Skip to content

GitLab

Commit cca0ceee authored by ltqin's avatar ltqin
Browse files

add tunning parameters

parent 1b9e6e11
Show whitespace changes
Inline Side-by-side
Showing with 22 additions and 22 deletions
device_operation/device_gemm_xdl_splitk_instance_f32_f32_f32_km_kn_mn.cpp
View file @ cca0ceee
...@@ -22,7 +22,7 @@ using S = ck::Sequence<Is...>; ...@@ -22,7 +22,7 @@ using S = ck::Sequence<Is...>;
using PassThrough = ck::tensor_operation::element_wise::PassThrough; using PassThrough = ck::tensor_operation::element_wise::PassThrough;
// Compilation parameters for a[k, m] * b[k, n] = c[m, n] // Compilation parameters for a[k, m] * b[k, n] = c[m, n]
/*using device_gemm_xdl_instance_f32_f32_f32_km_kn_mn = std::tuple< using device_gemm_xdl_instance_f32_f32_f32_km_kn_mn = std::tuple<
// clang-format off // clang-format off
//##########| AData| BData| CData| AccData| ALayout| BLayout| CLayout| A| B| C| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| CThreadTransfer| CThreadTransfer| ABlockLds| BBlockLds| //##########| AData| BData| CData| AccData| ALayout| BLayout| CLayout| A| B| C| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| CThreadTransfer| CThreadTransfer| ABlockLds| BBlockLds|
//##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| SrcDstVectorDim| DstScalar| AddExtraM| AddExtraN| //##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| SrcDstVectorDim| DstScalar| AddExtraM| AddExtraN|
...@@ -38,12 +38,12 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough; ...@@ -38,12 +38,12 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
DeviceGemmSplitKXdl< F32, F32, F32, F32, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 4, 32, 32, 1, 2, S<1, 1, 1, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 1, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720> DeviceGemmSplitKXdl< F32, F32, F32, F32, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 4, 32, 32, 1, 2, S<1, 1, 1, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 1, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>
// clang-format on // clang-format on
>; >;
*/
template <> template <>
void add_device_splitk_gemm_instance<F32, F32, F32, Col, Row, Row>( void add_device_splitk_gemm_instance<F32, F32, F32, Col, Row, Row>(
std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& ) std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& device_op_instances)
{ {
/* using DeviceGemms = device_gemm_instance::device_gemm_xdl_instance_f32_f32_f32_km_kn_mn; using DeviceGemms = device_gemm_instance::device_gemm_xdl_instance_f32_f32_f32_km_kn_mn;
const auto device_gemms = DeviceGemms{}; const auto device_gemms = DeviceGemms{};
...@@ -53,7 +53,7 @@ void add_device_splitk_gemm_instance<F32, F32, F32, Col, Row, Row>( ...@@ -53,7 +53,7 @@ void add_device_splitk_gemm_instance<F32, F32, F32, Col, Row, Row>(
auto gemm = Gemm{}; auto gemm = Gemm{};
device_op_instances.push_back(std::make_unique<Gemm>(gemm)); device_op_instances.push_back(std::make_unique<Gemm>(gemm));
});*/ });
} }
} // namespace device_gemm_instance } // namespace device_gemm_instance
......
device_operation/device_gemm_xdl_splitk_instance_f32_f32_f32_km_nk_mn.cpp
View file @ cca0ceee
...@@ -22,7 +22,7 @@ using S = ck::Sequence<Is...>; ...@@ -22,7 +22,7 @@ using S = ck::Sequence<Is...>;
using PassThrough = ck::tensor_operation::element_wise::PassThrough; using PassThrough = ck::tensor_operation::element_wise::PassThrough;
// Compilation parameters for a[k, m] * b[n, k] = c[m, n] // Compilation parameters for a[k, m] * b[n, k] = c[m, n]
/*using device_gemm_xdl_instance_f32_f32_f32_km_nk_mn = std::tuple< using device_gemm_xdl_instance_f32_f32_f32_km_nk_mn = std::tuple<
// clang-format off // clang-format off
//##########| AData| BData| CData| AccData| ALayout| BLayout| CLayout| A| B| C| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| CThreadTransfer| CThreadTransfer| ABlockLds| BBlockLds| //##########| AData| BData| CData| AccData| ALayout| BLayout| CLayout| A| B| C| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| CThreadTransfer| CThreadTransfer| ABlockLds| BBlockLds|
//##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| SrcDstVectorDim| DstScalar| AddExtraM| AddExtraN| //##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| SrcDstVectorDim| DstScalar| AddExtraM| AddExtraN|
...@@ -38,12 +38,12 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough; ...@@ -38,12 +38,12 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
DeviceGemmSplitKXdl< F32, F32, F32, F32, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 4, 32, 32, 1, 2, S<1, 1, 1, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2> , S<0, 1, 3, 2> , 2, 1, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, 7, 1, true, true, 720> DeviceGemmSplitKXdl< F32, F32, F32, F32, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 4, 32, 32, 1, 2, S<1, 1, 1, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2> , S<0, 1, 3, 2> , 2, 1, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, 7, 1, true, true, 720>
// clang-format on // clang-format on
>; >;
*/
template <> template <>
void add_device_splitk_gemm_instance<F32, F32, F32, Col, Col, Row>( void add_device_splitk_gemm_instance<F32, F32, F32, Col, Col, Row>(
std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& ) std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& device_op_instances)
{ {
