From bcd2352d7fd99a2f6aab220fa0c3b3f3119a1a4c Mon Sep 17 00:00:00 2001 From: Michalis Spyrou Date: Thu, 21 May 2020 15:02:36 +0100 Subject: COMPMID-3391: Implement Async interfaces Change-Id: I8168cea5056ff48a0253ebb8c88ea549a3ea69a2 Signed-off-by: Michalis Spyrou Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3335 Tested-by: Arm Jenkins Comments-Addressed: Arm Jenkins Reviewed-by: Georgios Pinitas --- .../NEON/functions/NEGenerateProposalsLayer.h | 7 +-- .../runtime/NEON/functions/NEReductionOperation.h | 4 +- .../runtime/NEON/functions/NEReshapeLayer.h | 51 +++++++++++++++++++--- .../runtime/NEON/functions/NESoftmaxLayer.h | 8 ++-- 4 files changed, 56 insertions(+), 14 deletions(-) (limited to 'arm_compute/runtime/NEON/functions') diff --git a/arm_compute/runtime/NEON/functions/NEGenerateProposalsLayer.h b/arm_compute/runtime/NEON/functions/NEGenerateProposalsLayer.h index 7c470fbaf0..7260434606 100644 --- a/arm_compute/runtime/NEON/functions/NEGenerateProposalsLayer.h +++ b/arm_compute/runtime/NEON/functions/NEGenerateProposalsLayer.h @@ -23,18 +23,19 @@ */ #ifndef ARM_COMPUTE_NEGENERATEPROPOSALSLAYER_H #define ARM_COMPUTE_NEGENERATEPROPOSALSLAYER_H + #include "arm_compute/core/NEON/kernels/NEBoundingBoxTransformKernel.h" #include "arm_compute/core/NEON/kernels/NEDequantizationLayerKernel.h" #include "arm_compute/core/NEON/kernels/NEGenerateProposalsLayerKernel.h" #include "arm_compute/core/NEON/kernels/NEPadLayerKernel.h" #include "arm_compute/core/NEON/kernels/NEPermuteKernel.h" #include "arm_compute/core/NEON/kernels/NEQuantizationLayerKernel.h" -#include "arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h" #include "arm_compute/core/Types.h" #include "arm_compute/runtime/CPP/CPPScheduler.h" #include "arm_compute/runtime/CPP/functions/CPPBoxWithNonMaximaSuppressionLimit.h" #include "arm_compute/runtime/IFunction.h" #include "arm_compute/runtime/MemoryGroup.h" +#include "arm_compute/runtime/NEON/functions/NEReshapeLayer.h" #include "arm_compute/runtime/Tensor.h" namespace arm_compute @@ -112,9 +113,9 @@ private: // Neon kernels NEPermuteKernel _permute_deltas_kernel; - NEReshapeLayerKernel _flatten_deltas_kernel; + NEReshapeLayer _flatten_deltas; NEPermuteKernel _permute_scores_kernel; - NEReshapeLayerKernel _flatten_scores_kernel; + NEReshapeLayer _flatten_scores; NEComputeAllAnchorsKernel _compute_anchors_kernel; NEBoundingBoxTransformKernel _bounding_box_kernel; NEPadLayerKernel _pad_kernel; diff --git a/arm_compute/runtime/NEON/functions/NEReductionOperation.h b/arm_compute/runtime/NEON/functions/NEReductionOperation.h index abda4159ba..78e8b04dbb 100644 --- a/arm_compute/runtime/NEON/functions/NEReductionOperation.h +++ b/arm_compute/runtime/NEON/functions/NEReductionOperation.h @@ -28,8 +28,8 @@ #include "arm_compute/core/NEON/kernels/NEFillBorderKernel.h" #include "arm_compute/core/NEON/kernels/NEReductionOperationKernel.h" -#include "arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h" #include "arm_compute/core/Types.h" +#include "arm_compute/runtime/NEON/functions/NEReshapeLayer.h" #include "arm_compute/runtime/Tensor.h" namespace arm_compute @@ -76,7 +76,7 @@ private: MemoryGroup _memory_group; NEReductionOperationKernel _reduction_kernel; NEFillBorderKernel _fill_border_kernel; - NEReshapeLayerKernel _reshape_kernel; + NEReshapeLayer _reshape; Tensor _output_internal; size_t _window_split; int _reduction_axis; diff --git a/arm_compute/runtime/NEON/functions/NEReshapeLayer.h b/arm_compute/runtime/NEON/functions/NEReshapeLayer.h index d6643842d9..5a296a776d 100644 --- a/arm_compute/runtime/NEON/functions/NEReshapeLayer.h +++ b/arm_compute/runtime/NEON/functions/NEReshapeLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2019 ARM Limited. + * Copyright (c) 2017-2020 