diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_instance.cpp index f6ba6e1d89..5b55c8e15e 100644 --- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_instance.cpp @@ -34,27 +34,27 @@ using device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_instances = std::tuple< //##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| SrcDstVectorDim| DstScalar| //##########| | | | | | | | Operation| Operation| Operation| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | | PerVector| //##########| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 64, 64, 64, 4, 16, 32, 32, 2, 2, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 32, 4, 16, 32, 32, 2, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 32, 128, 4, 16, 32, 32, 1, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 64, 64, 32, 4, 16, 32, 32, 2, 1, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 64, 32, 64, 4, 16, 32, 32, 1, 2, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 8, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 8, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 8, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 16, 16, true, 7, 1> + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 64, 64, 64, 4, 16, 32, 32, 2, 2, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 32, 4, 16, 32, 32, 2, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 32, 128, 4, 16, 32, 32, 1, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 64, 64, 32, 4, 16, 32, 32, 2, 1, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 64, 32, 64, 4, 16, 32, 32, 1, 2, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 1, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 1, 2, 16, true, 7, 1> // clang-format on >; diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_instance.cpp index 892354ade8..9517e4577e 100644 --- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_instance.cpp @@ -34,27 +34,27 @@ using device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_instances = std::tuple< //##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| SrcDstVectorDim| DstScalar| //##########| | | | | | | | Operation| Operation| Operation| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | | PerVector| //##########| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 64, 64, 64, 4, 16, 32, 32, 2, 2, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 32, 4, 16, 32, 32, 2, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 32, 128, 4, 16, 32, 32, 1, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 64, 64, 32, 4, 16, 32, 32, 2, 1, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 64, 32, 64, 4, 16, 32, 32, 1, 2, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 8, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 8, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 8, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1> + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 64, 64, 64, 4, 16, 32, 32, 2, 2, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 32, 4, 16, 32, 32, 2, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 32, 128, 4, 16, 32, 32, 1, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 64, 64, 32, 4, 16, 32, 32, 2, 1, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 64, 32, 64, 4, 16, 32, 32, 1, 2, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Col, Col, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, 7, 1> // clang-format on >; diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_instance.cpp index fb5deb0a0c..43b9124406 100644 --- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_instance.cpp @@ -34,27 +34,27 @@ using device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_instances = std::tuple< //##########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| SrcDstVectorDim| DstScalar| //##########| | | | | | | | Operation| Operation| Operation| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | | PerVector| //##########| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 64, 64, 64, 4, 16, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 32, 4, 16, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 32, 128, 4, 16, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 64, 64, 32, 4, 16, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 64, 32, 64, 4, 16, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1>, - DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 16, 16, true, 7, 1> + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 64, 64, 64, 4, 16, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 32, 4, 16, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 32, 128, 4, 16, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 64, 64, 32, 4, 16, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 64, 32, 64, 4, 16, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, true, S<4, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 4, 16, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 4, 16, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 4, 16, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 4, 16, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 4, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 16, true, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 4, 16, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 4, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, true, 7, 1>, + DeviceBatchedGemmXdl< AData, BData, CData, AccData, Row, Row, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 4, 16, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 16, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, true, 7, 1> // clang-format on >;