Optimize CL DeconvolutionLayer-Part II: Add CLDirectDeconvolution function to be used by CLDeconvolution.

This is only a code refactoring (no optimizations have been added)

Change-Id: I78488f4aecfe1cce93c31dba31489dcee4c85c67
Signed-off-by: giuros01 <giuseppe.rossini@arm.com>
Reviewed-on: https://review.mlplatform.org/c/895
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

6 files changed, 370 insertions, 212 deletions

diff --git a/arm_compute/runtime/CL/CLFunctions.h b/arm_compute/runtime/CL/CLFunctions.h
index 46e43dc0a9..f1021843a0 100644
--- a/arm_compute/runtime/CL/CLFunctions.h
+++ b/arm_compute/runtime/CL/CLFunctions.h
@@ -61,6 +61,7 @@
 #include "arm_compute/runtime/CL/functions/CLDerivative.h"
 #include "arm_compute/runtime/CL/functions/CLDilate.h"
 #include "arm_compute/runtime/CL/functions/CLDirectConvolutionLayer.h"
+#include "arm_compute/runtime/CL/functions/CLDirectDeconvolutionLayer.h"
 #include "arm_compute/runtime/CL/functions/CLElementWiseUnaryLayer.h"
 #include "arm_compute/runtime/CL/functions/CLElementwiseOperations.h"
 #include "arm_compute/runtime/CL/functions/CLEqualizeHistogram.h"
diff --git a/arm_compute/runtime/CL/functions/CLDeconvolutionLayer.h b/arm_compute/runtime/CL/functions/CLDeconvolutionLayer.h
index 9c115f8b3d..b613708c50 100644
--- a/arm_compute/runtime/CL/functions/CLDeconvolutionLayer.h
+++ b/arm_compute/runtime/CL/functions/CLDeconvolutionLayer.h
@@ -24,13 +24,7 @@
 #ifndef __ARM_COMPUTE_CLDECONVOLUTIONLAYER_H__
 #define __ARM_COMPUTE_CLDECONVOLUTIONLAYER_H__
 
-#include "arm_compute/runtime/CL/functions/CLConvolutionLayer.h"
-#include "arm_compute/runtime/CL/functions/CLDeconvolutionLayerUpsample.h"
-
-#include "arm_compute/core/CPP/kernels/CPPFlipWeightsKernel.h"
-
-#include "arm_compute/runtime/CL/CLMemoryGroup.h"
-#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/functions/CLDirectDeconvolutionLayer.h"
 #include "arm_compute/runtime/IFunction.h"
 #include "arm_compute/runtime/IMemoryManager.h"
 
