diff options
| author | Michele Di Giorgio <michele.digiorgio@arm.com> | 2019-12-20 13:26:08 +0000 |
|---|---|---|
| committer | Michele Di Giorgio <michele.digiorgio@arm.com> | 2020-01-14 17:21:26 +0000 |
| commit | cbbed288a71f2f048123db3cf396361e5d66ce93 (patch) | |
| tree | bd48122e283272adb668a42097ea69ca3f0473a6 /src/core/CL/cl_kernels/convert_fc_weights.cl | |
| parent | 70d33bdfd36e1b44b0573189dca67ed7c63dd59e (diff) | |
| download | ComputeLibrary-cbbed288a71f2f048123db3cf396361e5d66ce93.tar.gz | |
COMPMID-2991: Add support for QASYMM8_SIGNED in CL kernels/functions - part 2
Adding support for QASYMM8_SIGNED to the following CL kernels/functions:
- CLActivationLayerKernel/CLActivationLayer
- CLComparisonKernel/CLComparison
- CLConvertFullyConnectedWeightsKernel/CLConvertFullyConnectedWeights
- CLDeconvolutionLayerUpsampleKernel/CLDeconvolutionLayerUpsample
- CLDepthToSpaceLayerKernel/CLDepthToSpaceLayer
- CLDequantizationLayerKernel/CLDequantizationLayer
- CLGEMMMatrixVectorMultiplyKernel
- CLNormalizePlanarYUVLayerKernel
- CLPReluLayer
- CLPixelWiseMultiplicationKernel/CLPixelWiseMultiplication
- CLPoolingLayerKernel/CLPoolingLayer
Change-Id: I874bbb7c2b08baa9c5ff4c9e6bc8778b42a6bec5
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/2539
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/CL/cl_kernels/convert_fc_weights.cl')
| -rw-r--r-- | src/core/CL/cl_kernels/convert_fc_weights.cl | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/src/core/CL/cl_kernels/convert_fc_weights.cl b/src/core/CL/cl_kernels/convert_fc_weights.cl index d47b733acd..db0873755e 100644 --- a/src/core/CL/cl_kernels/convert_fc_weights.cl +++ b/src/core/CL/cl_kernels/convert_fc_weights.cl @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018 ARM Limited. + * Copyright (c) 2018-2020 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -32,7 +32,7 @@ * @attention Data type can be passed using the -DDATA_TYPE compile flag, e.g. -DDATA_TYPE=float * @attention Original input tensor width*height and depth should be given as a preprocessor argument using -DFACTOR_1=size and -DFACTOR_2=size for NCHW and vice versa for NHWC. e.g. -DFACTOR_1=256 and -DFACTOR_2=128 * - * @param[in] src_ptr Pointer to the source image. Supported data types: U8, S8, QASYMM8, U16, S16, U32, S32, F16, F32 + * @param[in] src_ptr Pointer to the source image. Supported data types: All. * @param[in] src_stride_x Stride of the source image in X dimension (in bytes) * @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes) * @param[in] src_stride_y Stride of the source image in Y dimension (in bytes) |