ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-21 21:39:15 +00:00
Code Issues Packages Projects Releases Wiki Activity

Modularize ckProfiler operations (#514)

Browse Source 
* Re-structure ckProfiler source files

* Rename profiler.cpp to main.cpp

* Modularize ckProfiler operations

* Add description for profiler operations

* Use longer name to avoid name collision

* Use macro to delay expansion

* Use std::move() to avoid object copying

* Prohibit users from calling dtor

* Use macro to eliminate redundant code

* Make friend function hidden

* Add missing include directive <iostream>

* Fix wrong include directives

* Remove int8 from batchnorm-forward instances since it is not needed for forward training and could fail test

Co-authored-by: Qianfeng Zhang <Qianfeng.Zhang@amd.com>

[ROCm/composable_kernel commit: 8784a72e23]
This commit is contained in:
Po Yen Chen
2022-12-02 05:15:02 +08:00
committed by GitHub
parent 8e868bf880
commit 02db748e74
82 changed files with 346 additions and 273 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
profiler/src/profile_batched_gemm_gemm.cpp
View File

@@ -6,7 +6,8 @@
#include <initializer_list>
#include <cstdlib>
#include "profiler/include/profile_batched_gemm_gemm_impl.hpp"
#include "profiler/profile_batched_gemm_gemm_impl.hpp"
#include "profiler_operation_registry.hpp"
using F16 = ck::half_t;
using F32 = float;
@@ -14,6 +15,9 @@ using F32 = float;
using Row = ck::tensor_layout::gemm::RowMajor;
using Col = ck::tensor_layout::gemm::ColumnMajor;
#define OP_NAME "batched_gemm_gemm"
#define OP_DESC "Batched GEMM+GEMM"
int profile_batched_gemm_gemm(int argc, char* argv[])
{
enum struct GemmMatrixLayout
@@ -101,7 +105,7 @@ int profile_batched_gemm_gemm(int argc, char* argv[])
}
else
{
printf("arg1: tensor operation (batched_gemm_gemm: Batched_GEMM+Gemm)\n");
printf("arg1: tensor operation (" OP_NAME ": " OP_DESC ")\n");
printf("arg2: data type (1: fp16)\n");
printf("arg3: matrix layout (0: Relu(A0[m, k] * B0[n, k] + D0[m, n]) * B1[n, o] + D1[m, o] "
"= E1[m, o]; 1: Relu(A0[m, k] * B0[n, k] + D0[m, n]) * B1[o, n] + D1[m, o] = E1[m, "
@@ -179,3 +183,5 @@ int profile_batched_gemm_gemm(int argc, char* argv[])
return 0;
}
REGISTER_PROFILER_OPERATION(OP_NAME, OP_DESC, profile_batched_gemm_gemm);