ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-11 09:40:51 +00:00
Code Issues Packages Projects Releases Wiki Activity

replace the ENV macro with CK_ENV (#1296)

Browse Source 
[ROCm/composable_kernel commit: 1274861a9d]
This commit is contained in:
Illia Silin
2024-05-17 10:42:51 -07:00
committed by GitHub
parent 506dc5e284
commit 2026ce49e7
30 changed files with 65 additions and 65 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
include/ck/host_utility/flush_cache.hpp
View File

@@ -117,7 +117,7 @@ float launch_and_time_kernel_with_preprocess(const StreamConfig& stream_config,
#define MEDIAN 1
if(stream_config.time_kernel_)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
printf("%s: grid_dim {%u, %u, %u}, block_dim {%u, %u, %u} \n",
__func__,
@@ -142,7 +142,7 @@ float launch_and_time_kernel_with_preprocess(const StreamConfig& stream_config,
{
return 0.0;
}
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
printf("Start running %d times...\n", nrepeat);
}
@@ -186,7 +186,7 @@ float launch_and_time_kernel_with_preprocess(const StreamConfig& stream_config,
total_time += cur_time;
#endif
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "i: " << i << " cur_time: " << cur_time << std::endl;

8
include/ck/host_utility/kernel_launch.hpp
View File

@@ -20,7 +20,7 @@ float launch_and_time_kernel(const StreamConfig& stream_config,
#if CK_TIME_KERNEL
if(stream_config.time_kernel_)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
printf("%s: grid_dim {%u, %u, %u}, block_dim {%u, %u, %u} \n",
__func__,
@@ -41,7 +41,7 @@ float launch_and_time_kernel(const StreamConfig& stream_config,
}
const int nrepeat = stream_config.nrepeat_;
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
printf("Start running %d times...\n", nrepeat);
}
@@ -95,7 +95,7 @@ float launch_and_time_kernel_with_preprocess(const StreamConfig& stream_config,
#if CK_TIME_KERNEL
if(stream_config.time_kernel_)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
printf("%s: grid_dim {%u, %u, %u}, block_dim {%u, %u, %u} \n",
__func__,
@@ -117,7 +117,7 @@ float launch_and_time_kernel_with_preprocess(const StreamConfig& stream_config,
}
const int nrepeat = stream_config.nrepeat_;
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
printf("Start running %d times...\n", nrepeat);
}

2
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_multiple_d_gemm_multiple_d_xdl_cshuffle.hpp
View File

@@ -587,7 +587,7 @@ struct DeviceBatchedGemmMultipleDGemmMultipleD_Xdl_CShuffle
BatchStrideD1s,
BatchStrideE1}
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "a0_grid_desc_m_k_{" << a0_grid_desc_m_k_.GetLength(I0) << ", "
<< a0_grid_desc_m_k_.GetLength(I1) << "}" << std::endl;

2
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_reduce_xdl_cshuffle.hpp
View File

@@ -658,7 +658,7 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<0, ReduceO
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
{
std::cout << "arg.Batch_ = " << arg.Batch_ << std::endl;

2
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle.hpp
View File

@@ -719,7 +719,7 @@ struct DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle
static bool IsSupportedArgument(const Argument& arg)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
arg.Print();
}

2
include/ck/tensor_operation/gpu/device/impl/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk.hpp
View File

@@ -516,7 +516,7 @@ struct DeviceConv2dBwdDataXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
float ave_time = 0;
for(size_t i = 0; i < arg.a_grid_desc_k0_m_k1_container_.size(); i++)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
{
std::cout << "arg.a_grid_desc_k0_m_k1_container_{"

2
include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_bias_activation_add_nhwc_kyxc_nhwk.hpp
View File

@@ -644,7 +644,7 @@ struct
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << DeviceOp{}.GetTypeString() << std::endl;
std::cout << "N " << arg.Conv_N_ << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_bias_activation_nhwc_kyxc_nhwk.hpp
View File

@@ -614,7 +614,7 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << DeviceOp{}.GetTypeString() << std::endl;
std::cout << "N " << arg.Conv_N_ << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp
View File

@@ -579,7 +579,7 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_W
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << DeviceOp{}.GetTypeString() << std::endl;
std::cout << "N " << arg.Conv_N_ << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_nhwc_kyxc_nhwk.hpp
View File

@@ -431,7 +431,7 @@ struct DeviceConv2dFwdXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_conv3d_fwd_xdl_ndhwc_kzyxc_ndhwk.hpp
View File

@@ -401,7 +401,7 @@ struct DeviceConv3dFwdXdl_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "num_batches_of_GEMM = " << arg.num_subbatches_ << std::endl;
std::cout << "a_grid_desc_k0_m_k1{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)

