diff --git a/library/src/tensor_operation_instance/gpu/gemm_multi_abd/device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_kn_mn_common.hpp b/library/src/tensor_operation_instance/gpu/gemm_multi_abd/device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_kn_mn_common.hpp index 41118b4849..8d4c45ae82 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multi_abd/device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_kn_mn_common.hpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multi_abd/device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_kn_mn_common.hpp @@ -79,8 +79,8 @@ using device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_kn_mn_comp_instances = std::tup //###################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | DeviceGemmMultipleABD_Wmma_CShuffleV3< AsLayout, BsLayout, DsLayout, ELayout, AsDataType, BsDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 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<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>, DeviceGemmMultipleABD_Wmma_CShuffleV3< AsLayout, BsLayout, DsLayout, ELayout, AsDataType, BsDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 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<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>, - DeviceGemmMultipleABD_Wmma_CShuffleV3< AsLayout, BsLayout, DsLayout, ELayout, AsDataType, BsDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 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<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>, - DeviceGemmMultipleABD_Wmma_CShuffleV3< AsLayout, BsLayout, DsLayout, ELayout, AsDataType, BsDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 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<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1> + DeviceGemmMultipleABD_Wmma_CShuffleV3< AsLayout, BsLayout, DsLayout, ELayout, AsDataType, BsDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 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<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, BlkGemmPipeSched, BlockGemmPipelineVersion::v1>, + DeviceGemmMultipleABD_Wmma_CShuffleV3< AsLayout, BsLayout, DsLayout, ELayout, AsDataType, BsDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 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<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, BlkGemmPipeSched, BlockGemmPipelineVersion::v1> // clang-format on >; diff --git a/library/src/tensor_operation_instance/gpu/gemm_multi_abd/device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_nk_mn_common.hpp b/library/src/tensor_operation_instance/gpu/gemm_multi_abd/device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_nk_mn_common.hpp index 26ef35bcf8..0c2a34fbf8 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multi_abd/device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_nk_mn_common.hpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multi_abd/device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_nk_mn_common.hpp @@ -76,7 +76,7 @@ using device_gemm_wmma_multi_abd_bf16_i8_bf16_mk_nk_mn_comp_instances = std::tup //###################################| | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | | Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| | | | | //###################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | DeviceGemmMultipleABD_Wmma_CShuffleV3< AsLayout, BsLayout, DsLayout, ELayout, AsDataType, BsDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 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>, S<8, 8, 8>, BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>, - DeviceGemmMultipleABD_Wmma_CShuffleV3< AsLayout, BsLayout, DsLayout, ELayout, AsDataType, BsDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 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>, S<8, 8, 8>, BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1> + DeviceGemmMultipleABD_Wmma_CShuffleV3< AsLayout, BsLayout, DsLayout, ELayout, AsDataType, BsDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementOp, BElementOp, CDEElementOp, GemmSpec, 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>, S<8, 8, 8>, BlkGemmPipeSched, BlockGemmPipelineVersion::v1> // clang-format on >; } // namespace instance