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

Clean up conv example, Instances, profiler and test (#324)

Browse Source 
* convnd_fwd fp16 example

* update example

* update example

* update instance

* updating refernce conv

* update reference conv

* update conv fwd profiler

* update conv 1d and 3d instance

* update include path

* clean

* update profiler for conv bwd data and weight

* update conv bwd weight

* clean

* update conv example

* update profiler for conv bwd weight

* update ckprofiler for conv bwd data

* fix reference conv bwd data bug; update conv bwd data test

* update examples

* fix initialization issue

* update test for conv fwd

* clean

* clean

* remove test case too sensitive to error threshhold

* fix test

* clean

* fix build

* adding conv multiple d

* adding conv multiple D

* add matrix padder

* add gemm padding to convnd

* adding group conv

* update gemm multi-d

* refactor

* refactor

* refactor

* clean

* clean

* refactor

* refactor

* reorg

* add ds

* add bias

* clean

* add G

* adding group

* adding group

* adding group

* update Tensor

* clean

* update example

* update DeviceGemmMultipleD_Xdl_CShuffle

* update conv bwd-data and bwd-weight

* upate contraction example

* update gemm and batch gemm with e permute

* fix example build

* instance for grouped conv1d

* update example

* adding group conv instance

* update gemm bilinear instance

* update gemm+add+add+fastgelu instance

* update profiler

* update profiler

* update test

* update test and client example

* clean

* add grouped conv into profiler

* update profiler

* clean

* add test grouped conv, update all conv test to gtest

* update test

[ROCm/composable_kernel commit: 500fa99512]
This commit is contained in:
Chao Liu
2022-07-29 18:19:25 -05:00
committed by GitHub
parent 1450273dc5
commit 236f946292
373 changed files with 17544 additions and 17013 deletions
Download Patch File Download Diff File Expand all files Collapse all files

