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/NEGEMMConvolutionLayer.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/NEGEMMConvolutionLayer.h')
-rw-r--r-- | arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h | 11 |
1 files changed, 2 insertions, 9 deletions
diff --git a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h index ac5f4caa78..4ae8ee1fb3 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h +++ b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h @@ -36,6 +36,7 @@ #include "arm_compute/core/NEON/kernels/NEWeightsReshapeKernel.h" #include "arm_compute/core/Types.h" #include "arm_compute/runtime/MemoryGroup.h" +#include "arm_compute/runtime/NEON/AssemblyHelper.h" #include "arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h" #include "arm_compute/runtime/NEON/functions/NEGEMMLowpOutputStage.h" #include "arm_compute/runtime/Tensor.h" @@ -149,22 +150,14 @@ private: * @param[in] reshape_info (Optional) GEMM reshape info. If is_interleaved_transposed = true, this object must contain the information to understand how the matrix A and matrix B have been reshaped */ void configure_mm(const ITensor *input, const ITensor *weights, ITensor *output, bool is_interleaved, const GEMMReshapeInfo &reshape_info = GEMMReshapeInfo()); - /** Prepare the appropriate assembly optimized kernel - * - * @param[in] ci CPU information - * @param[in] M M parameter of matrix multiplication - * @param[in] N N parameter of matrix multiplication - * @param[in] K K parameter of matrix multiplication - */ - void configure_asm_mm(const struct CPUInfo &ci, int M, int N, int K); private: + AssemblyKernelGlueF32 _asm_glue; MemoryGroup _memory_group; NEIm2ColKernel _input_im2col_kernel; NEGEMMInterleave4x4Kernel _input_interleave_kernel; NEConvolutionLayerReshapeWeights _reshape_weights; NEGEMMMatrixMultiplyKernel _mm_kernel; - std::unique_ptr<NEGEMMAssemblyBaseKernel> _mm_optimised_kernel; NEGEMMLowpMatrixMultiplyCore _mm_gemmlowp; NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint _gemmlowp_output_stage; NECol2ImKernel _output_col2im_kernel; |