| // Copyright 2019 The Dawn Authors |
| // |
| // Licensed under the Apache License, Version 2.0 (the "License"); |
| // you may not use this file except in compliance with the License. |
| // You may obtain a copy of the License at |
| // |
| // http://www.apache.org/licenses/LICENSE-2.0 |
| // |
| // Unless required by applicable law or agreed to in writing, software |
| // distributed under the License is distributed on an "AS IS" BASIS, |
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| // See the License for the specific language governing permissions and |
| // limitations under the License. |
| |
| #include "tests/DawnTest.h" |
| |
| #include "common/Assert.h" |
| #include "common/Math.h" |
| #include "utils/ComboRenderPipelineDescriptor.h" |
| #include "utils/TextureFormatUtils.h" |
| #include "utils/WGPUHelpers.h" |
| |
| #include <cmath> |
| #include <type_traits> |
| |
| // An expectation for float buffer content that can correctly compare different NaN values and |
| // supports a basic tolerance for comparison of finite values. |
| class ExpectFloatWithTolerance : public detail::Expectation { |
| public: |
| ExpectFloatWithTolerance(std::vector<float> expected, float tolerance) |
| : mExpected(std::move(expected)), mTolerance(tolerance) { |
| } |
| |
| testing::AssertionResult Check(const void* data, size_t size) override { |
| ASSERT(size == sizeof(float) * mExpected.size()); |
| |
| const float* actual = static_cast<const float*>(data); |
| |
| for (size_t i = 0; i < mExpected.size(); ++i) { |
| float expectedValue = mExpected[i]; |
| float actualValue = actual[i]; |
| |
| if (!FloatsMatch(expectedValue, actualValue)) { |
| testing::AssertionResult result = testing::AssertionFailure() |
| << "Expected data[" << i << "] to be close to " |
| << expectedValue << ", actual " << actualValue |
| << std::endl; |
| return result; |
| } |
| } |
| return testing::AssertionSuccess(); |
| } |
| |
| private: |
| bool FloatsMatch(float expected, float actual) { |
| if (std::isnan(expected)) { |
| return std::isnan(actual); |
| } |
| |
| if (std::isinf(expected)) { |
| return std::isinf(actual) && std::signbit(expected) == std::signbit(actual); |
| } |
| |
| if (mTolerance == 0.0f) { |
| return expected == actual; |
| } |
| |
| float error = std::abs(expected - actual); |
| return error < mTolerance; |
| } |
| |
| std::vector<float> mExpected; |
| float mTolerance; |
| }; |
| |
| // An expectation for float16 buffers that can correctly compare NaNs (all NaNs are equivalent). |
| class ExpectFloat16 : public detail::Expectation { |
| public: |
| ExpectFloat16(std::vector<uint16_t> expected) : mExpected(std::move(expected)) { |
| } |
| |
| testing::AssertionResult Check(const void* data, size_t size) override { |
| ASSERT(size == sizeof(uint16_t) * mExpected.size()); |
| |
| const uint16_t* actual = static_cast<const uint16_t*>(data); |
| |
| for (size_t i = 0; i < mExpected.size(); ++i) { |
| uint16_t expectedValue = mExpected[i]; |
| uint16_t actualValue = actual[i]; |
| |
| if (!Floats16Match(expectedValue, actualValue)) { |
| testing::AssertionResult result = testing::AssertionFailure() |
| << "Expected data[" << i << "] to be " |
| << expectedValue << ", actual " << actualValue |
| << std::endl; |
| return result; |
| } |
| } |
| return testing::AssertionSuccess(); |
| } |
| |
| private: |
| bool Floats16Match(float expected, float actual) { |
| if (IsFloat16NaN(expected)) { |
| return IsFloat16NaN(actual); |
| } |
| |
| return expected == actual; |
| } |
| |
| std::vector<uint16_t> mExpected; |
| }; |
| |
| class TextureFormatTest : public DawnTest { |
| protected: |
| // Structure containing all the information that tests need to know about the format. |
| struct FormatTestInfo { |
| wgpu::TextureFormat format; |
| uint32_t texelByteSize; |
| wgpu::TextureComponentType type; |
| uint32_t componentCount; |
| }; |
| |
| // Returns a reprensentation of a format that can be used to contain the "uncompressed" values |