60
profiler/src/profile_gemm.cpp
View File

@@ -24,21 +24,27 @@ enum struct GemmDataType
INT8_INT8_INT8, // 3
};
static void print_helper_msg()
{
std::cout << "arg1: tensor operation (gemm: GEMM)\n"
<< "arg2: data type (0: fp32; 1: fp16; 2: bf16; 3: int8)\n"
<< "arg3: matrix layout (0: A[m, k] * B[k, n] = C[m, n];\n"
<< " 1: A[m, k] * B[n, k] = C[m, n];\n"
<< " 2: A[k, m] * B[k, n] = C[m, n];\n"
<< " 3: A[k, m] * B[n, k] = C[m, n])\n"
<< "arg4: verification (0: no; 1: yes)\n"
<< "arg5: initialization (0: no init; 1: integer value; 2: decimal value)\n"
<< "arg6: print tensor value (0: no; 1: yes)\n"
<< "arg7: time kernel (0: no, 1: yes)\n"
<< "arg8 to 13: M, N, K, StrideA, StrideB, StrideC\n"
<< std::endl;
}
int profile_gemm(int argc, char* argv[])
{
if(argc != 14)
{
printf("arg1: tensor operation (gemm: GEMM)\n");
printf("arg2: data type (0: fp32; 1: fp16; 2: bf16; 3: int8)\n");
printf("arg3: matrix layout (0: A[m, k] * B[k, n] = C[m, n];\n");
printf(" 1: A[m, k] * B[n, k] = C[m, n];\n");
printf(" 2: A[k, m] * B[k, n] = C[m, n];\n");
printf(" 3: A[k, m] * B[n, k] = C[m, n])\n");
printf("arg4: verification (0: no; 1: yes)\n");
printf("arg5: initialization (0: no init; 1: integer value; 2: decimal value)\n");
printf("arg6: print tensor value (0: no; 1: yes)\n");
printf("arg7: time kernel (0=no, 1=yes)\n");
printf("arg8 to 13: M, N, K, StrideA, StrideB, StrideC\n");
print_helper_msg();
exit(1);
}
@@ -109,67 +115,67 @@ int profile_gemm(int argc, char* argv[])
if(data_type == GemmDataType::F32_F32_F32 && layout == GemmMatrixLayout::MK_KN_MN)
{
return profile(F32{}, F32{}, F32{}, F32{}, Row{}, Row{}, Row{});
return profile(Row{}, Row{}, Row{}, F32{}, F32{}, F32{}, F32{});
}
else if(data_type == GemmDataType::F32_F32_F32 && layout == GemmMatrixLayout::MK_NK_MN)
{
return profile(F32{}, F32{}, F32{}, F32{}, Row{}, Col{}, Row{});
return profile(Row{}, Col{}, Row{}, F32{}, F32{}, F32{}, F32{});
}
else if(data_type == GemmDataType::F32_F32_F32 && layout == GemmMatrixLayout::KM_KN_MN)
{
return profile(F32{}, F32{}, F32{}, F32{}, Col{}, Row{}, Row{});
return profile(Col{}, Row{}, Row{}, F32{}, F32{}, F32{}, F32{});
}
else if(data_type == GemmDataType::F32_F32_F32 && layout == GemmMatrixLayout::KM_NK_MN)
{
return profile(F32{}, F32{}, F32{}, F32{}, Col{}, Col{}, Row{});
return profile(Col{}, Col{}, Row{}, F32{}, F32{}, F32{}, F32{});
}
else if(data_type == GemmDataType::F16_F16_F16 && layout == GemmMatrixLayout::MK_KN_MN)
{
return profile(F16{}, F16{}, F32{}, F16{}, Row{}, Row{}, Row{});
return profile(Row{}, Row{}, Row{}, F16{}, F16{}, F32{}, F16{});
}
else if(data_type == GemmDataType::F16_F16_F16 && layout == GemmMatrixLayout::MK_NK_MN)
{
return profile(F16{}, F16{}, F32{}, F16{}, Row{}, Col{}, Row{});
return profile(Row{}, Col{}, Row{}, F16{}, F16{}, F32{}, F16{});
}
else if(data_type == GemmDataType::F16_F16_F16 && layout == GemmMatrixLayout::KM_KN_MN)
{
return profile(F16{}, F16{}, F32{}, F16{}, Col{}, Row{}, Row{});
return profile(Col{}, Row{}, Row{}, F16{}, F16{}, F32{}, F16{});
}
else if(data_type == GemmDataType::F16_F16_F16 && layout == GemmMatrixLayout::KM_NK_MN)
{
return profile(F16{}, F16{}, F32{}, F16{}, Col{}, Col{}, Row{});
return profile(Col{}, Col{}, Row{}, F16{}, F16{}, F32{}, F16{});
}
else if(data_type == GemmDataType::BF16_BF16_BF16 && layout == GemmMatrixLayout::MK_KN_MN)
{
return profile(BF16{}, BF16{}, F32{}, BF16{}, Row{}, Row{}, Row{});
return profile(Row{}, Row{}, Row{}, BF16{}, BF16{}, F32{}, BF16{});
}
else if(data_type == GemmDataType::BF16_BF16_BF16 && layout == GemmMatrixLayout::MK_NK_MN)
{
return profile(BF16{}, BF16{}, F32{}, BF16{}, Row{}, Col{}, Row{});
return profile(Row{}, Col{}, Row{}, BF16{}, BF16{}, F32{}, BF16{});
}
else if(data_type == GemmDataType::BF16_BF16_BF16 && layout == GemmMatrixLayout::KM_KN_MN)
{
return profile(BF16{}, BF16{}, F32{}, BF16{}, Col{}, Row{}, Row{});
return profile(Col{}, Row{}, Row{}, BF16{}, BF16{}, F32{}, BF16{});
}
else if(data_type == GemmDataType::BF16_BF16_BF16 && layout == GemmMatrixLayout::KM_NK_MN)
{
return profile(BF16{}, BF16{}, F32{}, BF16{}, Col{}, Col{}, Row{});
return profile(Col{}, Col{}, Row{}, BF16{}, BF16{}, F32{}, BF16{});
}
else if(data_type == GemmDataType::INT8_INT8_INT8 && layout == GemmMatrixLayout::MK_KN_MN)
{
return profile(INT8{}, INT8{}, INT32{}, INT8{}, Row{}, Row{}, Row{});
return profile(Row{}, Row{}, Row{}, INT8{}, INT8{}, INT32{}, INT8{});
}
else if(data_type == GemmDataType::INT8_INT8_INT8 && layout == GemmMatrixLayout::MK_NK_MN)
{
return profile(INT8{}, INT8{}, INT32{}, INT8{}, Row{}, Col{}, Row{});
return profile(Row{}, Col{}, Row{}, INT8{}, INT8{}, INT32{}, INT8{});
}
else if(data_type == GemmDataType::INT8_INT8_INT8 && layout == GemmMatrixLayout::KM_KN_MN)
{
return profile(INT8{}, INT8{}, INT32{}, INT8{}, Col{}, Row{}, Row{});
return profile(Col{}, Row{}, Row{}, INT8{}, INT8{}, INT32{}, INT8{});
}
else if(data_type == GemmDataType::INT8_INT8_INT8 && layout == GemmMatrixLayout::KM_NK_MN)
{
return profile(INT8{}, INT8{}, INT32{}, INT8{}, Col{}, Col{}, Row{});
return profile(Col{}, Col{}, Row{}, INT8{}, INT8{}, INT32{}, INT8{});
}
else
{