blob: e916e6d5ee010b6c959c580a85423e6c93d42320 [file] [log] [blame] [edit]
// Copyright 2020 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 "common/Assert.h"
#include "tests/DawnTest.h"
#include "utils/ComboRenderPipelineDescriptor.h"
#include "utils/WGPUHelpers.h"
namespace {
constexpr wgpu::TextureFormat kDepthFormats[] = {
wgpu::TextureFormat::Depth32Float,
wgpu::TextureFormat::Depth24Plus,
wgpu::TextureFormat::Depth24PlusStencil8,
wgpu::TextureFormat::Depth16Unorm,
};
constexpr wgpu::TextureFormat kStencilFormats[] = {
wgpu::TextureFormat::Depth24PlusStencil8,
};
constexpr wgpu::CompareFunction kCompareFunctions[] = {
wgpu::CompareFunction::Never, wgpu::CompareFunction::Less,
wgpu::CompareFunction::LessEqual, wgpu::CompareFunction::Greater,
wgpu::CompareFunction::GreaterEqual, wgpu::CompareFunction::Equal,
wgpu::CompareFunction::NotEqual, wgpu::CompareFunction::Always,
};
// Test a "normal" ref value between 0 and 1; as well as negative and > 1 refs.
constexpr float kCompareRefs[] = {-0.1, 0.4, 1.2};
// Test 0, below the ref, equal to, above the ref, and 1.
const std::vector<float> kNormalizedTextureValues = {0.0, 0.3, 0.4, 0.5, 1.0};
// Test the limits, and some values in between.
const std::vector<uint32_t> kStencilValues = {0, 1, 38, 255};
} // anonymous namespace
class DepthStencilSamplingTest : public DawnTest {
protected:
enum class TestAspect {
Depth,
Stencil,
};
void SetUp() override {
DawnTest::SetUp();
wgpu::BufferDescriptor uniformBufferDesc;
uniformBufferDesc.usage = wgpu::BufferUsage::Uniform | wgpu::BufferUsage::CopyDst;
uniformBufferDesc.size = sizeof(float);
mUniformBuffer = device.CreateBuffer(&uniformBufferDesc);
}
void GenerateSamplingShader(const std::vector<TestAspect>& aspects,
const std::vector<uint32_t> components,
std::ostringstream& shaderSource,
std::ostringstream& shaderBody) {
shaderSource << "type StencilValues = array<u32, " << components.size() << ">;\n";
shaderSource << R"(
[[block]] struct DepthResult {
value : f32;
};
[[block]] struct StencilResult {
values : StencilValues;
};)";
shaderSource << "\n";
uint32_t index = 0;
for (TestAspect aspect : aspects) {
switch (aspect) {
case TestAspect::Depth:
shaderSource << "[[group(0), binding(" << 2 * index << ")]] var tex" << index
<< " : texture_depth_2d;\n";
shaderSource << "[[group(0), binding(" << 2 * index + 1
<< ")]] var<storage, read_write> result" << index
<< " : DepthResult;\n";
ASSERT(components.size() == 1 && components[0] == 0);
shaderBody << "\nresult" << index << ".value = textureLoad(tex" << index
<< ", vec2<i32>(0, 0), 0);";
break;
case TestAspect::Stencil:
shaderSource << "[[group(0), binding(" << 2 * index << ")]] var tex" << index
<< " : texture_2d<u32>;\n";
shaderSource << "[[group(0), binding(" << 2 * index + 1
<< ")]] var<storage, read_write> result" << index
<< " : StencilResult;\n";
shaderBody << "var texel = textureLoad(tex" << index
<< ", vec2<i32>(0, 0), 0);";
for (uint32_t i = 0; i < components.size(); ++i) {
shaderBody << "\nresult" << index << ".values[" << i << "] = texel["
<< components[i] << "];";
}
break;
}
index++;
}
}
wgpu::RenderPipeline CreateSamplingRenderPipeline(std::vector<TestAspect> aspects,
