ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-06-30 19:57:40 +00:00
Code Issues Packages Projects Releases Wiki Activity

Re-factor of BF16 instance tuples

Browse Source
This commit is contained in:
Wojciech Laskowski
2025-12-02 14:57:22 +00:00
parent a36fd78205
commit 4fd168ea67
54 changed files with 108 additions and 2626 deletions
Download Patch File Download Diff File Expand all files Collapse all files

37
library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_comp_instance.hpp
View File

@@ -52,43 +52,6 @@ static constexpr auto GemmMNKPadding = GemmSpecialization::MNKPadding;
static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
template <index_t NDimSpatial,
typename ALayout,
typename BLayout,
typename DsLayout,
typename ELayout,
ConvolutionForwardSpecialization ConvSpec,
typename DsDataTypes = Tuple<>,
typename OutElementOp = PassThrough>
using device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances = std::tuple<
// clang-format off
//########################################| NumDim| A| B| Ds| E| AData| BData| AccData| CShuffle| Ds| EData| A| B| CDE| ConvForward| GEMM| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MWmma| NWmma| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Pipeline scheduler | Pipeline version |
//########################################| Spatial| Layout| Layout| Layout| Layout| Type| Type| Type| DataType| DataType| Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Size| Block| Block| Block| | | WMMA| WMMA| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MWmmaPerWave| NWmmaPerWave| _MBlock_MWaveMPerWmma| ScalarPerVector| | |
//########################################| | | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerWmma| _NWaveNPerWmma| | |
//########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
// Compute friendly
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 256, 256, 32, 8, 8, 16, 16, 8, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>
#ifndef ONE_INSTANCE_PER_LIST
,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 256, 256, 32, 8, 8, 16, 16, 8, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 128, 256, 32, 8, 8, 16, 16, 4, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 256, 128, 32, 8, 8, 16, 16, 8, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 128, 128, 64, 8, 8, 16, 16, 4, 2, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 128, 64, 64, 8, 8, 16, 16, 4, 1, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 64, 128, 64, 8, 8, 16, 16, 2, 2, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 64, 64, 64, 8, 8, 16, 16, 2, 1, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
// "2x" instances
// DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 128, 128, 64, 16, 16, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1> // Assert broken
// "part 2" instances
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 256, 256, 32, 8, 8, 16, 16, 4, 8, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 2, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
// AGPR Spill when use permuted lds layout. so, use padding for these two.
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 224, 256, 64, 8, 8, 16, 16, 7, 4, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 2, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 256, 224, 64, 8, 8, 16, 16, 4, 7, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 2, 1, S<1, 64, 1, 4>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 128, 128, 64, 8, 8, 16, 16, 4, 2, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 32, 1, 8>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
#endif
// clang-format on
>;
template <index_t NDimSpatial,
typename ALayout,
typename BLayout,

45
library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_instance.hpp
View File

