diff options
author | Pablo Tello <pablo.tello@arm.com> | 2018-02-23 13:43:50 +0000 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:49:16 +0000 |
commit | eb82fd2aa786715c3b6a941dc6d6deac4ce8e2a0 (patch) | |
tree | 42cca378eed97c07348f28e1ec708d9c7ed531ce /arm_compute/runtime/NEON/functions/NEGEMM.h | |
parent | 8df6c452820719d201ee79596cde8445c2071db5 (diff) | |
download | ComputeLibrary-eb82fd2aa786715c3b6a941dc6d6deac4ce8e2a0.tar.gz |
COMPMID-881: RSH new arm_gemm interface.
Change-Id: I1e2a1a77097d8017c274af3f97eba6964f80f5fa
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/122592
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Diffstat (limited to 'arm_compute/runtime/NEON/functions/NEGEMM.h')
-rw-r--r-- | arm_compute/runtime/NEON/functions/NEGEMM.h | 29 |
1 files changed, 15 insertions, 14 deletions
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 <memory> 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<NEGEMMAssemblyBaseKernel> _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__ */ |