@@ -38,51 +32,16 @@
 
 namespace arm_compute
 {
-class ICLTensor;
-/** Function to run the deconvolution layer.
- *
- * Deconvolution Layer is the backward pass of Convolution Layer. First we transform the input depending on the stride and pad info and then perform a 1x1
- * convolution pass. Input stride defines how many zeroes we should put between each element of the input, pad is the amount of padding and finally a is a user
- * specified value where a < stride - 1, that increases the padding top and right of the input image.
- *
- *  The relation between input to output is as follows:
- *  \f[
- *       width\_output = (width\_input - 1) \cdot stride\_x - 2 \cdot padding\_x + kernel\_x
- *  \f]
- *  \f[
- *       height\_output = (height\_input - 1) \cdot stride\_y - 2 \cdot padding\_y + kernel\_y
- *  \f]
- *
- *  where:
- *      width_input is the size of the first input dimension.
- *      height_input is the size of the second input dimension.
- *      width_output is the size of the first output dimension.
- *      height_output is the size of the second output dimension.
- *      kernel_x and kernel_y are the convolution sizes in x and y.
- *      stride_x and stride_y is the input stride of the first and second dimension.
- *
- * The weights used by Deconvolution are supposed to be the same as the ones used for Convolution. Therefore, it will be necessary to use the weights in the
- * reverse order to perform an actual convolution. This is achieved by using the @ref CPPFlipWeightsKernel.
- *
- * This function calls the following OpenCL kernels/functions:
- *
- * -# @ref CLDeconvolutionLayerUpsample
- * -# @ref CLConvolutionLayer
+/** Basic function to compute the deconvolution layer. This function calls the following OpenCL kernels/functions:
  *
+ * -# @ref CLDirectDeconvolutionLayer
  */
 class CLDeconvolutionLayer : public IFunction
 {
 public:
-    /** Constructor */
+    /** Default constructor */
     CLDeconvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager = nullptr);
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLDeconvolutionLayer(const CLDeconvolutionLayer &) = delete;
-    /** Default move constructor */
-    CLDeconvolutionLayer(CLDeconvolutionLayer &&) = default;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLDeconvolutionLayer &operator=(const CLDeconvolutionLayer &) = delete;
-    /** Default move assignment operator */
-    CLDeconvolutionLayer &operator=(CLDeconvolutionLayer &&) = default;
+
     /** Set the input, weights, biases and output tensors.
      *
      * @deprecated This method is deprecated and will be removed in release 19.05
@@ -91,13 +50,13 @@ public:
      * @param[in]     weights            The 4d weights with dimensions [width, height, IFM, OFM]. Data type supported: Same as @p input.
      * @param[in]     bias               (Optional) The biases have one dimension. Data type supported: Same as @p input.
      * @param[out]    output             Output tensor. The output has the same number of dimensions as the @p input.
-     * @param[in]     info               Contains padding and policies to be used in the deconvolution, this is described in @ref PadStrideInfo.
+     * @param[in]     deconv_info        Contains padding and policies to be used in the deconvolution, this is described in @ref PadStrideInfo.
      * @param[in]     inner_border_right The number of zeros added to right edge of the input.
      * @param[in]     inner_border_top   The number of zeros added to top edge of the input.
      * @param[in]     weights_info       (Optional) Weights information needed for @ref CLConvolutionLayer, specifies if the weights tensor has been reshaped with @ref CLWeightsReshapeKernel.
      *
      */
-    void configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &info,
+    void configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &deconv_info,
                    unsigned int inner_border_right, unsigned int inner_border_top, const WeightsInfo &weights_info = WeightsInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref CLDeconvolutionLayer
      *
@@ -107,14 +66,14 @@ public:
      * @param[in] weights            The 4d weights info with dimensions [width, height, IFM, OFM]. Data type supported: Same as @p input.
      * @param[in] bias               (Optional) The biases have one dimension. Data type supported: Same as @p input.
      * @param[in] output             Output tensor info. The output has the same number of dimensions as the @p input.
-     * @param[in] info               Contains padding and policies to be used in the deconvolution, this is described in @ref PadStrideInfo.
+     * @param[in] deconv_info        Contains padding and policies to be used in the deconvolution, this is described in @ref PadStrideInfo.
      * @param[in] inner_border_right The number of zeros added to right edge of the input.
      * @param[in] inner_border_top   The number of zeros added to top edge of the input.
      * @param[in] weights_info       (Optional) Weights information needed for @ref CLConvolutionLayer, specifies if the weights tensor has been reshaped with @ref CLWeightsReshapeKernel.
      *
      * @return a status
      */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &info,
+    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &deconv_info,
                            unsigned int inner_border_right, unsigned int inner_border_top, const WeightsInfo &weights_info = WeightsInfo());
 
     /** Set the input, weights, biases and output tensors.
@@ -123,37 +82,32 @@ public:
      * @param[in]     weights      The 4d weights with dimensions [width, height, IFM, OFM]. Data type supported: Same as @p input.
      * @param[in]     bias         (Optional) The biases have one dimension. Data type supported: Same as @p input.
      * @param[out]    output       Output tensor. The output has the same number of dimensions as the @p input.
-     * @param[in]     info         Contains padding and policies to be used in the deconvolution, this is described in @ref PadStrideInfo.
+     * @param[in]     deconv_info  Contains padding and policies to be used in the deconvolution, this is described in @ref PadStrideInfo.
      * @param[in]     weights_info (Optional) Weights information needed for @ref CLConvolutionLayer, specifies if the weights tensor has been reshaped with @ref CLWeightsReshapeKernel.
      *
      */
-    void configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &info, const WeightsInfo &weights_info = WeightsInfo());
+    void configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &deconv_info, const WeightsInfo &weights_info = WeightsInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref CLDeconvolutionLayer
      *
      * @param[in] input        Input tensor info. 3 lower dimensions represent a single input, and an optional 4th dimension for batch of inputs. Data types supported: QASYMM8/F16/F32.
      * @param[in] weights      The 4d weights info with dimensions [width, height, IFM, OFM]. Data type supported: Same as @p input.
      * @param[in] bias         (Optional) The biases have one dimension. Data type supported: Same as @p input.
      * @param[in] output       Output tensor info. The output has the same number of dimensions as the @p input.
-     * @param[in] info         Contains padding and policies to be used in the deconvolution, this is described in @ref PadStrideInfo.
+     * @param[in] deconv_info  Contains padding and policies to be used in the deconvolution, this is described in @ref PadStrideInfo.
      * @param[in] weights_info (Optional) Weights information needed for @ref CLConvolutionLayer, specifies if the weights tensor has been reshaped with @ref CLWeightsReshapeKernel.
      *
      * @return a status
      */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &info, const WeightsInfo &weights_info = WeightsInfo());
+    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &deconv_info,
+                           const WeightsInfo &weights_info = WeightsInfo());
 