2
include/ck/tensor_operation/gpu/device/impl/device_convnd_bwd_data_nwc_kxc_nwk_dl.hpp
View File

@@ -1272,7 +1272,7 @@ struct DeviceConvNdBwdDataNwcKxcNwk_Dl
float ave_time = 0;
for(size_t i = 0; i < arg.a_grid_desc_k0_m_k1_container_.size(); i++)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "arg.a_grid_desc_k0_m_k1_container_{"
<< arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I0) << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_convnd_bwd_data_nwc_kxc_nwk_xdl.hpp
View File

@@ -1220,7 +1220,7 @@ struct DeviceConvNdBwdDataNwcKxcNwk_Xdl
float ave_time = 0;
for(size_t i = 0; i < arg.a_grid_desc_k0_m_k1_container_.size(); i++)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "arg.a_grid_desc_k0_m_k1{"
<< arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I0) << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp
View File

@@ -334,7 +334,7 @@ struct DeviceGemmDl : public DeviceGemm<ALayout,
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "arg.a_grid_desc_k0_m0_m1_k1_{"
<< arg.a_grid_desc_k0_m_k1_.GetLength(I0) << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_gemm_reduce_xdl_cshuffle.hpp
View File

@@ -510,7 +510,7 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<0, ReduceOperatio
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "arg.a_grid_desc_ak0_m_ak1_{"
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_layernorm_cshuffle.hpp
View File

@@ -514,7 +514,7 @@ struct DeviceGemmLayerNorm_Xdl_CShuffle : public BaseOperator
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "arg.a_grid_desc_ak0_m_ak1_{"
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_skip_b_lds.hpp
View File

@@ -299,7 +299,7 @@ struct DeviceGemmXdlSkipBLds : public DeviceGemm<ALayout,
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "

2
include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_dl.hpp
View File

@@ -553,7 +553,7 @@ struct DeviceGroupedGemmMultipleD_Dl : public DeviceGroupedGemm<ALayout,
for(std::size_t i = 0; i < arg.gemm_desc_kernel_arg_.size(); i++)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "group: " << i << " arg.a_grid_desc_k0_m_k1_{"
<< arg.gemm_desc_kernel_arg_[i].a_grid_desc_k0_m_k1_.GetLength(I0)

8
include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_two_stage.hpp
View File

@@ -468,7 +468,7 @@ struct DeviceGroupedGemmMultipleDSplitKXdlCShuffleTwoStage
gemm_kernel_args_[i].block_start_ = block_start;
gemm_kernel_args_[i].block_end_ = block_end;
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
index_t tiles = (block_end - block_start) / K_BATCH;
std::cout << "block_start: " << block_start << "\n"
@@ -495,7 +495,7 @@ struct DeviceGroupedGemmMultipleDSplitKXdlCShuffleTwoStage
arg.karg_.p_c_grid = p_workspace + offset;
index_t tiles = (arg.block_end_ - arg.block_start_) / arg.karg_.k_batch;
offset += tiles * MPerBlock * NPerBlock;
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "block_start: " << arg.block_start_ << "\n"
<< "block_end: " << arg.block_end_ << "\n"
@@ -819,7 +819,7 @@ struct DeviceGroupedGemmMultipleDSplitKXdlCShuffleTwoStage
if((ck::type_convert<ck::index_t>(arg.gemm_kernel_args_.size()) +
arg.skipped_group_count_) != arg.group_count_)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "The group count is not equal to sum of skipped groups "
"and kernel args size!"
@@ -836,7 +836,7 @@ struct DeviceGroupedGemmMultipleDSplitKXdlCShuffleTwoStage
bool group_arg_valid = GridwiseGemm::CheckValidity(gemm_arg);
if(not group_arg_valid)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "[" << __func__ << "] group id: " << i
<< " has invalid GridwiseGemm settings!" << std::endl;

2
include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_xdl_cshuffle_tile_loop.hpp
View File

@@ -620,7 +620,7 @@ struct DeviceGroupedGemmMultipleDXdlCShuffleTileLoop
GridwiseGemm::template CheckTensorTransfersValidity<ALayout, BLayout, ELayout>(
M, N, K)))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "The provided GEMM problem size (M,N,K) [" << M << "," << N << ","
<< K << "] are not supported by current template parameters!"

