COMPMID-1527 - Implementing ReorgLayer on OpenCL

Also extended tests on NEON

Change-Id: Icb0eced534e904ef807972dd3a31988f501bb02e
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/147095
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>

7 files changed, 158 insertions, 9 deletions

diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h
index 4750031603..4a6773a5f8 100644
--- a/arm_compute/core/CL/CLKernels.h
+++ b/arm_compute/core/CL/CLKernels.h
@@ -104,6 +104,7 @@
 #include "arm_compute/core/CL/kernels/CLROIPoolingLayerKernel.h"
 #include "arm_compute/core/CL/kernels/CLReductionOperationKernel.h"
 #include "arm_compute/core/CL/kernels/CLRemapKernel.h"
+#include "arm_compute/core/CL/kernels/CLReorgLayerKernel.h"
 #include "arm_compute/core/CL/kernels/CLReshapeLayerKernel.h"
 #include "arm_compute/core/CL/kernels/CLScaleKernel.h"
 #include "arm_compute/core/CL/kernels/CLScharr3x3Kernel.h"
diff --git a/arm_compute/core/CL/kernels/CLReorgLayerKernel.h b/arm_compute/core/CL/kernels/CLReorgLayerKernel.h
new file mode 100644
index 0000000000..922985422d
--- /dev/null
+++ b/arm_compute/core/CL/kernels/CLReorgLayerKernel.h
@@ -0,0 +1,79 @@
+/*
+ * Copyright (c) 2018 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_CLREORGLAYERKERNEL_H__
+#define __ARM_COMPUTE_CLREORGLAYERKERNEL_H__
+
+#include "arm_compute/core/CL/ICLKernel.h"
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+/** OpenCL kernel to perform a reorg layer */
+class CLReorgLayerKernel : public ICLKernel
+{
+public:
+    /** Default constructor */
+    CLReorgLayerKernel();
+    /** Prevent instances of this class from being copied (As this class contains pointers). */
+    CLReorgLayerKernel(const CLReorgLayerKernel &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers). */
+    CLReorgLayerKernel &operator=(const CLReorgLayerKernel &) = delete;
+    /** Allow instances of this class to be moved */
+    CLReorgLayerKernel(CLReorgLayerKernel &&) = default;
+    /** Allow instances of this class to be moved */
+    CLReorgLayerKernel &operator=(CLReorgLayerKernel &&) = default;
+    /** Initialize the kernel's input, output.
+     *
+     * @param[in]  input  Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32.
+     * @param[out] output Destination tensor with tensor shape:
+     *                    [width_input / stride, height_input / stride, channels_input * stride * stride, batch_size]. This means the output has
+     *                    the same number of input elements. Data types supported: same as @p input.
+     * @param[in]  stride Stride value to use for reorganizing the values in the output tensor.
+     *                    It defines the spatial distance between 2 consecutive pixels in the x and y direction
+     */
+    void configure(const ICLTensor *input, ICLTensor *output, int32_t stride);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLReorgLayerKernel
+     *
+     * @param[in] input  Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32.
+     * @param[in] output Destination tensor with tensor shape:
+     *                   [width_input / stride, height_input / stride, channels_input * stride * stride, batch_size]. This means the output has
+     *                   the same number of input elements. Data types supported: same as @p input. Data types supported: same as @p input.
+     * @param[in] stride Stride value to use for reorganizing the values in the output tensor
+     *                   It defines the spatial distance between 2 consecutive pixels in the x and y direction
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, int32_t stride);
+
+    // Inherited methods overridden:
+    void run(const Window &window, cl::CommandQueue &queue) override;
+
+private:
+    const ICLTensor *_input;
+    ICLTensor       *_output;
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_CLREORGLAYERKERNEL_H__ */
diff --git a/arm_compute/core/NEON/kernels/NEReorgLayerKernel.h b/arm_compute/core/NEON/kernels/NEReorgLayerKernel.h
index 323ab342b2..7e0fb4350d 100644
--- a/arm_compute/core/NEON/kernels/NEReorgLayerKernel.h
+++ b/arm_compute/core/NEON/kernels/NEReorgLayerKernel.h
@@ -55,7 +55,8 @@ public:
      *
      * @param[in]  input  Source tensor. Data type supported: U8/S8/U16/S16/QASYMM8/U32/S32/F16/F32
      * @param[out] output Destination tensor. Data type supported: Same as @p input
-     * @param[in]  stride Stride to be used during data re-organization
+     * @param[in]  stride Stride to be used during data re-organization.
+     *                    It defines the spatial distance between 2 consecutive pixels in the x and y direction
      */
     void configure(const ITensor *input, ITensor *output, int32_t stride);
 
@@ -64,6 +65,7 @@ public:
      * @param[in] input  Source tensor info. Data type supported: U8/S8/U16/S16/QASYMM8/U32/S32/F16/F32
      * @param[in] output Destination tensor info. Data type supported: Same as @p input
      * @param[in] stride Stride to be used during data re-organization
+     *                   It defines the spatial distance between 2 consecutive pixels in the x and y direction
      *
      * @return a status
      */
diff --git a/arm_compute/core/utils/misc/ShapeCalculator.h b/arm_compute/core/utils/misc/ShapeCalculator.h
index d2af844b2a..d72547ed07 100644
--- a/arm_compute/core/utils/misc/ShapeCalculator.h
+++ b/arm_compute/core/utils/misc/ShapeCalculator.h
@@ -61,17 +61,19 @@ inline TensorShape compute_permutation_output_shape(const ITensorInfo &input, co
 
 inline TensorShape compute_reorg_output_shape(const ITensorInfo &input, int32_t stride)
 {
-    ARM_COMPUTE_ERROR_ON(stride <= 0);
+    const size_t idx_width   = get_data_layout_dimension_index(input.data_layout(), DataLayoutDimension::WIDTH);
+    const size_t idx_height  = get_data_layout_dimension_index(input.data_layout(), DataLayoutDimension::HEIGHT);
+    const size_t idx_channel = get_data_layout_dimension_index(input.data_layout(), DataLayoutDimension::CHANNEL);
 
