ Error VulkanContext::swap_buffers() { } #if 0 - if (VK_KHR_incremental_present_enabled) { + if (is_device_extension_enabled(VK_KHR_incremental_present_enabled)) { // If using VK_KHR_incremental_present, we provide a hint of the region // that contains changed content relative to the previously-presented // image. The implementation can use this hint in order to save @@ -2218,7 +2392,7 @@ Error VulkanContext::swap_buffers() { #endif #if 0 - if (VK_GOOGLE_display_timing_enabled) { + if (is_device_extension_enabled(VK_GOOGLE_DISPLAY_TIMING_EXTENSION_NAME)) { VkPresentTimeGOOGLE ptime; if (prev_desired_present_time == 0) { // This must be the first present for this swapchain. @@ -2231,7 +2405,7 @@ Error VulkanContext::swap_buffers() { uint64_t curtime = getTimeInNanoseconds(); if (curtime == 0) { // Since we didn't find out the current time, don't give a - // desiredPresentTime: + // desiredPresentTime. ptime.desiredPresentTime = 0; } else { ptime.desiredPresentTime = curtime + (target_IPD >> 1); @@ -2248,13 +2422,11 @@ Error VulkanContext::swap_buffers() { /*swapchainCount*/ present.swapchainCount, /*pTimes*/ &ptime, }; - if (VK_GOOGLE_display_timing_enabled) { + if (is_device_extension_enabled(VK_GOOGLE_DISPLAY_TIMING_EXTENSION_NAME)) { present.pNext = &present_time; } } #endif - static int total_frames = 0; - total_frames++; // print_line("current buffer: " + itos(current_buffer)); err = fpQueuePresentKHR(present_queue, &present); @@ -2263,7 +2435,7 @@ Error VulkanContext::swap_buffers() { if (err == VK_ERROR_OUT_OF_DATE_KHR) { // Swapchain is out of date (e.g. the window was resized) and - // must be recreated: + // must be recreated. print_verbose("Vulkan: Swapchain is out of date, recreating."); resize_notify(); } else if (err == VK_SUBOPTIMAL_KHR) { @@ -2323,6 +2495,13 @@ RID VulkanContext::local_device_create() { queues[0].pQueuePriorities = queue_priorities; queues[0].flags = 0; + uint32_t enabled_extension_count = 0; + const char *enabled_extension_names[MAX_EXTENSIONS]; + ERR_FAIL_COND_V(enabled_device_extension_names.size() > MAX_EXTENSIONS, RID()); + for (const CharString &extension_name : enabled_device_extension_names) { + enabled_extension_names[enabled_extension_count++] = extension_name.ptr(); + } + VkDeviceCreateInfo sdevice = { /*sType =*/VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO, /*pNext */ nullptr, @@ -2332,7 +2511,7 @@ RID VulkanContext::local_device_create() { /*enabledLayerCount */ 0, /*ppEnabledLayerNames */ nullptr, /*enabledExtensionCount */ enabled_extension_count, - /*ppEnabledExtensionNames */ (const char *const *)extension_names, + /*ppEnabledExtensionNames */ (const char *const *)enabled_extension_names, /*pEnabledFeatures */ &physical_device_features, // If specific features are required, pass them in here. }; err = vkCreateDevice(gpu, &sdevice, nullptr, &ld.device); @@ -2397,7 +2576,7 @@ void VulkanContext::local_device_free(RID p_local_device) { } void VulkanContext::command_begin_label(VkCommandBuffer p_command_buffer, String p_label_name, const Color p_color) { - if (!enabled_debug_utils) { + if (!is_instance_extension_enabled(VK_EXT_DEBUG_UTILS_EXTENSION_NAME)) { return; } @@ -2414,7 +2593,7 @@ void VulkanContext::command_begin_label(VkCommandBuffer p_command_buffer, String } void VulkanContext::command_insert_label(VkCommandBuffer p_command_buffer, String p_label_name, const Color p_color) { - if (!enabled_debug_utils) { + if (!is_instance_extension_enabled(VK_EXT_DEBUG_UTILS_EXTENSION_NAME)) { return; } CharString cs = p_label_name.utf8(); @@ -2430,14 +2609,14 @@ void VulkanContext::command_insert_label(VkCommandBuffer p_command_buffer, Strin } void VulkanContext::command_end_label(VkCommandBuffer p_command_buffer) { - if (!enabled_debug_utils) { + if (!is_instance_extension_enabled(VK_EXT_DEBUG_UTILS_EXTENSION_NAME)) { return; } CmdEndDebugUtilsLabelEXT(p_command_buffer); } void VulkanContext::set_object_name(VkObjectType p_object_type, uint64_t p_object_handle, String p_object_name) { - if (!enabled_debug_utils) { + if (!is_instance_extension_enabled(VK_EXT_DEBUG_UTILS_EXTENSION_NAME)) { return; } CharString obj_data = p_object_name.utf8(); @@ -2463,7 +2642,7 @@ RenderingDevice::DeviceType VulkanContext::get_device_type() const { } String VulkanContext::get_device_api_version() const { - return vformat("%d.