3
0
Fork 0
forked from suyu/suyu

Merge pull request #6496 from ameerj/astc-fixes

astc: Various robustness enhancements for the gpu decoder
This commit is contained in:
bunnei 2021-06-24 21:47:05 -07:00 committed by GitHub
commit c805c0b395
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
5 changed files with 50 additions and 155 deletions

View file

@ -11,12 +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_BYTE_TO_16_BUFFER 5
#define BINDING_SWIZZLE_BUFFER 6
#define BINDING_OUTPUT_IMAGE 7
#define BINDING_SWIZZLE_BUFFER 2
#define BINDING_OUTPUT_IMAGE 3
#else // ^^^ Vulkan ^^^ // vvv OpenGL vvv
@ -26,10 +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_BYTE_TO_16_BUFFER 6
#define BINDING_OUTPUT_IMAGE 0
#endif
@ -76,19 +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_BYTE_TO_16_BUFFER, std430) readonly buffer REPLICATE_BYTE_TO_16 {
uint REPLICATE_BYTE_TO_16_TABLE[];
};
layout(binding = BINDING_OUTPUT_IMAGE, rgba8) uniform writeonly image2DArray dest_image;
@ -139,6 +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;
@ -207,8 +199,7 @@ uint Replicate(uint val, uint num_bits, uint to_bit) {
}
uvec4 ReplicateByteTo16(uvec4 value) {
return uvec4(REPLICATE_BYTE_TO_16_TABLE[value.x], REPLICATE_BYTE_TO_16_TABLE[value.y],
REPLICATE_BYTE_TO_16_TABLE[value.z], REPLICATE_BYTE_TO_16_TABLE[value.w]);
return value * 0x101;
}
uint ReplicateBitTo7(uint value) {
@ -236,7 +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);
}
@ -1327,6 +1318,9 @@ void main() {
offset += swizzle;
const ivec3 coord = ivec3(gl_GlobalInvocationID * uvec3(block_dims, 1));
if (any(greaterThanEqual(coord, imageSize(dest_image)))) {
return;
}
uint block_index =
pos.z * gl_WorkGroupSize.x * gl_WorkGroupSize.y + pos.y * gl_WorkGroupSize.x + pos.x;

View file

@ -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;
@ -79,12 +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_BYTE_TO_16_BUFFER = 6;
static constexpr GLuint BINDING_OUTPUT_IMAGE = 0;
const Extent2D tile_size{
@ -93,21 +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));
glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_BYTE_TO_16_BUFFER, astc_buffer.handle,
offsetof(AstcBufferData, replicate_byte_to_16),
sizeof(AstcBufferData::replicate_byte_to_16));
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);
@ -137,6 +118,12 @@ void UtilShaders::ASTCDecode(Image& image, const ImageBufferMap& map,
glDispatchCompute(num_dispatches_x, num_dispatches_y, image.info.resources.layers);
}
// Precautionary barrier to ensure the compute shader is done decoding prior to texture access.
// GL_TEXTURE_FETCH_BARRIER_BIT and GL_SHADER_IMAGE_ACCESS_BARRIER_BIT are used in a separate
// glMemoryBarrier call by the texture cache runtime
glMemoryBarrier(GL_UNIFORM_BARRIER_BIT | GL_COMMAND_BARRIER_BIT | GL_PIXEL_BUFFER_BARRIER_BIT |
GL_TEXTURE_UPDATE_BARRIER_BIT | GL_BUFFER_UPDATE_BARRIER_BIT |
GL_SHADER_STORAGE_BARRIER_BIT | GL_CLIENT_MAPPED_BUFFER_BARRIER_BIT);
program_manager.RestoreGuestCompute();
}