std::vector<uint32_t> components) {
wgpu::ShaderModule vsModule = utils::CreateShaderModule(device, R"(
[[stage(vertex)]] fn main() -> [[builtin(position)]] vec4<f32> {
return vec4<f32>(0.0, 0.0, 0.0, 1.0);
})");
utils::ComboRenderPipelineDescriptor pipelineDescriptor;
std::ostringstream shaderSource;
std::ostringstream shaderOutputStruct;
std::ostringstream shaderBody;
GenerateSamplingShader(aspects, components, shaderSource, shaderBody);
shaderSource << "[[stage(fragment)]] fn main() -> [[location(0)]] vec4<f32> {\n";
shaderSource << shaderBody.str() << "return vec4<f32>();\n }";
wgpu::ShaderModule fsModule = utils::CreateShaderModule(device, shaderSource.str().c_str());
pipelineDescriptor.vertex.module = vsModule;
pipelineDescriptor.cFragment.module = fsModule;
pipelineDescriptor.primitive.topology = wgpu::PrimitiveTopology::PointList;
pipelineDescriptor.cTargets[0].writeMask = wgpu::ColorWriteMask::None;
return device.CreateRenderPipeline(&pipelineDescriptor);
}
wgpu::ComputePipeline CreateSamplingComputePipeline(std::vector<TestAspect> aspects,
std::vector<uint32_t> components) {
std::ostringstream shaderSource;
std::ostringstream shaderBody;
GenerateSamplingShader(aspects, components, shaderSource, shaderBody);
shaderSource << "[[stage(compute), workgroup_size(1)]] fn main() { " << shaderBody.str()
<< "\n}";
wgpu::ShaderModule csModule = utils::CreateShaderModule(device, shaderSource.str().c_str());
wgpu::ComputePipelineDescriptor pipelineDescriptor;
pipelineDescriptor.compute.module = csModule;
pipelineDescriptor.compute.entryPoint = "main";
return device.CreateComputePipeline(&pipelineDescriptor);
}
wgpu::RenderPipeline CreateSamplingRenderPipeline(std::vector<TestAspect> aspects,
uint32_t componentIndex) {
return CreateSamplingRenderPipeline(std::move(aspects),
std::vector<uint32_t>{componentIndex});
}
wgpu::ComputePipeline CreateSamplingComputePipeline(std::vector<TestAspect> aspects,
uint32_t componentIndex) {
return CreateSamplingComputePipeline(std::move(aspects),
std::vector<uint32_t>{componentIndex});
}
wgpu::RenderPipeline CreateComparisonRenderPipeline() {
wgpu::ShaderModule vsModule = utils::CreateShaderModule(device, R"(
[[stage(vertex)]] fn main() -> [[builtin(position)]] vec4<f32> {
return vec4<f32>(0.0, 0.0, 0.0, 1.0);
})");
wgpu::ShaderModule fsModule = utils::CreateShaderModule(device, R"(
[[group(0), binding(0)]] var samp : sampler_comparison;
[[group(0), binding(1)]] var tex : texture_depth_2d;
[[block]] struct Uniforms {
compareRef : f32;
};
[[group(0), binding(2)]] var<uniform> uniforms : Uniforms;
[[stage(fragment)]] fn main() -> [[location(0)]] f32 {
return textureSampleCompare(tex, samp, vec2<f32>(0.5, 0.5), uniforms.compareRef);
})");
utils::ComboRenderPipelineDescriptor pipelineDescriptor;
pipelineDescriptor.vertex.module = vsModule;
pipelineDescriptor.cFragment.module = fsModule;
pipelineDescriptor.primitive.topology = wgpu::PrimitiveTopology::PointList;
pipelineDescriptor.cTargets[0].format = wgpu::TextureFormat::R32Float;
return device.CreateRenderPipeline(&pipelineDescriptor);
}
wgpu::ComputePipeline CreateComparisonComputePipeline() {