2
include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl.hpp
View File

@@ -514,7 +514,7 @@ struct DeviceGroupedGemm_Xdl : public DeviceGroupedGemm<ALayout,
for(std::size_t i = 0; i < arg.gemm_desc_kernel_arg_.size(); i++)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "group: " << i << " arg.a_grid_desc_ak0_m_ak1_{"
<< arg.gemm_desc_kernel_arg_[i].a_grid_desc_ak0_m_ak1_.GetLength(I0)

4
include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_splitk_cshuffle.hpp
View File

@@ -529,7 +529,7 @@ struct DeviceGroupedGemmXdlSplitKCShuffle : public DeviceGroupedGemmSplitK<ALayo
if((ck::type_convert<ck::index_t>(arg.gemm_kernel_args_.size()) +
arg.skipped_group_count_) != arg.group_count_)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "The group count is not equal to sum of skipped groups "
"and kernel args size!"
@@ -545,7 +545,7 @@ struct DeviceGroupedGemmXdlSplitKCShuffle : public DeviceGroupedGemmSplitK<ALayo
bool group_arg_valid = GridwiseGemm::CheckValidity(a);
if(not group_arg_valid)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "[" << __func__ << "] group id: " << i
<< " has invalid GridwiseGemm settings!" << std::endl;

20
include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3.hpp
View File

@@ -935,7 +935,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(!(karg.M % MPerBlock == 0))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg M value is not a multiple of MPerBlock! M: " << karg.M << " "
<< __FILE__ << ":" << __LINE__ << ", in function: " << __func__
@@ -952,7 +952,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(!(karg.N % NPerBlock == 0))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg N value is not a multiple of NPerBlock! N: " << karg.N << " "
<< __FILE__ << ":" << __LINE__ << ", in function: " << __func__
@@ -971,7 +971,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
auto K_t = karg.KBatch * KPerBlock;
if(!(karg.K % K_t == 0))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg K value is not a multiple of K_Batch * K0PerBlock * K1! K: "
<< karg.K << " " << __FILE__ << ":" << __LINE__
@@ -995,7 +995,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.K % ABlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg K (" << karg.K
<< ") value is not a multiple of ABlockTransferSrcScalarPerVector ("
@@ -1009,7 +1009,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.M % ABlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg M (" << karg.M
<< ") value is not a multiple of ABlockTransferSrcScalarPerVector ("
@@ -1024,7 +1024,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.N % BBlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg N (" << karg.N
<< ") value is not a multiple of BBlockTransferSrcScalarPerVector ("
@@ -1038,7 +1038,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.K % BBlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg K (" << karg.K
<< ") value is not a multiple of BBlockTransferSrcScalarPerVector ("
@@ -1053,7 +1053,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.N % CShuffleBlockTransferScalarPerVector_NPerBlock != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg N (" << karg.N
<< ") value is not a multiple of "
@@ -1069,7 +1069,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.M % CShuffleBlockTransferScalarPerVector_NPerBlock != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg M (" << karg.M
<< ") value is not a multiple of "
@@ -1084,7 +1084,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
if constexpr(is_same<remove_cvref_t<CDataType>, bhalf_t>::value)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << " KBatch: " << karg.KBatch << " > 1 is not support yet" << __FILE__
<< ":" << __LINE__ << ", in function: " << __func__ << std::endl;

18
include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_abd.hpp
View File

@@ -1113,7 +1113,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(!(karg.M % MPerBlock == 0))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg M value is not a multiple of MPerBlock! M: " << karg.M << " "
<< __FILE__ << ":" << __LINE__ << ", in function: " << __func__
@@ -1130,7 +1130,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(!(karg.N % NPerBlock == 0))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg N value is not a multiple of NPerBlock! N: " << karg.N << " "
<< __FILE__ << ":" << __LINE__ << ", in function: " << __func__
@@ -1149,7 +1149,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
auto K_t = karg.KBatch * KPerBlock;
if(!(karg.K % K_t == 0))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg K value is not a multiple of K_Batch * K0PerBlock * K1! K: "
<< karg.K << " " << __FILE__ << ":" << __LINE__
@@ -1173,7 +1173,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.K % ABlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg K (" << karg.K
<< ") value is not a multiple of ABlockTransferSrcScalarPerVector ("
@@ -1187,7 +1187,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.M % ABlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg M (" << karg.M
<< ") value is not a multiple of ABlockTransferSrcScalarPerVector ("
@@ -1202,7 +1202,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.N % BBlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg N (" << karg.N
<< ") value is not a multiple of BBlockTransferSrcScalarPerVector ("
@@ -1216,7 +1216,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.K % BBlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg K (" << karg.K
<< ") value is not a multiple of BBlockTransferSrcScalarPerVector ("
@@ -1231,7 +1231,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.N % CShuffleBlockTransferScalarPerVector_NPerBlock != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg N (" << karg.N
<< ") value is not a multiple of "
@@ -1247,7 +1247,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
{
if(karg.M % CShuffleBlockTransferScalarPerVector_NPerBlock != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg M (" << karg.M
<< ") value is not a multiple of "

20
include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r4r2.hpp
View File

@@ -446,7 +446,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
{
if(!(karg.M % MPerBlock == 0))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg M value is not a multiple of MPerBlock! M: " << karg.M << " "
<< __FILE__ << ":" << __LINE__ << ", in function: " << __func__
@@ -463,7 +463,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
{
if(!(karg.N % NPerBlock == 0))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg N value is not a multiple of NPerBlock! N: " << karg.N << " "
<< __FILE__ << ":" << __LINE__ << ", in function: " << __func__
@@ -482,7 +482,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
auto K_t = karg.k_batch * K0PerBlock * K1;
if(!(karg.K % K_t == 0))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg K value is not a multiple of K_Batch * K0PerBlock * K1! K: "
<< karg.K << " " << __FILE__ << ":" << __LINE__
@@ -496,7 +496,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
{
if(karg.K % ABlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg K (" << karg.K
<< ") value is not a multiple of ABlockTransferSrcScalarPerVector ("
@@ -510,7 +510,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
{
if(karg.M % ABlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg M (" << karg.M
<< ") value is not a multiple of ABlockTransferSrcScalarPerVector ("
@@ -525,7 +525,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
{
if(karg.N % BBlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg N (" << karg.N
<< ") value is not a multiple of BBlockTransferSrcScalarPerVector ("
@@ -539,7 +539,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
{
if(karg.K % BBlockTransferSrcScalarPerVector != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg K (" << karg.K
<< ") value is not a multiple of BBlockTransferSrcScalarPerVector ("
@@ -554,7 +554,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
{
if(karg.N % CBlockTransferScalarPerVector_NWaveNPerXDL != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg N (" << karg.N
<< ") value is not a multiple of "
@@ -569,7 +569,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
{
if(karg.M % CBlockTransferScalarPerVector_NWaveNPerXDL != 0)
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "Arg M (" << karg.M
<< ") value is not a multiple of "
@@ -584,7 +584,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
const auto num_k_loop = karg.K0Padded / K0PerBlock;
if(!GridwiseGemmPipe::IsSupported(num_k_loop))
{
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "The number of k loops (" << num_k_loop
<< ") value is not supported by GridwiseGemm Pipeline."

2
include/ck/utility/env.hpp
View File

@@ -124,7 +124,7 @@ struct EnvVar
#define CK_DECLARE_ENV_VAR_STR(name) CK_DECLARE_ENV_VAR(name, std::string, "")
#define ENV(name) \
#define CK_ENV(name) \
ck::env::name {}
template <class EnvVar>

2
profiler/include/profiler/profile_grouped_gemm_fixed_nk_impl.hpp
View File

@@ -88,7 +88,7 @@ bool profile_grouped_gemm_fixed_nk_impl(int do_verification,
c_m_n_host_results.push_back(
Tensor<CDataType>(f_host_tensor_descriptor(Ms[i], Ns[i], StrideCs[i], CLayout{})));
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "group: " << i << " a_m_k[" << i << "]:" << a_m_k[i].mDesc << ", b_k_n["
<< i << "]:" << b_k_n[i].mDesc << ", c_m_n_device_results[" << i

2
profiler/include/profiler/profile_grouped_gemm_impl.hpp
View File

@@ -87,7 +87,7 @@ bool profile_grouped_gemm_impl(int do_verification,
c_m_n_host_results.push_back(
Tensor<CDataType>(f_host_tensor_descriptor(Ms[i], Ns[i], StrideCs[i], CLayout{})));
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "group: " << i << " a_m_k[" << i << "]:" << a_m_k[i].mDesc << ", b_k_n["
<< i << "]:" << b_k_n[i].mDesc << ", c_m_n_device_results[" << i

2
profiler/include/profiler/profile_grouped_gemm_tile_loop_impl.hpp
View File

@@ -82,7 +82,7 @@ bool profile_grouped_gemm_tile_loop_impl(int do_verification,
Tensor<CDataType>(f_host_tensor_descriptor(Ms[i], Ns[i], StrideCs[i], CLayout{})));
c_m_n_host_results.push_back(
Tensor<CDataType>(f_host_tensor_descriptor(Ms[i], Ns[i], StrideCs[i], CLayout{})));
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "group: " << i << " a_m_k[" << i << "]:" << a_m_k[i].mDesc << ", b_k_n["
<< i << "]:" << b_k_n[i].mDesc << ", c_m_n_device_results[" << i

2
profiler/include/profiler/profile_grouped_gemm_two_stage_impl.hpp
View File

@@ -88,7 +88,7 @@ bool profile_grouped_gemm_two_stage_impl(int do_verification,
c_m_n_host_results.push_back(
Tensor<CDataType>(f_host_tensor_descriptor(Ms[i], Ns[i], StrideCs[i], CLayout{})));
if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
{
std::cout << "group: " << i << " a_m_k[" << i << "]:" << a_m_k[i].mDesc << ", b_k_n["
<< i << "]:" << b_k_n[i].mDesc << ", c_m_n_device_results[" << i