| // of the format. That the equivalent format with all channels 32bit-sized. |
| FormatTestInfo GetUncompressedFormatInfo(FormatTestInfo formatInfo) { |
| switch (formatInfo.type) { |
| case wgpu::TextureComponentType::Float: |
| return {wgpu::TextureFormat::RGBA32Float, 16, formatInfo.type, 4}; |
| case wgpu::TextureComponentType::Sint: |
| return {wgpu::TextureFormat::RGBA32Sint, 16, formatInfo.type, 4}; |
| case wgpu::TextureComponentType::Uint: |
| return {wgpu::TextureFormat::RGBA32Uint, 16, formatInfo.type, 4}; |
| default: |
| UNREACHABLE(); |
| } |
| } |
| |
| // Return a pipeline that can be used in a full-texture draw to sample from the texture in the |
| // bindgroup and output its decompressed values to the render target. |
| wgpu::RenderPipeline CreateSamplePipeline(FormatTestInfo sampleFormatInfo, |
| FormatTestInfo renderFormatInfo) { |
| utils::ComboRenderPipelineDescriptor2 desc; |
| |
| wgpu::ShaderModule vsModule = utils::CreateShaderModule(device, R"( |
| [[stage(vertex)]] |
| fn main([[builtin(vertex_index)]] VertexIndex : u32) -> [[builtin(position)]] vec4<f32> { |
| let pos : array<vec2<f32>, 3> = array<vec2<f32>, 3>( |
| vec2<f32>(-3.0, -1.0), |
| vec2<f32>( 3.0, -1.0), |
| vec2<f32>( 0.0, 2.0)); |
| |
| return vec4<f32>(pos[VertexIndex], 0.0, 1.0); |
| })"); |
| |
| // Compute the WGSL type of the texture's data. |
| const char* type = utils::GetWGSLColorTextureComponentType(sampleFormatInfo.format); |
| |
| std::ostringstream fsSource; |
| fsSource << "[[group(0), binding(0)]] var myTexture : texture_2d<" << type << ">;\n"; |
| fsSource << "struct FragmentOut {\n"; |
| fsSource << " [[location(0)]] color : vec4<" << type << ">;\n"; |
| fsSource << R"(}; |
| [[stage(fragment)]] |
| fn main([[builtin(position)]] FragCoord : vec4<f32>) -> FragmentOut { |
| var output : FragmentOut; |
| output.color = textureLoad(myTexture, vec2<i32>(FragCoord.xy), 0); |
| return output; |
| })"; |
| |
| wgpu::ShaderModule fsModule = utils::CreateShaderModule(device, fsSource.str().c_str()); |
| |
| desc.vertex.module = vsModule; |
| desc.cFragment.module = fsModule; |
| desc.cTargets[0].format = renderFormatInfo.format; |
| |
| return device.CreateRenderPipeline2(&desc); |
| } |
| |
| // The sampling test uploads the sample data in a texture with the sampleFormatInfo.format. |
| // It then samples from it and renders the results in a texture with the |
| // renderFormatInfo.format format. Finally it checks that the data rendered matches |
| // expectedRenderData, using the cutom expectation if present. |
| void DoSampleTest(FormatTestInfo sampleFormatInfo, |
| const void* sampleData, |
| size_t sampleDataSize, |
| FormatTestInfo renderFormatInfo, |
| const void* expectedRenderData, |
| size_t expectedRenderDataSize, |
| detail::Expectation* customExpectation) { |
| // The input data should contain an exact number of texels |
| ASSERT(sampleDataSize % sampleFormatInfo.texelByteSize == 0); |
| uint32_t width = sampleDataSize / sampleFormatInfo.texelByteSize; |
| |
| // The input data must be a multiple of 4 byte in length for WriteBuffer |
| ASSERT(sampleDataSize % 4 == 0); |
| ASSERT(expectedRenderDataSize % 4 == 0); |
| |
| // Create the texture we will sample from |
| wgpu::TextureDescriptor sampleTextureDesc; |
| sampleTextureDesc.usage = wgpu::TextureUsage::CopyDst | wgpu::TextureUsage::Sampled; |
| sampleTextureDesc.size = {width, 1, 1}; |
| sampleTextureDesc.format = sampleFormatInfo.format; |
| wgpu::Texture sampleTexture = device.CreateTexture(&sampleTextureDesc); |
| |
| wgpu::Buffer uploadBuffer = utils::CreateBufferFromData(device, sampleData, sampleDataSize, |
| wgpu::BufferUsage::CopySrc); |
| |
| // Create the texture that we will render results to |
| ASSERT(expectedRenderDataSize == width * renderFormatInfo.texelByteSize); |
| |
| wgpu::TextureDescriptor renderTargetDesc; |
| renderTargetDesc.usage = wgpu::TextureUsage::CopySrc | wgpu::TextureUsage::RenderAttachment; |
