From 861f0db548befac0cd5fb28fe2fa8ea1828c715d Mon Sep 17 00:00:00 2001 From: Michalis Spyrou Date: Mon, 26 Feb 2018 16:47:58 +0000 Subject: COMPMID-941 Add NEON broadcast multiply support Change-Id: I1f808c25750461bec9a28b2f6615fbd0f624117a Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/122262 Reviewed-by: Michele DiGiorgio Tested-by: Jenkins Reviewed-by: Anthony Barbier --- src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) (limited to 'src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp') diff --git a/src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp b/src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp index 5a474e4a83..cf6b984717 100644 --- a/src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp +++ b/src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -23,6 +23,7 @@ */ #include "arm_compute/runtime/NEON/functions/NEPixelWiseMultiplication.h" +#include "arm_compute/core/ITensor.h" #include "arm_compute/core/NEON/kernels/NEPixelWiseMultiplicationKernel.h" #include "support/ToolchainSupport.h" @@ -30,11 +31,21 @@ using namespace arm_compute; -void NEPixelWiseMultiplication::configure(const ITensor *input1, const ITensor *input2, ITensor *output, float scale, ConvertPolicy overflow_policy, RoundingPolicy rounding_policy) +void NEPixelWiseMultiplication::configure(ITensor *input1, ITensor *input2, ITensor *output, float scale, ConvertPolicy overflow_policy, RoundingPolicy rounding_policy) { auto k = arm_compute::support::cpp14::make_unique(); k->configure(input1, input2, output, scale, overflow_policy, rounding_policy); _kernel = std::move(k); + + if(output->info()->dimension(0) > 1) + { + ITensor *broadcasted_info = (input1->info()->dimension(0) == 1) ? input1 : input2; + + if(broadcasted_info->info()->dimension(0) == 1) + { + _border_handler.configure(broadcasted_info, _kernel->border_size(), BorderMode::REPLICATE); + } + } } Status NEPixelWiseMultiplication::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float scale, ConvertPolicy overflow_policy, RoundingPolicy rounding_policy) { -- cgit v1.2.1