wgpu::ShaderModule csModule = utils::CreateShaderModule(device, R"(
[[group(0), binding(0)]] var samp : sampler_comparison;
[[group(0), binding(1)]] var tex : texture_depth_2d;
[[block]] struct Uniforms {
compareRef : f32;
};
[[group(0), binding(2)]] var<uniform> uniforms : Uniforms;
[[block]] struct SamplerResult {
value : f32;
};
[[group(0), binding(3)]] var<storage, read_write> samplerResult : SamplerResult;
[[stage(compute), workgroup_size(1)]] fn main() {
samplerResult.value = textureSampleCompare(tex, samp, vec2<f32>(0.5, 0.5), uniforms.compareRef);
})");
wgpu::ComputePipelineDescriptor pipelineDescriptor;
pipelineDescriptor.compute.module = csModule;
pipelineDescriptor.compute.entryPoint = "main";
return device.CreateComputePipeline(&pipelineDescriptor);
}
wgpu::Texture CreateInputTexture(wgpu::TextureFormat format) {
wgpu::TextureDescriptor inputTextureDesc;
inputTextureDesc.usage =
wgpu::TextureUsage::TextureBinding | wgpu::TextureUsage::RenderAttachment;
inputTextureDesc.size = {1, 1, 1};
inputTextureDesc.format = format;
return device.CreateTexture(&inputTextureDesc);
}
wgpu::Texture CreateOutputTexture(wgpu::TextureFormat format) {
wgpu::TextureDescriptor outputTextureDesc;
outputTextureDesc.usage =
wgpu::TextureUsage::RenderAttachment | wgpu::TextureUsage::CopySrc;
outputTextureDesc.size = {1, 1, 1};
outputTextureDesc.format = format;
return device.CreateTexture(&outputTextureDesc);
}
wgpu::Buffer CreateOutputBuffer(uint32_t componentCount = 1) {
wgpu::BufferDescriptor outputBufferDesc;
outputBufferDesc.usage = wgpu::BufferUsage::Storage | wgpu::BufferUsage::CopySrc;
outputBufferDesc.size = sizeof(uint32_t) * componentCount;
return device.CreateBuffer(&outputBufferDesc);
}
void UpdateInputDepth(wgpu::CommandEncoder commandEncoder,
wgpu::Texture texture,
float depthValue) {
utils::ComboRenderPassDescriptor passDescriptor({}, texture.CreateView());
passDescriptor.cDepthStencilAttachmentInfo.clearDepth = depthValue;
wgpu::RenderPassEncoder pass = commandEncoder.BeginRenderPass(&passDescriptor);
pass.EndPass();
}
void UpdateInputStencil(wgpu::CommandEncoder commandEncoder,
wgpu::Texture texture,
uint8_t stencilValue) {
utils::ComboRenderPassDescriptor passDescriptor({}, texture.CreateView());
passDescriptor.cDepthStencilAttachmentInfo.clearStencil = stencilValue;
wgpu::RenderPassEncoder pass = commandEncoder.BeginRenderPass(&passDescriptor);
pass.EndPass();
}
template <typename T, typename CheckBufferFn>
void DoSamplingTestImpl(TestAspect aspect,
wgpu::RenderPipeline pipeline,
wgpu::TextureFormat format,
std::vector<T> textureValues,
uint32_t componentCount,
CheckBufferFn CheckBuffer) {
wgpu::Texture inputTexture = CreateInputTexture(format);
wgpu::TextureViewDescriptor inputViewDesc = {};
switch (aspect) {
case TestAspect::Depth:
inputViewDesc.aspect = wgpu::TextureAspect::DepthOnly;
break;
case TestAspect::Stencil:
inputViewDesc.aspect = wgpu::TextureAspect::StencilOnly;
break;
}
wgpu::Buffer outputBuffer = CreateOutputBuffer(componentCount);
wgpu::BindGroup bindGroup =
utils::MakeBindGroup(device, pipeline.GetBindGroupLayout(0),
{{0, inputTexture.CreateView(&inputViewDesc)}, {1, outputBuffer}});