@@ -57,7 +57,26 @@ template <index_t NDimSpatial,
ConvolutionForwardSpecialization ConvSpec,
typename DsDataTypes = Tuple<>,
typename OutElementOp = PassThrough>
using device_grouped_conv_fwd_wmma_cshufflev3_bf16_generic_instances = std::tuple<
using device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances = std::tuple<
// clang-format off
//########################################| NumDim| A| B| Ds| E| AData| BData| AccData| CShuffle| Ds| EData| A| B| CDE| ConvForward| GEMM| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MWmma| NWmma| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Pipeline scheduler | Pipeline version |
//########################################| Spatial| Layout| Layout| Layout| Layout| Type| Type| Type| DataType| DataType| Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Size| Block| Block| Block| | | WMMA| WMMA| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MWmmaPerWave| NWmmaPerWave| _MBlock_MWaveMPerWmma| ScalarPerVector| | |
//########################################| | | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerWmma| _NWaveNPerWmma| | |
//########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
// generic instance
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 64, 64, 64, 32, 8, 8, 16, 16, 4, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 1, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 1, 8, 1, 1, 1, S<1, 16, 1, 4>, 1, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>
// clang-format on
>;
template <index_t NDimSpatial,
typename ALayout,
typename BLayout,
typename DsLayout,
typename ELayout,
ConvolutionForwardSpecialization ConvSpec,
typename DsDataTypes = Tuple<>,
typename OutElementOp = PassThrough>
using device_grouped_conv_fwd_wmma_cshufflev3_bf16_gnhwc_instances = std::tuple<
// clang-format off
//########################################| NumDim| A| B| Ds| E| AData| BData| AccData| CShuffle| Ds| EData| A| B| CDE| ConvForward| GEMM| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MWmma| NWmma| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Pipeline scheduler | Pipeline version |
//########################################| Spatial| Layout| Layout| Layout| Layout| Type| Type| Type| DataType| DataType| Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Size| Block| Block| Block| | | WMMA| WMMA| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MWmmaPerWave| NWmmaPerWave| _MBlock_MWaveMPerWmma| ScalarPerVector| | |
@@ -126,30 +145,6 @@ using device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances = std::tuple<
// clang-format on
>;
template <index_t NDimSpatial,
typename ALayout,
typename BLayout,
typename DsLayout,
typename ELayout,
ConvolutionForwardSpecialization ConvSpec,
typename DsDataTypes = Tuple<>,
typename OutElementOp = PassThrough>
using device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances = std::tuple<
// clang-format off
//########################################| NumDim| A| B| Ds| E| AData| BData| AccData| CShuffle| Ds| EData| A| B| CDE| ConvForward| GEMM| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MWmma| NWmma| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Pipeline scheduler | Pipeline version |
//########################################| Spatial| Layout| Layout| Layout| Layout| Type| Type| Type| DataType| DataType| Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Size| Block| Block| Block| | | WMMA| WMMA| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MWmmaPerWave| NWmmaPerWave| _MBlock_MWaveMPerWmma| ScalarPerVector| | |
//########################################| | | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerWmma| _NWaveNPerWmma| | |
//########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 64, 64, 32, 8, 8, 16, 16, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 1, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 8, 1, 1, 1, S<1, 32, 1, 4>, 1, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>
#ifndef ONE_INSTANCE_PER_LIST
,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 64, 64, 32, 8, 8, 16, 16, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 1, 8, 1, 1, 1, S<1, 32, 1, 4>, 2, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 64, 64, 32, 8, 8, 16, 16, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 8, 1, 1, 1, S<1, 32, 1, 4>, 4, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 256, 64, 64, 32, 8, 8, 16, 16, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>
#endif
// clang-format on
>;
template <index_t NDimSpatial,
typename ALayout,
typename BLayout,

26
library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp
View File

@@ -49,32 +49,6 @@ static constexpr auto GemmMNKPadding = GemmSpecialization::MNKPadding;
static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
template <index_t NDimSpatial,
typename ALayout,
typename BLayout,
typename DsLayout,
typename ELayout,
ConvolutionForwardSpecialization ConvSpec,
BlockGemmPipelineScheduler BlkGemmPipeSched,
typename DsDataTypes = Tuple<>,
typename OutElementOp = PassThrough>
using device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances = std::tuple<
// clang-format off
//########################################| NumDim| A| B| Ds| E| AData| BData| AccData| CShuffle| Ds| EData| A| B| CDE| ConvForward| GEMM| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MWmma| NWmma| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Pipeline scheduler | Pipeline version |
//########################################| Spatial| Layout| Layout| Layout| Layout| Type| Type| Type| DataType| DataType| Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Size| Block| Block| Block| | | WMMA| WMMA| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MWmmaPerWave| NWmmaPerWave| _MBlock_MWaveMPerWmma| ScalarPerVector| | |
//########################################| | | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerWmma| _NWaveNPerWmma| | |
//########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
// Latency friendly
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 128, 32, 32, 64, 8, 8, 16, 16, 1, 1, S<8, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 16, 1, 8>, 2, BlkGemmPipeSched, BlockGemmPipelineVersion::v1>
#ifndef ONE_INSTANCE_PER_LIST
,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 64, 32, 16, 128, 8, 8, 16, 16, 1, 1, S<16, 4, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<16, 4, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 16, 1, 4>, 4, BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 64, 32, 16, 64, 8, 8, 16, 16, 1, 1, S<8, 8, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 8, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 16, 1, 4>, 4, BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3<NDimSpatial, ALayout, BLayout, DsLayout, ELayout, BF16, BF16, F32, BF16, DsDataTypes, BF16, PassThrough, PassThrough, OutElementOp, ConvSpec, GemmMNKPadding, 128, 32, 32, 64, 8, 8, 16, 16, 1, 1, S<8, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<8, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, 1, 1, S<1, 16, 1, 8>, 4, BlkGemmPipeSched, BlockGemmPipelineVersion::v1>
#endif
// clang-format on
>;
template <index_t NDimSpatial,
typename ALayout,
typename BLayout,

