diff options
| author | Michalis Spyrou <michalis.spyrou@arm.com> | 2020-10-21 00:04:14 +0100 |
|---|---|---|
| committer | Michalis Spyrou <michalis.spyrou@arm.com> | 2020-11-03 15:10:47 +0000 |
| commit | ebcebf1dee7f8314976b1e0cabd62b4cf893d765 (patch) | |
| tree | 95d3e691a0e88a3e213a1d30446a9224497f2055 /arm_compute/runtime/NEON/functions/NEGEMMLowpAssemblyMatrixMultiplyCore.h | |
| parent | da4b1b2055d96aaf73704eb9b0b82d74dc2d699c (diff) | |
| download | ComputeLibrary-ebcebf1dee7f8314976b1e0cabd62b4cf893d765.tar.gz | |
COMPMID-3638: Move NEON kernels
Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com>
Change-Id: Ieed3e4bc8be7fef80c90c5094599b477a56fc473
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4285
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'arm_compute/runtime/NEON/functions/NEGEMMLowpAssemblyMatrixMultiplyCore.h')
| -rw-r--r-- | arm_compute/runtime/NEON/functions/NEGEMMLowpAssemblyMatrixMultiplyCore.h | 23 |
1 files changed, 14 insertions, 9 deletions
diff --git a/arm_compute/runtime/NEON/functions/NEGEMMLowpAssemblyMatrixMultiplyCore.h b/arm_compute/runtime/NEON/functions/NEGEMMLowpAssemblyMatrixMultiplyCore.h index 9813b34661..961b1901e7 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMMLowpAssemblyMatrixMultiplyCore.h +++ b/arm_compute/runtime/NEON/functions/NEGEMMLowpAssemblyMatrixMultiplyCore.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2019 Arm Limited. + * Copyright (c) 2017-2020 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -24,7 +24,6 @@ #ifndef ARM_COMPUTE_NEGEMMLOWPASSEMBLYMATRIXMULTIPLYCORE_H #define ARM_COMPUTE_NEGEMMLOWPASSEMBLYMATRIXMULTIPLYCORE_H -#include "arm_compute/core/NEON/INEKernel.h" #include "arm_compute/runtime/IFunction.h" #include "arm_compute/runtime/IMemoryManager.h" #include "arm_compute/runtime/MemoryGroup.h" @@ -37,6 +36,9 @@ namespace arm_compute { // Forward declarations class ITensor; +class NEGEMMInterleave4x4Kernel; +class NEGEMMTranspose1xWKernel; +class NEGEMMLowpMatrixMultiplyKernel; /** Basic function to execute matrix multiply assembly kernels. */ class NEGEMMLowpAssemblyMatrixMultiplyCore : public IFunction @@ -44,6 +46,9 @@ class NEGEMMLowpAssemblyMatrixMultiplyCore : public IFunction public: /** Constructor */ NEGEMMLowpAssemblyMatrixMultiplyCore(std::shared_ptr<IMemoryManager> memory_manager = nullptr); + /** Destructor */ + ~NEGEMMLowpAssemblyMatrixMultiplyCore(); + /** Initialise the kernel's inputs, output * * @param[in] a First input tensor (Matrix A). Data type supported: U8, S8. @@ -57,13 +62,13 @@ public: void run() override; private: - MemoryGroup _memory_group; - NEGEMMAssemblyDispatch _asm_glue; - std::unique_ptr<INEKernel> _mm_kernel; - std::unique_ptr<INEKernel> _mtx_a_reshape_kernel; - std::unique_ptr<INEKernel> _mtx_b_reshape_kernel; - Tensor _tmp_a; - Tensor _tmp_b; + MemoryGroup _memory_group; + NEGEMMAssemblyDispatch _asm_glue; + std::unique_ptr<NEGEMMLowpMatrixMultiplyKernel> _mm_kernel; + std::unique_ptr<NEGEMMInterleave4x4Kernel> _mtx_a_reshape_kernel; + std::unique_ptr<NEGEMMTranspose1xWKernel> _mtx_b_reshape_kernel; + Tensor _tmp_a; + Tensor _tmp_b; }; } // namespace arm_compute #endif /*ARM_COMPUTE_NEGEMMLOWPASSEMBLYMATRIXMULTIPLYCORE_H */ |