for (size_t i = 0; i < textureValues.size(); ++i) {
// Set the input depth texture to the provided texture value
wgpu::CommandEncoder commandEncoder = device.CreateCommandEncoder();
switch (aspect) {
case TestAspect::Depth:
UpdateInputDepth(commandEncoder, inputTexture, textureValues[i]);
break;
case TestAspect::Stencil:
UpdateInputStencil(commandEncoder, inputTexture, textureValues[i]);
break;
}
// Render into the output texture
{
utils::BasicRenderPass renderPass =
utils::CreateBasicRenderPass(device, 1, 1, wgpu::TextureFormat::RGBA8Unorm);
wgpu::RenderPassEncoder pass =
commandEncoder.BeginRenderPass(&renderPass.renderPassInfo);
pass.SetPipeline(pipeline);
pass.SetBindGroup(0, bindGroup);
pass.Draw(1);
pass.EndPass();
}
wgpu::CommandBuffer commands = commandEncoder.Finish();
queue.Submit(1, &commands);
CheckBuffer(textureValues[i], outputBuffer);
}
}
template <typename T, typename CheckBufferFn>
void DoSamplingTestImpl(TestAspect aspect,
wgpu::ComputePipeline pipeline,
wgpu::TextureFormat format,
std::vector<T> textureValues,
uint32_t componentCount,
CheckBufferFn CheckBuffer) {
wgpu::Texture inputTexture = CreateInputTexture(format);
wgpu::TextureViewDescriptor inputViewDesc = {};
switch (aspect) {
case TestAspect::Depth:
inputViewDesc.aspect = wgpu::TextureAspect::DepthOnly;
break;
case TestAspect::Stencil:
inputViewDesc.aspect = wgpu::TextureAspect::StencilOnly;
break;
}
wgpu::Buffer outputBuffer = CreateOutputBuffer(componentCount);
wgpu::BindGroup bindGroup =
utils::MakeBindGroup(device, pipeline.GetBindGroupLayout(0),
{{0, inputTexture.CreateView(&inputViewDesc)}, {1, outputBuffer}});
for (size_t i = 0; i < textureValues.size(); ++i) {
// Set the input depth texture to the provided texture value
wgpu::CommandEncoder commandEncoder = device.CreateCommandEncoder();
switch (aspect) {
case TestAspect::Depth:
UpdateInputDepth(commandEncoder, inputTexture, textureValues[i]);
break;
case TestAspect::Stencil:
UpdateInputStencil(commandEncoder, inputTexture, textureValues[i]);
break;
}
// Sample into the output buffer
{
wgpu::ComputePassEncoder pass = commandEncoder.BeginComputePass();
pass.SetPipeline(pipeline);
pass.SetBindGroup(0, bindGroup);
pass.Dispatch(1);
pass.EndPass();
}
wgpu::CommandBuffer commands = commandEncoder.Finish();
queue.Submit(1, &commands);
CheckBuffer(textureValues[i], outputBuffer);
}
}
template <typename T>
void DoSamplingTest(TestAspect aspect,
wgpu::RenderPipeline pipeline,
wgpu::TextureFormat format,
std::vector<T> textureValues,
T tolerance = {}) {
DoSamplingTestImpl(aspect, pipeline, format, textureValues, 1,
[this, tolerance](T expected, wgpu::Buffer buffer) {
EXPECT_BUFFER(buffer, 0, sizeof(T),
new ::detail::ExpectEq<T>(expected, tolerance));
});
}
template <typename T>
void DoSamplingTest(TestAspect aspect,
wgpu::ComputePipeline pipeline,
wgpu::TextureFormat format,
std::vector<T> textureValues,
T tolerance = {}) {
DoSamplingTestImpl(aspect, pipeline, format, textureValues, 1,
[this, tolerance](T expected, wgpu::Buffer buffer) {
EXPECT_BUFFER(buffer, 0, sizeof(T),
new ::detail::ExpectEq<T>(expected, tolerance));
});
}
class ExtraStencilComponentsExpectation : public detail::Expectation {
using StencilData = std::array<uint32_t, 4>;
public:
ExtraStencilComponentsExpectation(uint32_t expected) : mExpected(expected) {
}
~ExtraStencilComponentsExpectation() override = default;
testing::AssertionResult Check(const void* rawData, size_t size) override {
ASSERT(size == sizeof(StencilData));
const uint32_t* data = static_cast<const uint32_t*>(rawData);
StencilData ssss = {mExpected, mExpected, mExpected, mExpected};
StencilData s001 = {mExpected, 0, 0, 1};
if (memcmp(data, ssss.data(), size) == 0 || memcmp(data, s001.data(), size) == 0) {
return testing::AssertionSuccess();
}
return testing::AssertionFailure() << "Expected stencil data to be "
<< "(" << ssss[0] << ", " << ssss[1] << ", "
<< ssss[2] << ", " << ssss[3] << ") or "
<< "(" << s001[0] << ", " << s001[1] << ", "
<< s001[2] << ", " << s001[3] << "). Got "
<< "(" << data[0] << ", " << data[1] << ", "
<< data[2] << ", " << data[3] << ").";
}
private:
uint32_t mExpected;
};
void DoSamplingExtraStencilComponentsRenderTest(TestAspect aspect,
wgpu::TextureFormat format,
std::vector<uint8_t> textureValues) {
DoSamplingTestImpl(aspect,
CreateSamplingRenderPipeline({TestAspect::Stencil}, {0, 1, 2, 3}),
format, textureValues, 4, [&](uint32_t expected, wgpu::Buffer buffer) {
EXPECT_BUFFER(buffer, 0, 4 * sizeof(uint32_t),
new ExtraStencilComponentsExpectation(expected));
});
}
void DoSamplingExtraStencilComponentsComputeTest(TestAspect aspect,
wgpu::TextureFormat format,
std::vector<uint8_t> textureValues) {
DoSamplingTestImpl(aspect,
CreateSamplingComputePipeline({TestAspect::Stencil}, {0, 1, 2, 3}),
format, textureValues, 4, [&](uint32_t expected, wgpu::Buffer buffer) {
EXPECT_BUFFER(buffer, 0, 4 * sizeof(uint32_t),
new ExtraStencilComponentsExpectation(expected));
});
}
static bool CompareFunctionPasses(float compareRef,
wgpu::CompareFunction compare,
float textureValue) {
switch (compare) {
case wgpu::CompareFunction::Never:
return false;
case wgpu::CompareFunction::Less:
return compareRef < textureValue;
case wgpu::CompareFunction::LessEqual:
return compareRef <= textureValue;
case wgpu::CompareFunction::Greater:
return compareRef > textureValue;
case wgpu::CompareFunction::GreaterEqual:
return compareRef >= textureValue;
case wgpu::CompareFunction::Equal:
return compareRef == textureValue;
case wgpu::CompareFunction::NotEqual:
return compareRef != textureValue;
case wgpu::CompareFunction::Always:
return true;
default:
return false;
}
}
void DoDepthCompareRefTest(wgpu::RenderPipeline pipeline,
wgpu::TextureFormat format,
float compareRef,
wgpu::CompareFunction compare,
std::vector<float> textureValues) {
queue.WriteBuffer(mUniformBuffer, 0, &compareRef, sizeof(float));
wgpu::SamplerDescriptor samplerDesc;
samplerDesc.compare = compare;
wgpu::Sampler sampler = device.CreateSampler(&samplerDesc);
wgpu::Texture inputTexture = CreateInputTexture(format);
wgpu::TextureViewDescriptor inputViewDesc = {};
inputViewDesc.aspect = wgpu::TextureAspect::DepthOnly;
wgpu::BindGroup bindGroup =
utils::MakeBindGroup(device, pipeline.GetBindGroupLayout(0),
{
{0, sampler},
{1, inputTexture.CreateView(&inputViewDesc)},
{2, mUniformBuffer},