     // Inherited methods overridden:
     void run() override;
     void prepare() override;
 
 private:
-    CLMemoryGroup                _memory_group;
-    CLDeconvolutionLayerUpsample _scale_f;
-    CLConvolutionLayer           _conv_f;
-    CPPFlipWeightsKernel         _flip_weights;
-    CLTensor                     _scaled_output;
-    ICLTensor                   *_original_weights;
-    CLTensor                     _weights_flipped;
-    bool                         _is_prepared;
+    std::shared_ptr<IMemoryManager> _memory_manager;
+    std::unique_ptr<IFunction>      _function;
 };
 }
 #endif /* __ARM_COMPUTE_CLDECONVOLUTIONLAYER_H__ */
diff --git a/arm_compute/runtime/CL/functions/CLDirectDeconvolutionLayer.h b/arm_compute/runtime/CL/functions/CLDirectDeconvolutionLayer.h
new file mode 100644
index 0000000000..936263d635
--- /dev/null
+++ b/arm_compute/runtime/CL/functions/CLDirectDeconvolutionLayer.h
@@ -0,0 +1,131 @@
+/*
+ * Copyright (c) 2019 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __ARM_COMPUTE_CLDIRECTDECONVOLUTIONLAYER_H__
+#define __ARM_COMPUTE_CLDIRECTDECONVOLUTIONLAYER_H__
+
+#include "arm_compute/runtime/CL/functions/CLConvolutionLayer.h"
+#include "arm_compute/runtime/CL/functions/CLDeconvolutionLayerUpsample.h"
+#include "arm_compute/runtime/CL/functions/CLTranspose.h"
+
+#include "arm_compute/core/CPP/kernels/CPPFlipWeightsKernel.h"
+
+#include "arm_compute/runtime/CL/CLMemoryGroup.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/IFunction.h"
+#include "arm_compute/runtime/IMemoryManager.h"
+
+#include <memory>
+
+namespace arm_compute
+{
+class ICLTensor;
+/** Function to run the deconvolution layer.
+ *
+ * Deconvolution Layer is the backward pass of Convolution Layer. First we transform the input depending on the stride and pad info and then perform a 1x1
+ * convolution pass. Input stride defines how many zeroes we should put between each element of the input and pad is the amount of padding.
+ *
+ *  The relation between input to output is as follows:
+ *  \f[
+ *       width\_output = (width\_input - 1) \cdot stride\_x - 2 \cdot padding\_x + kernel\_x
+ *  \f]
+ *  \f[
+ *       height\_output = (height\_input - 1) \cdot stride\_y - 2 \cdot padding\_y + kernel\_y
+ *  \f]
+ *
+ *  where:
+ *      width_input is the size of the first input dimension.
+ *      height_input is the size of the second input dimension.
+ *      width_output is the size of the first output dimension.
+ *      height_output is the size of the second output dimension.
+ *      kernel_x and kernel_y are the convolution sizes in x and y.
+ *      stride_x and stride_y is the input stride of the first and second dimension.
+ *
+ * The weights used by Deconvolution are supposed to be the same as the ones used for Convolution. Therefore, it will be necessary to use the weights in the
+ * reverse order to perform an actual convolution. This is achieved by using the @ref CPPFlipWeightsKernel.
+ *
+ * This function calls the following OpenCL kernels/functions:
+ *
+ * -# @ref CLDeconvolutionLayerUpsample
+ * -# @ref CLConvolutionLayer
+ *
+ * And the following CPP kernels:
+ * -# @ref CPPFlipWeightsKernel
+ *
+ */
+class CLDirectDeconvolutionLayer : public IFunction
+{
+public:
+    /** Constructor */
+    CLDirectDeconvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager = nullptr);
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLDirectDeconvolutionLayer(const CLDirectDeconvolutionLayer &) = delete;
+    /** Default move constructor */
+    CLDirectDeconvolutionLayer(CLDirectDeconvolutionLayer &&) = default;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLDirectDeconvolutionLayer &operator=(const CLDirectDeconvolutionLayer &) = delete;
