From b57d7e857c8aebf0c9cca4c604c92b4f41c1bae6 Mon Sep 17 00:00:00 2001 From: qin letao Date: Tue, 25 Feb 2025 09:03:36 +0000 Subject: [PATCH] split file for compile --- ...ultiple_d_xdl_cshuffle_v3_b_preshuffle.hpp | 4 + ...m_xdl_cshuffle_v3_multi_d_b_preshuffle.hpp | 8 +- ...mm_multiply_multiply_weight_preshuffle.hpp | 231 +++++++++++++++++- .../CMakeLists.txt | 37 +++ ..._mfma16x16_mn_compute_default_instance.cpp | 5 - ...16_mn_compute_default_instance_padding.cpp | 33 +++ ...16_mk_mfma_mn_compute_default_instance.cpp | 5 - ...ma_mn_compute_default_instance_padding.cpp | 33 +++ ...f8_bf16_mk_mfma_mn_p1_default_instance.cpp | 6 - ...mk_mfma_mn_p1_default_instance_padding.cpp | 34 +++ ...bf16_mk_mfma_mn_p1_default_instance_v2.cpp | 6 - ...mfma_mn_p1_default_instance_v2_padding.cpp | 34 +++ ...f8_bf16_mk_mfma_mn_p2_default_instance.cpp | 6 - ...mk_mfma_mn_p2_default_instance_padding.cpp | 34 +++ ...bf16_mk_mfma_mn_p2_default_instance_v2.cpp | 6 - ...mfma_mn_p2_default_instance_v2_padding.cpp | 34 +++ ...f8_bf16_mk_mfma_mn_p3_default_instance.cpp | 6 - ...mk_mfma_mn_p3_default_instance_padding.cpp | 34 +++ ...bf16_mk_mfma_mn_p3_default_instance_v2.cpp | 6 - ...mfma_mn_p3_default_instance_v2_padding.cpp | 34 +++ ..._mfma16x16_mn_compute_default_instance.cpp | 5 - ...16_mn_compute_default_instance_padding.cpp | 33 +++ ...16_mk_mfma_mn_compute_default_instance.cpp | 5 - ...ma_mn_compute_default_instance_padding.cpp | 33 +++ ..._f8_f16_mk_mfma_mn_p1_default_instance.cpp | 6 - ...mk_mfma_mn_p1_default_instance_padding.cpp | 34 +++ ..._f16_mk_mfma_mn_p1_default_instance_v2.cpp | 6 - ...mfma_mn_p1_default_instance_v2_padding.cpp | 34 +++ ..._f8_f16_mk_mfma_mn_p2_default_instance.cpp | 6 - ...mk_mfma_mn_p2_default_instance_padding.cpp | 34 +++ ..._f16_mk_mfma_mn_p2_default_instance_v2.cpp | 6 - ...mfma_mn_p2_default_instance_v2_padding.cpp | 34 +++ ..._f8_f16_mk_mfma_mn_p3_default_instance.cpp | 6 - ...mk_mfma_mn_p3_default_instance_padding.cpp | 34 +++ ..._f16_mk_mfma_mn_p3_default_instance_v2.cpp | 6 - ...mfma_mn_p3_default_instance_v2_padding.cpp | 34 +++ 36 files changed, 815 insertions(+), 97 deletions(-) create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_padding.cpp create mode 100644 library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2_padding.cpp diff --git a/include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3_b_preshuffle.hpp b/include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3_b_preshuffle.hpp index 215aaef278..448b745f40 100644 --- a/include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3_b_preshuffle.hpp +++ b/include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3_b_preshuffle.hpp @@ -526,6 +526,10 @@ struct DeviceGemmMultiD_Xdl_CShuffle_V3_BPreshuffle return false; } + if(GemmSpec == GemmSpecialization::KPadding && arg.KBatch > 2){ + return false; + } + if(arg.N % NPerBlock != 0 || (arg.K % KPerBlock != 0 && GemmSpec != GemmSpecialization::KPadding)) { diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d_b_preshuffle.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d_b_preshuffle.hpp index 0ecf70f5b6..8b6af69b0d 100644 --- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d_b_preshuffle.hpp +++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d_b_preshuffle.hpp @@ -350,7 +350,7 @@ struct GridwiseGemmMultiD_xdl_cshuffle_v3_b_preshuffle } __host__ __device__ static auto - MakeBGridDescriptor_Preshuffled(index_t N0, index_t K0, index_t KPad) + MakeBGridDescriptor_Preshuffled(index_t N0, index_t K0, index_t KPad, index_t KBatch) { using GemmSpecialization = tensor_operation::device::GemmSpecialization; // if K padding @@ -364,7 +364,7 @@ struct GridwiseGemmMultiD_xdl_cshuffle_v3_b_preshuffle make_tuple( NWave * K0 * NkSwizzleNumber, K0 * NkSwizzleNumber, NkSwizzleNumber, I1)); - auto K0new = CalculateBK0Shuffled(KPad); + auto K0new = CalculateBK0Shuffled(KPad * KBatch); return transform_tensor_descriptor( b_grid_desc_raw, @@ -1162,7 +1162,7 @@ struct GridwiseGemmMultiD_xdl_cshuffle_v3_b_preshuffle problem.M, problem.MPadded, problem.K, problem.KPadded, problem.StrideA, problem.AK0); const auto b_grid_desc_bpreshuffled = MakeBGridDescriptor_Preshuffled( - problem.BN0Shuffled, problem.BK0Shuffled, problem.KPadded); + problem.BN0Shuffled, problem.BK0Shuffled, problem.KPadded, problem.KBatch); const auto c_grid_desc_m_n = MakeCGridDescriptor_M_N( problem.M, problem.MPadded, problem.N, problem.NPadded, problem.StrideC); @@ -1599,7 +1599,7 @@ struct GridwiseGemmMultiD_xdl_cshuffle_v3_b_preshuffle problem.M, problem.MPadded, problem.K, problem.KPadded, problem.StrideA, problem.AK0); const auto b_grid_desc_bpreshuffled = MakeBGridDescriptor_Preshuffled( - problem.BN0Shuffled, problem.BK0Shuffled, problem.KPadded); + problem.BN0Shuffled, problem.BK0Shuffled, problem.KPadded, problem.KBatch); const auto c_grid_desc_m_n = MakeCGridDescriptor_M_N( problem.M, problem.MPadded, problem.N, problem.NPadded, problem.StrideC); diff --git a/library/include/ck/library/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle.hpp b/library/include/ck/library/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle.hpp index 90da3ad0fe..0d2b2116f7 100644 --- a/library/include/ck/library/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle.hpp +++ b/library/include/ck/library/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle.hpp @@ -115,6 +115,103 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_m PassThrough, MultiplyMultiply>>>& instances); +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); #endif #if(defined(CK_ENABLE_BF16) || defined(CK_ENABLE_FP8)) @@ -215,6 +312,104 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_ PassThrough, MultiplyMultiply>>>& instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& + instances); #endif template > op_ptrs; -// TODO: Add MFMA layout into tensor layout + // TODO: Add MFMA layout into tensor layout #if(defined(CK_ENABLE_F16) || defined(CK_ENABLE_FP8)) if constexpr(is_same_v && is_same_v && is_same_v) @@ -277,6 +472,23 @@ struct