32
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward.hpp
View File

@@ -870,18 +870,10 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
{
add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
op_ptrs);
// add_device_grouped_conv2d_fwd_wmma_cshufflev3_large_tensor_nhwgc_gkyxc_nhwgk_bf16_instances(
// op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
op_ptrs);
}
#endif
#ifdef CK_ENABLE_INT8
@@ -939,14 +931,6 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_16x16_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_comp_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_intra_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_inter_instances(
op_ptrs);
}
#endif
}
@@ -1094,18 +1078,10 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
{
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instances(
op_ptrs);
// add_device_grouped_conv3d_fwd_wmma_cshufflev3_large_tensor_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
// op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_merged_groups_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instances(
op_ptrs);
}
#endif
#ifdef CK_ENABLE_INT8
@@ -1153,14 +1129,6 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x16_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_comp_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_intra_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_inter_instances(
op_ptrs);
}
#endif
}

16
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_bias_clamp.hpp
View File

@@ -285,18 +285,10 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
{
add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
op_ptrs);
// add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_large_tensor_nhwgc_gkyxc_nhwgk_bf16_instances(
// op_ptrs);
add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
op_ptrs);
}
#endif
#ifdef CK_ENABLE_FP16
@@ -334,18 +326,10 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
{
add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instances(
op_ptrs);
// add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_large_tensor_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
// op_ptrs);
add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instances(
op_ptrs);
}
#endif
#ifdef CK_ENABLE_FP16

113
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_bias_clamp_wmma_cshufflev3.inc
View File

@@ -24,20 +24,6 @@ void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_
PassThrough,
AddClamp>>>& instances);
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances);
// void
// add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_large_tensor_nhwgc_gkyxc_nhwgk_bf16_instances(
// std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
@@ -67,48 +53,6 @@ void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_nhwg
PassThrough,
AddClamp>>>& instances);
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances);
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances);
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances);
void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
@@ -123,20 +67,6 @@ void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhw
PassThrough,
AddClamp>>>& instances);
void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances);
// void
// add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_large_tensor_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
// std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
@@ -165,49 +95,6 @@ void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_ndhw
PassThrough,
PassThrough,
AddClamp>>>& instances);
void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances);
void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances);
void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances);
#endif
#ifdef CK_ENABLE_FP16

16
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_clamp.hpp
View File

@@ -281,18 +281,10 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
{
add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
op_ptrs);
// add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_large_tensor_nhwgc_gkyxc_nhwgk_bf16_instances(
// op_ptrs);
add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
op_ptrs);
add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
op_ptrs);
}
#endif
#ifdef CK_ENABLE_FP16
@@ -330,18 +322,10 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
{
add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instances(
op_ptrs);
// add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_large_tensor_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
// op_ptrs);
add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_merged_groups_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instances(
op_ptrs);
add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instances(
op_ptrs);
}
#endif
#ifdef CK_ENABLE_FP16

113
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_clamp_wmma_cshufflev3.inc
View File

@@ -24,20 +24,6 @@ void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_
PassThrough,
Clamp>>>& instances);
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances);
// void
// add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_large_tensor_nhwgc_gkyxc_nhwgk_bf16_instances(
// std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
@@ -67,48 +53,6 @@ void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_merged_groups_nhwgc_gky
PassThrough,
Clamp>>>& instances);
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances);
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances);
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances);
void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
@@ -123,20 +67,6 @@ void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf
PassThrough,
Clamp>>>& instances);
void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances);
// void
// add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_large_tensor_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
// std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
@@ -165,49 +95,6 @@ void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_merged_groups_ndhwgc_gk
PassThrough,
PassThrough,
Clamp>>>& instances);
void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances);
void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances);
void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances);
#endif
#ifdef CK_ENABLE_FP16

65
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_comp_wmma_cshufflev3.inc
View File

@@ -9,22 +9,6 @@ namespace device {
namespace instance {
// grouped conv2d forward, NHWGC/GKYXC/NHWGK
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_FP16
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
@@ -74,39 +58,6 @@ void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_comp_in
PassThrough>>>& instances);
#endif // CK_ENABLE_FP16
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
// grouped conv3d forward, NDHWGC/GKZYXC/NDHWGK
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_FP16
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_f16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
@@ -140,22 +91,6 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_comp
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
} // namespace instance
} // namespace device
} // namespace tensor_operation

65
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_mem_inter_wmma_cshufflev3.inc
View File