+    /** Default move assignment operator */
+    CLDirectDeconvolutionLayer &operator=(CLDirectDeconvolutionLayer &&) = default;
+    /** Set the input, weights, biases and output tensors.
+     *
+     * @param[in,out] input        Input tensor. 3 lower dimensions represent a single input, and an optional 4th dimension for batch of inputs. Data types supported: QASYMM8/F16/F32.
+     * @param[in]     weights      The 4d weights with dimensions [width, height, IFM, OFM]. Data type supported: Same as @p input.
+     * @param[in]     bias         (Optional) The biases have one dimension. Data type supported: Same as @p input.
+     * @param[out]    output       Output tensor. The output has the same number of dimensions as the @p input.
+     * @param[in]     info         Contains padding and policies to be used in the deconvolution, this is decribed in @ref PadStrideInfo.
+     * @param[in]     weights_info (Optional) Weights information needed for @ref CLConvolutionLayer, specifies if the weights tensor has been reshaped with @ref CLWeightsReshapeKernel.
+     *
+     */
+    void configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &info, const WeightsInfo &weights_info = WeightsInfo());
+    /** Static function to check if given info will lead to a valid configuration of @ref CLDirectDeconvolutionLayer
+     *
+     * @param[in] input        Input tensor info. 3 lower dimensions represent a single input, and an optional 4th dimension for batch of inputs. Data types supported: QASYMM8/F16/F32.
+     * @param[in] weights      The 4d weights info with dimensions [width, height, IFM, OFM]. Data type supported: Same as @p input.
+     * @param[in] bias         (Optional) The biases have one dimension. Data type supported: Same as @p input.
+     * @param[in] output       Output tensor info. The output has the same number of dimensions as the @p input.
+     * @param[in] info         Contains padding and policies to be used in the deconvolution, this is decribed in @ref PadStrideInfo.
+     * @param[in] weights_info (Optional) Weights information needed for @ref CLConvolutionLayer, specifies if the weights tensor has been reshaped with @ref CLWeightsReshapeKernel.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &info,
+                           const WeightsInfo &weights_info = WeightsInfo());
+
+    // Inherited methods overridden:
+    void run() override;
+    void prepare() override;
+
+private:
+    CLMemoryGroup                _memory_group;
+    CLDeconvolutionLayerUpsample _scale_f;
+    CLConvolutionLayer           _conv_f;
+    CPPFlipWeightsKernel         _flip_weights;
+
+    CLTensor   _scaled_output;
+    ICLTensor *_original_weights;
+    CLTensor   _weights_flipped;
+
+    bool _is_prepared;
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_CLDECONVOLUTIONLAYER_H__ */
diff --git a/src/runtime/CL/functions/CLDeconvolutionLayer.cpp b/src/runtime/CL/functions/CLDeconvolutionLayer.cpp
index 9da02c10ad..2c17473fc7 100644
--- a/src/runtime/CL/functions/CLDeconvolutionLayer.cpp
+++ b/src/runtime/CL/functions/CLDeconvolutionLayer.cpp
@@ -23,188 +23,62 @@
  */
 #include "arm_compute/runtime/CL/functions/CLDeconvolutionLayer.h"
 
-#include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Utils.h"
 #include "arm_compute/core/Validate.h"
 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
+#include "arm_compute/core/utils/quantization/AsymmHelpers.h"
 #include "arm_compute/runtime/CL/CLScheduler.h"
-#include "arm_compute/runtime/CPP/CPPScheduler.h"
 
+#include <cmath>
 #include <memory>
 #include <tuple>
 
 using namespace arm_compute;
 using namespace arm_compute::misc::shape_calculator;
 
