From eb82fd2aa786715c3b6a941dc6d6deac4ce8e2a0 Mon Sep 17 00:00:00 2001 From: Pablo Tello Date: Fri, 23 Feb 2018 13:43:50 +0000 Subject: COMPMID-881: RSH new arm_gemm interface. Change-Id: I1e2a1a77097d8017c274af3f97eba6964f80f5fa Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/122592 Tested-by: Jenkins Reviewed-by: Anthony Barbier --- arm_compute/runtime/NEON/functions/NEGEMM.h | 29 +++++++++++++++-------------- 1 file changed, 15 insertions(+), 14 deletions(-) (limited to 'arm_compute/runtime/NEON/functions/NEGEMM.h') diff --git a/arm_compute/runtime/NEON/functions/NEGEMM.h b/arm_compute/runtime/NEON/functions/NEGEMM.h index f2b6ef77bd..5279995be4 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMM.h +++ b/arm_compute/runtime/NEON/functions/NEGEMM.h @@ -25,7 +25,6 @@ #define __ARM_COMPUTE_NEGEMM_H__ #include "arm_compute/core/NEON/kernels/NEFillBorderKernel.h" -#include "arm_compute/core/NEON/kernels/NEGEMMAssemblyBaseKernel.h" #include "arm_compute/core/NEON/kernels/NEGEMMInterleave4x4Kernel.h" #include "arm_compute/core/NEON/kernels/NEGEMMMatrixAdditionKernel.h" #include "arm_compute/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.h" @@ -35,6 +34,8 @@ #include "arm_compute/runtime/MemoryGroup.h" #include "arm_compute/runtime/Tensor.h" +#include "arm_compute/runtime/NEON/AssemblyHelper.h" + #include namespace arm_compute @@ -73,19 +74,19 @@ public: void run() override; private: - MemoryGroup _memory_group; - NEGEMMInterleave4x4Kernel _interleave_kernel; - NEGEMMTranspose1xWKernel _transpose_kernel; - NEGEMMMatrixMultiplyKernel _mm_kernel; - std::unique_ptr _mm_optimised_kernel; - NEGEMMMatrixAdditionKernel _ma_kernel; - Tensor _tmp_a; - Tensor _tmp_b; - Tensor _workspace; - bool _run_vector_matrix_multiplication; - bool _run_addition; - bool _is_first_run; - bool _reshape_b_only_on_first_run; + MemoryGroup _memory_group; + NEGEMMInterleave4x4Kernel _interleave_kernel; + NEGEMMTranspose1xWKernel _transpose_kernel; + NEGEMMMatrixMultiplyKernel _mm_kernel; + AssemblyKernelGlueF32 _asm_glue; + NEGEMMMatrixAdditionKernel _ma_kernel; + Tensor _tmp_a; + Tensor _tmp_b; + Tensor _workspace; + bool _run_vector_matrix_multiplication; + bool _run_addition; + bool _is_first_run; + bool _reshape_b_only_on_first_run; }; } #endif /*__ARM_COMPUTE_NEGEMM_H__ */ -- cgit v1.2.1