| renderTargetDesc.size = {width, 1, 1}; |
| renderTargetDesc.format = renderFormatInfo.format; |
| |
| wgpu::Texture renderTarget = device.CreateTexture(&renderTargetDesc); |
| |
| // Create the readback buffer for the data in renderTarget |
| wgpu::BufferDescriptor readbackBufferDesc; |
| readbackBufferDesc.usage = wgpu::BufferUsage::CopyDst | wgpu::BufferUsage::CopySrc; |
| readbackBufferDesc.size = expectedRenderDataSize; |
| wgpu::Buffer readbackBuffer = device.CreateBuffer(&readbackBufferDesc); |
| |
| // Prepare objects needed to sample from texture in the renderpass |
| wgpu::RenderPipeline pipeline = CreateSamplePipeline(sampleFormatInfo, renderFormatInfo); |
| wgpu::BindGroup bindGroup = utils::MakeBindGroup(device, pipeline.GetBindGroupLayout(0), |
| {{0, sampleTexture.CreateView()}}); |
| |
| // Encode commands for the test that fill texture, sample it to render to renderTarget then |
| // copy renderTarget in a buffer so we can read it easily. |
| wgpu::CommandEncoder encoder = device.CreateCommandEncoder(); |
| |
| { |
| wgpu::ImageCopyBuffer bufferView = utils::CreateImageCopyBuffer(uploadBuffer, 0, 256); |
| wgpu::ImageCopyTexture textureView = |
| utils::CreateImageCopyTexture(sampleTexture, 0, {0, 0, 0}); |
| wgpu::Extent3D extent{width, 1, 1}; |
| encoder.CopyBufferToTexture(&bufferView, &textureView, &extent); |
| } |
| |
| utils::ComboRenderPassDescriptor renderPassDesc({renderTarget.CreateView()}); |
| wgpu::RenderPassEncoder renderPass = encoder.BeginRenderPass(&renderPassDesc); |
| renderPass.SetPipeline(pipeline); |
| renderPass.SetBindGroup(0, bindGroup); |
| renderPass.Draw(3); |
| renderPass.EndPass(); |
| |
| { |
| wgpu::ImageCopyBuffer bufferView = utils::CreateImageCopyBuffer(readbackBuffer, 0, 256); |
| wgpu::ImageCopyTexture textureView = |
| utils::CreateImageCopyTexture(renderTarget, 0, {0, 0, 0}); |
| wgpu::Extent3D extent{width, 1, 1}; |
| encoder.CopyTextureToBuffer(&textureView, &bufferView, &extent); |
| } |
| |
| wgpu::CommandBuffer commands = encoder.Finish(); |
| queue.Submit(1, &commands); |
| |
| // For floats use a special expectation that understands how to compare NaNs and support a |
| // tolerance. |
| if (customExpectation != nullptr) { |
| AddBufferExpectation(__FILE__, __LINE__, readbackBuffer, 0, expectedRenderDataSize, |
| customExpectation); |
| } else { |
| EXPECT_BUFFER_U32_RANGE_EQ(static_cast<const uint32_t*>(expectedRenderData), |
| readbackBuffer, 0, |
| expectedRenderDataSize / sizeof(uint32_t)); |
| } |
| } |
| |
| template <typename Data> |
| std::vector<Data> ExpandDataTo4Component(const std::vector<Data>& originalData, |
| uint32_t originalComponentCount, |
| const std::array<Data, 4>& defaultValues) { |
| std::vector<Data> result; |
| |
| for (size_t i = 0; i < originalData.size() / originalComponentCount; i++) { |
| for (size_t component = 0; component < 4; component++) { |
| if (component < originalComponentCount) { |
| result.push_back(originalData[i * originalComponentCount + component]); |
| } else { |
| result.push_back(defaultValues[component]); |
| } |
| } |
| } |
| |
| return result; |
| } |
| |
| // Helper functions used to run tests that convert the typeful test objects to typeless void* |
| |
| template <typename TextureData, typename RenderData> |
| void DoFormatSamplingTest(FormatTestInfo formatInfo, |
| const std::vector<TextureData>& textureData, |
| const std::vector<RenderData>& expectedRenderData, |
| detail::Expectation* customExpectation = nullptr) { |
| FormatTestInfo renderFormatInfo = GetUncompressedFormatInfo(formatInfo); |
| |
| // Expand the expected data to be 4 component wide with the default sampling values of |
| // (0, 0, 0, 1) |
| std::array<RenderData, 4> defaultValues = {RenderData(0), RenderData(0), RenderData(0), |
| RenderData(1)}; |
| std::vector<RenderData> expandedRenderData = |
| ExpandDataTo4Component(expectedRenderData, formatInfo.componentCount, defaultValues); |
| |
| DoSampleTest(formatInfo, textureData.data(), textureData.size() * sizeof(TextureData), |
