From fc004494c9cde94fe7b5d67199faf2596e223cf3 Mon Sep 17 00:00:00 2001 From: Pablo Tello Date: Fri, 23 Mar 2018 11:40:05 +0000 Subject: COMPMID-959: Removed Interleave blocked kernel. Change-Id: I775eecbc39da583aae2eb4e033c5930dfc402899 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/125684 Tested-by: Jenkins Reviewed-by: Georgios Pinitas --- arm_compute/core/NEON/NEKernels.h | 1 - .../NEON/kernels/NEGEMMInterleaveBlockedKernel.h | 79 ---------------------- .../NEON/functions/NEGEMMLowpMatrixMultiplyCore.h | 3 +- 3 files changed, 1 insertion(+), 82 deletions(-) delete mode 100644 arm_compute/core/NEON/kernels/NEGEMMInterleaveBlockedKernel.h (limited to 'arm_compute') diff --git a/arm_compute/core/NEON/NEKernels.h b/arm_compute/core/NEON/NEKernels.h index 7ec74eaccd..31f4881ef5 100644 --- a/arm_compute/core/NEON/NEKernels.h +++ b/arm_compute/core/NEON/NEKernels.h @@ -62,7 +62,6 @@ #include "arm_compute/core/NEON/kernels/NEFloorKernel.h" #include "arm_compute/core/NEON/kernels/NEGEMMAssemblyBaseKernel.h" #include "arm_compute/core/NEON/kernels/NEGEMMInterleave4x4Kernel.h" -#include "arm_compute/core/NEON/kernels/NEGEMMInterleaveBlockedKernel.h" #include "arm_compute/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.h" #include "arm_compute/core/NEON/kernels/NEGEMMLowpOffsetContributionKernel.h" #include "arm_compute/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel.h" diff --git a/arm_compute/core/NEON/kernels/NEGEMMInterleaveBlockedKernel.h b/arm_compute/core/NEON/kernels/NEGEMMInterleaveBlockedKernel.h deleted file mode 100644 index f5a1121b04..0000000000 --- a/arm_compute/core/NEON/kernels/NEGEMMInterleaveBlockedKernel.h +++ /dev/null @@ -1,79 +0,0 @@ -/* - * Copyright (c) 2017-2018 ARM Limited. - * - * SPDX-License-Identifier: MIT - * - * Permission is hereby granted, free of charge, to any person obtaining a copy - * of this software and associated documentation files (the "Software"), to - * deal in the Software without restriction, including without limitation the - * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or - * sell copies of the Software, and to permit persons to whom the Software is - * furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in all - * copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE - * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, - * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE - * SOFTWARE. - */ -#ifndef __ARM_COMPUTE_NEGEMMINTERLEAVEBLOCKEDKERNEL_H__ -#define __ARM_COMPUTE_NEGEMMINTERLEAVEBLOCKEDKERNEL_H__ - -#include "arm_compute/core/NEON/INESimpleKernel.h" - -namespace arm_compute -{ -class ITensor; - -/** NEON kernel to interleave the elements of a matrix - * - * Interleave_Blocked copies a block of values at a time instead of just one. The main use of this is the gemmlowp with the "dot product" - * instruction, where each operation consumes 4 values, so we need to copy blocks of 4 values. - * - */ -class NEGEMMInterleaveBlockedKernel : public INESimpleKernel -{ -public: - const char *name() const override - { - return "NEGEMMInterleaveBlockedKernel"; - } - /* Constructor */ - NEGEMMInterleaveBlockedKernel(); - /** Initialise the kernel's input and output. - * - * @param[in] input Input tensor. Data types supported: U8 - * @param[out] output Output tensor which stores the interleaved matrix. Data type supported: same as @p input. - * @param[in] block_height The height of the blocks to be interleaved. - * @param[in] block_width The width of the blocks to be interleaved. - * @param[in] transpose True if transpose operation must be performed, false otherwise. - */ - void configure(const ITensor *input, ITensor *output, unsigned int block_height, unsigned int block_width, bool transpose); - /** Static function to check if given info will lead to a valid configuration of @ref NEGEMMInterleaveBlockedKernel - * - * @param[in] input Input tensor. Data types supported: U8 - * @param[in] output Output tensor which stores the interleaved matrix. Data type supported: same as @p input. - * @param[in] block_height The height of the blocks to be interleaved. - * @param[in] block_width The width of the blocks to be interleaved. - * @param[in] transpose True if transpose operation must be performed, false otherwise. - * - * @return a status - */ - static Status validate(const ITensorInfo *input, const ITensorInfo *output, unsigned int block_height, unsigned int block_width, bool transpose); - - // Inherited methods overridden: - void run(const Window &window, const ThreadInfo &info) override; - -private: - unsigned int _block_height; - unsigned int _block_width; - bool _transpose; -}; - -} // namespace arm_compute -#endif /*__ARM_COMPUTE_NEGEMMINTERLEAVEBLOCKEDKERNEL_H__*/ diff --git a/arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h b/arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h index 95776f829a..5e05e93855 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h +++ b/arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h @@ -48,7 +48,6 @@ class ITensor; * * otherwise if the DOT product instruction is available: * - * -# @ref NEGEMMInterleaveBlockedKernel * -# @ref NEGEMMLowpOffsetContributionKernel * */ @@ -85,7 +84,7 @@ public: */ static Status validate(const ITensorInfo *a, const ITensorInfo *b, const ITensorInfo *output, const GEMMInfo &gemm_info = GEMMInfo()); - // Inherited methods overridden: + // Inherited methods overridden void run() override; private: -- cgit v1.2.1