From ace20ba4a4774ae3c42f2ef5566c7113f3b980b3 Mon Sep 17 00:00:00 2001 From: ameerj <52414509+ameerj@users.noreply.github.com> Date: Sat, 19 Jun 2021 10:56:13 -0400 Subject: [PATCH] astc_decoder.comp: Remove unnecessary LUT SSBOs We can move them to instead be compile time constants within the shader. --- src/video_core/host_shaders/astc_decoder.comp | 35 ++++----- .../renderer_opengl/util_shaders.cpp | 21 +----- .../renderer_vulkan/vk_compute_pass.cpp | 74 +++---------------- src/video_core/textures/astc.cpp | 8 +- src/video_core/textures/astc.h | 9 +-- 5 files changed, 34 insertions(+), 113 deletions(-) diff --git a/src/video_core/host_shaders/astc_decoder.comp b/src/video_core/host_shaders/astc_decoder.comp index 71327e233..c37f15bfd 100644 --- a/src/video_core/host_shaders/astc_decoder.comp +++ b/src/video_core/host_shaders/astc_decoder.comp @@ -11,11 +11,8 @@ #define UNIFORM(n) #define BINDING_INPUT_BUFFER 0 #define BINDING_ENC_BUFFER 1 -#define BINDING_6_TO_8_BUFFER 2 -#define BINDING_7_TO_8_BUFFER 3 -#define BINDING_8_TO_8_BUFFER 4 -#define BINDING_SWIZZLE_BUFFER 5 -#define BINDING_OUTPUT_IMAGE 6 +#define BINDING_SWIZZLE_BUFFER 2 +#define BINDING_OUTPUT_IMAGE 3 #else // ^^^ Vulkan ^^^ // vvv OpenGL vvv @@ -25,9 +22,6 @@ #define BINDING_SWIZZLE_BUFFER 0 #define BINDING_INPUT_BUFFER 1 #define BINDING_ENC_BUFFER 2 -#define BINDING_6_TO_8_BUFFER 3 -#define BINDING_7_TO_8_BUFFER 4 -#define BINDING_8_TO_8_BUFFER 5 #define BINDING_OUTPUT_IMAGE 0 #endif @@ -74,16 +68,6 @@ layout(binding = BINDING_INPUT_BUFFER, std430) readonly buffer InputBufferU32 { layout(binding = BINDING_ENC_BUFFER, std430) readonly buffer EncodingsValues { EncodingData encoding_values[]; }; -// ASTC Precompiled tables -layout(binding = BINDING_6_TO_8_BUFFER, std430) readonly buffer REPLICATE_6_BIT_TO_8 { - uint REPLICATE_6_BIT_TO_8_TABLE[]; -}; -layout(binding = BINDING_7_TO_8_BUFFER, std430) readonly buffer REPLICATE_7_BIT_TO_8 { - uint REPLICATE_7_BIT_TO_8_TABLE[]; -}; -layout(binding = BINDING_8_TO_8_BUFFER, std430) readonly buffer REPLICATE_8_BIT_TO_8 { - uint REPLICATE_8_BIT_TO_8_TABLE[]; -}; layout(binding = BINDING_OUTPUT_IMAGE, rgba8) uniform writeonly image2DArray dest_image; @@ -134,6 +118,19 @@ const uint REPLICATE_4_BIT_TO_6_TABLE[16] = const uint REPLICATE_5_BIT_TO_6_TABLE[32] = uint[](0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 33, 35, 37, 39, 41, 43, 45, 47, 49, 51, 53, 55, 57, 59, 61, 63); +const uint REPLICATE_6_BIT_TO_8_TABLE[64] = + uint[](0, 4, 8, 12, 16, 20, 24, 28, 32, 36, 40, 44, 48, 52, 56, 60, 65, 69, 73, 77, 81, 85, 89, + 93, 97, 101, 105, 109, 113, 117, 121, 125, 130, 134, 138, 142, 146, 150, 154, 158, 162, + 166, 170, 174, 178, 182, 186, 190, 195, 199, 203, 207, 211, 215, 219, 223, 227, 231, 235, + 239, 243, 247, 251, 255); +const uint REPLICATE_7_BIT_TO_8_TABLE[128] = + uint[](0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, + 46, 48, 50, 52, 54, 56, 58, 60, 62, 64, 66, 68, 70, 72, 74, 76, 78, 80, 82, 84, 86, 88, + 90, 92, 94, 96, 98, 100, 102, 104, 106, 108, 110, 112, 114, 116, 118, 120, 122, 124, 126, + 129, 131, 133, 135, 137, 139, 141, 143, 145, 147, 149, 151, 153, 155, 157, 159, 161, 163, + 165, 167, 169, 171, 173, 175, 177, 179, 181, 183, 185, 187, 189, 191, 193, 195, 197, 199, + 201, 203, 205, 207, 209, 211, 213, 215, 217, 219, 221, 223, 225, 227, 229, 231, 233, 235, + 237, 239, 241, 243, 245, 247, 249, 251, 253, 255); // Input ASTC texture globals uint current_index = 0; @@ -230,7 +227,7 @@ uint FastReplicateTo8(uint value, uint num_bits) { case 