@@ -9,22 +9,6 @@ namespace device {
namespace instance {
// grouped conv2d forward, NHWGC/GKYXC/NHWGK
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_FP16
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
@@ -74,39 +58,6 @@ void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_mem_int
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
// grouped conv3d forward, NDHWGC/GKZYXC/NDHWGK
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_FP16
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_f16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
@@ -140,22 +91,6 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_mem_
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
} // namespace instance
} // namespace device
} // namespace tensor_operation

65
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_mem_intra_wmma_cshufflev3.inc
View File

@@ -9,22 +9,6 @@ namespace device {
namespace instance {
// grouped conv2d forward, NHWGC/GKYXC/NHWGK
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_FP16
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
@@ -74,39 +58,6 @@ void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_mem_int
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
// grouped conv3d forward, NDHWGC/GKZYXC/NDHWGK
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_FP16
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_f16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
@@ -140,22 +91,6 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_mem_
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
} // namespace instance
} // namespace device
} // namespace tensor_operation

55
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_wmma_cshufflev3.inc
View File

@@ -105,20 +105,6 @@ void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instan
PassThrough,
PassThrough,
PassThrough>>>& instances);
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_FP16
@@ -261,20 +247,6 @@ void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_instan
PassThrough,
PassThrough,
PassThrough>>>& instances);
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
@@ -342,20 +314,6 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_ins
PassThrough,
PassThrough,
PassThrough>>>& instances);
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
#ifdef CK_ENABLE_FP16
@@ -534,19 +492,6 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_ins
PassThrough,
PassThrough,
PassThrough>>>& instances);
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances);
#endif
} // namespace instance

15
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/CMakeLists.txt
View File

@@ -117,7 +117,6 @@ set(GROUPED_CONV2D_FWD
# NHWGC, GKYXC, NHWGK
wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_instance.cpp
wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instance.cpp
wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_16x16_instance.cpp
wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_instance.cpp
# NGCHW, GKYXC, NGKHW
@@ -125,7 +124,6 @@ set(GROUPED_CONV2D_FWD
wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_f16_instance.cpp
wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_int8_instance.cpp
# NGCHW, GKCYX, NGKHW
wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_16x16_instance.cpp
wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_16x16_instance.cpp
# merged groups
@@ -139,20 +137,15 @@ set(GROUPED_CONV2D_FWD
#mem
# NHWGC, GKYXC, NHWGK intra
wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_intra_instance.cpp
# NHWGC, GKYXC, NHWGK inter
wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instance.cpp
wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_inter_instance.cpp
# NGCHW, GKCYX, NGKHW intra
wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_intra_instance.cpp
wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_mem_intra_instance.cpp
# NGCHW, GKCYX, NGKHW inter
wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_inter_instance.cpp
wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_mem_inter_instance.cpp
#comp
# NHWGC, GKYXC, NHWGK
wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_comp_instance.cpp
wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_comp_instance.cpp
# NGCHW, GKCYX, NGKHW
@@ -202,14 +195,6 @@ generate_sharded_instantiations(
OUTPUT_DIR ${GENERATED_DIR}/xdl/comp
)
set(GENERATED_DIR ${CMAKE_CURRENT_BINARY_DIR}/generated)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_comp_instances
TEMPLATE_FILE wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_comp_instance.in
NUM_SHARDS 16
SRC_LIST GROUPED_CONV2D_FWD
OUTPUT_DIR ${GENERATED_DIR}/wmma/comp
)
set(GENERATED_DIR ${CMAKE_CURRENT_BINARY_DIR}/generated)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_int8_mem_inter_instances
TEMPLATE_FILE xdl/mem/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_int8_mem_inter_instance.in

44
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_comp_instance.in
View File

@@ -1,44 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_comp_instance.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/utility/filter_tuple.hpp"
namespace ck::tensor_operation::device::instance {
using device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_comp_instances =
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>;
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
template <int Shards, int ShardIndex>
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_comp_instances_shard(
[[maybe_unused]] device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_comp_instances&
instances)
{
add_device_operation_instances(
instances,
ck::util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>,
Shards,
ShardIndex>{});
}
} // namespace ck::tensor_operation::device::instance

66
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
View File

@@ -1,66 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_comp_instance.hpp"
#include "ck/host_utility/device_prop.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwdDefault>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwd1x1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwd1x1S1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwdOddC>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

48
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_gnhwc_gkyxc_gnhwk_bf16_instance.cpp
View File