DeviceOperationInstanceFactory< add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances( op_ptrs); + + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_padding( + op_ptrs); + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_padding( + op_ptrs); + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_padding( + op_ptrs); + + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_v2_padding( + op_ptrs); + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_v2_padding( + op_ptrs); + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_v2_padding( + op_ptrs); + + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances_padding( + op_ptrs); } } #endif @@ -304,6 +516,23 @@ struct DeviceOperationInstanceFactory< add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances( op_ptrs); + + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_padding( + op_ptrs); + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_padding( + op_ptrs); + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_padding( + op_ptrs); + + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_v2_padding( + op_ptrs); + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_v2_padding( + op_ptrs); + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_v2_padding( + op_ptrs); + + add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances_padding( + op_ptrs); } } #endif diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/CMakeLists.txt b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/CMakeLists.txt index 943b2bf4c7..19d07b795e 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/CMakeLists.txt +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/CMakeLists.txt @@ -19,6 +19,24 @@ list(APPEND GEMM_MULTIPLY_MULTIPLY_WEIGHT_PRESHUFFLE_INSTANCES device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2.cpp device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance.cpp device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance.cpp + + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_padding.cpp + + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_padding.cpp + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_padding.cpp ) set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") @@ -39,4 +57,23 @@ set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_ set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") + +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") + +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") +set_source_files_properties(device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1") + add_instance_library(device_gemm_multiply_multiply_weight_preshuffle_instance ${GEMM_MULTIPLY_MULTIPLY_WEIGHT_PRESHUFFLE_INSTANCES}) diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance.cpp index 36ed181f7a..71383f5dc1 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance.cpp @@ -25,11 +25,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_ instances, device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_instances< GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_instances< - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_padding.cpp new file mode 100644 index 0000000000..1bcdf9f40d --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_padding.cpp @@ -0,0 +1,33 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_instances< + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance.cpp index 1dde4e5606..e3ff079d99 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance.cpp @@ -25,11 +25,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma1 instances, device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_instances< GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_instances< - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_padding.cpp new file mode 100644 index 0000000000..cefa7288da --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_padding.cpp @@ -0,0 +1,33 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_instances< + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance.cpp index e1202518ac..6e9b3ea172 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_ device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_instances< v1, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_instances< - v1, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_padding.cpp new file mode 100644 index 0000000000..823d2869b8 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_instances< + v1, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2.cpp index 488e32e020..cc543b19c1 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_ device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_instances< v2, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_instances< - v2, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2_padding.cpp new file mode 100644 index 0000000000..79f7410141 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_instances< + v2, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance.cpp index 211beb3ea8..8557d0c80e 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_ device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_instances< v1, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_instances< - v1, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_padding.cpp new file mode 100644 index 0000000000..43e6596bb8 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_instances< + v1, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2.cpp index 8a65443f0b..9fcce478e7 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_ device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_instances< v2, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_instances< - v2, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2_padding.cpp new file mode 100644 index 0000000000..f291d27b5c --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_instances< + v2, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance.cpp index d96514844b..84c2c70e35 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_ device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_instances< v1, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_instances< - v1, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_padding.cpp