mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-19 22:39:03 +00:00
MIGraphX hipRTC fix (#1923)
* 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:
@@ -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))));
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user