-    const DataLayout   data_layout = input.data_layout();
-    const unsigned int width_idx   = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
-    const unsigned int height_idx  = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
-    const unsigned int channel_idx = get_data_layout_dimension_index(data_layout, DataLayoutDimension::CHANNEL);
+    ARM_COMPUTE_ERROR_ON(stride <= 0);
+    ARM_COMPUTE_ERROR_ON_MSG((input.tensor_shape()[idx_width] % stride != 0), "The width of the input tensor must be a multiple of stride");
+    ARM_COMPUTE_ERROR_ON_MSG((input.tensor_shape()[idx_height] % stride != 0), "The height of the input tensor must be a multiple of stride");
 
     TensorShape output_shape{ input.tensor_shape() };
-    output_shape.set(width_idx, input.tensor_shape()[width_idx] / stride);
-    output_shape.set(height_idx, input.tensor_shape()[height_idx] / stride);
-    output_shape.set(channel_idx, input.tensor_shape()[channel_idx] * stride * stride);
+
+    output_shape.set(idx_width, output_shape[idx_width] / stride);
+    output_shape.set(idx_height, output_shape[idx_height] / stride);
+    output_shape.set(idx_channel, output_shape[idx_channel] * stride * stride);
 
     return output_shape;
 }
diff --git a/arm_compute/runtime/CL/CLFunctions.h b/arm_compute/runtime/CL/CLFunctions.h
index d0c62c3426..2139d5dad3 100644
--- a/arm_compute/runtime/CL/CLFunctions.h
+++ b/arm_compute/runtime/CL/CLFunctions.h
@@ -104,6 +104,7 @@
 #include "arm_compute/runtime/CL/functions/CLROIPoolingLayer.h"
 #include "arm_compute/runtime/CL/functions/CLReductionOperation.h"
 #include "arm_compute/runtime/CL/functions/CLRemap.h"
+#include "arm_compute/runtime/CL/functions/CLReorgLayer.h"
 #include "arm_compute/runtime/CL/functions/CLReshapeLayer.h"
 #include "arm_compute/runtime/CL/functions/CLScale.h"
 #include "arm_compute/runtime/CL/functions/CLScharr3x3.h"
diff --git a/arm_compute/runtime/CL/functions/CLReorgLayer.h b/arm_compute/runtime/CL/functions/CLReorgLayer.h
new file mode 100644
index 0000000000..88559a83f2
--- /dev/null
+++ b/arm_compute/runtime/CL/functions/CLReorgLayer.h
@@ -0,0 +1,62 @@
+/*
+ * Copyright (c) 2018 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_CLREORGLAYER_H__
+#define __ARM_COMPUTE_CLREORGLAYER_H__
+
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/ICLSimpleFunction.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+class CLReorgLayer : public ICLSimpleFunction
+{
+public:
+    /** Initialise the function's source and destination.
+     *
+     * @param[in]  input  Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32.
+     * @param[out] output Destination tensor with tensor shape:
+     *                    [width_input / stride, height_input / stride, channels_input * stride * stride, batch_size]. This means the output has
+     *                    the same number of input elements. Data types supported: same as @p input.
+     * @param[in]  stride Stride value to use for reorganizing the values in the output tensor.
+     *                    It defines the spatial distance between 2 consecutive pixels in the x and y direction
+     *
+     */
+    void configure(ICLTensor *input, ICLTensor *output, int32_t stride);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLReorgLayer
+     *
+     * @param[in] input  Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32.
+     * @param[in] output Destination tensor with tensor shape:
+     *                   [width_input / stride, height_input / stride, channels_input * stride * stride, batch_size]. This means the output has
+     *                   the same number of input elements. Data types supported: same as @p input. Data types supported: same as @p input.
+     * @param[in] stride Stride value to use for reorganizing the values in the output tensor
+     *                   It defines the spatial distance between 2 consecutive pixels in the x and y direction
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, int32_t stride);
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_CLREORGLAYER_H__ */
diff --git a/arm_compute/runtime/NEON/functions/NEReorgLayer.h b/arm_compute/runtime/NEON/functions/NEReorgLayer.h
index f29b2f1964..a73752832b 100644
--- a/arm_compute/runtime/NEON/functions/NEReorgLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEReorgLayer.h
@@ -41,6 +41,7 @@ public:
      * @param[in]  input  First tensor input. Data type supported: U8/S8/QASYMM8//U16/S16/U32/S32/F16/F32
      * @param[out] output Output tensor. Data type supported: Same as @p input
      * @param[in]  stride Stride to be used during data re-organization
+     *                    It defines the spatial distance between 2 consecutive pixels in the x and y direction
      */
     void configure(const ITensor *input, ITensor *output, int32_t stride);
 
@@ -49,6 +50,7 @@ public:
      * @param[in] input  First tensor info. Data type supported: U8/S8/QASYMM8//U16/S16/U32/S32/F16/F32
      * @param[in] output Output tensor info. Data type supported: Same as @p input
      * @param[in] stride Stride to be used during data re-organization
+     *                   It defines the spatial distance between 2 consecutive pixels in the x and y direction
      *
      * @return a status
      */