| renderFormatInfo, expandedRenderData.data(), |
| expandedRenderData.size() * sizeof(RenderData), customExpectation); |
| } |
| |
| template <typename TextureData> |
| void DoFloatFormatSamplingTest(FormatTestInfo formatInfo, |
| const std::vector<TextureData>& textureData, |
| const std::vector<float>& expectedRenderData, |
| float floatTolerance = 0.0f) { |
| // Expand the expected data to be 4 component wide with the default sampling values of |
| // (0, 0, 0, 1) |
| std::array<float, 4> defaultValues = {0.0f, 0.0f, 0.0f, 1.0f}; |
| std::vector<float> expandedRenderData = |
| ExpandDataTo4Component(expectedRenderData, formatInfo.componentCount, defaultValues); |
| |
| // Use a special expectation that understands how to compare NaNs and supports a tolerance. |
| DoFormatSamplingTest(formatInfo, textureData, expectedRenderData, |
| new ExpectFloatWithTolerance(expandedRenderData, floatTolerance)); |
| } |
| |
| template <typename TextureData, typename RenderData> |
| void DoFormatRenderingTest(FormatTestInfo formatInfo, |
| const std::vector<TextureData>& textureData, |
| const std::vector<RenderData>& expectedRenderData, |
| detail::Expectation* customExpectation = nullptr) { |
| FormatTestInfo sampleFormatInfo = GetUncompressedFormatInfo(formatInfo); |
| |
| // Expand the sampling texture data to contain garbage data for unused components to check |
| // that they don't influence the rendering result. |
| std::array<TextureData, 4> garbageValues; |
| garbageValues.fill(13); |
| std::vector<TextureData> expandedTextureData = |
| ExpandDataTo4Component(textureData, formatInfo.componentCount, garbageValues); |
| |
| DoSampleTest(sampleFormatInfo, expandedTextureData.data(), |
| expandedTextureData.size() * sizeof(TextureData), formatInfo, |
| expectedRenderData.data(), expectedRenderData.size() * sizeof(RenderData), |
| customExpectation); |
| } |
| |
| // Below are helper functions for types that are very similar to one another so the logic is |
| // shared. |
| |
| template <typename T> |
| void DoUnormTest(FormatTestInfo formatInfo) { |
| static_assert(!std::is_signed<T>::value && std::is_integral<T>::value, ""); |
| ASSERT(sizeof(T) * formatInfo.componentCount == formatInfo.texelByteSize); |
| ASSERT(formatInfo.type == wgpu::TextureComponentType::Float); |
| |
| T maxValue = std::numeric_limits<T>::max(); |
| std::vector<T> textureData = {0, 1, maxValue, maxValue}; |
| std::vector<float> uncompressedData = {0.0f, 1.0f / maxValue, 1.0f, 1.0f}; |
| |
| DoFormatSamplingTest(formatInfo, textureData, uncompressedData); |
| DoFormatRenderingTest(formatInfo, uncompressedData, textureData); |
| } |
| |
| template <typename T> |
| void DoSnormTest(FormatTestInfo formatInfo) { |
| static_assert(std::is_signed<T>::value && std::is_integral<T>::value, ""); |
| ASSERT(sizeof(T) * formatInfo.componentCount == formatInfo.texelByteSize); |
| ASSERT(formatInfo.type == wgpu::TextureComponentType::Float); |
| |
| T maxValue = std::numeric_limits<T>::max(); |
| T minValue = std::numeric_limits<T>::min(); |
| std::vector<T> textureData = {0, 1, -1, maxValue, minValue, T(minValue + 1), 0, 0}; |
| std::vector<float> uncompressedData = { |
| 0.0f, 1.0f / maxValue, -1.0f / maxValue, 1.0f, -1.0f, -1.0f, 0.0f, 0.0f}; |
| |
| DoFloatFormatSamplingTest(formatInfo, textureData, uncompressedData, 0.0001f / maxValue); |
| // Snorm formats aren't renderable because they are not guaranteed renderable in Vulkan |
| } |
| |
| template <typename T> |
| void DoUintTest(FormatTestInfo formatInfo) { |
| static_assert(!std::is_signed<T>::value && std::is_integral<T>::value, ""); |
| ASSERT(sizeof(T) * formatInfo.componentCount == formatInfo.texelByteSize); |
| ASSERT(formatInfo.type == wgpu::TextureComponentType::Uint); |
| |
| T maxValue = std::numeric_limits<T>::max(); |
| std::vector<T> textureData = {0, 1, maxValue, maxValue}; |
| std::vector<uint32_t> uncompressedData = {0, 1, maxValue, maxValue}; |
| |
