diff options
author | Giorgio Arena <giorgio.arena@arm.com> | 2017-07-21 10:08:48 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-09-17 14:16:42 +0100 |
commit | a2611815f278334c801094d095901d36e111c3f9 (patch) | |
tree | 1a2bf86b37614ce5473669b283dc37c708dd4957 /arm_compute | |
parent | a629da13544c361ba9580a233109b2418fd6bb73 (diff) | |
download | ComputeLibrary-a2611815f278334c801094d095901d36e111c3f9.tar.gz |
COMPMID-417 NEON/CL MeanStdDev bugfix using FillBorderKernel
Change-Id: Ic48ba7f69783d0e1e80611264e2bc67d1732436e
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/81293
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Diffstat (limited to 'arm_compute')
-rw-r--r-- | arm_compute/core/CL/kernels/CLMeanStdDevKernel.h | 3 | ||||
-rw-r--r-- | arm_compute/core/NEON/kernels/NEMeanStdDevKernel.h | 3 | ||||
-rw-r--r-- | arm_compute/runtime/CL/functions/CLMeanStdDev.h | 10 | ||||
-rw-r--r-- | arm_compute/runtime/NEON/functions/NEMeanStdDev.h | 10 |
4 files changed, 18 insertions, 8 deletions
diff --git a/arm_compute/core/CL/kernels/CLMeanStdDevKernel.h b/arm_compute/core/CL/kernels/CLMeanStdDevKernel.h index 9f30f76e1b..2c09590dab 100644 --- a/arm_compute/core/CL/kernels/CLMeanStdDevKernel.h +++ b/arm_compute/core/CL/kernels/CLMeanStdDevKernel.h @@ -63,12 +63,15 @@ public: // Inherited methods overridden: void run(const Window &window, cl::CommandQueue &queue) override; + BorderSize border_size() const override; + private: const ICLImage *_input; float *_mean; float *_stddev; cl::Buffer *_global_sum; cl::Buffer *_global_sum_squared; + BorderSize _border_size; }; } #endif /* __ARM_COMPUTE_CLMEANSTDDEVKERNEL_H__ */ diff --git a/arm_compute/core/NEON/kernels/NEMeanStdDevKernel.h b/arm_compute/core/NEON/kernels/NEMeanStdDevKernel.h index 83407ccb7d..9c72b20d58 100644 --- a/arm_compute/core/NEON/kernels/NEMeanStdDevKernel.h +++ b/arm_compute/core/NEON/kernels/NEMeanStdDevKernel.h @@ -64,6 +64,8 @@ public: // Inherited methods overridden: void run(const Window &window) override; + BorderSize border_size() const override; + private: const IImage *_input; float *_mean; @@ -71,6 +73,7 @@ private: uint64_t *_global_sum; uint64_t *_global_sum_squared; std::mutex _mtx; + BorderSize _border_size; }; } #endif /* __ARM_COMPUTE_NEMEANSTDDEVKERNEL_H__ */ diff --git a/arm_compute/runtime/CL/functions/CLMeanStdDev.h b/arm_compute/runtime/CL/functions/CLMeanStdDev.h index e33bcdd779..7622138236 100644 --- a/arm_compute/runtime/CL/functions/CLMeanStdDev.h +++ b/arm_compute/runtime/CL/functions/CLMeanStdDev.h @@ -25,6 +25,7 @@ #define __ARM_COMPUTE_CLMEANSTDDEV_H__ #include "arm_compute/core/CL/OpenCL.h" +#include "arm_compute/core/CL/kernels/CLFillBorderKernel.h" #include "arm_compute/core/CL/kernels/CLMeanStdDevKernel.h" #include "arm_compute/runtime/IFunction.h" @@ -38,17 +39,18 @@ public: CLMeanStdDev(); /** Initialise the kernel's inputs and outputs. * - * @param[in] input Input image. Data types supported: U8. - * @param[out] mean Output average pixel value. - * @param[out] stddev (Optional)Output standard deviation of pixel values. + * @param[in, out] input Input image. Data types supported: U8. (Written to only for border filling) + * @param[out] mean Output average pixel value. + * @param[out] stddev (Optional)Output standard deviation of pixel values. */ - void configure(const ICLImage *input, float *mean, float *stddev = nullptr); + void configure(ICLImage *input, float *mean, float *stddev = nullptr); // Inherited methods overridden: void run() override; private: CLMeanStdDevKernel _mean_stddev_kernel; /**< Kernel that standard deviation calculation. */ + CLFillBorderKernel _fill_border_kernel; /**< Kernel that fills the border with zeroes. */ cl::Buffer _global_sum; /**< Variable that holds the global sum among calls in order to ease reduction */ cl::Buffer _global_sum_squared; /**< Variable that holds the global sum of squared values among calls in order to ease reduction */ }; diff --git a/arm_compute/runtime/NEON/functions/NEMeanStdDev.h b/arm_compute/runtime/NEON/functions/NEMeanStdDev.h index 3770b2a270..e8bf8dfa3d 100644 --- a/arm_compute/runtime/NEON/functions/NEMeanStdDev.h +++ b/arm_compute/runtime/NEON/functions/NEMeanStdDev.h @@ -24,6 +24,7 @@ #ifndef __ARM_COMPUTE_NEMEANSTDDEV_H__ #define __ARM_COMPUTE_NEMEANSTDDEV_H__ +#include "arm_compute/core/NEON/kernels/NEFillBorderKernel.h" #include "arm_compute/core/NEON/kernels/NEMeanStdDevKernel.h" #include "arm_compute/core/Types.h" #include "arm_compute/runtime/IFunction.h" @@ -44,17 +45,18 @@ public: NEMeanStdDev(); /** Initialise the kernel's inputs and outputs. * - * @param[in] input Input image. Data type supported: U8. - * @param[out] mean Output average pixel value. - * @param[out] stddev (Optional) Output standard deviation of pixel values. + * @param[in, out] input Input image. Data types supported: U8. (Written to only for border filling) + * @param[out] mean Output average pixel value. + * @param[out] stddev (Optional) Output standard deviation of pixel values. */ - void configure(const IImage *input, float *mean, float *stddev = nullptr); + void configure(IImage *input, float *mean, float *stddev = nullptr); // Inherited methods overridden: void run() override; private: NEMeanStdDevKernel _mean_stddev_kernel; /**< Kernel that standard deviation calculation. */ + NEFillBorderKernel _fill_border_kernel; /**< Kernel that fills tensor's borders with zeroes. */ uint64_t _global_sum; /**< Variable that holds the global sum among calls in order to ease reduction */ uint64_t _global_sum_squared; /**< Variable that holds the global sum of squared values among calls in order to ease reduction */ }; |