diff options
author | Gian Marco Iodice <gianmarco.iodice@arm.com> | 2021-04-16 15:08:59 +0100 |
---|---|---|
committer | Gian Marco Iodice <gianmarco.iodice@arm.com> | 2021-07-02 15:56:45 +0000 |
commit | 8155c0253c00aa9e26651361460c66feb39829a6 (patch) | |
tree | 41dacc432d4d1f1daa32d20d15e5120c11b9fa56 /arm_compute/core/KernelDescriptors.h | |
parent | 2eb5d16b839cbc28c6cb7f0de7a0bf15290b425a (diff) | |
download | ComputeLibrary-8155c0253c00aa9e26651361460c66feb39829a6.tar.gz |
Rework OpenCL Depthwise Convolution
- Remove dedicated kernels for NCHW. Now we only use NHWC with permute
- Remove specialized kernels for 3x3 NHWC
- Simplify CLDepthwiseConvolutionLayer.cpp to call just the native
implementation for both floating-point and quantized data types
- Develop two parametric opencl kernels for depthwise convolution layer NHWC
(floating-point and quantized)
- Add support to export the weights to cl_image
- Extend test for depthwise convolution on opencl
Resolves COMPMID-4417
Change-Id: Ibe533f79c2860f9cac8e921895d5a8f947753a5c
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5893
Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'arm_compute/core/KernelDescriptors.h')
-rw-r--r-- | arm_compute/core/KernelDescriptors.h | 14 |
1 files changed, 5 insertions, 9 deletions
diff --git a/arm_compute/core/KernelDescriptors.h b/arm_compute/core/KernelDescriptors.h index 6c1fc74b1e..a6e5c3372e 100644 --- a/arm_compute/core/KernelDescriptors.h +++ b/arm_compute/core/KernelDescriptors.h @@ -96,16 +96,12 @@ struct GEMMKernelInfo GEMMLowpOutputStageInfo output_stage{}; /**< GEMMLowp output stage information */ }; -/** Descriptor used by the depthwise convolution kernels */ -struct DWCKernelInfo +/** Compute descriptor used by the depthwise convolution native kernel */ +struct DWCComputeKernelInfo { - ActivationLayerInfo activation_info{}; /**< Activation function to perform after the depthwise convolution */ -}; - -/** Descriptor used by the depthwise convolution kernels to retrieve the number of output elements processed by each thread */ -struct DWCWeightsKernelInfo -{ - unsigned int n0{ 0 }; /**< Number of columns processed by each thread */ + unsigned int n0{ 0 }; /**< Number of columns processed by each thread */ + unsigned int m0{ 0 }; /**< Number of rows processed by each thread */ + bool export_weights_to_cl_image{ false }; /**< Export the weights to cl_image */ }; /** Descriptor used by the softmax kernels */ |