-CLDeconvolutionLayer::CLDeconvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager) // NOLINT
-    : _memory_group(std::move(memory_manager)),
-      _scale_f(),
-      _conv_f(),
-      _flip_weights(),
-      _scaled_output(),
-      _original_weights(nullptr),
-      _weights_flipped(),
-      _is_prepared(false)
+CLDeconvolutionLayer::CLDeconvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager)
+    : _memory_manager(std::move(memory_manager)), _function()
 {
 }
 
-Status CLDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &info,
-                                      unsigned int inner_border_right, unsigned int inner_border_top, const WeightsInfo &weights_info)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, weights, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::F16, DataType::F32);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_LAYOUT(input, weights);
-
-    const DataLayout data_layout = input->data_layout();
-
-    const size_t idx_w = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
-    const size_t idx_h = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
-    const size_t idx_c = get_data_layout_dimension_index(data_layout, DataLayoutDimension::CHANNEL);
-
-    ARM_COMPUTE_RETURN_ERROR_ON(weights->dimension(idx_w) != weights->dimension(idx_h));
-    ARM_COMPUTE_RETURN_ERROR_ON(weights->dimension(idx_w) < 1);
-    ARM_COMPUTE_RETURN_ERROR_ON(!info.padding_is_symmetric());
-
-    const unsigned int stride_x = info.stride().first;
-    const unsigned int stride_y = info.stride().second;
-
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(inner_border_right > stride_x - 1, "inner_border_right must be smaller than stride_x");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(inner_border_top > stride_y - 1, "inner_border_top must be smaller than stride_y");
-
-    auto out_dims = deconvolution_output_dimensions(input->dimension(idx_w), input->dimension(idx_h), weights->dimension(idx_w), weights->dimension(idx_h),
-                                                    info.pad().first, info.pad().second, stride_x, stride_y);
-
-    const TensorShape output_shape = compute_deconvolution_output_shape(out_dims, *input, *weights);
-
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output, weights);
-
-    if(bias != nullptr)
-    {
-        if(is_data_type_quantized_asymmetric(input->data_type()))
-        {
-            ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(bias, 1, DataType::S32);
-        }
-        else
-        {
-            ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, bias);
-        }
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_LAYOUT(input, bias);
-    }
-
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(idx_w) != output_shape[idx_w], "Output's width is invalid.");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(idx_h) != output_shape[idx_h], "Output's height is invalid.");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(idx_c) != output_shape[idx_c], "Output's depth is invalid.");
-
-    unsigned int        padx            = 0;
-    unsigned int        pady            = 0;
-    const TensorShape   scale_out_shape = compute_deconvolution_upsampled_shape(*input, *weights, stride_x, stride_y, inner_border_right, inner_border_top, out_dims, padx, pady);
-    TensorInfo          scale_out_info(input->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(scale_out_shape).set_data_layout(data_layout));
-    const PadStrideInfo conv_info(1, 1, 0, 0, 0, 0, DimensionRoundingType::CEIL);
-
-    ARM_COMPUTE_RETURN_ON_ERROR(CLDeconvolutionLayerUpsample::validate(input, &scale_out_info, BorderSize(inner_border_right, inner_border_top), info));
-    ARM_COMPUTE_RETURN_ON_ERROR(CLConvolutionLayer::validate(&scale_out_info, weights, bias, output, conv_info, weights_info));
-
-    return Status{};
-}
-
-void CLDeconvolutionLayer::configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &info,
+void CLDeconvolutionLayer::configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &deconv_info,
                                      unsigned int inner_border_right, unsigned int inner_border_top, const WeightsInfo &weights_info)
 {
+    ARM_COMPUTE_UNUSED(inner_border_right, inner_border_top);
     ARM_COMPUTE_ERROR_ON_NULLPTR(input, weights, output);
+    auto f = arm_compute::support::cpp14::make_unique<CLDirectDeconvolutionLayer>();
+    f->configure(input, weights, bias, output, deconv_info, weights_info);
+    _function = std::move(f);
+}
 