7: return REPLICATE_7_BIT_TO_8_TABLE[value]; case 8: - return REPLICATE_8_BIT_TO_8_TABLE[value]; + return value; } return Replicate(value, num_bits, 8); } diff --git a/src/video_core/renderer_opengl/util_shaders.cpp b/src/video_core/renderer_opengl/util_shaders.cpp index d57998cdc..7e32f49ca 100644 --- a/src/video_core/renderer_opengl/util_shaders.cpp +++ b/src/video_core/renderer_opengl/util_shaders.cpp @@ -69,7 +69,8 @@ UtilShaders::UtilShaders(ProgramManager& program_manager_) swizzle_table_buffer.Create(); astc_buffer.Create(); glNamedBufferStorage(swizzle_table_buffer.handle, sizeof(swizzle_table), &swizzle_table, 0); - glNamedBufferStorage(astc_buffer.handle, sizeof(ASTC_BUFFER_DATA), &ASTC_BUFFER_DATA, 0); + glNamedBufferStorage(astc_buffer.handle, sizeof(ASTC_ENCODINGS_VALUES), &ASTC_ENCODINGS_VALUES, + 0); } UtilShaders::~UtilShaders() = default; @@ -79,11 +80,6 @@ void UtilShaders::ASTCDecode(Image& image, const ImageBufferMap& map, static constexpr GLuint BINDING_SWIZZLE_BUFFER = 0; static constexpr GLuint BINDING_INPUT_BUFFER = 1; static constexpr GLuint BINDING_ENC_BUFFER = 2; - - static constexpr GLuint BINDING_6_TO_8_BUFFER = 3; - static constexpr GLuint BINDING_7_TO_8_BUFFER = 4; - static constexpr GLuint BINDING_8_TO_8_BUFFER = 5; - static constexpr GLuint BINDING_OUTPUT_IMAGE = 0; const Extent2D tile_size{ @@ -92,18 +88,7 @@ void UtilShaders::ASTCDecode(Image& image, const ImageBufferMap& map, }; program_manager.BindHostCompute(astc_decoder_program.handle); glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_SWIZZLE_BUFFER, swizzle_table_buffer.handle); - glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_ENC_BUFFER, astc_buffer.handle, - offsetof(AstcBufferData, encoding_values), - sizeof(AstcBufferData::encoding_values)); - glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_6_TO_8_BUFFER, astc_buffer.handle, - offsetof(AstcBufferData, replicate_6_to_8), - sizeof(AstcBufferData::replicate_6_to_8)); - glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_7_TO_8_BUFFER, astc_buffer.handle, - offsetof(AstcBufferData, replicate_7_to_8), - sizeof(AstcBufferData::replicate_7_to_8)); - glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_8_TO_8_BUFFER, astc_buffer.handle, - offsetof(AstcBufferData, replicate_8_to_8), - sizeof(AstcBufferData::replicate_8_to_8)); + glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_ENC_BUFFER, astc_buffer.handle); glFlushMappedNamedBufferRange(map.buffer, map.offset, image.guest_size_bytes); glUniform2ui(1, tile_size.width, tile_size.height); diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.cpp b/src/video_core/renderer_vulkan/vk_compute_pass.cpp index 123bed794..205cd3b05 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pass.cpp @@ -30,19 +30,16 @@ namespace Vulkan { using Tegra::Texture::SWIZZLE_TABLE; -using Tegra::Texture::ASTC::EncodingsValues; +using Tegra::Texture::ASTC::ASTC_ENCODINGS_VALUES; using namespace Tegra::Texture::ASTC; namespace { constexpr u32 ASTC_BINDING_INPUT_BUFFER = 0; constexpr u32 ASTC_BINDING_ENC_BUFFER = 1; -constexpr u32 ASTC_BINDING_6_TO_8_BUFFER = 2; -constexpr u32 ASTC_BINDING_7_TO_8_BUFFER = 3; -constexpr u32 ASTC_BINDING_8_TO_8_BUFFER = 4; -constexpr u32 ASTC_BINDING_SWIZZLE_BUFFER = 5; -constexpr u32 ASTC_BINDING_OUTPUT_IMAGE = 6; -constexpr size_t ASTC_NUM_BINDINGS = 7; +constexpr u32 ASTC_BINDING_SWIZZLE_BUFFER = 2; +constexpr u32 ASTC_BINDING_OUTPUT_IMAGE = 3; +constexpr size_t ASTC_NUM_BINDINGS = 4; VkPushConstantRange BuildComputePushConstantRange(std::size_t size) { return { @@ -87,27 +84,6 @@ std::array BuildASTCDescriptorS .