mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-05 22:22:27 +00:00
2b684f0a7d
* [CK][Examples] Extending support for rdna3/4 in following examples: -example_gemm_xdl_splitk_reduce_multi_d_fp16 -example_gemm_xdl_splitk_reduce_multi_d_bf16 -example_gemm_xdl_splitk_reduce_bf16A_i8B -example_gemm_xdl_splitk_reduce_bfp16 -example_splitk_gemm_bias_e_permute_xdl_fp32 -example_gemm_add_multiply_xdl_fp16 -example_complex_contraction_bilinear_xdl_fp32 -example_grouped_gemm_lower_triangle_scale_softmax_gemm_permute_xdl_fp16 -example_batched_gemm_bias_e_permute_xdl_fp16 -example_gemm_xdl_fp16 -example_gemm_xdl_fp16_av2 -example_gemm_xdl_wavelet_fp16 -example_gemm_add_add_fastgelu_xdl_bf16 -example_gemm_add_add_fastgelu_xdl_fp16 -example_gemm_add_add_fastgelu_xdl_fp32 -example_grouped_gemm_xdl_fp32 -example_grouped_gemm_xdl_fp16 -example_grouped_gemm_xdl_bf16 -example_cgemm_xdl_bf16 -example_cgemm_xdl_fp16 Signed-off-by: Michal Kulikowski <Michal.Kulikowski@amd.com> * [CK][Examples] Extending support for rdna3/4 in following examples: -example_gemm_xdl_splitk_reduce_multi_d_fp16 -example_gemm_xdl_splitk_reduce_multi_d_bf16 -example_gemm_xdl_splitk_reduce_bf16A_i8B -example_gemm_xdl_splitk_reduce_bfp16 -example_splitk_gemm_bias_e_permute_xdl_fp32 -example_gemm_add_multiply_xdl_fp16 -example_complex_contraction_bilinear_xdl_fp32 -example_grouped_gemm_lower_triangle_scale_softmax_gemm_permute_xdl_fp16 -example_batched_gemm_bias_e_permute_xdl_fp16 -example_gemm_xdl_fp16 -example_gemm_xdl_fp16_av2 -example_gemm_xdl_wavelet_fp16 -example_gemm_add_add_fastgelu_xdl_bf16 -example_gemm_add_add_fastgelu_xdl_fp16 -example_gemm_add_add_fastgelu_xdl_fp32 -example_grouped_gemm_xdl_fp32 -example_grouped_gemm_xdl_fp16 -example_grouped_gemm_xdl_bf16 -example_cgemm_xdl_bf16 -example_cgemm_xdl_fp16 Signed-off-by: Michal Kulikowski <Michal.Kulikowski@amd.com> --------- Signed-off-by: Michal Kulikowski <Michal.Kulikowski@amd.com>
197 lines
28 KiB
C++
197 lines
28 KiB
C++
// SPDX-License-Identifier: MIT
|
|
// Copyright (c) 2018-2025, Advanced Micro Devices, Inc. All rights reserved.
|
|
|
|
#pragma once
|
|
|
|
#include "ck/ck.hpp"
|
|
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
|
|
#include "ck/tensor_operation/gpu/device/impl/device_contraction_multiple_d_xdl_cshuffle.hpp"
|
|
|
|
using F16 = ck::half_t;
|
|
using BF16 = ck::bhalf_t;
|
|
using F32 = float;
|
|
using F64 = double;
|
|
|
|
template <ck::index_t... Is>
|
|
using S = ck::Sequence<Is...>;
|
|
|
|
static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
|
|
|
|
// Generic instances for fp32, fp16 and bf16 data types.
|
|
template <ck::index_t NumDimM,
|
|
ck::index_t NumDimN,
|
|
ck::index_t NumDimK,
|
|
typename ADataType,
|
|
typename BDataType,
|
|
typename AccDataType,
|
|
typename CShuffleDataType,
|
|
typename DsDataType,
|
|
typename EDataType,
|
|
typename ComputeDataType,
|
|
typename AElementOp,
|
|
typename BElementOp,
|
|
typename CDEElementOp>
|
|
// clang-format off
|
|
using DeviceOpInstanceKK_Generic = ck::tensor_operation::device::
|
|
//#####################################| NumDimM| NumDimN| NumDimK| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Compute|
|
|
//#####################################| | | | Type| Type| Type| DataType| Type| Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| Data|
|
|
//#####################################| | | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| Type|
|
|
//#####################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
|
DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 1, 256, 256, 128, 16, 4, 4, 16, 16, 8, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 4, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 4, 1, 1, 1, S<1, 16, 1, 16>, 2, ComputeDataType>;
|
|
// clang-format on
|
|
|
|
template <ck::index_t NumDimM,
|
|
ck::index_t NumDimN,
|
|
ck::index_t NumDimK,
|
|
typename ADataType,
|
|
typename BDataType,
|
|
typename AccDataType,
|
|
typename CShuffleDataType,
|
|
typename DsDataType,
|
|
typename EDataType,
|
|
typename ComputeDataType,
|
|
typename AElementOp,
|
|
typename BElementOp,
|
|
typename CDEElementOp>
|
|
// clang-format off
|
|
using DeviceOpInstanceKN_Generic = ck::tensor_operation::device::
|
|
//#####################################| NumDimM| NumDimN| NumDimK| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Compute|
|
|
//#####################################| | | | Type| Type| Type| DataType| Type| Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| Data|
|
|
//#####################################| | | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| Type|
|
|
//#####################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
|
DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 1, 256, 256, 128, 16, 4, 1, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 4, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 1, 0, 1, 1, S<1, 16, 1, 16>, 4, ComputeDataType>;
|
|
// clang-format on
|
|
|
|
template <ck::index_t NumDimM,
|
|
ck::index_t NumDimN,
|
|
ck::index_t NumDimK,
|
|
typename ADataType,
|
|
typename BDataType,
|
|
typename AccDataType,
|
|
typename CShuffleDataType,
|
|
typename DsDataType,
|
|
typename EDataType,
|
|
typename ComputeDataType,
|
|
typename AElementOp,
|
|
typename BElementOp,
|
|
typename CDEElementOp>
|
|
// clang-format off
|
|
using DeviceOpInstanceMK_Generic = ck::tensor_operation::device::
|
|
//#####################################| NumDimM| NumDimN| NumDimK| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Compute|
|
|
//#####################################| | | | Type| Type| Type| DataType| Type| Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| Data|
|
|
//#####################################| | | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| Type|
|
|
//#####################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
|
DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 1, 256, 256, 128, 16, 1, 4, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 1, 0, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 4, 1, 1, 1, S<1, 16, 1, 16>, 4, ComputeDataType>;
|
|
// clang-format on
|
|
|
|
template <ck::index_t NumDimM,
|
|
ck::index_t NumDimN,
|
|
ck::index_t NumDimK,
|
|
typename ADataType,
|
|
typename BDataType,
|
|
typename AccDataType,
|
|
typename CShuffleDataType,
|
|
typename DsDataType,
|
|
typename EDataType,
|
|
typename ComputeDataType,
|
|
typename AElementOp,
|
|
typename BElementOp,
|
|
typename CDEElementOp>
|
|
// clang-format off
|
|
using DeviceOpInstanceMN_Generic = ck::tensor_operation::device::
|
|
//#####################################| NumDimM| NumDimN| NumDimK| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Compute|
|
|
//#####################################| | | | Type| Type| Type| DataType| Type| Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| Data|
|
|
//#####################################| | | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| Type|
|
|
//#####################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
|
DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 1, 256, 256, 128, 16, 1, 1, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 1, 0, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 1, 0, 1, 1, S<1, 16, 1, 16>, 4, ComputeDataType>;
|
|
// clang-format on
|
|
|
|
// Fp64 instances.
