ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-16 19:09:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add an option to change the number of warm-up cycles and iterations. (#1124)

Browse Source 
* allow setting the number of warmup cycles and iterations for profiler

* fix the gemm_splitk and grouped_gemm examples

[ROCm/composable_kernel commit: 886d9eeb99]
This commit is contained in:
Illia Silin
2024-01-09 09:43:08 -08:00
committed by GitHub
parent 70e816fd12
commit 264de3901a
9 changed files with 119 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
profiler/src/profile_gemm_splitk.cpp
View File

@@ -33,7 +33,7 @@ enum struct GemmDataType
int profile_gemm_splitk(int argc, char* argv[])
{
if(argc != 15)
if(argc != 15 && argc != 17)
{
printf("arg1: tensor operation (" OP_NAME ": " OP_DESC ")\n");
printf("arg2: data type (0: fp32; 1: fp16; 2: bf16; 3: int8; 4: f8@f16; 5: f16@f8; 6: f16, "
@@ -48,6 +48,9 @@ int profile_gemm_splitk(int argc, char* argv[])
printf("arg7: time kernel (0=no, 1=yes)\n");
printf("arg8 to 13: M, N, K, StrideA, StrideB, StrideC\n");
printf("arg14: split k into mulitiple batch\n");
printf("optional:\n");
printf("arg15: number of warm-up cycles (default 1)\n");
printf("arg16: number of iterations (default 10)\n");
exit(1);
}
@@ -67,6 +70,14 @@ int profile_gemm_splitk(int argc, char* argv[])
const int StrideC = std::stoi(argv[13]);
const int KBatch = std::stoi(argv[14]);
int n_warmup = 1;
int n_iter = 10;
if(argc == 17)
{
n_warmup = std::stoi(argv[15]);
n_iter = std::stoi(argv[16]);
}
using F32 = float;
using F16 = ck::half_t;
#if defined CK_ENABLE_FP8
@@ -117,7 +128,9 @@ int profile_gemm_splitk(int argc, char* argv[])
(StrideA < 0) ? DefaultStrideA : StrideA,
(StrideB < 0) ? DefaultStrideB : StrideB,
(StrideC < 0) ? DefaultStrideC : StrideC,
KBatch);
KBatch,
n_warmup,
n_iter);
return pass ? 0 : 1;
};