| DoFormatSamplingTest(formatInfo, textureData, uncompressedData); |
| DoFormatRenderingTest(formatInfo, uncompressedData, textureData); |
| } |
| |
| template <typename T> |
| void DoSintTest(FormatTestInfo formatInfo) { |
| static_assert(std::is_signed<T>::value && std::is_integral<T>::value, ""); |
| ASSERT(sizeof(T) * formatInfo.componentCount == formatInfo.texelByteSize); |
| ASSERT(formatInfo.type == wgpu::TextureComponentType::Sint); |
| |
| T maxValue = std::numeric_limits<T>::max(); |
| T minValue = std::numeric_limits<T>::min(); |
| std::vector<T> textureData = {0, 1, maxValue, minValue}; |
| std::vector<int32_t> uncompressedData = {0, 1, maxValue, minValue}; |
| |
| DoFormatSamplingTest(formatInfo, textureData, uncompressedData); |
| DoFormatRenderingTest(formatInfo, uncompressedData, textureData); |
| } |
| |
| void DoFloat32Test(FormatTestInfo formatInfo) { |
| ASSERT(sizeof(float) * formatInfo.componentCount == formatInfo.texelByteSize); |
| ASSERT(formatInfo.type == wgpu::TextureComponentType::Float); |
| |
| std::vector<float> textureData = {+0.0f, -0.0f, 1.0f, 1.0e-29f, |
| 1.0e29f, NAN, INFINITY, -INFINITY}; |
| |
| DoFloatFormatSamplingTest(formatInfo, textureData, textureData); |
| DoFormatRenderingTest(formatInfo, textureData, textureData, |
| new ExpectFloatWithTolerance(textureData, 0.0f)); |
| } |
| |
| void DoFloat16Test(FormatTestInfo formatInfo) { |
| ASSERT(sizeof(int16_t) * formatInfo.componentCount == formatInfo.texelByteSize); |
| ASSERT(formatInfo.type == wgpu::TextureComponentType::Float); |
| |
| std::vector<float> uncompressedData = {+0.0f, -0.0f, 1.0f, 1.01e-4f, |
| 1.0e4f, NAN, INFINITY, -INFINITY}; |
| std::vector<uint16_t> textureData; |
| for (float value : uncompressedData) { |
| textureData.push_back(Float32ToFloat16(value)); |
| } |
| |
| DoFloatFormatSamplingTest(formatInfo, textureData, uncompressedData, 1.0e-5f); |
| |
| // Use a special expectation that knows that all Float16 NaNs are equivalent. |
| DoFormatRenderingTest(formatInfo, uncompressedData, textureData, |
| new ExpectFloat16(textureData)); |
| } |
| }; |
| |
| // Test the R8Unorm format |
| TEST_P(TextureFormatTest, R8Unorm) { |
| DoUnormTest<uint8_t>({wgpu::TextureFormat::R8Unorm, 1, wgpu::TextureComponentType::Float, 1}); |
| } |
| |
| // Test the RG8Unorm format |
| TEST_P(TextureFormatTest, RG8Unorm) { |
| DoUnormTest<uint8_t>({wgpu::TextureFormat::RG8Unorm, 2, wgpu::TextureComponentType::Float, 2}); |
| } |
| |
| // Test the RGBA8Unorm format |
| TEST_P(TextureFormatTest, RGBA8Unorm) { |
| DoUnormTest<uint8_t>( |
| {wgpu::TextureFormat::RGBA8Unorm, 4, wgpu::TextureComponentType::Float, 4}); |
| } |
| |
| // Test the BGRA8Unorm format |
| TEST_P(TextureFormatTest, BGRA8Unorm) { |
| // TODO(crbug.com/dawn/596): BGRA is unsupported on OpenGL ES; add workaround or validation |
| DAWN_SKIP_TEST_IF(IsOpenGLES()); |
| uint8_t maxValue = std::numeric_limits<uint8_t>::max(); |
| std::vector<uint8_t> textureData = {maxValue, 1, 0, maxValue}; |
| std::vector<float> uncompressedData = {0.0f, 1.0f / maxValue, 1.0f, 1.0f}; |
| DoFormatSamplingTest({wgpu::TextureFormat::BGRA8Unorm, 4, wgpu::TextureComponentType::Float, 4}, |
| textureData, uncompressedData); |
| DoFormatRenderingTest( |
| {wgpu::TextureFormat::BGRA8Unorm, 4, wgpu::TextureComponentType::Float, 4}, |
| uncompressedData, textureData); |
| } |
| |
| // Test the R8Snorm format |
| TEST_P(TextureFormatTest, R8Snorm) { |
| DoSnormTest<int8_t>({wgpu::TextureFormat::R8Snorm, 1, wgpu::TextureComponentType::Float, 1}); |
| } |
| |
| // Test the RG8Snorm format |
| TEST_P(TextureFormatTest, RG8Snorm) { |
| DoSnormTest<int8_t>({wgpu::TextureFormat::RG8Snorm, 2, wgpu::TextureComponentType::Float, 2}); |
| } |
| |
| // Test the RGBA8Snorm format |
| TEST_P(TextureFormatTest, RGBA8Snorm) { |
| DoSnormTest<int8_t>({wgpu::TextureFormat::RGBA8Snorm, 4, wgpu::TextureComponentType::Float, 4}); |
| } |
| |
| // Test the R8Uint format |
| TEST_P(TextureFormatTest, R8Uint) { |