|
|
template <ck::index_t NumDimM,
|
|
ck::index_t NumDimN,
|
|
ck::index_t NumDimK,
|
|
typename ADataType,
|
|
typename BDataType,
|
|
typename AccDataType,
|
|
typename CShuffleDataType,
|
|
typename DsDataType,
|
|
typename EDataType,
|
|
typename ComputeDataType,
|
|
typename AElementOp,
|
|
typename BElementOp,
|
|
typename CDEElementOp>
|
|
// clang-format off
|
|
using DeviceOpInstanceKK_FP64 = ck::tensor_operation::device::
|
|
//#####################################| NumDimM| NumDimN| NumDimK| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Compute|
|
|
//#####################################| | | | Type| Type| Type| DataType| Type| Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| Data|
|
|
//#####################################| | | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| Type|
|
|
//#####################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
|
DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 1, 256, 128, 128, 16, 2, 2, 16, 16, 4, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, 1, 1, 1, S<1, 16, 1, 16>, 1, ComputeDataType>;
|
|
// clang-format on
|
|
|
|
template <ck::index_t NumDimM,
|
|
ck::index_t NumDimN,
|
|
ck::index_t NumDimK,
|
|
typename ADataType,
|
|
typename BDataType,
|
|
typename AccDataType,
|
|
typename CShuffleDataType,
|
|
typename DsDataType,
|
|
typename EDataType,
|
|
typename ComputeDataType,
|
|
typename AElementOp,
|
|
typename BElementOp,
|
|
typename CDEElementOp>
|
|
// clang-format off
|
|
using DeviceOpInstanceKN_FP64 = ck::tensor_operation::device::
|
|
//#####################################| NumDimM| NumDimN| NumDimK| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Compute|
|
|
//#####################################| | | | Type| Type| Type| DataType| Type| Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| Data|
|
|
//#####################################| | | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| Type|
|
|
//#####################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
|
DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 1, 256, 128, 128, 16, 2, 1, 16, 16, 4, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 1, 0, 1, 1, S<1, 16, 1, 16>, 1, ComputeDataType>;
|
|
// clang-format on
|
|
|
|
template <ck::index_t NumDimM,
|
|
ck::index_t NumDimN,
|
|
ck::index_t NumDimK,
|
|
typename ADataType,
|
|
typename BDataType,
|
|
typename AccDataType,
|
|
typename CShuffleDataType,
|
|
typename DsDataType,
|
|
typename EDataType,
|
|
typename ComputeDataType,
|
|
typename AElementOp,
|
|
typename BElementOp,
|
|
typename CDEElementOp>
|
|
// clang-format off
|
|
using DeviceOpInstanceMK_FP64 = ck::tensor_operation::device::
|
|
//#####################################| NumDimM| NumDimN| NumDimK| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Compute|
|
|
//#####################################| | | | Type| Type| Type| DataType| Type| Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| Data|
|
|
//#####################################| | | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| Type|
|
|
//#####################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
|
DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 1, 256, 128, 128, 16, 1, 2, 16, 16, 4, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 1, 0, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, 1, 1, 1, S<1, 16, 1, 16>, 1, ComputeDataType>;
|
|
// clang-format on
|
|
|
|
template <ck::index_t NumDimM,
|
|
ck::index_t NumDimN,
|
|
ck::index_t NumDimK,
|
|
typename ADataType,
|
|
typename BDataType,
|
|
typename AccDataType,
|
|
typename CShuffleDataType,
|
|
typename DsDataType,
|
|
typename EDataType,
|
|
typename ComputeDataType,
|
|
typename AElementOp,
|
|
typename BElementOp,
|
|
typename CDEElementOp>
|
|
// clang-format off
|
|
using DeviceOpInstanceMN_FP64 = ck::tensor_operation::device::
|
|
//#####################################| NumDimM| NumDimN| NumDimK| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Compute|
|
|
//#####################################| | | | Type| Type| Type| DataType| Type| Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| Data|
|
|
//#####################################| | | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| Type|
|
|
//#####################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
|
DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 1, 256, 128, 128, 16, 1, 1, 16, 16, 4, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 1, 0, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 1, 0, 1, 1, S<1, 16, 1, 16>, 1, ComputeDataType>;
|
|
// clang-format on
|