ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-13 09:45:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add splitk gemm fp16 @ fp16 with fp8 compute instances (#983)

Browse Source 
* Add ComputeType

* Update for compatibility

* Add instances

* Update profiler api
This commit is contained in:
Rostyslav Geyyer
2023-10-13 16:27:11 -05:00
committed by GitHub
parent 2ce9b56c64
commit fa753f27ba
10 changed files with 398 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
profiler/include/profiler/profile_gemm_splitk_impl.hpp
View File

@@ -30,7 +30,8 @@ template <typename ADataType,
typename CDataType,
typename ALayout,
typename BLayout,
typename CLayout>
typename CLayout,
typename ComputeType = CDataType>
bool profile_gemm_splitk_impl(int do_verification,
int init_method,
bool do_log,
@@ -103,7 +104,8 @@ bool profile_gemm_splitk_impl(int do_verification,
CDataType,
AElementOp,
BElementOp,
CElementOp>;
CElementOp,
ComputeType>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
@@ -120,7 +122,8 @@ bool profile_gemm_splitk_impl(int do_verification,
AccDataType,
AElementOp,
BElementOp,
CElementOp>;
CElementOp,
ComputeType>;
auto ref_gemm = ReferenceGemmInstance{};
auto ref_invoker = ref_gemm.MakeInvoker();