@@ -24,39 +24,39 @@ void add_device_grouped_conv2d_fwd_wmma_cshufflev3_gnhwc_gkyxc_gnhwk_bf16_instan
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<2,
GNHWC,
GKYXC,
Empty_Tuple,
GNHWK,
ConvFwdDefault>{});
device_grouped_conv_fwd_wmma_cshufflev3_bf16_gnhwc_instances<2,
GNHWC,
GKYXC,
Empty_Tuple,
GNHWK,
ConvFwdDefault>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<2,
GNHWC,
GKYXC,
Empty_Tuple,
GNHWK,
ConvFwd1x1P0>{});
device_grouped_conv_fwd_wmma_cshufflev3_bf16_gnhwc_instances<2,
GNHWC,
GKYXC,
Empty_Tuple,
GNHWK,
ConvFwd1x1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<2,
GNHWC,
GKYXC,
Empty_Tuple,
GNHWK,
ConvFwd1x1S1P0>{});
device_grouped_conv_fwd_wmma_cshufflev3_bf16_gnhwc_instances<2,
GNHWC,
GKYXC,
Empty_Tuple,
GNHWK,
ConvFwd1x1S1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<2,
GNHWC,
GKYXC,
Empty_Tuple,
GNHWK,
ConvFwdOddC>{});
device_grouped_conv_fwd_wmma_cshufflev3_bf16_gnhwc_instances<2,
GNHWC,
GKYXC,
Empty_Tuple,
GNHWK,
ConvFwdOddC>{});
}
} // namespace instance

54
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_16x16_instance.cpp
View File

@@ -1,54 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwd1x1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwd1x1S1P0>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

36
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_bf16_instance.cpp
View File

@@ -24,12 +24,36 @@ void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_bf16_instan
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_generic_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwd1x1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwd1x1S1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwdOddC>{});
}
} // namespace instance

56
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instance.cpp
View File

@@ -1,56 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwdDefault>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwd1x1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwd1x1S1P0>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

39
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_inter_instance.cpp
View File

@@ -1,39 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault,
Interwave>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

39
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_intra_instance.cpp
View File

@@ -1,39 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault,
Intrawave>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

69
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instance.cpp
View File

@@ -1,69 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwdDefault,
Interwave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwd1x1P0,
Interwave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwd1x1S1P0,
Interwave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwdOddC,
Interwave>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

69
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
View File

@@ -1,69 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwdDefault,
Intrawave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwd1x1P0,
Intrawave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwd1x1S1P0,
Intrawave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Empty_Tuple,
NHWGK,
ConvFwdOddC,
Intrawave>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

4
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/CMakeLists.txt
View File

@@ -39,11 +39,7 @@ add_instance_library(device_grouped_conv2d_fwd_bias_clamp_instance
# WMMA CSHUFFLE V3
wmma/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
wmma/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instance.cpp
wmma/merged_groups/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
wmma/mem/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
wmma/mem/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instance.cpp
wmma/comp/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
wmma/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_instance.cpp
wmma/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_16x16_instance.cpp

63
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/comp/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
View File

@@ -1,63 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_comp_instance.hpp"
#include "ck/host_utility/device_prop.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwdDefault,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwd1x1P0,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwd1x1S1P0,
Tuple<BF16>,
AddClamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

62
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instance.cpp
View File

@@ -1,62 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwdDefault,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwd1x1P0,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwd1x1S1P0,
Tuple<BF16>,
AddClamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

65
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/mem/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instance.cpp
View File

@@ -1,65 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwdDefault,
Interwave,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwd1x1P0,
Interwave,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwd1x1S1P0,
Interwave,
Tuple<BF16>,
AddClamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

65
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/mem/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
View File

@@ -1,65 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwdDefault,
Intrawave,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwd1x1P0,
Intrawave,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<NHWGK>,
NHWGK,
ConvFwd1x1S1P0,
Intrawave,
Tuple<BF16>,
AddClamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

4
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/CMakeLists.txt
View File

@@ -39,11 +39,7 @@ add_instance_library(device_grouped_conv2d_fwd_clamp_instance
# WMMA CSHUFFLE V3
wmma/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
wmma/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instance.cpp
wmma/merged_groups/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
wmma/mem/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
wmma/mem/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instance.cpp
wmma/comp/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
wmma/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_instance.cpp
wmma/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_16x16_instance.cpp

63
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/comp/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
View File

@@ -1,63 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_comp_instance.hpp"
#include "ck/host_utility/device_prop.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwdDefault,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwd1x1P0,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwd1x1S1P0,
Tuple<>,
Clamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