/* using DeviceGemms = device_gemm_instance::device_gemm_xdl_instance_f32_f32_f32_km_nk_mn; using DeviceGemms = device_gemm_instance::device_gemm_xdl_instance_f32_f32_f32_km_nk_mn;
const auto device_gemms = DeviceGemms{}; const auto device_gemms = DeviceGemms{};
...@@ -53,7 +53,7 @@ void add_device_splitk_gemm_instance<F32, F32, F32, Col, Col, Row>( ...@@ -53,7 +53,7 @@ void add_device_splitk_gemm_instance<F32, F32, F32, Col, Col, Row>(
auto gemm = Gemm{}; auto gemm = Gemm{};
device_op_instances.push_back(std::make_unique<Gemm>(gemm)); device_op_instances.push_back(std::make_unique<Gemm>(gemm));
});*/ });
} }
} // namespace device_gemm_instance } // namespace device_gemm_instance
......
device_operation/device_gemm_xdl_splitk_instance_f32_f32_f32_mk_kn_mn.cpp
View file @ cca0ceee
...@@ -28,15 +28,15 @@ using device_gemm_xdl_instance_f32_f32_f32_mk_kn_mn = std::tuple< ...@@ -28,15 +28,15 @@ using device_gemm_xdl_instance_f32_f32_f32_mk_kn_mn = std::tuple<
//##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| SrcDstVectorDim| DstScalar| AddExtraM| AddExtraN| //##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| SrcDstVectorDim| DstScalar| AddExtraM| AddExtraN|
//##########| | | | | | | | Operation| Operation| Operation| | | | | | | | Wave| Wave| Lengths_K0_N_K1| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| Lengths_K0_N_K1| Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerVector| | | //##########| | | | | | | | Operation| Operation| Operation| | | | | | | | Wave| Wave| Lengths_K0_N_K1| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| Lengths_K0_N_K1| Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerVector| | |
//##########| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | //##########| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 96, 128, 4, 8, 16, 16, 3, 4, S<1, 1, 3, 4>, S<1, 4, 32, 2>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 8>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720> DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 96, 128, 4, 8, 16, 16, 3, 4, S<1, 1, 3, 4>, S<1, 4, 32, 2>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 8>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>,
/* DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 4, 32, 32, 4, 2, S<1, 1, 4, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>, DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 4, 32, 32, 4, 2, S<1, 1, 4, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>,
DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 4, 32, 32, 2, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 4, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 4, 4, 7, 1, true, true, 720>, DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 4, 32, 32, 2, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 4, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 4, 4, 7, 1, true, true, 720>,
DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 4, 32, 32, 4, 2, S<1, 1, 4, 4>, S<1, 4, 32, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 4, 4>, S<1, 4, 32, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 4, 4, 7, 1, true, true, 720>, DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 4, 32, 32, 4, 2, S<1, 1, 4, 4>, S<1, 4, 32, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 4, 4>, S<1, 4, 32, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 4, 4, 7, 1, true, true, 720>,
DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 4, 32, 32, 2, 2, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>, DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 4, 32, 32, 2, 2, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>,
DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 4, 32, 32, 2, 2, S<1, 1, 4, 4>, S<1, 4, 32, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 4>, S<1, 4, 32, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>, DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 4, 32, 32, 2, 2, S<1, 1, 4, 4>, S<1, 4, 32, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 4>, S<1, 4, 32, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>,
DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 4, 32, 32, 2, 2, S<1, 1, 2, 4>, S<1, 4, 32, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 4, 4>, S<1, 4, 32, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 4, 4, 7, 1, true, true, 720>, DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 4, 32, 32, 2, 2, S<1, 1, 2, 4>, S<1, 4, 32, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 4, 4>, S<1, 4, 32, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 4, 4, 7, 1, true, true, 720>,
DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 4, 32, 32, 2, 1, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 1, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 1, 4, 7, 1, true, true, 720>, DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 4, 32, 32, 2, 1, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 1, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 1, 4, 7, 1, true, true, 720>,
DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 4, 32, 32, 1, 2, S<1, 1, 1, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>*/ DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 4, 32, 32, 1, 2, S<1, 1, 1, 4>, S<1, 4, 64, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 2, 4>, S<1, 4, 64, 1>, S<0, 1, 3, 2>, S<0, 1, 3, 2>, 2, 2, 4, 7, 1, true, true, 720>
>; >;
template <> template <>
......