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT, .pImmutableSamplers = nullptr, }, - { - .binding = ASTC_BINDING_6_TO_8_BUFFER, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .descriptorCount = 1, - .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT, - .pImmutableSamplers = nullptr, - }, - { - .binding = ASTC_BINDING_7_TO_8_BUFFER, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .descriptorCount = 1, - .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT, - .pImmutableSamplers = nullptr, - }, - { - .binding = ASTC_BINDING_8_TO_8_BUFFER, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .descriptorCount = 1, - .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT, - .pImmutableSamplers = nullptr, - }, { .binding = ASTC_BINDING_SWIZZLE_BUFFER, .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, @@ -155,30 +131,6 @@ BuildASTCPassDescriptorUpdateTemplateEntry() { .offset = ASTC_BINDING_ENC_BUFFER * sizeof(DescriptorUpdateEntry), .stride = sizeof(DescriptorUpdateEntry), }, - { - .dstBinding = ASTC_BINDING_6_TO_8_BUFFER, - .dstArrayElement = 0, - .descriptorCount = 1, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .offset = ASTC_BINDING_6_TO_8_BUFFER * sizeof(DescriptorUpdateEntry), - .stride = sizeof(DescriptorUpdateEntry), - }, - { - .dstBinding = ASTC_BINDING_7_TO_8_BUFFER, - .dstArrayElement = 0, - .descriptorCount = 1, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .offset = ASTC_BINDING_7_TO_8_BUFFER * sizeof(DescriptorUpdateEntry), - .stride = sizeof(DescriptorUpdateEntry), - }, - { - .dstBinding = ASTC_BINDING_8_TO_8_BUFFER, - .dstArrayElement = 0, - .descriptorCount = 1, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .offset = ASTC_BINDING_8_TO_8_BUFFER * sizeof(DescriptorUpdateEntry), - .stride = sizeof(DescriptorUpdateEntry), - }, { .dstBinding = ASTC_BINDING_SWIZZLE_BUFFER, .dstArrayElement = 0, @@ -400,7 +352,7 @@ ASTCDecoderPass::ASTCDecoderPass(const Device& device_, VKScheduler& scheduler_, ASTCDecoderPass::~ASTCDecoderPass() = default; void ASTCDecoderPass::MakeDataBuffer() { - constexpr size_t TOTAL_BUFFER_SIZE = sizeof(ASTC_BUFFER_DATA) + sizeof(SWIZZLE_TABLE); + constexpr size_t TOTAL_BUFFER_SIZE = sizeof(ASTC_ENCODINGS_VALUES) + sizeof(SWIZZLE_TABLE); data_buffer = device.GetLogical().CreateBuffer(VkBufferCreateInfo{ .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO, .pNext = nullptr, @@ -414,9 +366,10 @@ void ASTCDecoderPass::MakeDataBuffer() { data_buffer_commit = memory_allocator.Commit(data_buffer, MemoryUsage::Upload); const auto staging_ref = staging_buffer_pool.Request(TOTAL_BUFFER_SIZE, MemoryUsage::Upload); - std::memcpy(staging_ref.mapped_span.data(), &ASTC_BUFFER_DATA, sizeof(ASTC_BUFFER_DATA)); + std::memcpy(staging_ref.mapped_span.data(), &ASTC_ENCODINGS_VALUES, + sizeof(ASTC_ENCODINGS_VALUES)); // Tack on the swizzle table at the end of the buffer - std::memcpy(staging_ref.mapped_span.data() + sizeof(ASTC_BUFFER_DATA), &SWIZZLE_TABLE, + std::memcpy(staging_ref.mapped_span.data() + sizeof(ASTC_ENCODINGS_VALUES), &SWIZZLE_TABLE, sizeof(SWIZZLE_TABLE)); scheduler.Record([src = staging_ref.buffer, offset = staging_ref.offset, dst = *data_buffer, @@ -486,15 +439,8 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map, update_descriptor_queue.Acquire(); update_descriptor_queue.AddBuffer(map.buffer, input_offset, image.guest_size_bytes - swizzle.buffer_offset); - update_descriptor_queue.AddBuffer(*data_buffer, offsetof(AstcBufferData, encoding_values), - sizeof(AstcBufferData::encoding_values)); - update_descriptor_queue.AddBuffer(*data_buffer, offsetof(AstcBufferData, replicate_6_to_8), - sizeof(AstcBufferData::replicate_6_to_8)); - update_descriptor_queue.AddBuffer(*data_buffer, offsetof(AstcBufferData, replicate_7_to_8), - sizeof(AstcBufferData::replicate_7_to_8)); - update_descriptor_queue.AddBuffer(*data_buffer, offsetof(AstcBufferData, replicate_8_to_8), - sizeof(AstcBufferData::replicate_8_to_8)); - update_descriptor_queue.AddBuffer(*data_buffer, sizeof(AstcBufferData), + update_descriptor_queue.AddBuffer(*data_buffer, 0, sizeof(ASTC_ENCODINGS_VALUES)); + update_descriptor_queue.AddBuffer(*data_buffer, sizeof(ASTC_ENCODINGS_VALUES), sizeof(SWIZZLE_TABLE)); update_descriptor_queue.AddImage(image.StorageImageView(swizzle.level)); diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp index b6e2022f2..7b756ba41 100644 --- a/src/video_core/textures/astc.cpp +++ b/src/video_core/textures/astc.cpp @@ -269,7 +269,7 @@ static void DecodeQuintBlock(InputBitStream& bits, IntegerEncodedVector& result, static void DecodeIntegerSequence(IntegerEncodedVector& result, InputBitStream& bits, u32 maxRange, u32 nValues) { // Determine encoding parameters - IntegerEncodedValue val = EncodingsValues[maxRange]; + IntegerEncodedValue val = ASTC_ENCODINGS_VALUES[maxRange]; // Start decoding u32 nValsDecoded = 0; @@ -310,7 +310,7 @@ struct TexelWeightParams { nIdxs *= 2; } - return EncodingsValues[m_MaxWeight].GetBitLength(nIdxs); + return ASTC_ENCODINGS_VALUES[m_MaxWeight].GetBitLength(nIdxs); } u32 GetNumWeightValues() const { @@ -755,12 +755,12 @@ static void DecodeColorValues(u32* out, std::span data, const u32* modes, co // figure out the max value for each of them... u32 range = 256; while (--range > 0) { - IntegerEncodedValue val = EncodingsValues[range]; + IntegerEncodedValue val = ASTC_ENCODINGS_VALUES[range]; u32 bitLength = val.GetBitLength(nValues); if (bitLength <= nBitsForColorData) { // Find the smallest possible range that matches the given encoding while (--range > 0) { - IntegerEncodedValue newval = EncodingsValues[range]; + IntegerEncodedValue newval = ASTC_ENCODINGS_VALUES[range]; if (!newval.MatchesEncoding(val)) { break; } diff --git a/src/video_core/textures/astc.h b/src/video_core/textures/astc.h index 441e8eb04..0229ae122 100644 --- a/src/video_core/textures/astc.h +++ b/src/video_core/textures/astc.h @@ -77,7 +77,7 @@ constexpr std::array MakeEncodedValues() { return encodings; } -constexpr std::array EncodingsValues = MakeEncodedValues(); +constexpr std::array ASTC_ENCODINGS_VALUES = MakeEncodedValues(); // Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)] // is the same as [(num_bits - 1):0] and repeats all the way down. @@ -120,13 +120,6 @@ constexpr auto REPLICATE_6_BIT_TO_8_TABLE = MakeReplicateTable(); constexpr auto REPLICATE_7_BIT_TO_8_TABLE = MakeReplicateTable(); constexpr auto REPLICATE_8_BIT_TO_8_TABLE = MakeReplicateTable(); -struct AstcBufferData { - decltype(EncodingsValues) encoding_values = EncodingsValues; - decltype(REPLICATE_6_BIT_TO_8_TABLE) replicate_6_to_8 = REPLICATE_6_BIT_TO_8_TABLE; - decltype(REPLICATE_7_BIT_TO_8_TABLE) replicate_7_to_8 = REPLICATE_7_BIT_TO_8_TABLE; - decltype(REPLICATE_8_BIT_TO_8_TABLE) replicate_8_to_8 = REPLICATE_8_BIT_TO_8_TABLE; -} constexpr ASTC_BUFFER_DATA; - void Decompress(std::span data, uint32_t width, uint32_t height, uint32_t depth, uint32_t block_width, uint32_t block_height, std::span output);