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -24,8 +24,11 @@ #ifndef ARM_COMPUTE_NERESHAPELAYER_H #define ARM_COMPUTE_NERESHAPELAYER_H +#include "arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h" #include "arm_compute/core/Types.h" -#include "arm_compute/runtime/NEON/INESimpleFunctionNoBorder.h" +#include "arm_compute/runtime/IFunction.h" +#include "arm_compute/runtime/NEON/INEOperator.h" +#include "arm_compute/runtime/Types.h" namespace arm_compute { @@ -33,24 +36,62 @@ namespace arm_compute class ITensor; /** Basic function to run @ref NEReshapeLayerKernel */ -class NEReshapeLayer : public INESimpleFunctionNoBorder +class NEReshapeLayer : public IFunction { public: /** Initialise the kernel's inputs and outputs * - * @param[in] input First tensor input. Data type supported: All + * @param[in] input Input tensor. Data type supported: All * @param[out] output Output tensor. Data type supported: Same as @p input */ void configure(const ITensor *input, ITensor *output); /** Static function to check if given info will lead to a valid configuration of @ref NEReshapeLayer * - * @param[in] input First tensor info. Data type supported: All + * @param[in] input Input tensor info. Data type supported: All * @param[in] output Output tensor info. Data type supported: Same as @p input * * @return a status */ static Status validate(const ITensorInfo *input, const ITensorInfo *output); + + // Inherited methods overridden: + void run() override; + +private: + const ITensor *_input + { + nullptr + }; + ITensor *_output{ nullptr }; + std::unique_ptr _kernel{ nullptr }; +}; + +namespace experimental +{ +/** Basic function to run @ref NEReshapeLayerKernel */ +class NEReshapeLayer : public INEOperator +{ +public: + /** Initialise the kernel's inputs and outputs + * + * @param[in] input Input tensor info. Data type supported: All + * @param[out] output Output info. Data type supported: Same as @p input + */ + void configure(const ITensorInfo *input, ITensorInfo *output); + + /** Static function to check if given info will lead to a valid configuration of @ref NEReshapeLayer + * + * @param[in] input Input tensor info. Data type supported: All + * @param[in] output Output tensor info. Data type supported: Same as @p input + * + * @return a status + */ + static Status validate(const ITensorInfo *input, const ITensorInfo *output); + + // Inherited methods overridden: + MemoryRequirements workspace() const override; }; +} // namespace experimental } // namespace arm_compute #endif /*ARM_COMPUTE_NERESHAPELAYER_H */ diff --git a/arm_compute/runtime/NEON/functions/NESoftmaxLayer.h b/arm_compute/runtime/NEON/functions/NESoftmaxLayer.h index c5c83d8b5a..51d981de44 100644 --- a/arm_compute/runtime/NEON/functions/NESoftmaxLayer.h +++ b/arm_compute/runtime/NEON/functions/NESoftmaxLayer.h @@ -25,11 +25,11 @@ #define ARM_COMPUTE_NESOFTMAXLAYER_H #include "arm_compute/core/NEON/kernels/NEFillBorderKernel.h" -#include "arm_compute/core/NEON/kernels/NEFlattenLayerKernel.h" -#include "arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h" #include "arm_compute/core/NEON/kernels/NESoftmaxLayerKernel.h" #include "arm_compute/runtime/IFunction.h" #include "arm_compute/runtime/MemoryGroup.h" +#include "arm_compute/runtime/NEON/functions/NEFlattenLayer.h" +#include "arm_compute/runtime/NEON/functions/NEReshapeLayer.h" #include "arm_compute/runtime/Tensor.h" namespace arm_compute @@ -119,9 +119,9 @@ private: MemoryGroup _memory_group; NELogits1DMaxKernel _max_kernel; NELogits1DSoftmaxKernel _softmax_kernel; - std::unique_ptr _flat_or_reshape_kernel_ptr; + std::unique_ptr _flat_or_reshape_ptr; NEFillBorderKernel _fill_border_kernel; - NEReshapeLayerKernel _reshape_kernel; + NEReshapeLayer _reshape; Tensor _max; Tensor _tmp; Tensor _input_flattened; -- cgit v1.2.1