| DoUintTest<uint8_t>({wgpu::TextureFormat::R8Uint, 1, wgpu::TextureComponentType::Uint, 1}); |
| } |
| |
| // Test the RG8Uint format |
| TEST_P(TextureFormatTest, RG8Uint) { |
| DoUintTest<uint8_t>({wgpu::TextureFormat::RG8Uint, 2, wgpu::TextureComponentType::Uint, 2}); |
| } |
| |
| // Test the RGBA8Uint format |
| TEST_P(TextureFormatTest, RGBA8Uint) { |
| DoUintTest<uint8_t>({wgpu::TextureFormat::RGBA8Uint, 4, wgpu::TextureComponentType::Uint, 4}); |
| } |
| |
| // Test the R16Uint format |
| TEST_P(TextureFormatTest, R16Uint) { |
| DoUintTest<uint16_t>({wgpu::TextureFormat::R16Uint, 2, wgpu::TextureComponentType::Uint, 1}); |
| } |
| |
| // Test the RG16Uint format |
| TEST_P(TextureFormatTest, RG16Uint) { |
| DoUintTest<uint16_t>({wgpu::TextureFormat::RG16Uint, 4, wgpu::TextureComponentType::Uint, 2}); |
| } |
| |
| // Test the RGBA16Uint format |
| TEST_P(TextureFormatTest, RGBA16Uint) { |
| DoUintTest<uint16_t>({wgpu::TextureFormat::RGBA16Uint, 8, wgpu::TextureComponentType::Uint, 4}); |
| } |
| |
| // Test the R32Uint format |
| TEST_P(TextureFormatTest, R32Uint) { |
| DoUintTest<uint32_t>({wgpu::TextureFormat::R32Uint, 4, wgpu::TextureComponentType::Uint, 1}); |
| } |
| |
| // Test the RG32Uint format |
| TEST_P(TextureFormatTest, RG32Uint) { |
| DoUintTest<uint32_t>({wgpu::TextureFormat::RG32Uint, 8, wgpu::TextureComponentType::Uint, 2}); |
| } |
| |
| // Test the RGBA32Uint format |
| TEST_P(TextureFormatTest, RGBA32Uint) { |
| DoUintTest<uint32_t>( |
| {wgpu::TextureFormat::RGBA32Uint, 16, wgpu::TextureComponentType::Uint, 4}); |
| } |
| |
| // Test the R8Sint format |
| TEST_P(TextureFormatTest, R8Sint) { |
| DoSintTest<int8_t>({wgpu::TextureFormat::R8Sint, 1, wgpu::TextureComponentType::Sint, 1}); |
| } |
| |
| // Test the RG8Sint format |
| TEST_P(TextureFormatTest, RG8Sint) { |
| DoSintTest<int8_t>({wgpu::TextureFormat::RG8Sint, 2, wgpu::TextureComponentType::Sint, 2}); |
| } |
| |
| // Test the RGBA8Sint format |
| TEST_P(TextureFormatTest, RGBA8Sint) { |
| DoSintTest<int8_t>({wgpu::TextureFormat::RGBA8Sint, 4, wgpu::TextureComponentType::Sint, 4}); |
| } |
| |
| // Test the R16Sint format |
| TEST_P(TextureFormatTest, R16Sint) { |
| DoSintTest<int16_t>({wgpu::TextureFormat::R16Sint, 2, wgpu::TextureComponentType::Sint, 1}); |
| } |
| |
| // Test the RG16Sint format |
| TEST_P(TextureFormatTest, RG16Sint) { |
| DoSintTest<int16_t>({wgpu::TextureFormat::RG16Sint, 4, wgpu::TextureComponentType::Sint, 2}); |
| } |
| |
| // Test the RGBA16Sint format |
| TEST_P(TextureFormatTest, RGBA16Sint) { |
| DoSintTest<int16_t>({wgpu::TextureFormat::RGBA16Sint, 8, wgpu::TextureComponentType::Sint, 4}); |
| } |
| |
| // Test the R32Sint format |
| TEST_P(TextureFormatTest, R32Sint) { |
| DoSintTest<int32_t>({wgpu::TextureFormat::R32Sint, 4, wgpu::TextureComponentType::Sint, 1}); |
| } |
| |
| // Test the RG32Sint format |
| TEST_P(TextureFormatTest, RG32Sint) { |
| DoSintTest<int32_t>({wgpu::TextureFormat::RG32Sint, 8, wgpu::TextureComponentType::Sint, 2}); |
| } |
| |
| // Test the RGBA32Sint format |
| TEST_P(TextureFormatTest, RGBA32Sint) { |
| DoSintTest<int32_t>({wgpu::TextureFormat::RGBA32Sint, 16, wgpu::TextureComponentType::Sint, 4}); |
| } |
| |
| // Test the R32Float format |
| TEST_P(TextureFormatTest, R32Float) { |
| DoFloat32Test({wgpu::TextureFormat::R32Float, 4, wgpu::TextureComponentType::Float, 1}); |
| } |
| |
| // Test the RG32Float format |
| TEST_P(TextureFormatTest, RG32Float) { |
| DoFloat32Test({wgpu::TextureFormat::RG32Float, 8, wgpu::TextureComponentType::Float, 2}); |
| } |
| |
| // Test the RGBA32Float format |
| TEST_P(TextureFormatTest, RGBA32Float) { |
| DoFloat32Test({wgpu::TextureFormat::RGBA32Float, 16, wgpu::TextureComponentType::Float, 4}); |
| } |
| |
| // Test the R16Float format |
| TEST_P(TextureFormatTest, R16Float) { |
| // TODO(https://crbug.com/swiftshader/147) Rendering INFINITY isn't handled correctly by |
| // swiftshader |
| DAWN_SKIP_TEST_IF(IsVulkan() && IsSwiftshader() || IsANGLE()); |
