COMPMID-1343: Add grouping support to CLCol2ImKernel

Change-Id: I5188a2163e7341f1915d98c21464fea13a9a7faf
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/143330
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>

2 files changed, 13 insertions, 5 deletions

diff --git a/arm_compute/core/CL/kernels/CLCol2ImKernel.h b/arm_compute/core/CL/kernels/CLCol2ImKernel.h
index 94f21b1ebc..5c047ca091 100644
--- a/arm_compute/core/CL/kernels/CLCol2ImKernel.h
+++ b/arm_compute/core/CL/kernels/CLCol2ImKernel.h
@@ -70,18 +70,20 @@ public:
      * @param[out] output         The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
      *                            while the rest represent batch of outputs. Data types supported: Same as @p input
      * @param[in]  convolved_dims Output convolved dimensions.
+     * @param[in]  num_groups     (Optional) Number of groups when performing a grouped convolution
      */
-    void configure(const ICLTensor *input, ICLTensor *output, std::pair<unsigned int, unsigned int> convolved_dims);
+    void configure(const ICLTensor *input, ICLTensor *output, std::pair<unsigned int, unsigned int> convolved_dims, unsigned int num_groups = 1);
     /** Static function to check if given info will lead to a valid configuration of @ref CLCol2ImKernel
      *
      * @param[in] input          The input tensor to convert. Data types supported: QASYMM8/F16/F32
      * @param[in] output         The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
      *                           while the rest represent batch of outputs. Data types supported: Same as @p input
      * @param[in] convolved_dims Output convolved dimensions.
+     * @param[in] num_groups     (Optional) Number of groups when performing a grouped convolution
      *
      * @return a status
      */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *output, std::pair<unsigned int, unsigned int> convolved_dims);
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, std::pair<unsigned int, unsigned int> convolved_dims, unsigned int num_groups = 1);
 
     // Inherited methods overridden:
     void run(const Window &window, cl::CommandQueue &queue) override;
diff --git a/arm_compute/core/utils/misc/ShapeCalculator.h b/arm_compute/core/utils/misc/ShapeCalculator.h
index fc6abf95f3..8a00c22306 100644
--- a/arm_compute/core/utils/misc/ShapeCalculator.h
+++ b/arm_compute/core/utils/misc/ShapeCalculator.h
@@ -140,13 +140,19 @@ inline TensorShape compute_reductionB_shape(const ITensorInfo &a)
 
     return shape_vector_sum_row;
 }
-inline TensorShape compute_col2im_shape(const ITensorInfo &input, std::pair<unsigned int, unsigned int> convolved_dims)
+inline TensorShape compute_col2im_shape(const ITensorInfo &input, std::pair<unsigned int, unsigned int> convolved_dims, unsigned int num_groups = 1)
 {
+    ARM_COMPUTE_ERROR_ON(num_groups == 0);
+    ARM_COMPUTE_ERROR_ON(input.tensor_shape()[1] != (convolved_dims.first * convolved_dims.second));
+    ARM_COMPUTE_ERROR_ON((num_groups > 1) && input.tensor_shape()[2] != num_groups);
+
     TensorShape col2im_shape{ input.tensor_shape() };
-    col2im_shape.shift_right(1);
     col2im_shape.set(0, convolved_dims.first);
     col2im_shape.set(1, convolved_dims.second);
-    col2im_shape.set(2, input.tensor_shape()[0]);
+    col2im_shape.set(2, input.tensor_shape()[0] * num_groups);
+
+    const unsigned int batch_idx = (num_groups == 1) ? 2 : 3;
+    col2im_shape.set(3, input.tensor_shape()[batch_idx]);
 
     return col2im_shape;
 }