-    const unsigned int stride_x = info.stride().first;
-    const unsigned int stride_y = info.stride().second;
-
-    const DataLayout data_layout = input->info()->data_layout();
-
-    const size_t idx_w = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
-    const size_t idx_h = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
-
-    _original_weights = weights;
-    _weights_flipped.allocator()->init(weights->info()->clone()->set_data_layout(data_layout));
-    _flip_weights.configure(weights, &_weights_flipped);
-
-    auto out_dims = deconvolution_output_dimensions(input->info()->dimension(idx_w), input->info()->dimension(idx_h), weights->info()->dimension(idx_w), weights->info()->dimension(idx_h),
-                                                    info.pad().first, info.pad().second, stride_x, stride_y);
-
-    const TensorShape output_shape = compute_deconvolution_output_shape(out_dims, *input->info(), *weights->info());
-
-    // Output auto initialization if not yet initialized
-    auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(output_shape).set_data_layout(data_layout));
-
-    // Perform validation step
-    ARM_COMPUTE_ERROR_THROW_ON(CLDeconvolutionLayer::validate(input->info(), weights->info(), bias == nullptr ? nullptr : bias->info(), output->info(), info, inner_border_right, inner_border_top));
-
-    _is_prepared = weights_info.retain_internal_weights();
-
-    _memory_group.manage(&_scaled_output);
-
-    // Find the upsampled dimensions and the padding needed for the convolution with stride 1 in order to match output shape
-    unsigned int      padx            = 0;
-    unsigned int      pady            = 0;
-    const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*input->info(), *weights->info(), stride_x, stride_y, inner_border_right, inner_border_top, out_dims, padx, pady);
-
-    TensorInfo scale_out_info(scale_out_shape, 1, input->info()->data_type(), input->info()->quantization_info());
-    scale_out_info.set_data_layout(data_layout);
-    _scaled_output.allocator()->init(scale_out_info);
-
-    // configure scale function
-    const PadStrideInfo upsample_info(stride_x, stride_y, padx / 2, pady / 2);
-    _scale_f.configure(input, &_scaled_output, BorderSize(inner_border_top, inner_border_right), upsample_info);
-
-    // setup the function to convolve the upscaled output
-    const PadStrideInfo conv_info(1, 1, 0, 0, 0, 0, DimensionRoundingType::CEIL);
-    _conv_f.configure(&_scaled_output, &_weights_flipped, bias, output, conv_info, weights_info);
-    _scaled_output.allocator()->allocate();
+Status CLDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &deconv_info,
+                                      unsigned int inner_border_right, unsigned int inner_border_top, const WeightsInfo &weights_info)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, weights, output);
+    ARM_COMPUTE_UNUSED(inner_border_right, inner_border_top);
+    ARM_COMPUTE_RETURN_ON_ERROR(CLDirectDeconvolutionLayer::validate(input, weights, bias, output, deconv_info, weights_info));
+    return Status{};
 }
 
-void CLDeconvolutionLayer::configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &info,
+void CLDeconvolutionLayer::configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &deconv_info,
                                      const WeightsInfo &weights_info)
 {
-    configure(input, weights, bias, output, info, 0, 0, weights_info);
+    configure(input, weights, bias, output, deconv_info, 0, 0, weights_info);
 }
 
-Status CLDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &info,
+Status CLDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &deconv_info,
                                       const WeightsInfo &weights_info)
 {
-    return CLDeconvolutionLayer::validate(input, weights, bias, output, info, 0, 0, weights_info);
+    return CLDeconvolutionLayer::validate(input, weights, bias, output, deconv_info, 0, 0, weights_info);
 }
 
 void CLDeconvolutionLayer::run()
 {
     prepare();
-
-    _memory_group.acquire();
-
-    _scale_f.run();
-    _conv_f.run();
-
-    _memory_group.release();
+    _function->run();
 }
 
