diff options
| author | Frank Lei <frank.lei@arm.com> | 2018-02-01 14:47:14 +0800 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:47:40 +0000 |
| commit | 4406fd6cc4abded564d3791324e1f48bdfd34273 (patch) | |
| tree | 22fe402fe9ac7ca338df49e9eccd6eb1587ae875 /arm_compute/core/GLES_COMPUTE | |
| parent | 898d399a0f62c15612a52df4bff5018e783214e4 (diff) | |
| download | ComputeLibrary-4406fd6cc4abded564d3791324e1f48bdfd34273.tar.gz | |
APPBROWSER-391: Fix GLES COMPUTE alignment issues
APPBROWSER-402: Performance optimization for squeezenet/xray model
Change-Id: If31b186b99a6d6087164019fe94d3ac9279e3204
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/119526
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Diffstat (limited to 'arm_compute/core/GLES_COMPUTE')
4 files changed, 21 insertions, 5 deletions
diff --git a/arm_compute/core/GLES_COMPUTE/IGCTensor.h b/arm_compute/core/GLES_COMPUTE/IGCTensor.h index ab4e57e0ce..7329864b85 100644 --- a/arm_compute/core/GLES_COMPUTE/IGCTensor.h +++ b/arm_compute/core/GLES_COMPUTE/IGCTensor.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017 ARM Limited. + * Copyright (c) 2017-2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -75,6 +75,18 @@ public: */ virtual GLuint gc_buffer() const = 0; + /** Flag indicating whether the tensor has been left aligned by a kernel and therefore needs shifting. + * + * @return True if the tensor is left aligned. + */ + bool needs_shifting() const; + /** Set the flag indicating whether or not a tensor needs shifting. + * + * @param[in] needs_shifting Indicates if the tensor is left aligned or not. + * + */ + void set_needs_shifting(bool needs_shifting); + protected: /** Method to be implemented by the child class to map the SSBO. * @@ -92,6 +104,7 @@ protected: private: uint8_t *_mapping; + bool _needs_shifting; }; using IGCImage = IGCTensor; diff --git a/arm_compute/core/GLES_COMPUTE/kernels/GCDepthConcatenateLayerKernel.h b/arm_compute/core/GLES_COMPUTE/kernels/GCDepthConcatenateLayerKernel.h index ce220cc564..06a54dd0b3 100644 --- a/arm_compute/core/GLES_COMPUTE/kernels/GCDepthConcatenateLayerKernel.h +++ b/arm_compute/core/GLES_COMPUTE/kernels/GCDepthConcatenateLayerKernel.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017 ARM Limited. + * Copyright (c) 2017-2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -71,6 +71,7 @@ private: IGCTensor *_output; int _top_bottom; int _left_right; + int _depth_offset; }; } #endif /* __ARM_COMPUTE_GCDEPTHCONCATENATEKERNEL_H__ */ diff --git a/arm_compute/core/GLES_COMPUTE/kernels/GCScaleKernel.h b/arm_compute/core/GLES_COMPUTE/kernels/GCScaleKernel.h index 3f936f6cb9..bdbc50c735 100644 --- a/arm_compute/core/GLES_COMPUTE/kernels/GCScaleKernel.h +++ b/arm_compute/core/GLES_COMPUTE/kernels/GCScaleKernel.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -24,7 +24,7 @@ #ifndef __ARM_COMPUTE_GCSCALEKERNEL_H__ #define __ARM_COMPUTE_GCSCALEKERNEL_H__ -#include "arm_compute/core/GLES_COMPUTE/IGCSimple2DKernel.h" +#include "arm_compute/core/GLES_COMPUTE/IGCSimple3DKernel.h" #include "arm_compute/core/Types.h" namespace arm_compute @@ -32,7 +32,7 @@ namespace arm_compute class IGCTensor; /** Interface for the scale kernel */ -class GCScaleKernel : public IGCSimple2DKernel +class GCScaleKernel : public IGCSimple3DKernel { public: /** Initialise the kernel's inputs, output and interpolation policy @@ -47,6 +47,7 @@ public: void configure(const IGCTensor *input, IGCTensor *output, InterpolationPolicy policy, bool border_undefined, SamplingPolicy sampling_policy = SamplingPolicy::CENTER); // Inherited methods overridden: + void run(const Window &window) override; BorderSize border_size() const override; }; } // namespace arm_compute diff --git a/arm_compute/core/GLES_COMPUTE/kernels/GCTensorShiftKernel.h b/arm_compute/core/GLES_COMPUTE/kernels/GCTensorShiftKernel.h index 5f108764b4..452caae146 100644 --- a/arm_compute/core/GLES_COMPUTE/kernels/GCTensorShiftKernel.h +++ b/arm_compute/core/GLES_COMPUTE/kernels/GCTensorShiftKernel.h @@ -78,6 +78,7 @@ public: private: IGCTensor *_input; gles::NDRange _lws; + int _left_padding; }; } #endif /*__ARM_COMPUTE_GCTENSORSHIFTKERNEL_H__ */ |