From 4b90865ab985d571f70c60583cdfb8c7a65f1670 Mon Sep 17 00:00:00 2001 From: Gian Marco Iodice Date: Thu, 18 Oct 2018 10:21:02 +0100 Subject: COMPMID-1413 - Improve the performance of GEMMLowp with 8 bit dot product on OpenCL COMPMID-1424 - Add dot product support for CLDepthwise QASYMM8 3x3 NHWC non-unit stride With this patch we are able to improve the performance of MobileNet v1-qasymm8 by 37 % Tried to use the dot product instruction in CLDepthwise QASYMM8 3x3 NHWC non-unit stride but I have not seen any benefit (maybe because we have few arithemtic operation and we do not have more load instructions). However Depthwise convolution has been improved by 30% Change-Id: Id768a99c2e53a04276707e427af5d0ec93419ada Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/155082 Tested-by: bsgcomp Reviewed-by: Georgios Pinitas --- .../core/CL/kernels/CLGEMMLowpOffsetContributionKernel.h | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'arm_compute/core/CL/kernels/CLGEMMLowpOffsetContributionKernel.h') diff --git a/arm_compute/core/CL/kernels/CLGEMMLowpOffsetContributionKernel.h b/arm_compute/core/CL/kernels/CLGEMMLowpOffsetContributionKernel.h index 871b97c1d7..e6b79176b5 100644 --- a/arm_compute/core/CL/kernels/CLGEMMLowpOffsetContributionKernel.h +++ b/arm_compute/core/CL/kernels/CLGEMMLowpOffsetContributionKernel.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017 ARM Limited. + * Copyright (c) 2017-2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -58,16 +58,18 @@ public: CLGEMMLowpOffsetContributionKernel &operator=(CLGEMMLowpOffsetContributionKernel &&) = default; /** Initialise the kernel's input and output. * - * @param[in, out] mm_result Input tensor containing the result of @ref CLGEMMLowpMatrixMultiplyKernel. Data type supported: S32 + * @param[in, out] mm_result Input tensor containing the result of @ref CLGEMMLowpMatrixMultiplyKernel * @param[in] vector_sum_col Input row-vector of sums of all the entries in each column of matrix B. * Note: vector_sum_col can be a nullptr in case a_offset = 0. Data type supported: same as @p mm_result * @param[in] vector_sum_row Input row-vector of sums of all the entries in each row of matrix A. * Note: vector_sum_row can be a nullptr in case b_offset = 0. Data type supported: same as @p mm_result + * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the addition of biases is not required. + * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. * @param[in] k Number of matrix A columns or Matrix B rows * @param[in] a_offset Offset to be added to each element of the matrix A. * @param[in] b_offset Offset to be added to each element of the matrix B. */ - void configure(ICLTensor *mm_result, const ICLTensor *vector_sum_col, const ICLTensor *vector_sum_row, int32_t k, int32_t a_offset, int32_t b_offset); + void configure(ICLTensor *mm_result, const ICLTensor *vector_sum_col, const ICLTensor *vector_sum_row, const ICLTensor *bias, int32_t k, int32_t a_offset, int32_t b_offset); /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMLowpOffsetContributionKernel * * @param[in] mm_result Input tensor containing the result of @ref CLGEMMLowpOffsetContributionKernel. Data type supported: S32 @@ -75,12 +77,14 @@ public: * Note: vector_sum_col can be a nullptr in case a_offset = 0. Data type supported: same as @p mm_result * @param[in] vector_sum_row Input row-vector of sums of all the entries in each row of matrix A. * Note: vector_sum_row can be a nullptr in case b_offset = 0. Data type supported: same as @p mm_result + * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the addition of biases is not required. + * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. * @param[in] a_offset Offset to be added to each element of the matrix A. * @param[in] b_offset Offset to be added to each element of the matrix B. * * @return a status */ - static Status validate(const ITensorInfo *mm_result, const ITensorInfo *vector_sum_col, const ITensorInfo *vector_sum_row, int32_t a_offset, int32_t b_offset); + static Status validate(const ITensorInfo *mm_result, const ITensorInfo *vector_sum_col, const ITensorInfo *vector_sum_row, const ITensorInfo *bias, int32_t a_offset, int32_t b_offset); // Inherited methods overridden: void run(const Window &window, cl::CommandQueue &queue) override; @@ -89,6 +93,7 @@ private: const ICLTensor *_vector_sum_col; const ICLTensor *_vector_sum_row; ICLTensor *_mm_result; + const ICLTensor *_bias; }; } // namespace arm_compute -- cgit v1.2.1