 void CLDeconvolutionLayer::prepare()
 {
-    if(!_is_prepared)
-    {
-        ARM_COMPUTE_ERROR_ON(!_original_weights->is_used());
-
-        // Run weights flipping and mark original weights tensor as unused
-        _weights_flipped.allocator()->allocate();
-        _weights_flipped.map(true);
-        _original_weights->map(CLScheduler::get().queue(), true);
-        CPPScheduler::get().schedule(&_flip_weights, Window::DimZ);
-        _weights_flipped.unmap();
-        _original_weights->unmap(CLScheduler::get().queue());
-        _original_weights->mark_as_unused();
-
-        // Prepare convolution
-        _conv_f.prepare();
-
-        if(!_weights_flipped.is_used())
-        {
-            _weights_flipped.allocator()->free();
-        }
-
-        _is_prepared = true;
-    }
+    _function->prepare();
 }
diff --git a/src/runtime/CL/functions/CLDirectDeconvolutionLayer.cpp b/src/runtime/CL/functions/CLDirectDeconvolutionLayer.cpp
new file mode 100644
index 0000000000..c01588a164
--- /dev/null
+++ b/src/runtime/CL/functions/CLDirectDeconvolutionLayer.cpp
@@ -0,0 +1,198 @@
+/*
+ * Copyright (c) 2019 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/runtime/CL/functions/CLDirectDeconvolutionLayer.h"
+
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/Utils.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/core/utils/misc/ShapeCalculator.h"
+#include "arm_compute/runtime/CL/CLScheduler.h"
+#include "arm_compute/runtime/CPP/CPPScheduler.h"
+#include "utils/TypePrinter.h"
+
+#include <memory>
+#include <tuple>
+
+namespace arm_compute
+{
+using namespace arm_compute::misc::shape_calculator;
+
+CLDirectDeconvolutionLayer::CLDirectDeconvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager) // NOLINT
+    : _memory_group(std::move(memory_manager)),
+      _scale_f(),
+      _conv_f(),
+      _flip_weights(),
+      _scaled_output(),
+      _original_weights(nullptr),
+      _weights_flipped(),
+      _is_prepared(false)
+{
+}
+
+Status CLDirectDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &info,
+                                            const WeightsInfo &weights_info)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, weights, output);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_LAYOUT(input, weights);
+
+    const DataLayout data_layout = input->data_layout();
+
+    const size_t idx_w = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
+    const size_t idx_h = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
+    const size_t idx_c = get_data_layout_dimension_index(data_layout, DataLayoutDimension::CHANNEL);
+
+    ARM_COMPUTE_RETURN_ERROR_ON(weights->dimension(idx_w) != weights->dimension(idx_h));
+    ARM_COMPUTE_RETURN_ERROR_ON(weights->dimension(idx_w) < 1);
+    ARM_COMPUTE_RETURN_ERROR_ON(!info.padding_is_symmetric());
+
+    const unsigned int stride_x = info.stride().first;
+    const unsigned int stride_y = info.stride().second;
+
+    auto out_dims = deconvolution_output_dimensions(input->dimension(idx_w), input->dimension(idx_h), weights->dimension(idx_w), weights->dimension(idx_h),
+                                                    info.pad().first, info.pad().second, stride_x, stride_y);
+
+    const TensorShape output_shape = compute_deconvolution_output_shape(out_dims, *input, *weights);
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output, weights);
+
+    if(bias != nullptr)
+    {
+        if(is_data_type_quantized_asymmetric(input->data_type()))
+        {
+            ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(bias, 1, DataType::S32);
+        }
+        else
+        {
+            ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, bias);
+        }
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_LAYOUT(input, bias);
+    }
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(idx_w) != output_shape[idx_w], "Output's width is invalid.");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(idx_h) != output_shape[idx_h], "Output's height is invalid.");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(idx_c) != output_shape[idx_c], "Output's depth is invalid.");
+
+    unsigned int        padx            = 0;
+    unsigned int        pady            = 0;
+    const TensorShape   scale_out_shape = compute_deconvolution_upsampled_shape(*input, *weights, stride_x, stride_y, 0, 0, out_dims, padx, pady);
+    TensorInfo          scale_out_info(input->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(scale_out_shape).set_data_layout(data_layout));
+    const PadStrideInfo conv_info(1, 1, 0, 0, 0, 0, DimensionRoundingType::CEIL);
+
+    ARM_COMPUTE_RETURN_ON_ERROR(CLDeconvolutionLayerUpsample::validate(input, &scale_out_info, BorderSize(), info));
+    ARM_COMPUTE_RETURN_ON_ERROR(CLConvolutionLayer::validate(&scale_out_info, weights, bias, output, conv_info, weights_info));