62
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instance.cpp
View File

@@ -1,62 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwdDefault,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwd1x1P0,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwd1x1S1P0,
Tuple<>,
Clamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

65
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/mem/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instance.cpp
View File

@@ -1,65 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwdDefault,
Interwave,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwd1x1P0,
Interwave,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwd1x1S1P0,
Interwave,
Tuple<>,
Clamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

65
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/mem/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
View File

@@ -1,65 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwdDefault,
Intrawave,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwd1x1P0,
Intrawave,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<2,
NHWGC,
GKYXC,
Tuple<>,
NHWGK,
ConvFwd1x1S1P0,
Intrawave,
Tuple<>,
Clamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

33
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/CMakeLists.txt
View File

@@ -73,21 +73,18 @@ set(GROUPED_CONV3D_FWD
wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_gndhwc_gkzyxc_gndhwk_int8_instance.cpp
wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_f16_instance.cpp
wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instance.cpp
wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_f16_16x16_instance.cpp
wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_int8_instance.cpp
wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x16_instance.cpp
wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_16x16_instance.cpp
wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_instance.cpp
wmma/merged_groups/device_grouped_conv3d_fwd_wmma_cshufflev3_merged_groups_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
wmma/merged_groups/device_grouped_conv3d_fwd_wmma_cshufflev3_merged_groups_ndhwgc_gkzyxc_ndhwgk_f16_instance.cpp
wmma/merged_groups/device_grouped_conv3d_fwd_wmma_cshufflev3_merged_groups_ngcdhw_gkczyx_ngkdhw_bf16_instance.cpp
wmma/merged_groups/device_grouped_conv3d_fwd_wmma_cshufflev3_merged_groups_ngcdhw_gkczyx_ngkdhw_f16_instance.cpp
wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instance.cpp
wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_f16_mem_inter_instance.cpp
wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instance.cpp
wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_f16_mem_intra_instance.cpp
)
# Add generated files for sharded instantiations.
@@ -200,13 +197,6 @@ generate_sharded_instantiations(
OUTPUT_DIR ${GENERATED_DIR}/wmma
)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_inter_instances
TEMPLATE_FILE wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_inter_instance.in
NUM_SHARDS 16
SRC_LIST GROUPED_CONV3D_FWD
OUTPUT_DIR ${GENERATED_DIR}/wmma/mem
)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_mem_inter_instances
TEMPLATE_FILE wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_mem_inter_instance.in
@@ -215,13 +205,6 @@ generate_sharded_instantiations(
OUTPUT_DIR ${GENERATED_DIR}/wmma/mem
)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_intra_instances
TEMPLATE_FILE wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_intra_instance.in
NUM_SHARDS 16
SRC_LIST GROUPED_CONV3D_FWD
OUTPUT_DIR ${GENERATED_DIR}/wmma/mem
)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_mem_intra_instances
TEMPLATE_FILE wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_mem_intra_instance.in
@@ -230,13 +213,6 @@ generate_sharded_instantiations(
OUTPUT_DIR ${GENERATED_DIR}/wmma/mem
)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances
TEMPLATE_FILE wmma/comp/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.in
NUM_SHARDS 16
SRC_LIST GROUPED_CONV3D_FWD
OUTPUT_DIR ${GENERATED_DIR}/wmma/comp
)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_f16_comp_instances
TEMPLATE_FILE wmma/comp/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_f16_comp_instance.in
@@ -244,13 +220,6 @@ generate_sharded_instantiations(
SRC_LIST GROUPED_CONV3D_FWD
OUTPUT_DIR ${GENERATED_DIR}/wmma/comp
)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_comp_instances
TEMPLATE_FILE wmma/comp/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_comp_instance.in
NUM_SHARDS 16
SRC_LIST GROUPED_CONV3D_FWD
OUTPUT_DIR ${GENERATED_DIR}/wmma/comp
)
generate_sharded_instantiations(
INSTANCES_NAME device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_comp_instances
TEMPLATE_FILE wmma/comp/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_comp_instance.in

65
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/comp/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.in
View File

@@ -1,65 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_comp_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/utility/filter_tuple.hpp"
namespace ck::tensor_operation::device::instance {
using device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances =
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>;
template <int Shards, int ShardIndex>
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances_shard(
device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances& instances)
{
add_device_operation_instances(
instances,
util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwdDefault>,
Shards,
ShardIndex>{});
add_device_operation_instances(
instances,
util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwd1x1P0>,
Shards,
ShardIndex>{});
add_device_operation_instances(
instances,
util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwd1x1S1P0>,
Shards,
ShardIndex>{});
}
} // namespace ck::tensor_operation::device::instance

64
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/comp/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_comp_instance.in
View File

@@ -1,64 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_comp_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/utility/filter_tuple.hpp"
namespace ck::tensor_operation::device::instance {
using device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_comp_instances =
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>;
template <int Shards, int ShardIndex>
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_comp_instances_shard(
device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_comp_instances& instances)
{
add_device_operation_instances(
instances,
util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwdDefault>,
Shards,
ShardIndex>{});
add_device_operation_instances(
instances,
util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwd1x1P0>,
Shards,
ShardIndex>{});
add_device_operation_instances(
instances,
util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwd1x1S1P0>,
Shards,
ShardIndex>{});
}
} // namespace ck::tensor_operation::device::instance

36
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_gndhwc_gkzyxc_gndhwk_bf16_instance.cpp
View File

@@ -25,28 +25,28 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_gndhwc_gkzyxc_gndhwk_bf16_ins
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<3,
GNDHWC,
GKZYXC,
Empty_Tuple,
GNDHWK,
ConvFwdDefault>{});
device_grouped_conv_fwd_wmma_cshufflev3_bf16_gnhwc_instances<3,
GNDHWC,
GKZYXC,
Empty_Tuple,
GNDHWK,
ConvFwdDefault>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<3,
GNDHWC,
GKZYXC,
Empty_Tuple,
GNDHWK,
ConvFwd1x1P0>{});
device_grouped_conv_fwd_wmma_cshufflev3_bf16_gnhwc_instances<3,
GNDHWC,
GKZYXC,
Empty_Tuple,
GNDHWK,
ConvFwd1x1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<3,
GNDHWC,
GKZYXC,
Empty_Tuple,
GNDHWK,
ConvFwd1x1S1P0>{});
device_grouped_conv_fwd_wmma_cshufflev3_bf16_gnhwc_instances<3,
GNDHWC,
GKZYXC,
Empty_Tuple,
GNDHWK,
ConvFwd1x1S1P0>{});
}
} // namespace instance

55
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instance.cpp
View File

@@ -1,55 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwdDefault>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwd1x1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwd1x1S1P0>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

16
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x16_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_instance.cpp
View File

@@ -8,7 +8,7 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x16_instances(
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,
GKCZYX,
@@ -24,7 +24,7 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
@@ -33,7 +33,7 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
@@ -41,12 +41,20 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x
ConvFwd1x1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwd1x1S1P0>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwdOddC>{});
}
} // namespace instance

6
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_instance.in
View File

@@ -26,7 +26,7 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_ins
{
add_device_operation_instances(
instances,
util::filter_tuple_by_modulo_t<device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<3,
util::filter_tuple_by_modulo_t<device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
@@ -37,7 +37,7 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_ins
add_device_operation_instances(
instances,
util::filter_tuple_by_modulo_t<device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<3,
util::filter_tuple_by_modulo_t<device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
@@ -48,7 +48,7 @@ void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_ins
add_device_operation_instances(
instances,
util::filter_tuple_by_modulo_t<device_grouped_conv_fwd_wmma_cshufflev3_bf16_instances<3,
util::filter_tuple_by_modulo_t<device_grouped_conv_fwd_wmma_cshufflev3_bf16_ngchw_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,

58
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instance.cpp
View File

@@ -1,58 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwdDefault,
Interwave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwd1x1P0,
Interwave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwd1x1S1P0,
Interwave>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

58
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instance.cpp
View File

@@ -1,58 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwdDefault,
Intrawave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwd1x1P0,
Intrawave>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Empty_Tuple,
NDHWGK,
ConvFwd1x1S1P0,
Intrawave>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

66
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_inter_instance.in
View File

@@ -1,66 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/utility/filter_tuple.hpp"
namespace ck::tensor_operation::device::instance {
using device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_inter_instances =
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>;
template <int Shards, int ShardIndex>
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_inter_instances_shard(
device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_inter_instances&
instances)
{
add_device_operation_instances(
instances,
ck::util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwdDefault,
Interwave>,
Shards,
ShardIndex>{});
add_device_operation_instances(
instances,
ck::util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwd1x1P0,
Interwave>,
Shards,
ShardIndex>{});
add_device_operation_instances(
instances,
ck::util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwd1x1S1P0,
Interwave>,
Shards,
ShardIndex>{});
}
} // namespace ck::tensor_operation::device::instance

