ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 21:21:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

replace hipMemcpy with hipMemcpyWithStream (#734)

Browse Source
This commit is contained in:
who who who
2023-06-02 05:23:41 +08:00
committed by GitHub
parent 9eae73df9b
commit e2ebc8e795
5 changed files with 26 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
include/ck/tensor_operation/gpu/device/device_grouped_gemm_softmax_gemm_permute_xdl_cshuffle.hpp
View File

@@ -611,10 +611,11 @@ struct DeviceGroupedGemmSoftmaxGemmPermute_Xdl_CShuffle
some_has_main_k_block_loop |= y;
}
hipGetErrorString(hipMemcpy(arg.p_workspace_,
arg.group_kernel_args_.data(),
arg.group_kernel_args_.size() * sizeof(GroupKernelArg),
hipMemcpyHostToDevice));
hipGetErrorString(hipMemcpyWithStream(arg.p_workspace_,
arg.group_kernel_args_.data(),
arg.group_kernel_args_.size() * sizeof(GroupKernelArg),
hipMemcpyHostToDevice,
stream_config.stream_id_));
float ave_time = 0;

11
include/ck/tensor_operation/gpu/device/impl/device_grouped_contraction_multiple_d_xdl_cshuffle.hpp
View File

@@ -652,11 +652,12 @@ struct DeviceGroupedContractionMultipleD_Xdl_CShuffle
}
}
hipGetErrorString(hipMemcpy(arg.p_workspace_,
arg.contraction_multi_d_kernel_args_.data(),
arg.contraction_multi_d_kernel_args_.size() *
sizeof(ContractionMultiDKernelArg),
hipMemcpyHostToDevice));
hipGetErrorString(hipMemcpyWithStream(arg.p_workspace_,
arg.contraction_multi_d_kernel_args_.data(),
arg.contraction_multi_d_kernel_args_.size() *
sizeof(ContractionMultiDKernelArg),
hipMemcpyHostToDevice,
stream_config.stream_id_));
float ave_time = 0;

9
include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_dl.hpp
View File

@@ -597,10 +597,11 @@ struct DeviceGroupedGemmMultipleD_Dl : public DeviceGroupedGemm<ALayout,
}
}
hipGetErrorString(hipMemcpy(arg.p_workspace_,
arg.gemm_desc_kernel_arg_.data(),
arg.gemm_desc_kernel_arg_.size() * sizeof(GemmKernelArg),
hipMemcpyHostToDevice));
hipGetErrorString(hipMemcpyWithStream(arg.p_workspace_,
arg.gemm_desc_kernel_arg_.data(),
arg.gemm_desc_kernel_arg_.size() * sizeof(GemmKernelArg),
hipMemcpyHostToDevice,
stream_config.stream_id_));
auto launch_kernel = [&](auto has_main_k_block_loop,
auto has_double_tail_k_block_loop) {

9
include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl.hpp
View File

@@ -549,10 +549,11 @@ struct DeviceGroupedGemm_Xdl : public DeviceGroupedGemm<ALayout,
}
hipGetErrorString(
hipMemcpy(arg.p_workspace_,
arg.gemm_desc_kernel_arg_.data(),
arg.gemm_desc_kernel_arg_.size() * sizeof(GemmBiasTransKernelArg),
hipMemcpyHostToDevice));
hipMemcpyWithStream(arg.p_workspace_,
arg.gemm_desc_kernel_arg_.data(),
arg.gemm_desc_kernel_arg_.size() * sizeof(GemmBiasTransKernelArg),
hipMemcpyHostToDevice,
stream_config.stream_id_));
float ave_time = 0;

9
include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_splitk_cshuffle.hpp
View File

@@ -406,10 +406,11 @@ struct DeviceGroupedGemmXdlSplitKCShuffle : public DeviceGroupedGemmSplitK<ALayo
}
}
hip_check_error(hipMemcpy(arg.p_workspace_,
arg.gemm_kernel_args_.data(),
arg.gemm_kernel_args_.size() * sizeof(GemmTransKernelArg),
hipMemcpyHostToDevice));
hip_check_error(hipMemcpyWithStream(arg.p_workspace_,
arg.gemm_kernel_args_.data(),
arg.gemm_kernel_args_.size() * sizeof(GemmTransKernelArg),
hipMemcpyHostToDevice,
stream_config.stream_id_));
float ave_time = 0;