View file

@ -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_BYTE_TO_16_BUFFER = 5;
constexpr u32 ASTC_BINDING_SWIZZLE_BUFFER = 6;
constexpr u32 ASTC_BINDING_OUTPUT_IMAGE = 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 {
@ -71,7 +68,7 @@ std::array<VkDescriptorSetLayoutBinding, 2> BuildInputOutputDescriptorSetBinding
}};
}
std::array<VkDescriptorSetLayoutBinding, 8> BuildASTCDescriptorSetBindings() {
std::array<VkDescriptorSetLayoutBinding, ASTC_NUM_BINDINGS> BuildASTCDescriptorSetBindings() {
return {{
{
.binding = ASTC_BINDING_INPUT_BUFFER,
@ -87,34 +84,6 @@ std::array<VkDescriptorSetLayoutBinding, 8> BuildASTCDescriptorSetBindings() {
.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_BYTE_TO_16_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,
@ -143,7 +112,8 @@ VkDescriptorUpdateTemplateEntryKHR BuildInputOutputDescriptorUpdateTemplate() {
};
}
std::array<VkDescriptorUpdateTemplateEntryKHR, 8> BuildASTCPassDescriptorUpdateTemplateEntry() {
std::array<VkDescriptorUpdateTemplateEntryKHR, ASTC_NUM_BINDINGS>
BuildASTCPassDescriptorUpdateTemplateEntry() {
return {{
{
.dstBinding = ASTC_BINDING_INPUT_BUFFER,
@ -161,38 +131,6 @@ std::array<VkDescriptorUpdateTemplateEntryKHR, 8> BuildASTCPassDescriptorUpdateT
.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_BYTE_TO_16_BUFFER,
.dstArrayElement = 0,
.descriptorCount = 1,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.offset = ASTC_BINDING_BYTE_TO_16_BUFFER * sizeof(DescriptorUpdateEntry),
.stride = sizeof(DescriptorUpdateEntry),
},
{
.dstBinding = ASTC_BINDING_SWIZZLE_BUFFER,
.dstArrayElement = 0,
@ -222,15 +160,6 @@ struct AstcPushConstants {
u32 block_height_mask;
};
struct AstcBufferData {
decltype(SWIZZLE_TABLE) swizzle_table_buffer = SWIZZLE_TABLE;
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;
decltype(REPLICATE_BYTE_TO_16_TABLE) replicate_byte_to_16 = REPLICATE_BYTE_TO_16_TABLE;
} constexpr ASTC_BUFFER_DATA;
} // Anonymous namespace
VKComputePass::VKComputePass(const Device& device, VKDescriptorPool& descriptor_pool,
@ -423,7 +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,
@ -437,9 +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,
@ -509,18 +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,
offsetof(AstcBufferData, replicate_byte_to_16),
sizeof(AstcBufferData::replicate_byte_to_16));
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));
@ -569,6 +489,7 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, image_barrier);
});
scheduler.Finish();
}
} // namespace Vulkan

View file

@ -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 {
@ -551,6 +551,8 @@ static void FillError(std::span<u32> outBuf, u32 blockWidth, u32 blockHeight) {
}
}
}
static constexpr auto REPLICATE_BYTE_TO_16_TABLE = MakeReplicateTable<u32, 8, 16>();
static constexpr u32 ReplicateByteTo16(std::size_t value) {
return REPLICATE_BYTE_TO_16_TABLE[value];
}
@ -753,12 +755,12 @@ static void DecodeColorValues(u32* out, std::span<u8> 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;
}

View file

@ -77,7 +77,7 @@ constexpr std::array<IntegerEncodedValue, 256> MakeEncodedValues() {
return encodings;
}
constexpr std::array<IntegerEncodedValue, 256> EncodingsValues = MakeEncodedValues();
constexpr std::array<IntegerEncodedValue, 256> 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.
@ -116,19 +116,10 @@ constexpr auto MakeReplicateTable() {
return table;
}
constexpr auto REPLICATE_BYTE_TO_16_TABLE = MakeReplicateTable<u32, 8, 16>();
constexpr auto REPLICATE_6_BIT_TO_8_TABLE = MakeReplicateTable<u32, 6, 8>();
constexpr auto REPLICATE_7_BIT_TO_8_TABLE = MakeReplicateTable<u32, 7, 8>();
constexpr auto REPLICATE_8_BIT_TO_8_TABLE = MakeReplicateTable<u32, 8, 8>();
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;
decltype(REPLICATE_BYTE_TO_16_TABLE) replicate_byte_to_16 = REPLICATE_BYTE_TO_16_TABLE;
} constexpr ASTC_BUFFER_DATA;
void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
uint32_t block_width, uint32_t block_height, std::span<uint8_t> output);