ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-19 22:39:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

MIGraphX hipRTC fix (#1923)

Browse Source 
* fixed hiprtc compilation issues from new additions, removed clashing mixed precision functionality from codegen(ignore the whole file)

* fixed device op error: misplaced header guard

* restrict virtual function use in device_gemm_multiple_d file for codegen hiprtc compilation

* add CK_CODE_GEN_RTC flag for compilation, since this flag has wider coverage for hiprtc compilation

* fixed conditional error in amd_ck_fp8.hpp

* Add MaskOutUpperTriangle as a problem parameter to
BatchedGemmSoftmaxGemm and disable tests with
MaskOutUpperTriangle==True.

Signed-off-by: Mirza Halilcevic <mirza.halilcevic@amd.com>

---------

Signed-off-by: Mirza Halilcevic <mirza.halilcevic@amd.com>
Co-authored-by: Mirza Halilcevic <mirza.halilcevic@amd.com>
This commit is contained in:
arai713
2025-03-03 07:55:05 -08:00
committed by GitHub
parent ef16010273
commit fd06ed926c
12 changed files with 90 additions and 48 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
codegen/test/batched_gemm_softmax_gemm.cpp
View File

@@ -42,7 +42,7 @@ TEST_CASE(test_problem_kernel)
prob.K = 1024;
prob.O = 1024;
prob.TransB = true;
check_all<half> check1, check2;
check_all<half> check;
auto a = to_gpu(generate_buffer<half>(1024 * 1024, 0));
auto b = to_gpu(generate_buffer<half>(1024 * 1024, 1));
auto b1 = to_gpu(generate_buffer<half>(1024 * 1024, 2));
@@ -77,10 +77,8 @@ TEST_CASE(test_problem_kernel)
k.launch(nullptr, grid_size * block_size, block_size)(
a.data(), b.data(), b1.data(), c.data());
if(solution.GetTemplateParameter<bool>("MaskOutUpperTriangle"))
CHECK(report(solution, check1(rtc::from_gpu(c))));
else
CHECK(report(solution, check2(rtc::from_gpu(c))));
// NOTE: Solutions where MaskOutUpperTriangle is True don't produce consistent results
CHECK(report(solution, check(rtc::from_gpu(c))));
}
}

1
codegen/test/rtc/src/compile_kernel.cpp
View File

@@ -279,6 +279,7 @@ static kernel hiprtc_compile_kernel(const std::vector<src_file>& srcs, compile_o
{
options.flags += " -I. -O3";
options.flags += " -std=c++17";
options.flags += " -DCK_CODE_GEN_RTC";
options.flags += " --offload-arch=" + get_device_name();
auto cos = compile_hip_src_with_hiprtc(srcs, options);
if(cos.size() != 1)