%d.%d", vulkan_major, vulkan_minor, vulkan_patch); + return vformat("%d.%d.%d", VK_API_VERSION_MAJOR(device_api_version), VK_API_VERSION_MINOR(device_api_version), VK_API_VERSION_PATCH(device_api_version)); } String VulkanContext::get_device_pipeline_cache_uuid() const { @@ -2471,12 +2650,12 @@ String VulkanContext::get_device_pipeline_cache_uuid() const { } DisplayServer::VSyncMode VulkanContext::get_vsync_mode(DisplayServer::WindowID p_window) const { - ERR_FAIL_COND_V_MSG(!windows.has(p_window), DisplayServer::VSYNC_ENABLED, "Could not get VSync mode for window with WindowID " + itos(p_window) + " because it does not exist."); + ERR_FAIL_COND_V_MSG(!windows.has(p_window), DisplayServer::VSYNC_ENABLED, "Could not get V-Sync mode for window with WindowID " + itos(p_window) + " because it does not exist."); return windows[p_window].vsync_mode; } void VulkanContext::set_vsync_mode(DisplayServer::WindowID p_window, DisplayServer::VSyncMode p_mode) { - ERR_FAIL_COND_MSG(!windows.has(p_window), "Could not set VSync mode for window with WindowID " + itos(p_window) + " because it does not exist."); + ERR_FAIL_COND_MSG(!windows.has(p_window), "Could not set V-Sync mode for window with WindowID " + itos(p_window) + " because it does not exist."); windows[p_window].vsync_mode = p_mode; _update_swap_chain(&windows[p_window]); } @@ -2498,7 +2677,7 @@ VulkanContext::~VulkanContext() { vkDestroySemaphore(device, image_ownership_semaphores[i], nullptr); } } - if (inst_initialized && enabled_debug_utils) { + if (inst_initialized && is_instance_extension_enabled(VK_EXT_DEBUG_UTILS_EXTENSION_NAME)) { DestroyDebugUtilsMessengerEXT(inst, dbg_messenger, nullptr); } if (inst_initialized && dbg_debug_report != VK_NULL_HANDLE) { diff --git a/drivers/vulkan/vulkan_context.h b/drivers/vulkan/vulkan_context.h index 35e7ce7db8..7044064e12 100644 --- a/drivers/vulkan/vulkan_context.h +++ b/drivers/vulkan/vulkan_context.h @@ -1,32 +1,32 @@ -/*************************************************************************/ -/* vulkan_context.h */ -/*************************************************************************/ -/* This file is part of: */ -/* GODOT ENGINE */ -/* https://godotengine.org */ -/*************************************************************************/ -/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ -/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ -/* */ -/* Permission is hereby granted, free of charge, to any person obtaining */ -/* a copy of this software and associated documentation files (the */ -/* "Software"), to deal in the Software without restriction, including */ -/* without limitation the rights to use, copy, modify, merge, publish, */ -/* distribute, sublicense, and/or sell copies of the Software, and to */ -/* permit persons to whom the Software is furnished to do so, subject to */ -/* the following conditions: */ -/* */ -/* The above copyright notice and this permission notice shall be */ -/* included in all copies or substantial portions of the Software. */ -/* */ -/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ -/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ -/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ -/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ -/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ -/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ -/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/*************************************************************************/ +/**************************************************************************/ +/* vulkan_context.h */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ #ifndef VULKAN_CONTEXT_H #define VULKAN_CONTEXT_H @@ -34,6 +34,7 @@ #include "core/error/error_list.h" #include "core/os/mutex.h" #include "core/string/ustring.h" +#include "core/templates/hash_map.h" #include "core/templates/rb_map.h" #include "core/templates/rid_owner.h" #include "servers/display_server.h" @@ -70,12 +71,14 @@ public: }; struct VRSCapabilities { - bool pipeline_vrs_supported; // We can specify our fragment rate on a pipeline level - bool primitive_vrs_supported; // We can specify our fragment rate on each drawcall - bool attachment_vrs_supported; // We can provide a density map attachment on our framebuffer + bool pipeline_vrs_supported; // We can specify our fragment rate on a pipeline level. + bool primitive_vrs_supported; // We can specify our fragment rate on each drawcall. + bool attachment_vrs_supported; // We can provide a density map attachment on our framebuffer. Size2i min_texel_size; Size2i max_texel_size; + + Size2i texel_size; // The texel size we'll use }; struct ShaderCapabilities { @@ -107,10 +110,7 @@ private: bool device_initialized = false; bool inst_initialized = false; - // Vulkan 1.0 doesn't return version info so we assume this by default until we know otherwise - uint32_t vulkan_major = 1; - uint32_t vulkan_minor = 0; - uint32_t vulkan_patch = 0; + uint32_t instance_api_version = VK_API_VERSION_1_0; SubgroupCapabilities subgroup_capabilities; MultiviewCapabilities multiview_capabilities; VRSCapabilities vrs_capabilities; @@ -182,18 +182,15 @@ private: int command_buffer_count = 1; // Extensions. + static bool instance_extensions_initialized; + static HashMap<CharString, bool> requested_instance_extensions; + HashSet<CharString> enabled_instance_extension_names; + static bool device_extensions_initialized; + static HashMap<CharString, bool> requested_device_extensions; + HashSet<CharString> enabled_device_extension_names; bool VK_KHR_incremental_present_enabled = true; bool VK_GOOGLE_display_timing_enabled = true; - uint32_t enabled_extension_count = 0; - const char *extension_names[MAX_EXTENSIONS]; - bool enabled_debug_utils = false; - - /** - * True if VK_EXT_debug_report extension is used. VK_EXT_debug_report is deprecated but it is - * still used if VK_EXT_debug_utils is not available. - */ - bool enabled_debug_report = false; PFN_vkCreateDebugUtilsMessengerEXT CreateDebugUtilsMessengerEXT = nullptr; PFN_vkDestroyDebugUtilsMessengerEXT DestroyDebugUtilsMessengerEXT = nullptr; @@ -222,7 +219,8 @@ private: VkDebugReportCallbackEXT dbg_debug_report = VK_NULL_HANDLE; Error _obtain_vulkan_version(); - Error _initialize_extensions(); + Error _initialize_instance_extensions(); + Error _initialize_device_extensions(); Error _check_capabilities(); VkBool32 _check_layers(uint32_t check_count, const char *const *check_names, uint32_t layer_count, VkLayerProperties *layers); @@ -257,6 +255,8 @@ private: Error _create_swap_chain(); Error _create_semaphores(); + Vector<VkAttachmentReference> _convert_VkAttachmentReference2(uint32_t p_count, const VkAttachmentReference2 *p_refs); + protected: virtual const char *_get_platform_surface_extension() const = 0; @@ -266,12 +266,15 @@ protected: Error _get_preferred_validation_layers(uint32_t *count, const char *const **names); + virtual VkExtent2D _compute_swapchain_extent(const VkSurfaceCapabilitiesKHR &p_surf_capabilities, int *p_window_width, int *p_window_height) const; + public: - // Extension calls - VkResult vkCreateRenderPass2KHR(VkDevice