new file mode 100644 index 0000000000..a0ede75177 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_instances< + v1, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2.cpp index a4e4cf43c4..0933b1fe18 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_ device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_instances< v2, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_instances< - v2, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2_padding.cpp new file mode 100644 index 0000000000..80af299827 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + BF16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_instances< + v2, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance.cpp index fd54667b5b..05529f9cdd 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance.cpp @@ -25,11 +25,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16 instances, device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_instances< GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_instances< - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_padding.cpp new file mode 100644 index 0000000000..cd102ddf88 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_padding.cpp @@ -0,0 +1,33 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_instances< + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance.cpp index 1fba2f6b28..c123a0fdd8 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance.cpp @@ -25,11 +25,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_m instances, device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_instances< GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_instances< - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_padding.cpp new file mode 100644 index 0000000000..8223d818aa --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_padding.cpp @@ -0,0 +1,33 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_instances< + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance.cpp index 57155e26ef..cb15688e6f 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_m device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_instances< v2, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_instances< - v2, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_padding.cpp new file mode 100644 index 0000000000..4adb814ba0 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_instances< + v2, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2.cpp index e5dc8e3de5..c5a8448b59 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_m device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_instances< v1, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_instances< - v1, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2_padding.cpp new file mode 100644 index 0000000000..c49422f7b9 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_instances< + v1, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance.cpp index 3dab48d954..c9ab9c1071 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_m device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_instances< v1, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_instances< - v1, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_padding.cpp new file mode 100644 index 0000000000..6a55f67c2c --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_instances< + v1, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2.cpp index 192f45f99b..bb83bacb35 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_m device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_instances< v2, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_instances< - v2, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2_padding.cpp new file mode 100644 index 0000000000..ea6376eeff --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_instances< + v2, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance.cpp index 989f6a9669..fb43347ceb 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_m device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_instances< v1, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_instances< - v1, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_padding.cpp new file mode 100644 index 0000000000..c9c371d0e9 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_instances< + v1, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2.cpp index cee44f8501..c8ff03d6ef 100644 --- a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2.cpp +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2.cpp @@ -26,12 +26,6 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_m device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_instances< v2, GemmDefault>{}); - - add_device_operation_instances( - instances, - device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_instances< - v2, - GemmKPadding>{}); } } // namespace instance diff --git a/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2_padding.cpp b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2_padding.cpp new file mode 100644 index 0000000000..68e8e57d47 --- /dev/null +++ b/library/src/tensor_operation_instance/gpu/gemm_multiply_multiply_weight_preshuffle/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16/device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2_padding.cpp @@ -0,0 +1,34 @@ +// SPDX-License-Identifier: MIT +// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved. + +#include "device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn.hpp" + +namespace ck { +namespace tensor_operation { +namespace device { +namespace instance { + +void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_v2_padding( + std::vector, + Row, + F8, + F8, + Tuple, + F16, + PassThrough, + PassThrough, + MultiplyMultiply>>>& instances) +{ + add_device_operation_instances( + instances, + device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_instances< + v2, + GemmKPadding>{}); +} + +} // namespace instance +} // namespace device +} // namespace tensor_operation +} // namespace ck