adding compute pass

This commit is contained in:
2025-09-10 17:46:53 +02:00
parent 287f18402e
commit ff24a19480
17 changed files with 426 additions and 82 deletions

View File

@@ -33,7 +33,14 @@ const char* hlsl_source = HLSL_SOURCE(
int main(void)
{
PulseBackend backend = PulseLoadBackend(PULSE_BACKEND_D3D11, PULSE_SHADER_FORMAT_HLSL_BIT, PULSE_HIGH_DEBUG);
PulseFlags shader_format;
#ifdef PULSE_D3D11_COMPILER_UNAVAILABLE
shader_format = PULSE_SHADER_FORMAT_DXBC_BIT;
#else
shader_format = PULSE_SHADER_FORMAT_HLSL_BIT;
#endif
PulseBackend backend = PulseLoadBackend(PULSE_BACKEND_D3D11, shader_format, PULSE_HIGH_DEBUG);
PulseSetDebugCallback(backend, DebugCallBack);
PulseDevice device = PulseCreateDevice(backend, NULL, 0);
@@ -45,16 +52,30 @@ int main(void)
// GPU computations
{
PulseComputePipelineCreateInfo info = { 0 };
info.code_size = strlen(hlsl_source);
info.code = (const uint8_t*)hlsl_source;
#ifdef PULSE_D3D11_COMPILER_UNAVAILABLE
const uint8_t shader_bytecode[] = {
#include "shader.cso.h"
};
info.code_size = sizeof(shader_bytecode);
info.code = shader_bytecode;
#else
info.code_size = strlen(hlsl_source);
info.code = (const uint8_t*)hlsl_source;
#endif
info.entrypoint = "CSMain";
info.format = PULSE_SHADER_FORMAT_HLSL_BIT;
info.format = shader_format;
info.num_readwrite_storage_buffers = 1;
PulseComputePipeline pipeline = PulseCreateComputePipeline(device, &info);
PulseFence fence = PulseCreateFence(device);
PulseCommandList cmd = PulseRequestCommandList(device, PULSE_COMMAND_LIST_GENERAL);
PulseComputePass pass = PulseBeginComputePass(cmd);
PulseBindStorageBuffers(pass, &buffer, 1);
PulseBindComputePipeline(pass, pipeline);
PulseDispatchComputations(pass, 16, 1, 1);
PulseEndComputePass(pass);
PulseSubmitCommandList(device, cmd, fence);
PulseWaitForFences(device, &fence, 1, true);
@@ -63,6 +84,42 @@ int main(void)
PulseDestroyComputePipeline(device, pipeline);
}
// Get result and read it on CPU
{
PulseBufferCreateInfo staging_buffer_create_info = { 0 };
staging_buffer_create_info.size = BUFFER_SIZE;
staging_buffer_create_info.usage = PULSE_BUFFER_USAGE_TRANSFER_UPLOAD | PULSE_BUFFER_USAGE_TRANSFER_DOWNLOAD;
PulseBuffer staging_buffer = PulseCreateBuffer(device, &staging_buffer_create_info);
PulseFence fence = PulseCreateFence(device);
PulseCommandList cmd = PulseRequestCommandList(device, PULSE_COMMAND_LIST_TRANSFER_ONLY);
PulseBufferRegion src_region = { 0 };
src_region.buffer = buffer;
src_region.size = BUFFER_SIZE;
PulseBufferRegion dst_region = { 0 };
dst_region.buffer = staging_buffer;
dst_region.size = BUFFER_SIZE;
PulseCopyBufferToBuffer(cmd, &src_region, &dst_region);
PulseSubmitCommandList(device, cmd, fence);
PulseWaitForFences(device, &fence, 1, true);
void* ptr;
PulseMapBuffer(staging_buffer, PULSE_MAP_READ, &ptr);
for(uint32_t i = 0; i < BUFFER_SIZE / sizeof(uint32_t); i++)
printf("%d, ", ((int32_t*)ptr)[i]);
puts("");
PulseUnmapBuffer(staging_buffer);
PulseDestroyBuffer(device, staging_buffer);
PulseReleaseCommandList(device, cmd);
PulseDestroyFence(device, fence);
}
PulseDestroyBuffer(device, buffer);
PulseDestroyDevice(device);

