ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-19 04:19:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

CK pk_i4_t test failures fix (SWDEV-518629) (#2075)

Browse Source 
* fix pk_i4_v3 tests failures in Unbuntu env.

* fix pk_i4_t tests failure on Unbuntu issues.

* some fixed.

---------

Co-authored-by: mtgu0705 <mtgu@amd.com>

[ROCm/composable_kernel commit: 56378f810f]
This commit is contained in:
Mingtao Gu
2025-04-14 16:58:57 +08:00
committed by GitHub
parent 1d62f5edb9
commit 3cbda9a11c
13 changed files with 148 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
example/01_gemm/gemm_xdl_bf16_pk_i4_v3.cpp
View File

@@ -133,7 +133,7 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
std::cout << "c_m_n: " << c_m_n_host_result.mDesc << std::endl;
DeviceMem a_m_k_device_buf(sizeof(ADataType) * a_m_k.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_permute.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_permute.mDesc.GetElementSpaceSize() / 2);
DeviceMem c_m_n_device_buf(sizeof(CDataType) * c_m_n_device_result.mDesc.GetElementSpaceSize());
// weight permute
@@ -192,14 +192,20 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
b_element_op,
c_element_op);
if(!gemm.IsSupportedArgument(argument) || ck::get_device_name() != "gfx942" ||
ck::get_device_name() != "gfx950")
if(!gemm.IsSupportedArgument(argument))
{
std::cerr << gemm.GetTypeString() << " does not support this problem" << std::endl;
return true;
}
if(!(ck::get_device_name() == "gfx942" || ck::get_device_name() == "gfx950"))
{
std::cout << "This kernel support gfx942 and gfx950 only" << std::endl;
return true;
}
bool pass = true;
if(config.do_verification)
{

12
example/01_gemm/gemm_xdl_fp16_pk_i4_v3.cpp
View File

@@ -134,7 +134,7 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
std::cout << "c_m_n: " << c_m_n_host_result.mDesc << std::endl;
DeviceMem a_m_k_device_buf(sizeof(ADataType) * a_m_k.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_permute.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_permute.mDesc.GetElementSpaceSize() / 2);
DeviceMem c_m_n_device_buf(sizeof(CDataType) * c_m_n_device_result.mDesc.GetElementSpaceSize());
// weight permute
@@ -242,14 +242,20 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
b_element_op,
c_element_op);
if(!gemm.IsSupportedArgument(argument) || ck::get_device_name() != "gfx942" ||
ck::get_device_name() != "gfx950")
if(!gemm.IsSupportedArgument(argument))
{
std::cerr << gemm.GetTypeString() << " does not support this problem" << std::endl;
return true;
}
if(!(ck::get_device_name() == "gfx942" || ck::get_device_name() == "gfx950"))
{
std::cout << "This kernel support gfx942 and gfx950 only" << std::endl;
return true;
}
bool pass = true;
if(config.do_verification)
{

12
example/01_gemm/gemm_xdl_fp16_pk_i4_v3_b_scale.cpp
View File

@@ -161,7 +161,7 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
std::cout << "c_m_n: " << c_m_n_host_result.mDesc << std::endl;
DeviceMem a_m_k_device_buf(sizeof(ADataType) * a_m_k.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_permute.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_permute.mDesc.GetElementSpaceSize() / 2);
DeviceMem b1_scale_device_buf(sizeof(BScaleDataType) * b1_k_n.mDesc.GetElementSpaceSize());
DeviceMem c_m_n_device_buf(sizeof(CDataType) * c_m_n_device_result.mDesc.GetElementSpaceSize());
@@ -274,14 +274,20 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
b_element_op,
c_element_op);
if(!gemm.IsSupportedArgument(argument) || ck::get_device_name() != "gfx942" ||
ck::get_device_name() != "gfx950")
if(!gemm.IsSupportedArgument(argument))
{
std::cerr << gemm.GetTypeString() << " does not support this problem" << std::endl;
return true;
}
if(!(ck::get_device_name() == "gfx942" || ck::get_device_name() == "gfx950"))
{
std::cout << "This kernel support gfx942 and gfx950 only" << std::endl;
return true;
}
bool pass = true;
if(config.do_verification)
{

13
example/01_gemm/gemm_xdl_fp8_pk_i4_bpreshuffle_v3.cpp
View File

@@ -152,7 +152,8 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
std::cout << "c_m_n: " << c_m_n_host_result.mDesc << std::endl;
DeviceMem a_m_k_device_buf(sizeof(ADataType) * a_m_k.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_preshuffled.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_preshuffled.mDesc.GetElementSpaceSize() /
2);
DeviceMem c_m_n_device_buf(sizeof(CDataType) * c_m_n_device_result.mDesc.GetElementSpaceSize());
// do GEMM
@@ -261,14 +262,20 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
b_element_op,
c_element_op);
if(!gemm.IsSupportedArgument(argument) || ck::get_device_name() != "gfx942" ||
ck::get_device_name() != "gfx950")
if(!gemm.IsSupportedArgument(argument))
{
std::cerr << gemm.GetTypeString() << " does not support this problem" << std::endl;
return true;
}
if(!(ck::get_device_name() == "gfx942" || ck::get_device_name() == "gfx950"))
{
std::cout << "This kernel support gfx942 and gfx950 only" << std::endl;
return true;
}
bool pass = true;
if(config.do_verification)
{