66
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/mem/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_intra_instance.in
View File

@@ -1,66 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/utility/filter_tuple.hpp"
namespace ck::tensor_operation::device::instance {
using device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_intra_instances =
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>;
template <int Shards, int ShardIndex>
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_intra_instances_shard(
device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_mem_intra_instances&
instances)
{
add_device_operation_instances(
instances,
ck::util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwdDefault,
Intrawave>,
Shards,
ShardIndex>{});
add_device_operation_instances(
instances,
ck::util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwd1x1P0,
Intrawave>,
Shards,
ShardIndex>{});
add_device_operation_instances(
instances,
ck::util::filter_tuple_by_modulo_t<
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NGCDHW,
GKCZYX,
Empty_Tuple,
NGKDHW,
ConvFwd1x1S1P0,
Intrawave>,
Shards,
ShardIndex>{});
}
} // namespace ck::tensor_operation::device::instance

4
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_bias_clamp/CMakeLists.txt
View File

@@ -34,11 +34,7 @@ set(GROUPED_CONV3D_FWD
# WMMA CSHUFFLE V3
wmma/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
wmma/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instance.cpp
wmma/merged_groups/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
wmma/mem/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instance.cpp
wmma/mem/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instance.cpp
wmma/comp/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.cpp
wmma/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_fp16_instance.cpp
wmma/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_fp16_16x16_instance.cpp

62
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_bias_clamp/wmma/comp/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.cpp
View File

@@ -1,62 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_comp_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/host_utility/device_prop.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwdDefault,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwd1x1P0,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwd1x1S1P0,
Tuple<BF16>,
AddClamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

61
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_bias_clamp/wmma/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instance.cpp
View File

@@ -1,61 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwdDefault,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwd1x1P0,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwd1x1S1P0,
Tuple<BF16>,
AddClamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

64
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_bias_clamp/wmma/mem/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instance.cpp
View File

@@ -1,64 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwdDefault,
Interwave,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwd1x1P0,
Interwave,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwd1x1S1P0,
Interwave,
Tuple<BF16>,
AddClamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

64
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_bias_clamp/wmma/mem/device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instance.cpp
View File

@@ -1,64 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_bias_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
BF16,
BF16,
Tuple<BF16>,
BF16,
PassThrough,
PassThrough,
AddClamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwdDefault,
Intrawave,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwd1x1P0,
Intrawave,
Tuple<BF16>,
AddClamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<NDHWGK>,
NDHWGK,
ConvFwd1x1S1P0,
Intrawave,
Tuple<BF16>,
AddClamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

4
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_clamp/CMakeLists.txt
View File

@@ -34,11 +34,7 @@ set(GROUPED_CONV3D_FWD
# WMMA CSHUFFLE V3
wmma/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
wmma/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instance.cpp
wmma/merged_groups/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_merged_groups_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
wmma/mem/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instance.cpp
wmma/mem/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instance.cpp
wmma/comp/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.cpp
wmma/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_fp16_instance.cpp
wmma/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_fp16_16x16_instance.cpp

62
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_clamp/wmma/comp/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.cpp
View File

@@ -1,62 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_comp_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/host_utility/device_prop.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwdDefault,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwd1x1P0,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_comp_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwd1x1S1P0,
Tuple<>,
Clamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

61
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_clamp/wmma/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instance.cpp
View File

@@ -1,61 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwdDefault,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwd1x1P0,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_16x16_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwd1x1S1P0,
Tuple<>,
Clamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

64
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_clamp/wmma/mem/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instance.cpp
View File

@@ -1,64 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwdDefault,
Interwave,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwd1x1P0,
Interwave,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwd1x1S1P0,
Interwave,
Tuple<>,
Clamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

64
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_clamp/wmma/mem/device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instance.cpp
View File

@@ -1,64 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_wmma_cshufflev3_mem_instance.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_grouped_conv3d_fwd_clamp_wmma_cshufflev3_ndhwgc_gkzyxc_ndhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
BF16,
BF16,
Tuple<>,
BF16,
PassThrough,
PassThrough,
Clamp>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwdDefault,
Intrawave,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwd1x1P0,
Intrawave,
Tuple<>,
Clamp>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_wmma_cshufflev3_bf16_mem_instances<3,
NDHWGC,
GKZYXC,
Tuple<>,
NDHWGK,
ConvFwd1x1S1P0,
Intrawave,
Tuple<>,
Clamp>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck