device_grouped_conv_bwd_weight_dl.hpp File Reference

device_grouped_conv_bwd_weight_dl.hpp File Reference#

Composable Kernel: device_grouped_conv_bwd_weight_dl.hpp File Reference
Classes | Namespaces | Functions
device_grouped_conv_bwd_weight_dl.hpp File Reference
#include <iostream>
#include <numeric>
#include <sstream>
#include "ck/utility/common_header.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_grouped_conv_bwd_weight.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_utils.hpp"
#include "ck/tensor_operation/gpu/device/convolution_backward_weight_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_dl_v1r3.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"

Go to the source code of this file.

Classes

struct  ck::tensor_operation::device::DeviceGroupedConvBwdWeight_Dl< NDimSpatial, InLayout, WeiLayout, OutLayout, InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvBackwardWeightSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >
struct  ck::tensor_operation::device::DeviceGroupedConvBwdWeight_Dl< NDimSpatial, InLayout, WeiLayout, OutLayout, InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvBackwardWeightSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::Argument
struct  ck::tensor_operation::device::DeviceGroupedConvBwdWeight_Dl< NDimSpatial, InLayout, WeiLayout, OutLayout, InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvBackwardWeightSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::Invoker

Namespaces

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

Functions

template<typename GridwiseGemm, typename FloatAB, typename FloatC, typename AGridDesc_B_K0_M0_M1_K1, typename BGridDesc_B_K0_N0_N1_K1, typename CGridDesc_M0_M10_M11_N0_N10_N11, typename Block2CTileMap, typename ComputePtrOffsetOfBatch, bool HasMainKBlockLoop, bool HasDoubleTailKBlockLoop>
__global__ void ck::tensor_operation::device::kernel_batched_gemm_dlops_bwd_weight (const FloatAB *__restrict__ p_a_grid, const FloatAB *__restrict__ p_b_grid, FloatC *__restrict__ p_c_grid, const index_t batch_count, const AGridDesc_B_K0_M0_M1_K1 a_grid_desc_kbatch_k0_m0_m1_k1, const BGridDesc_B_K0_N0_N1_K1 b_grid_desc_kbatch_k0_n0_n1_k1, const CGridDesc_M0_M10_M11_N0_N10_N11 c_grid_desc_m0_m10_m11_n0_n10_n11, const Block2CTileMap block_2_ctile_map, const ComputePtrOffsetOfBatch compute_ptr_offset_of_batch)