44
Examples/D3D11/shader.cso.h git.filemode.normal_file
View File

@@ -0,0 +1,44 @@
0x44, 0x58, 0x42, 0x43, 0x73, 0x34, 0xc8, 0x1f, 0xdb, 0x93, 0x0e, 0xcb,
0x41, 0xcd, 0x2e, 0x23, 0xde, 0x7c, 0x23, 0x9b, 0x01, 0x00, 0x00, 0x00,
0x08, 0x02, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00, 0x34, 0x00, 0x00, 0x00,
0xd0, 0x00, 0x00, 0x00, 0xe0, 0x00, 0x00, 0x00, 0xf0, 0x00, 0x00, 0x00,
0x6c, 0x01, 0x00, 0x00, 0x52, 0x44, 0x45, 0x46, 0x94, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
0x3c, 0x00, 0x00, 0x00, 0x00, 0x05, 0x53, 0x43, 0x00, 0x01, 0x00, 0x00,
0x61, 0x00, 0x00, 0x00, 0x52, 0x44, 0x31, 0x31, 0x3c, 0x00, 0x00, 0x00,
0x18, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x28, 0x00, 0x00, 0x00,
0x24, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x5c, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
0x01, 0x00, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00,
0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x73, 0x73, 0x62, 0x6f,
0x00, 0x4d, 0x69, 0x63, 0x72, 0x6f, 0x73, 0x6f, 0x66, 0x74, 0x20, 0x28,
0x52, 0x29, 0x20, 0x48, 0x4c, 0x53, 0x4c, 0x20, 0x53, 0x68, 0x61, 0x64,
0x65, 0x72, 0x20, 0x43, 0x6f, 0x6d, 0x70, 0x69, 0x6c, 0x65, 0x72, 0x20,
0x39, 0x2e, 0x32, 0x39, 0x2e, 0x39, 0x35, 0x32, 0x2e, 0x33, 0x31, 0x31,
0x31, 0x00, 0xab, 0xab, 0x49, 0x53, 0x47, 0x4e, 0x08, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x08, 0x00, 0x00, 0x00, 0x4f, 0x53, 0x47, 0x4e,
0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x08, 0x00, 0x00, 0x00,
0x53, 0x48, 0x45, 0x58, 0x74, 0x00, 0x00, 0x00, 0x50, 0x00, 0x05, 0x00,
0x1d, 0x00, 0x00, 0x00, 0x6a, 0x08, 0x00, 0x01, 0x9c, 0x08, 0x00, 0x04,
0x00, 0xe0, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x33, 0x33, 0x00, 0x00,
0x5f, 0x00, 0x00, 0x02, 0x32, 0x00, 0x02, 0x00, 0x68, 0x00, 0x00, 0x02,
0x01, 0x00, 0x00, 0x00, 0x9b, 0x00, 0x00, 0x04, 0x10, 0x00, 0x00, 0x00,
0x10, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x26, 0x00, 0x00, 0x06,
0x00, 0xd0, 0x00, 0x00, 0x12, 0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00,
0x0a, 0x00, 0x02, 0x00, 0x1a, 0x00, 0x02, 0x00, 0xa4, 0x00, 0x00, 0x07,
0xf2, 0xe0, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x10, 0x00,
0x00, 0x00, 0x00, 0x00, 0x06, 0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00,
0x3e, 0x00, 0x00, 0x01, 0x53, 0x54, 0x41, 0x54, 0x94, 0x00, 0x00, 0x00,
0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00

View File

@@ -2,6 +2,7 @@ target("D3D11Example")
add_deps("pulse_gpu")
if is_plat("linux") then
set_extension(".x86_64")
add_defines("PULSE_D3D11_COMPILER_UNAVAILABLE")
end
add_files("*.c")
target_end()