From db63b9c431264c9ef612e69a66b13a07b8f54786 Mon Sep 17 00:00:00 2001 From: Gian Marco Iodice Date: Thu, 17 Jan 2019 09:47:04 +0000 Subject: COMPMID-1698: Implementing CLGEMMLowpMatrixMultiplyReshapedKernel Change-Id: Ia4db21b394a0b9235393202ce3c00b11cceb94ea Reviewed-on: https://review.mlplatform.org/568 Tested-by: Arm Jenkins Reviewed-by: Michele Di Giorgio --- arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h') diff --git a/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h b/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h index 72d91070f8..4345ff267b 100644 --- a/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h +++ b/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h @@ -25,6 +25,7 @@ #define __ARM_COMPUTE_CLGEMMLOWPMATRIXMULTIPLYCORE_H__ #include "arm_compute/core/CL/kernels/CLGEMMLowpMatrixMultiplyKernel.h" +#include "arm_compute/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMLowpOffsetContributionKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMLowpOffsetContributionOutputStageKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMLowpReductionKernel.h" @@ -43,7 +44,8 @@ class ICLTensor; * * -# @ref CLGEMMReshapeLHSMatrixKernel (if the output tensor is a matrix) * -# @ref CLGEMMReshapeRHSMatrixKernel (if the output tensor is a matrix) - * -# @ref CLGEMMLowpMatrixMultiplyKernel + * -# @ref CLGEMMLowpMatrixMultiplyKernel (if the input matrix is a vector or for Midgard architectures) + * -# @ref CLGEMMLowpMatrixMultiplyReshapedKernel (if the input matrix is not a vector and if the GPU architecture is not Midgard) * -# @ref CLGEMMLowpMatrixAReductionKernel (if the offset of matrix B is not 0) * -# @ref CLGEMMLowpMatrixBReductionKernel (if the offset of matrix A is not 0) * -# @ref CLGEMMLowpOffsetContributionKernel (if gemm_info.gemmlowp_output_stage == NONE) @@ -101,6 +103,7 @@ public: private: CLMemoryGroup _memory_group; CLGEMMLowpMatrixMultiplyKernel _mm_kernel; + CLGEMMLowpMatrixMultiplyReshapedKernel _mm_reshaped_kernel; CLGEMMReshapeLHSMatrixKernel _mtx_a_reshape_kernel; CLGEMMReshapeRHSMatrixKernel _mtx_b_reshape_kernel; CLGEMMLowpMatrixAReductionKernel _mtx_a_reduction_kernel; @@ -115,10 +118,10 @@ private: const ICLTensor *_original_b; int32_t _a_offset; int32_t _b_offset; - bool _is_interleaved_transposed; + bool _is_gemm_reshaped; bool _reshape_b_only_on_first_run; bool _is_prepared; bool _fuse_output_stage; }; } -#endif /*__ARM_COMPUTE_CLGEMMLOWPMATRIXMULTIPLYCORE_H__ */ +#endif /*__ARM_COMPUTE_CLGEMMLOWPMATRIXMULTIPLYCORE_H__ */ \ No newline at end of file -- cgit v1.2.1