});
wgpu::Texture outputTexture = CreateOutputTexture(wgpu::TextureFormat::R32Float);
for (float textureValue : textureValues) {
// Set the input depth texture to the provided texture value
wgpu::CommandEncoder commandEncoder = device.CreateCommandEncoder();
UpdateInputDepth(commandEncoder, inputTexture, textureValue);
// Render into the output texture
{
utils::ComboRenderPassDescriptor passDescriptor({outputTexture.CreateView()});
wgpu::RenderPassEncoder pass = commandEncoder.BeginRenderPass(&passDescriptor);
pass.SetPipeline(pipeline);
pass.SetBindGroup(0, bindGroup);
pass.Draw(1);
pass.EndPass();
}
wgpu::CommandBuffer commands = commandEncoder.Finish();
queue.Submit(1, &commands);
EXPECT_TEXTURE_EQ(CompareFunctionPasses(compareRef, compare, textureValue) ? 1.f : 0.f,
outputTexture, {0, 0});
}
}
void DoDepthCompareRefTest(wgpu::ComputePipeline pipeline,
wgpu::TextureFormat format,
float compareRef,
wgpu::CompareFunction compare,
std::vector<float> textureValues) {
queue.WriteBuffer(mUniformBuffer, 0, &compareRef, sizeof(float));
wgpu::SamplerDescriptor samplerDesc;
samplerDesc.compare = compare;
wgpu::Sampler sampler = device.CreateSampler(&samplerDesc);
wgpu::Texture inputTexture = CreateInputTexture(format);
wgpu::TextureViewDescriptor inputViewDesc = {};
inputViewDesc.aspect = wgpu::TextureAspect::DepthOnly;
wgpu::Buffer outputBuffer = CreateOutputBuffer();
wgpu::BindGroup bindGroup =
utils::MakeBindGroup(device, pipeline.GetBindGroupLayout(0),
{{0, sampler},
{1, inputTexture.CreateView(&inputViewDesc)},
{2, mUniformBuffer},
{3, outputBuffer}});
for (float textureValue : textureValues) {
// Set the input depth texture to the provided texture value
wgpu::CommandEncoder commandEncoder = device.CreateCommandEncoder();
UpdateInputDepth(commandEncoder, inputTexture, textureValue);
// Sample into the output buffer
{
wgpu::ComputePassEncoder pass = commandEncoder.BeginComputePass();
pass.SetPipeline(pipeline);
pass.SetBindGroup(0, bindGroup);
pass.Dispatch(1);
pass.EndPass();
}
wgpu::CommandBuffer commands = commandEncoder.Finish();
queue.Submit(1, &commands);
float float0 = 0.f;
float float1 = 1.f;
float* expected =
CompareFunctionPasses(compareRef, compare, textureValue) ? &float1 : &float0;
EXPECT_BUFFER_U32_EQ(*reinterpret_cast<uint32_t*>(expected), outputBuffer, 0);
}
}
private:
wgpu::Buffer mUniformBuffer;
};
// Test that sampling a depth texture with a render/compute pipeline works
TEST_P(DepthStencilSamplingTest, SampleDepth) {
for (wgpu::TextureFormat format : kDepthFormats) {
float tolerance = 0.0f;
if (format == wgpu::TextureFormat::Depth16Unorm) {
tolerance = 0.001f;
}
// Test 0, between [0, 1], and 1.
DoSamplingTest(TestAspect::Depth, CreateSamplingRenderPipeline({TestAspect::Depth}, 0),
format, kNormalizedTextureValues, tolerance);
DoSamplingTest(TestAspect::Depth, CreateSamplingComputePipeline({TestAspect::Depth}, 0),
format, kNormalizedTextureValues, tolerance);
}
}
// Test that sampling a stencil texture with a render/compute pipeline works
TEST_P(DepthStencilSamplingTest, SampleStencil) {
// TODO(crbug.com/dawn/593): This test requires glTextureView, which is unsupported on GLES.
