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>

1 files changed, 1 insertions, 1 deletions

diff --git a/tests/datasets/DepthwiseConvolutionLayerDataset.h b/tests/datasets/DepthwiseConvolutionLayerDataset.h
index 53b5248374..3b17910eac 100644
--- a/tests/datasets/DepthwiseConvolutionLayerDataset.h
+++ b/tests/datasets/DepthwiseConvolutionLayerDataset.h
@@ -121,7 +121,7 @@ public:
     SmallDepthwiseConvolutionLayerDataset()
     {
         add_config(TensorShape(7U, 7U, 1U), Size2D(1U, 1U), PadStrideInfo(1, 1, 0, 0));
-        add_config(TensorShape(23U, 27U, 5U), Size2D(3U, 5U), PadStrideInfo(2, 1, 0, 0));
+        add_config(TensorShape(3U, 3U, 2U), Size2D(2U, 2U), PadStrideInfo(1, 1, 0, 0));
         add_config(TensorShape(33U, 27U, 7U), Size2D(7U, 3U), PadStrideInfo(3, 2, 1, 0));
         // Asymmetric padding
         add_config(TensorShape(33U, 27U, 7U), Size2D(5U, 7U), PadStrideInfo(3, 2, 1, 1, 2, 0, DimensionRoundingType::FLOOR));