device_operation/device_gemm_xdl_splitk_instance_f32_f32_f32_mk_nk_mn.cpp
View file @ cca0ceee
...@@ -22,7 +22,7 @@ using S = ck::Sequence<Is...>; ...@@ -22,7 +22,7 @@ using S = ck::Sequence<Is...>;
using PassThrough = ck::tensor_operation::element_wise::PassThrough; using PassThrough = ck::tensor_operation::element_wise::PassThrough;
// Compilation parameters for a[m, k] * b[n, k] = c[m, n] // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
/*using device_gemm_xdl_instance_f32_f32_f32_mk_nk_mn = std::tuple< using device_gemm_xdl_instance_f32_f32_f32_mk_nk_mn = std::tuple<
// clang-format off // clang-format off
//##########| AData| BData| CData| AccData| ALayout| BLayout| CLayout| A| B| C| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| CThreadTransfer| CThreadTransfer| ABlockLds| BBlockLds| //##########| AData| BData| CData| AccData| ALayout| BLayout| CLayout| A| B| C| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| CThreadTransfer| CThreadTransfer| ABlockLds| BBlockLds|
//##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| SrcDstVectorDim| DstScalar| AddExtraM| AddExtraN| //##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ThreadSlice| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| SrcDstVectorDim| DstScalar| AddExtraM| AddExtraN|
...@@ -43,12 +43,12 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough; ...@@ -43,12 +43,12 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, 64, 32, 64, 4, 4, 32, 32, 1, 2, S<1, 1, 2, 4>, S<1, 4, 16, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 4, 4>, S<1, 4, 16, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, 7, 1, true, true, 720> DeviceGemmSplitKXdl< F32, F32, F32, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, 64, 32, 64, 4, 4, 32, 32, 1, 2, S<1, 1, 2, 4>, S<1, 4, 16, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, S<1, 1, 4, 4>, S<1, 4, 16, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 4, 4, 7, 1, true, true, 720>
// clang-format on // clang-format on
>; >;
*/
template <> template <>
void add_device_splitk_gemm_instance<F32, F32, F32, Row, Col, Row>( void add_device_splitk_gemm_instance<F32, F32, F32, Row, Col, Row>(
std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& ) std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& device_op_instances)
{ {
/* using DeviceGemms = device_gemm_instance::device_gemm_xdl_instance_f32_f32_f32_mk_nk_mn; using DeviceGemms = device_gemm_instance::device_gemm_xdl_instance_f32_f32_f32_mk_nk_mn;
const auto device_gemms = DeviceGemms{}; const auto device_gemms = DeviceGemms{};
...@@ -58,7 +58,7 @@ void add_device_splitk_gemm_instance<F32, F32, F32, Row, Col, Row>( ...@@ -58,7 +58,7 @@ void add_device_splitk_gemm_instance<F32, F32, F32, Row, Col, Row>(
auto gemm = Gemm{}; auto gemm = Gemm{};
device_op_instances.push_back(std::make_unique<Gemm>(gemm)); device_op_instances.push_back(std::make_unique<Gemm>(gemm));
});*/ });
} }
} // namespace device_gemm_instance } // namespace device_gemm_instance
......
device_operation/include/device_gemm_instance.hpp
View file @ cca0ceee
...@@ -9,6 +9,10 @@ namespace tensor_operation { ...@@ -9,6 +9,10 @@ namespace tensor_operation {
namespace device { namespace device {
namespace device_gemm_instance { namespace device_gemm_instance {
using DeviceGemmNoOpPtr = DeviceGemmPtr<ck::tensor_operation::element_wise::PassThrough,
ck::tensor_operation::element_wise::PassThrough,
ck::tensor_operation::element_wise::PassThrough>;
template <typename ADataType, template <typename ADataType,
typename BDataType, typename BDataType,
typename CDataType, typename CDataType,
......
device_operation/include/device_gemm_xdl_instance.hpp
View file @ cca0ceee
...@@ -5,10 +5,6 @@ namespace tensor_operation { ...@@ -5,10 +5,6 @@ namespace tensor_operation {
namespace device { namespace device {
namespace device_gemm_instance { namespace device_gemm_instance {
using DeviceGemmNoOpPtr = DeviceGemmPtr<ck::tensor_operation::element_wise::PassThrough,
ck::tensor_operation::element_wise::PassThrough,
ck::tensor_operation::element_wise::PassThrough>;
template <> template <>
void add_device_splitk_gemm_instance<float, void add_device_splitk_gemm_instance<float,
float, float,
......
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