Skip to content

GitLab

Commit 0a929502 authored by Po-Yen, Chen's avatar Po-Yen, Chen
Browse files

Remove tailing underscore in public attribute name

parent f4ea00fc
Show whitespace changes
Inline Side-by-side
Showing with 69 additions and 69 deletions
include/ck/tensor_operation/gpu/device/impl/device_cgemm_4gemm_xdl_cshuffle.hpp
View file @ 0a929502
...@@ -172,12 +172,12 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -172,12 +172,12 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
{ {
using Parent = typename GridwiseGemm::Argument; using Parent = typename GridwiseGemm::Argument;
Argument(const ADataType* p_a_grid_real, Argument(const ADataType* p_a_grid_real_,
const ADataType* p_a_grid_imag, const ADataType* p_a_grid_imag_,
const BDataType* p_b_grid_real, const BDataType* p_b_grid_real_,
const BDataType* p_b_grid_imag, const BDataType* p_b_grid_imag_,
CDataType* p_c_grid_real, CDataType* p_c_grid_real_,
CDataType* p_c_grid_imag, CDataType* p_c_grid_imag_,
CDataType* p_workspace, CDataType* p_workspace,
index_t M_, index_t M_,
index_t N_, index_t N_,
...@@ -196,40 +196,40 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -196,40 +196,40 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
GridwiseGemm::CalculateKPadded(K_), GridwiseGemm::CalculateKPadded(K_),
GridwiseGemm::CalculateAK0(K_), GridwiseGemm::CalculateAK0(K_),
GridwiseGemm::CalculateBK0(K_)), GridwiseGemm::CalculateBK0(K_)),
p_a_grid_real_{p_a_grid_real}, p_a_grid_real{p_a_grid_real_},
p_a_grid_imag_{p_a_grid_imag}, p_a_grid_imag{p_a_grid_imag_},
p_b_grid_real_{p_b_grid_real}, p_b_grid_real{p_b_grid_real_},
p_b_grid_imag_{p_b_grid_imag}, p_b_grid_imag{p_b_grid_imag_},
p_c_grid_real_{p_c_grid_real}, p_c_grid_real{p_c_grid_real_},
p_c_grid_imag_{p_c_grid_imag}, p_c_grid_imag{p_c_grid_imag_},
p_aux_grid_{p_workspace} p_aux_grid{p_workspace}
{ {
const index_t grid_size = std::get<1>(GridwiseGemm::CalculateGridSize(M_, N_)); const index_t grid_size = std::get<1>(GridwiseGemm::CalculateGridSize(M_, N_));
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value) if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{ {
c_grid_desc_m_ = c_grid_desc_m =
DeviceOp::MakeDescriptor_M({M_, N_}, {StrideC_, I1}, grid_size, BlockSize); DeviceOp::MakeDescriptor_M({M_, N_}, {StrideC_, I1}, grid_size, BlockSize);
} }
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value) else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{ {
c_grid_desc_m_ = c_grid_desc_m =
DeviceOp::MakeDescriptor_M({M_, N_}, {I1, StrideC_}, grid_size, BlockSize); DeviceOp::MakeDescriptor_M({M_, N_}, {I1, StrideC_}, grid_size, BlockSize);
} }
p_aux_2_grid_ = p_workspace + GetCElementSpaceSize(M_, N_, StrideC_); p_aux_2_grid = p_workspace + GetCElementSpaceSize(M_, N_, StrideC_);
} }
// private: // private:
const ADataType* p_a_grid_real_; const ADataType* p_a_grid_real;
const ADataType* p_a_grid_imag_; const ADataType* p_a_grid_imag;
const BDataType* p_b_grid_real_; const BDataType* p_b_grid_real;
const BDataType* p_b_grid_imag_; const BDataType* p_b_grid_imag;
CDataType* p_c_grid_real_; CDataType* p_c_grid_real;
CDataType* p_c_grid_imag_; CDataType* p_c_grid_imag;
CDataType* p_aux_grid_; CDataType* p_aux_grid;
CDataType* p_aux_2_grid_; CDataType* p_aux_2_grid;
CGridDesc_M c_grid_desc_m_; CGridDesc_M c_grid_desc_m;
}; };
// Invoker // Invoker
...@@ -303,9 +303,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -303,9 +303,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
karg.p_a_grid_real_, karg.p_a_grid_real,
karg.p_b_grid_real_, karg.p_b_grid_real,
karg.p_aux_grid_, karg.p_aux_grid,
karg); karg);
ave_time += launch_and_time_kernel(stream_config, ave_time += launch_and_time_kernel(stream_config,
...@@ -313,9 +313,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -313,9 +313,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
karg.p_a_grid_imag_, karg.p_a_grid_imag,
karg.p_b_grid_imag_, karg.p_b_grid_imag,
karg.p_aux_2_grid_, karg.p_aux_2_grid,
karg); karg);
// c_real = aux - aux_2 // c_real = aux - aux_2
...@@ -325,11 +325,11 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -325,11 +325,11 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
make_tuple(karg.c_grid_desc_m_, karg.c_grid_desc_m_), make_tuple(karg.c_grid_desc_m, karg.c_grid_desc_m),
make_tuple(karg.c_grid_desc_m_), make_tuple(karg.c_grid_desc_m),
make_tuple(const_cast<const CDataType*>(karg.p_aux_grid_), make_tuple(const_cast<const CDataType*>(karg.p_aux_grid),
