device_batched_gemm_softmax_gemm_xdl_cshuffle.hpp File Reference

device_batched_gemm_softmax_gemm_xdl_cshuffle.hpp File Reference#

Composable Kernel: device_batched_gemm_softmax_gemm_xdl_cshuffle.hpp File Reference
Classes | Namespaces | Functions
device_batched_gemm_softmax_gemm_xdl_cshuffle.hpp File Reference
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_batched_gemm_softmax_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/masking_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_batched_gemm_softmax_gemm_xdl_cshuffle_v1.hpp"

Go to the source code of this file.

Classes

struct  ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle< ALayout, BLayout, B1Layout, CLayout, ADataType, BDataType, B1DataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock, Gemm1KPerBlock, AK1, BK1, B1K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, Gemm1NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, B1BlockTransferThreadClusterLengths_BK0_N_BK1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_BK1, B1BlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskOutUpperTriangle, LoopSched >
struct  ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle< ALayout, BLayout, B1Layout, CLayout, ADataType, BDataType, B1DataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock, Gemm1KPerBlock, AK1, BK1, B1K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, Gemm1NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, B1BlockTransferThreadClusterLengths_BK0_N_BK1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_BK1, B1BlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskOutUpperTriangle, LoopSched >::ComputeBasePtrOfStridedBatch
struct  ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle< ALayout, BLayout, B1Layout, CLayout, ADataType, BDataType, B1DataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock, Gemm1KPerBlock, AK1, BK1, B1K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, Gemm1NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, B1BlockTransferThreadClusterLengths_BK0_N_BK1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_BK1, B1BlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskOutUpperTriangle, LoopSched >::Argument
struct  ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle< ALayout, BLayout, B1Layout, CLayout, ADataType, BDataType, B1DataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock, Gemm1KPerBlock, AK1, BK1, B1K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, Gemm1NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, B1BlockTransferThreadClusterLengths_BK0_N_BK1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_BK1, B1BlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskOutUpperTriangle, LoopSched >::Invoker
struct  ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle< ALayout, BLayout, B1Layout, CLayout, ADataType, BDataType, B1DataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock, Gemm1KPerBlock, AK1, BK1, B1K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, Gemm1NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, B1BlockTransferThreadClusterLengths_BK0_N_BK1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_BK1, B1BlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskOutUpperTriangle, LoopSched >::Descriptor< ADesc, BDesc, B1Desc, CDesc >

Namespaces

namespace  ck
namespace  ck::tensor_operation
namespace  ck::tensor_operation::device

Functions

template<typename GridwiseGemm, typename FloatAB, typename FloatC, typename AElementwiseOperation, typename BElementwiseOperation, typename AccElementwiseOperation, typename B1ElementwiseOperation, typename CElementwiseOperation, typename AGridDesc_AK0_M_AK1, typename BGridDesc_BK0_N_BK1, typename B1GridDesc_BK0_N_BK1, typename CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock, typename Block2CTileMap, typename ComputeBasePtrOfStridedBatch, typename C0MatrixMask, bool HasMainKBlockLoop>
__global__ void ck::tensor_operation::device::kernel_batched_gemm_softmax_gemm_xdl_cshuffle_v1 (const FloatAB *__restrict__ p_a_grid, const FloatAB *__restrict__ p_b_grid, const FloatAB *__restrict__ p_b1_grid, FloatC *__restrict__ p_c_grid, const AElementwiseOperation a_element_op, const BElementwiseOperation b_element_op, const AccElementwiseOperation acc_element_op, const B1ElementwiseOperation b1_element_op, const CElementwiseOperation c_element_op, const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1, const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1, const B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1, const CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock c_grid_desc_mblock_mperblock_nblock_nperblock, const Block2CTileMap block_2_ctile_map, const index_t batch_count, const ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch, const C0MatrixMask c0_matrix_mask)