device, const VkRenderPassCreateInfo2 *pCreateInfo, const VkAllocationCallbacks *pAllocator, VkRenderPass *pRenderPass); + // Extension calls. + bool supports_renderpass2() const { return is_device_extension_enabled(VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME); } + VkResult vkCreateRenderPass2KHR(VkDevice p_device, const VkRenderPassCreateInfo2 *p_create_info, const VkAllocationCallbacks *p_allocator, VkRenderPass *p_render_pass); - uint32_t get_vulkan_major() const { return vulkan_major; }; - uint32_t get_vulkan_minor() const { return vulkan_minor; }; + uint32_t get_vulkan_major() const { return VK_API_VERSION_MAJOR(device_api_version); }; + uint32_t get_vulkan_minor() const { return VK_API_VERSION_MINOR(device_api_version); }; const SubgroupCapabilities &get_subgroup_capabilities() const { return subgroup_capabilities; }; const MultiviewCapabilities &get_multiview_capabilities() const { return multiview_capabilities; }; const VRSCapabilities &get_vrs_capabilities() const { return vrs_capabilities; }; @@ -287,6 +290,16 @@ public: static void set_vulkan_hooks(VulkanHooks *p_vulkan_hooks) { vulkan_hooks = p_vulkan_hooks; }; + static void register_requested_instance_extension(const CharString &extension_name, bool p_required); + bool is_instance_extension_enabled(const CharString &extension_name) const { + return enabled_instance_extension_names.has(extension_name); + } + + static void register_requested_device_extension(const CharString &extension_name, bool p_required); + bool is_device_extension_enabled(const CharString &extension_name) const { + return enabled_device_extension_names.has(extension_name); + } + void window_resize(DisplayServer::WindowID p_window_id, int p_width, int p_height); int window_get_width(DisplayServer::WindowID p_window = 0); int window_get_height(DisplayServer::WindowID p_window = 0); diff --git a/drivers/vulkan/vulkan_hooks.h b/drivers/vulkan/vulkan_hooks.h index dec2042d0d..eaa52658e4 100644 --- a/drivers/vulkan/vulkan_hooks.h +++ b/drivers/vulkan/vulkan_hooks.h @@ -1,32 +1,32 @@ -/*************************************************************************/ -/* vulkan_hooks.h */ -/*************************************************************************/ -/* This file is part of: */ -/* GODOT ENGINE */ -/* https://godotengine.org */ -/*************************************************************************/ -/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ -/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ -/* */ -/* Permission is hereby granted, free of charge, to any person obtaining */ -/* a copy of this software and associated documentation files (the */ -/* "Software"), to deal in the Software without restriction, including */ -/* without limitation the rights to use, copy, modify, merge, publish, */ -/* distribute, sublicense, and/or sell copies of the Software, and to */ -/* permit persons to whom the Software is furnished to do so, subject to */ -/* the following conditions: */ -/* */ -/* The above copyright notice and this permission notice shall be */ -/* included in all copies or substantial portions of the Software. */ -/* */ -/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ -/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ -/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ -/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ -/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ -/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ -/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/*************************************************************************/ +/**************************************************************************/ +/* vulkan_hooks.h */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ #ifndef VULKAN_HOOKS_H #define VULKAN_HOOKS_H |