COMPMID-1009 Support 4x4 output tile for Winograd Filter Transform on OpenCL.

Change-Id: I68c6453e0f192de659582404f109a89616b9fbb9
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/124811
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

1 files changed, 5 insertions, 11 deletions

diff --git a/arm_compute/core/utils/misc/ShapeCalculator.h b/arm_compute/core/utils/misc/ShapeCalculator.h
index 5344ce7e74..383fc6cda6 100644
--- a/arm_compute/core/utils/misc/ShapeCalculator.h
+++ b/arm_compute/core/utils/misc/ShapeCalculator.h
@@ -196,23 +196,17 @@ inline TensorShape compute_fully_connected_reshaped_weights_shape(const ITensorI
     return output_shape;
 }
 
-inline TensorShape compute_winograd_filter_transform_shape(const ITensorInfo &input)
+inline TensorShape compute_winograd_filter_transform_shape(const ITensorInfo &input, const Size2D &output_tile)
 {
-    // COMPMID-984 (giaiod01)
     TensorShape tensor_shape{ input.tensor_shape() };
 
+    tensor_shape.remove_dimension(get_data_layout_dimension_index(input.data_layout(), DataLayoutDimension::WIDTH));
+    tensor_shape.set(Window::DimY, input.dimension(2));
+    tensor_shape.set(Window::DimZ, (output_tile.width == 2) ? 16 : 36);
+
     if(input.data_layout() == DataLayout::NCHW)
     {
-        tensor_shape.remove_dimension(0);
         tensor_shape.set(Window::DimX, input.dimension(3));
-        tensor_shape.set(Window::DimY, input.dimension(2));
-        tensor_shape.set(Window::DimZ, 16);
-    }
-    else
-    {
-        tensor_shape.remove_dimension(1);
-        tensor_shape.set(Window::DimY, input.dimension(2));
-        tensor_shape.set(Window::DimZ, 16);
     }
 
     return tensor_shape;