| |
| DoFloat16Test({wgpu::TextureFormat::R16Float, 2, wgpu::TextureComponentType::Float, 1}); |
| } |
| |
| // Test the RG16Float format |
| TEST_P(TextureFormatTest, RG16Float) { |
| // TODO(https://crbug.com/swiftshader/147) Rendering INFINITY isn't handled correctly by |
| // swiftshader |
| DAWN_SKIP_TEST_IF(IsVulkan() && IsSwiftshader() || IsANGLE()); |
| |
| DoFloat16Test({wgpu::TextureFormat::RG16Float, 4, wgpu::TextureComponentType::Float, 2}); |
| } |
| |
| // Test the RGBA16Float format |
| TEST_P(TextureFormatTest, RGBA16Float) { |
| // TODO(https://crbug.com/swiftshader/147) Rendering INFINITY isn't handled correctly by |
| // swiftshader |
| DAWN_SKIP_TEST_IF(IsVulkan() && IsSwiftshader() || IsANGLE()); |
| |
| DoFloat16Test({wgpu::TextureFormat::RGBA16Float, 8, wgpu::TextureComponentType::Float, 4}); |
| } |
| |
| // Test the RGBA8Unorm format |
| TEST_P(TextureFormatTest, RGBA8UnormSrgb) { |
| uint8_t maxValue = std::numeric_limits<uint8_t>::max(); |
| std::vector<uint8_t> textureData = {0, 1, maxValue, 64, 35, 68, 152, 168}; |
| |
| std::vector<float> uncompressedData; |
| for (size_t i = 0; i < textureData.size(); i += 4) { |
| uncompressedData.push_back(SRGBToLinear(textureData[i + 0] / float(maxValue))); |
| uncompressedData.push_back(SRGBToLinear(textureData[i + 1] / float(maxValue))); |
| uncompressedData.push_back(SRGBToLinear(textureData[i + 2] / float(maxValue))); |
| // Alpha is linear for sRGB formats |
| uncompressedData.push_back(textureData[i + 3] / float(maxValue)); |
| } |
| |
| DoFloatFormatSamplingTest( |
| {wgpu::TextureFormat::RGBA8UnormSrgb, 4, wgpu::TextureComponentType::Float, 4}, textureData, |
| uncompressedData, 1.0e-3); |
| DoFormatRenderingTest( |
| {wgpu::TextureFormat::RGBA8UnormSrgb, 4, wgpu::TextureComponentType::Float, 4}, |
| uncompressedData, textureData); |
| } |
| |
| // Test the BGRA8UnormSrgb format |
| TEST_P(TextureFormatTest, BGRA8UnormSrgb) { |
| // TODO(cwallez@chromium.org): This format doesn't exist in OpenGL, emulate it using |
| // RGBA8UnormSrgb and swizzling / shader twiddling |
| DAWN_SKIP_TEST_IF(IsOpenGL()); |
| DAWN_SKIP_TEST_IF(IsOpenGLES()); |
| |
| uint8_t maxValue = std::numeric_limits<uint8_t>::max(); |
| std::vector<uint8_t> textureData = {0, 1, maxValue, 64, 35, 68, 152, 168}; |
| |
| std::vector<float> uncompressedData; |
| for (size_t i = 0; i < textureData.size(); i += 4) { |
| // Note that R and B are swapped |
| uncompressedData.push_back(SRGBToLinear(textureData[i + 2] / float(maxValue))); |
| uncompressedData.push_back(SRGBToLinear(textureData[i + 1] / float(maxValue))); |
| uncompressedData.push_back(SRGBToLinear(textureData[i + 0] / float(maxValue))); |
| // Alpha is linear for sRGB formats |
| uncompressedData.push_back(textureData[i + 3] / float(maxValue)); |
| } |
| |
| DoFloatFormatSamplingTest( |
| {wgpu::TextureFormat::BGRA8UnormSrgb, 4, wgpu::TextureComponentType::Float, 4}, textureData, |
| uncompressedData, 1.0e-3); |
| DoFormatRenderingTest( |
| {wgpu::TextureFormat::BGRA8UnormSrgb, 4, wgpu::TextureComponentType::Float, 4}, |
| uncompressedData, textureData); |
| } |
| |
| // Test the RGB10A2Unorm format |
| TEST_P(TextureFormatTest, RGB10A2Unorm) { |
| auto MakeRGB10A2 = [](uint32_t r, uint32_t g, uint32_t b, uint32_t a) -> uint32_t { |
| ASSERT((r & 0x3FF) == r); |
| ASSERT((g & 0x3FF) == g); |
| ASSERT((b & 0x3FF) == b); |
| ASSERT((a & 0x3) == a); |
| return r | g << 10 | b << 20 | a << 30; |
| }; |
| |
| std::vector<uint32_t> textureData = {MakeRGB10A2(0, 0, 0, 0), MakeRGB10A2(1023, 1023, 1023, 1), |
| MakeRGB10A2(243, 576, 765, 2), MakeRGB10A2(0, 0, 0, 3)}; |
| // clang-format off |
| std::vector<float> uncompressedData = { |
| 0.0f, 0.0f, 0.0f, 0.0f, |
| 1.0f, 1.0f, 1.0f, 1 / 3.0f, |
| 243 / 1023.0f, 576 / 1023.0f, 765 / 1023.0f, 2 / 3.0f, |
| 0.0f, 0.0f, 0.0f, 1.0f |
| }; |
| // clang-format on |
| |
| DoFloatFormatSamplingTest( |
| {wgpu::TextureFormat::RGB10A2Unorm, 4, wgpu::TextureComponentType::Float, 4}, textureData, |