+
+    return Status{};
+}
+
+void CLDirectDeconvolutionLayer::configure(ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &info,
+                                           const WeightsInfo &weights_info)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input, weights, output);
+
+    const unsigned int stride_x = info.stride().first;
+    const unsigned int stride_y = info.stride().second;
+
+    const DataLayout data_layout = input->info()->data_layout();
+
+    const size_t idx_w = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
+    const size_t idx_h = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
+
+    _original_weights = weights;
+    _weights_flipped.allocator()->init(weights->info()->clone()->set_data_layout(data_layout));
+    _flip_weights.configure(weights, &_weights_flipped);
+
+    auto out_dims = deconvolution_output_dimensions(input->info()->dimension(idx_w), input->info()->dimension(idx_h), weights->info()->dimension(idx_w), weights->info()->dimension(idx_h),
+                                                    info.pad().first, info.pad().second, stride_x, stride_y);
+
+    const TensorShape output_shape = compute_deconvolution_output_shape(out_dims, *input->info(), *weights->info());
+
+    // Output auto initialization if not yet initialized
+    auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(output_shape).set_data_layout(data_layout));
+
+    // Perform validation step
+    ARM_COMPUTE_ERROR_THROW_ON(CLDirectDeconvolutionLayer::validate(input->info(), weights->info(), bias == nullptr ? nullptr : bias->info(), output->info(), info));
+
+    _is_prepared = weights_info.retain_internal_weights();
+
+    _memory_group.manage(&_scaled_output);
+
+    // Find the upsampled dimensions and the padding needed for the convolution with stride 1 in order to match output shape
+    unsigned int      padx            = 0;
+    unsigned int      pady            = 0;
+    const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*input->info(), *weights->info(), stride_x, stride_y, 0, 0, out_dims, padx, pady);
+
+    TensorInfo scale_out_info(scale_out_shape, 1, input->info()->data_type(), input->info()->quantization_info());
+    scale_out_info.set_data_layout(data_layout);
+    _scaled_output.allocator()->init(scale_out_info);
+
+    // configure scale function
+    const PadStrideInfo upsample_info(stride_x, stride_y, padx / 2, pady / 2);
+    _scale_f.configure(input, &_scaled_output, BorderSize(), upsample_info);
+
+    // setup the function to convolve the upscaled output
+    const PadStrideInfo conv_info(1, 1, 0, 0, 0, 0, DimensionRoundingType::CEIL);
+    _conv_f.configure(&_scaled_output, &_weights_flipped, bias, output, conv_info, weights_info);
+    _scaled_output.allocator()->allocate();
+}
+
+void CLDirectDeconvolutionLayer::run()
+{
+    prepare();
+
+    _memory_group.acquire();
+
+    _scale_f.run();
+    _conv_f.run();
+
+    _memory_group.release();
+}
+
+void CLDirectDeconvolutionLayer::prepare()
+{
+    if(!_is_prepared)
+    {
+        ARM_COMPUTE_ERROR_ON(!_original_weights->is_used());
+
+        // Run weights flipping and mark original weights tensor as unused
+        _weights_flipped.allocator()->allocate();
+        _weights_flipped.map(true);
+        _original_weights->map(CLScheduler::get().queue(), true);
+        CPPScheduler::get().schedule(&_flip_weights, Window::DimZ);
+        _weights_flipped.unmap();
+        _original_weights->unmap(CLScheduler::get().queue());
+        _original_weights->mark_as_unused();
+
+        // Prepare convolution
+        _conv_f.prepare();
+
+        if(!_weights_flipped.is_used())
+        {
+            _weights_flipped.allocator()->free();
+        }
+
+        _is_prepared = true;
+    }
+}
+} // namespace arm_compute
diff --git a/tests/validation/CL/DeconvolutionLayer.cpp b/tests/validation/CL/DeconvolutionLayer.cpp
index 31852c8eb6..958a0e438a 100644
--- a/tests/validation/CL/DeconvolutionLayer.cpp
+++ b/tests/validation/CL/DeconvolutionLayer.cpp
@@ -119,7 +119,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(zi
                                                 1U,
                                                 0U,
                                             })),
-    framework::dataset::make("Expected", { false, false, false, false, false, true })),
+    framework::dataset::make("Expected", { false, false, false, false, true, true })),
     input_info, weights_info, bias_info, output_info, pad_info, ax, ay, expected)
 {
     bool is_valid = bool(CLDeconvolutionLayer::validate(&input_info.clone()->set_is_resizable(false), &weights_info.clone()->set_is_resizable(false), &bias_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), pad_info, ax, ay));