12
example/01_gemm/gemm_xdl_fp8_pk_i4_v3.cpp
View File

@@ -132,7 +132,7 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
std::cout << "c_m_n: " << c_m_n_host_result.mDesc << std::endl;
DeviceMem a_m_k_device_buf(sizeof(ADataType) * a_m_k.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_permute.mDesc.GetElementSpaceSize());
DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n_permute.mDesc.GetElementSpaceSize() / 2);
DeviceMem c_m_n_device_buf(sizeof(CDataType) * c_m_n_device_result.mDesc.GetElementSpaceSize());
// weight permute
@@ -240,14 +240,20 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
b_element_op,
c_element_op);
if(!gemm.IsSupportedArgument(argument) || ck::get_device_name() != "gfx942" ||
ck::get_device_name() != "gfx950")
if(!gemm.IsSupportedArgument(argument))
{
std::cerr << gemm.GetTypeString() << " does not support this problem" << std::endl;
return true;
}
if(!(ck::get_device_name() == "gfx942" || ck::get_device_name() == "gfx950"))
{
std::cout << "This kernel support gfx942 and gfx950 only" << std::endl;
return true;
}
bool pass = true;
if(config.do_verification)
{

3
example/24_batched_gemm/run_batched_gemm_example_fp16int4_b_scale.inc
View File

@@ -212,7 +212,8 @@ bool run_batched_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
std::cout << "c_g_m_n: " << c_g_m_n_host_result.mDesc << std::endl;
DeviceMem a_g_m_k_device_buf(sizeof(ADataType) * a_g_m_k.mDesc.GetElementSpaceSize());
DeviceMem b_g_k_n_device_buf(sizeof(BDataType) * b_g_k_n_permute.mDesc.GetElementSpaceSize());
DeviceMem b_g_k_n_device_buf(sizeof(BDataType) * b_g_k_n_permute.mDesc.GetElementSpaceSize() /
2);
DeviceMem b1_g_scale_device_buf(sizeof(BScaleDataType) * b1_g_k_n.mDesc.GetElementSpaceSize());
DeviceMem c_g_m_n_device_buf(sizeof(CDataType) *
c_g_m_n_device_result.mDesc.GetElementSpaceSize());

11
example/65_gemm_multiply_multiply/moe_gemm1_xdl_pk_i4.cpp
View File

@@ -301,7 +301,7 @@ int main(int argc, char* argv[])
DeviceMem expert_ids_dev(sizeof(ck::index_t) * expert_ids.mDesc.GetElementSpaceSize());
DeviceMem max_token_id_dev(sizeof(ck::index_t) * max_token_id.mDesc.GetElementSpaceSize());
DeviceMem a0_device_buf(sizeof(A0DataType) * a0_t_k.mDesc.GetElementSpaceSize());
DeviceMem b0_device_buf(sizeof(B0DataType) * b0_e_n_k.mDesc.GetElementSpaceSize());
DeviceMem b0_device_buf(sizeof(B0DataType) * b0_e_n_k.mDesc.GetElementSpaceSize() / 2);
DeviceMem d0_device_buf(sizeof(D0DataType) * d0_t_n.mDesc.GetElementSpaceSize());
DeviceMem d1_device_buf(sizeof(D1DataType) * d1_e_n.mDesc.GetElementSpaceSize());
DeviceMem e_device_buf(sizeof(EDataType) * e_t_n_device_result.mDesc.GetElementSpaceSize());
@@ -440,13 +440,18 @@ int main(int argc, char* argv[])
b_element_op,
cde_element_op);
if(!device_op.IsSupportedArgument(argument) ||
!(ck::get_device_name() == "gfx942" || ck::get_device_name() == "gfx950"))
if(!device_op.IsSupportedArgument(argument))
{
throw std::runtime_error(
"wrong! device_gemm with the specified compilation parameters does "
"not support this GEMM problem");
}
if(!(ck::get_device_name() == "gfx942" || ck::get_device_name() == "gfx950"))
{
std::cout << "This kernel support gfx942 and gfx950 only" << std::endl;
}
if(time_kernel)
{
float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});

11
example/65_gemm_multiply_multiply/moe_gemm2_xdl_pk_i4.cpp
View File

@@ -298,7 +298,7 @@ int main(int argc, char* argv[])
DeviceMem expert_ids_dev(sizeof(ck::index_t) * expert_ids.mDesc.GetElementSpaceSize());
DeviceMem max_token_id_dev(sizeof(ck::index_t) * max_token_id.mDesc.GetElementSpaceSize());
DeviceMem a0_device_buf(sizeof(A0DataType) * a0_t_k_k.mDesc.GetElementSpaceSize());
DeviceMem b0_device_buf(sizeof(B0DataType) * b0_e_n_k.mDesc.GetElementSpaceSize());
DeviceMem b0_device_buf(sizeof(B0DataType) * b0_e_n_k.mDesc.GetElementSpaceSize() / 2);
DeviceMem d0_device_buf(sizeof(D0DataType) * d0_t_n.mDesc.GetElementSpaceSize());
DeviceMem d1_device_buf(sizeof(D1DataType) * d1_e_n.mDesc.GetElementSpaceSize());
DeviceMem d2_device_buf(sizeof(D2DataType) * d2_e_n.mDesc.GetElementSpaceSize());
@@ -407,13 +407,18 @@ int main(int argc, char* argv[])
b_element_op,
cde_element_op);
if(!device_op.IsSupportedArgument(argument) ||
!(ck::get_device_name() == "gfx942" || ck::get_device_name() == "gfx950"))
if(!device_op.IsSupportedArgument(argument))
{
throw std::runtime_error(
"wrong! device_gemm with the specified compilation parameters does "
"not support this GEMM problem");
}
if(!(ck::get_device_name() == "gfx942" || ck::get_device_name() == "gfx950"))
{
std::cout << "This kernel support gfx942 and gfx950 only" << std::endl;
}
if(time_kernel)
{
// not result correct here because output buf not setzero