Repository navigation
Expand file tree
/
Copy pathmain_webgpu_f16.cpp
More file actions
35 lines (34 loc) · 1.51 KB
/
Copy pathmain_webgpu_f16.cpp
File metadata and controls
35 lines (34 loc) · 1.51 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
#include "webgpu_feature_test.h"
namespace {
auto run()->void {
auto input{gpu_test::buffer(4, wgpu::BufferUsage::Storage | wgpu::BufferUsage::CopyDst)};
auto output{gpu_test::buffer(4, wgpu::BufferUsage::Storage | wgpu::BufferUsage::CopySrc)};
float const value{1.5F};
gpu_test::device.GetQueue().WriteBuffer(input, 0, &value, sizeof(value));
auto module{gpu_test::shader(R"(
enable f16;
@group(0) @binding(0) var<storage, read> input: f32;
@group(0) @binding(1) var<storage, read_write> output: f32;
@compute @workgroup_size(1) fn main() {
output = f32(f16(input) * 2.0h + 0.5h);
}
)")};
wgpu::ComputePipelineDescriptor descriptor{.compute{.module = module, .entryPoint = "main"}};
auto pipeline{gpu_test::device.CreateComputePipeline(&descriptor)};
wgpu::BindGroupEntry entries[]{{.binding = 0, .buffer = input}, {.binding = 1, .buffer = output}};
wgpu::BindGroupDescriptor group_descriptor{.layout = pipeline.GetBindGroupLayout(0), .entryCount = 2, .entries = entries};
auto group{gpu_test::device.CreateBindGroup(&group_descriptor)};
auto encoder{gpu_test::device.CreateCommandEncoder()};
auto pass{encoder.BeginComputePass()};
pass.SetPipeline(pipeline);
pass.SetBindGroup(0, group);
pass.DispatchWorkgroups(1);
pass.End();
gpu_test::submit(encoder);
gpu_test::verify(output, 4, [](void const *data) { return *static_cast<float const *>(data) == 3.5F; },
"f16 shader arithmetic verified");
}
}
auto main()->int {
return gpu_test::start(wgpu::FeatureName::ShaderF16, run);
}