DAWN_TEST_UNSUPPORTED_IF(IsOpenGLES());
for (wgpu::TextureFormat format : kStencilFormats) {
DoSamplingTest(TestAspect::Stencil, CreateSamplingRenderPipeline({TestAspect::Stencil}, 0),
format, kStencilValues);
DoSamplingTest(TestAspect::Stencil, CreateSamplingComputePipeline({TestAspect::Stencil}, 0),
format, kStencilValues);
}
}
// Test that sampling a depth/stencil texture at components 1, 2, and 3 yield 0, 0, and 1
// respectively
TEST_P(DepthStencilSamplingTest, SampleExtraComponents) {
// TODO(crbug.com/dawn/593): This test requires glTextureView, which is unsupported on GLES.
DAWN_TEST_UNSUPPORTED_IF(IsOpenGLES());
DoSamplingExtraStencilComponentsRenderTest(
TestAspect::Stencil, wgpu::TextureFormat::Depth24PlusStencil8, {uint8_t(42), uint8_t(37)});
DoSamplingExtraStencilComponentsComputeTest(
TestAspect::Stencil, wgpu::TextureFormat::Depth24PlusStencil8, {uint8_t(42), uint8_t(37)});
}
// Test sampling both depth and stencil with a render/compute pipeline works.
TEST_P(DepthStencilSamplingTest, SampleDepthAndStencilRender) {
// TODO(crbug.com/dawn/593): This test requires glTextureView, which is unsupported on GLES.
DAWN_TEST_UNSUPPORTED_IF(IsOpenGLES());
wgpu::SamplerDescriptor samplerDesc;
wgpu::Sampler sampler = device.CreateSampler(&samplerDesc);
wgpu::Texture inputTexture = CreateInputTexture(wgpu::TextureFormat::Depth24PlusStencil8);
wgpu::TextureViewDescriptor depthViewDesc = {};
depthViewDesc.aspect = wgpu::TextureAspect::DepthOnly;
wgpu::TextureViewDescriptor stencilViewDesc = {};
stencilViewDesc.aspect = wgpu::TextureAspect::StencilOnly;
// With render pipeline
{
wgpu::RenderPipeline pipeline =
CreateSamplingRenderPipeline({TestAspect::Depth, TestAspect::Stencil}, 0);
wgpu::Buffer depthOutput = CreateOutputBuffer();
wgpu::Buffer stencilOutput = CreateOutputBuffer();
wgpu::BindGroup bindGroup =
utils::MakeBindGroup(device, pipeline.GetBindGroupLayout(0),
{
{0, inputTexture.CreateView(&depthViewDesc)},
{1, depthOutput},
{2, inputTexture.CreateView(&stencilViewDesc)},
{3, stencilOutput},
});
wgpu::CommandEncoder commandEncoder = device.CreateCommandEncoder();
// Initialize both depth and stencil aspects.
utils::ComboRenderPassDescriptor passDescriptor({}, inputTexture.CreateView());
passDescriptor.cDepthStencilAttachmentInfo.clearDepth = 0.43f;
passDescriptor.cDepthStencilAttachmentInfo.clearStencil = 31;
wgpu::RenderPassEncoder pass = commandEncoder.BeginRenderPass(&passDescriptor);
pass.EndPass();
// Render into the output textures
{
utils::BasicRenderPass renderPass =
utils::CreateBasicRenderPass(device, 1, 1, wgpu::TextureFormat::RGBA8Unorm);
wgpu::RenderPassEncoder pass =
commandEncoder.BeginRenderPass(&renderPass.renderPassInfo);
pass.SetPipeline(pipeline);
pass.SetBindGroup(0, bindGroup);
pass.Draw(1);
pass.EndPass();
}
wgpu::CommandBuffer commands = commandEncoder.Finish();
queue.Submit(1, &commands);
uint32_t expectedValueU32 = 0;
memcpy(&expectedValueU32, &passDescriptor.cDepthStencilAttachmentInfo.clearDepth,
sizeof(float));
EXPECT_BUFFER_U32_EQ(expectedValueU32, depthOutput, 0);
expectedValueU32 = 0;
memcpy(&expectedValueU32, &passDescriptor.cDepthStencilAttachmentInfo.clearStencil,
sizeof(uint8_t));
EXPECT_BUFFER_U32_EQ(expectedValueU32, stencilOutput, 0);
}
// With compute pipeline
{
wgpu::ComputePipeline pipeline =
CreateSamplingComputePipeline({TestAspect::Depth, TestAspect::Stencil}, 0);
wgpu::Buffer depthOutput = CreateOutputBuffer();
wgpu::Buffer stencilOutput = CreateOutputBuffer();
wgpu::BindGroup bindGroup =
utils::MakeBindGroup(device, pipeline.GetBindGroupLayout(0),
{{0, inputTexture.CreateView(&depthViewDesc)},
{1, depthOutput},
{2, inputTexture.CreateView(&stencilViewDesc)},
{3, stencilOutput}});
wgpu::CommandEncoder commandEncoder = device.CreateCommandEncoder();
// Initialize both depth and stencil aspects.
utils::ComboRenderPassDescriptor passDescriptor({}, inputTexture.CreateView());
passDescriptor.cDepthStencilAttachmentInfo.clearDepth = 0.43f;
passDescriptor.cDepthStencilAttachmentInfo.clearStencil = 31;
wgpu::RenderPassEncoder pass = commandEncoder.BeginRenderPass(&passDescriptor);
pass.EndPass();
// Sample into the output buffers
{
wgpu::ComputePassEncoder pass = commandEncoder.BeginComputePass();
pass.SetPipeline(pipeline);
pass.SetBindGroup(0, bindGroup);
pass.Dispatch(1);
pass.EndPass();
}
wgpu::CommandBuffer commands = commandEncoder.Finish();
queue.Submit(1, &commands);
uint32_t expectedValueU32 = 0;
memcpy(&expectedValueU32, &passDescriptor.cDepthStencilAttachmentInfo.clearDepth,
sizeof(float));
EXPECT_BUFFER_U32_EQ(expectedValueU32, depthOutput, 0);
expectedValueU32 = 0;
memcpy(&expectedValueU32, &passDescriptor.cDepthStencilAttachmentInfo.clearStencil,
sizeof(uint8_t));
EXPECT_BUFFER_U32_EQ(expectedValueU32, stencilOutput, 0);
}
}
// Test that sampling in a render pipeline with all of the compare functions works.
TEST_P(DepthStencilSamplingTest, CompareFunctionsRender) {
// Initialization via renderPass loadOp doesn't work on Mac Intel.
DAWN_SUPPRESS_TEST_IF(IsMetal() && IsIntel());
wgpu::RenderPipeline pipeline = CreateComparisonRenderPipeline();
for (wgpu::TextureFormat format : kDepthFormats) {
// Test does not account for precision issues when comparison testing Depth16Unorm.
if (format == wgpu::TextureFormat::Depth16Unorm) {
continue;
}
// Test a "normal" ref value between 0 and 1; as well as negative and > 1 refs.
for (float compareRef : kCompareRefs) {
// Test 0, below the ref, equal to, above the ref, and 1.
for (wgpu::CompareFunction f : kCompareFunctions) {
DoDepthCompareRefTest(pipeline, format, compareRef, f, kNormalizedTextureValues);
}
}
}
}
DAWN_INSTANTIATE_TEST(DepthStencilSamplingTest,
D3D12Backend(),
MetalBackend(),
OpenGLBackend(),
OpenGLESBackend(),
VulkanBackend());