| uncompressedData, 1.0e-5); |
| DoFormatRenderingTest( |
| {wgpu::TextureFormat::RGB10A2Unorm, 4, wgpu::TextureComponentType::Float, 4}, |
| uncompressedData, textureData); |
| } |
| |
| // Test the RG11B10Ufloat format |
| TEST_P(TextureFormatTest, RG11B10Ufloat) { |
| constexpr uint32_t kFloat11Zero = 0; |
| constexpr uint32_t kFloat11Infinity = 0x7C0; |
| constexpr uint32_t kFloat11Nan = 0x7C1; |
| constexpr uint32_t kFloat11One = 0x3C0; |
| |
| constexpr uint32_t kFloat10Zero = 0; |
| constexpr uint32_t kFloat10Infinity = 0x3E0; |
| constexpr uint32_t kFloat10Nan = 0x3E1; |
| constexpr uint32_t kFloat10One = 0x1E0; |
| |
| auto MakeRG11B10 = [](uint32_t r, uint32_t g, uint32_t b) { |
| ASSERT((r & 0x7FF) == r); |
| ASSERT((g & 0x7FF) == g); |
| ASSERT((b & 0x3FF) == b); |
| return r | g << 11 | b << 22; |
| }; |
| |
| // Test each of (0, 1, INFINITY, NaN) for each component but never two with the same value at a |
| // time. |
| std::vector<uint32_t> textureData = { |
| MakeRG11B10(kFloat11Zero, kFloat11Infinity, kFloat10Nan), |
| MakeRG11B10(kFloat11Infinity, kFloat11Nan, kFloat10One), |
| MakeRG11B10(kFloat11Nan, kFloat11One, kFloat10Zero), |
| MakeRG11B10(kFloat11One, kFloat11Zero, kFloat10Infinity), |
| }; |
| |
| // This is one of the only 3-channel formats, so we don't have specific testing for them. Alpha |
| // should always be sampled as 1 |
| // clang-format off |
| std::vector<float> uncompressedData = { |
| 0.0f, INFINITY, NAN, 1.0f, |
| INFINITY, NAN, 1.0f, 1.0f, |
| NAN, 1.0f, 0.0f, 1.0f, |
| 1.0f, 0.0f, INFINITY, 1.0f |
| }; |
| // clang-format on |
| |
| DoFloatFormatSamplingTest( |
| {wgpu::TextureFormat::RG11B10Ufloat, 4, wgpu::TextureComponentType::Float, 4}, textureData, |
| uncompressedData); |
| // This format is not renderable. |
| } |
| |
| // Test the RGB9E5Ufloat format |
| TEST_P(TextureFormatTest, RGB9E5Ufloat) { |
| // RGB9E5 is different from other floating point formats because the mantissa doesn't index in |
| // the window defined by the exponent but is instead treated as a pure multiplier. There is |
| // also no Infinity or NaN. The OpenGL 4.6 spec has the best explanation I've found in section |
| // 8.25 "Shared Exponent Texture Color Conversion": |
| // |
| // red = reduint * 2^(expuint - B - N) = reduint * 2^(expuint - 24) |
| // |
| // Where reduint and expuint are the integer values when considering the E5 as a 5bit uint, and |
| // the r9 as a 9bit uint. B the number of bits of the mantissa (9), and N the offset for the |
| // exponent (15). |
| |
| float smallestExponent = std::pow(2.0f, -24.0f); |
| float largestExponent = std::pow(2.0f, float(31 - 24)); |
| |
| auto MakeRGB9E5 = [](uint32_t r, uint32_t g, uint32_t b, uint32_t e) { |
| ASSERT((r & 0x1FF) == r); |
| ASSERT((g & 0x1FF) == g); |
| ASSERT((b & 0x1FF) == b); |
| ASSERT((e & 0x1F) == e); |
| return r | g << 9 | b << 18 | e << 27; |
| }; |
| |
| // Test the smallest largest, and "1" exponents |
| std::vector<uint32_t> textureData = { |
| MakeRGB9E5(0, 1, 2, 0b00000), |
| MakeRGB9E5(2, 1, 0, 0b11111), |
| MakeRGB9E5(0, 1, 2, 0b11000), |
| }; |
| |
| // This is one of the only 3-channel formats, so we don't have specific testing for them. Alpha |
| // should always be sampled as 1 |
| // clang-format off |
| std::vector<float> uncompressedData = { |
| 0.0f, smallestExponent, 2.0f * smallestExponent, 1.0f, |
| 2.0f * largestExponent, largestExponent, 0.0f, 1.0f, |
| 0.0f, 1.0f, 2.0f, 1.0f, |
| }; |
| // clang-format on |
| |
| DoFloatFormatSamplingTest( |
| {wgpu::TextureFormat::RGB9E5Ufloat, 4, wgpu::TextureComponentType::Float, 4}, textureData, |
| uncompressedData); |
| // This format is not renderable. |
| } |
| |
| // TODO(cwallez@chromium.org): Add tests for depth-stencil formats when we know if they are copyable |
| // in WebGPU. |
| |
| DAWN_INSTANTIATE_TEST(TextureFormatTest, |
| D3D12Backend(), |
| MetalBackend(), |
| OpenGLBackend(), |
| OpenGLESBackend(), |
| VulkanBackend()); |