const_cast<const CDataType*>(karg.p_aux_2_grid_)), const_cast<const CDataType*>(karg.p_aux_2_grid)),
make_tuple(karg.p_c_grid_real_), make_tuple(karg.p_c_grid_real),
Subtract{}); Subtract{});
ave_time += launch_and_time_kernel(stream_config, ave_time += launch_and_time_kernel(stream_config,
...@@ -337,9 +337,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -337,9 +337,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
karg.p_a_grid_real_, karg.p_a_grid_real,
karg.p_b_grid_imag_, karg.p_b_grid_imag,
karg.p_aux_grid_, karg.p_aux_grid,
karg); karg);
ave_time += launch_and_time_kernel(stream_config, ave_time += launch_and_time_kernel(stream_config,
...@@ -347,9 +347,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -347,9 +347,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
karg.p_a_grid_imag_, karg.p_a_grid_imag,
karg.p_b_grid_real_, karg.p_b_grid_real,
karg.p_aux_2_grid_, karg.p_aux_2_grid,
karg); karg);
// c_imag = aux + aux_2 // c_imag = aux + aux_2
...@@ -359,11 +359,11 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -359,11 +359,11 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
make_tuple(karg.c_grid_desc_m_, karg.c_grid_desc_m_), make_tuple(karg.c_grid_desc_m, karg.c_grid_desc_m),
make_tuple(karg.c_grid_desc_m_), make_tuple(karg.c_grid_desc_m),
make_tuple(const_cast<const CDataType*>(karg.p_aux_grid_), make_tuple(const_cast<const CDataType*>(karg.p_aux_grid),
const_cast<const CDataType*>(karg.p_aux_2_grid_)), const_cast<const CDataType*>(karg.p_aux_2_grid)),
make_tuple(karg.p_c_grid_imag_), make_tuple(karg.p_c_grid_imag),
Add{}); Add{});
} }
else else
...@@ -375,9 +375,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -375,9 +375,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
karg.p_a_grid_real_, karg.p_a_grid_real,
karg.p_b_grid_real_, karg.p_b_grid_real,
karg.p_aux_grid_, karg.p_aux_grid,
karg); karg);
ave_time += launch_and_time_kernel(stream_config, ave_time += launch_and_time_kernel(stream_config,
...@@ -385,9 +385,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -385,9 +385,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
karg.p_a_grid_imag_, karg.p_a_grid_imag,
karg.p_b_grid_imag_, karg.p_b_grid_imag,
karg.p_aux_2_grid_, karg.p_aux_2_grid,
karg); karg);
// c_real = aux - aux_2 // c_real = aux - aux_2
...@@ -397,11 +397,11 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -397,11 +397,11 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
make_tuple(karg.c_grid_desc_m_, karg.c_grid_desc_m_), make_tuple(karg.c_grid_desc_m, karg.c_grid_desc_m),
make_tuple(karg.c_grid_desc_m_), make_tuple(karg.c_grid_desc_m),
make_tuple(const_cast<const CDataType*>(karg.p_aux_grid_), make_tuple(const_cast<const CDataType*>(karg.p_aux_grid),
const_cast<const CDataType*>(karg.p_aux_2_grid_)), const_cast<const CDataType*>(karg.p_aux_2_grid)),
make_tuple(karg.p_c_grid_real_), make_tuple(karg.p_c_grid_real),
Subtract{}); Subtract{});
ave_time += launch_and_time_kernel(stream_config, ave_time += launch_and_time_kernel(stream_config,
...@@ -409,9 +409,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -409,9 +409,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
karg.p_a_grid_real_, karg.p_a_grid_real,
karg.p_b_grid_imag_, karg.p_b_grid_imag,
karg.p_aux_grid_, karg.p_aux_grid,
karg); karg);
ave_time += launch_and_time_kernel(stream_config, ave_time += launch_and_time_kernel(stream_config,
...@@ -419,9 +419,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -419,9 +419,9 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
karg.p_a_grid_imag_, karg.p_a_grid_imag,
karg.p_b_grid_real_, karg.p_b_grid_real,
karg.p_aux_2_grid_, karg.p_aux_2_grid,
karg); karg);
// c_imag = aux + aux_2 // c_imag = aux + aux_2
...@@ -431,11 +431,11 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle ...@@ -431,11 +431,11 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
dim3(gdx, gdy, gdz), dim3(gdx, gdy, gdz),
dim3(BlockSize), dim3(BlockSize),
0, 0,
make_tuple(karg.c_grid_desc_m_, karg.c_grid_desc_m_), make_tuple(karg.c_grid_desc_m, karg.c_grid_desc_m),
make_tuple(karg.c_grid_desc_m_), make_tuple(karg.c_grid_desc_m),
make_tuple(const_cast<const CDataType*>(karg.p_aux_grid_), make_tuple(const_cast<const CDataType*>(karg.p_aux_grid),
const_cast<const CDataType*>(karg.p_aux_2_grid_)), const_cast<const CDataType*>(karg.p_aux_2_grid)),
make_tuple(karg.p_c_grid_imag_), make_tuple(karg.p_c_grid_imag),
Add{}); Add{});
} }
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment