COMPMID-1528: Add ReorgLayer on NEON

Change-Id: I44369b4a716767163e2233b7d87bff300c523383
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/146314
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>

5 files changed, 172 insertions, 0 deletions

diff --git a/arm_compute/core/NEON/NEKernels.h b/arm_compute/core/NEON/NEKernels.h
index 8664c7732c..bbeb604140 100644
--- a/arm_compute/core/NEON/NEKernels.h
+++ b/arm_compute/core/NEON/NEKernels.h
@@ -102,6 +102,7 @@
 #include "arm_compute/core/NEON/kernels/NEROIPoolingLayerKernel.h"
 #include "arm_compute/core/NEON/kernels/NEReductionOperationKernel.h"
 #include "arm_compute/core/NEON/kernels/NERemapKernel.h"
+#include "arm_compute/core/NEON/kernels/NEReorgLayerKernel.h"
 #include "arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h"
 #include "arm_compute/core/NEON/kernels/NEScaleKernel.h"
 #include "arm_compute/core/NEON/kernels/NEScharr3x3Kernel.h"
diff --git a/arm_compute/core/NEON/kernels/NEReorgLayerKernel.h b/arm_compute/core/NEON/kernels/NEReorgLayerKernel.h
new file mode 100644
index 0000000000..323ab342b2
--- /dev/null
+++ b/arm_compute/core/NEON/kernels/NEReorgLayerKernel.h
@@ -0,0 +1,96 @@
+/*
+ * 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_NEREORGLAYERKERNEL_H__
+#define __ARM_COMPUTE_NEREORGLAYERKERNEL_H__
+
+#include "arm_compute/core/NEON/INEKernel.h"
+
+namespace arm_compute
+{
+// Forward declarations
+class ITensor;
+
+/** Interface for the kernel to perform tensor re-organization */
+class NEReorgLayerKernel : public INEKernel
+{
+public:
+    const char *name() const override
+    {
+        return "NEReorgLayerKernel";
+    }
+    /** Default constructor */
+    NEReorgLayerKernel();
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    NEReorgLayerKernel(const NEReorgLayerKernel &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    NEReorgLayerKernel &operator=(const NEReorgLayerKernel &) = delete;
+    /** Default Move Constructor. */
+    NEReorgLayerKernel(NEReorgLayerKernel &&) = default;
+    /** Default move assignment operator */
+    NEReorgLayerKernel &operator=(NEReorgLayerKernel &&) = default;
+    /** Default destructor */
+    ~NEReorgLayerKernel() = default;
+    /** Set the input and output of the kernel
+     *
+     * @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
+     */
+    void configure(const ITensor *input, ITensor *output, int32_t stride);
+
+    /** Static function to check if given info will lead to a valid configuration of @ref NEReshapeLayerKernel
+     *
+     * @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
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, int32_t stride);
+
+    // Inherited methods overridden:
+    void run(const Window &window, const ThreadInfo &info) override;
+
+private:
+    /** Template function to run the reorg
+     *
+     * @param[in] window Region on which to execute the kernel. (Must be a valid region of the window returned by window()).
+     */
+    template <typename T>
+    void run_reorg(const Window &window);
+
+    /** Common signature for all the specialised reorg functions
+     *
+     * @param[in] window Region on which to execute the kernel.
+     */
+    using ReorgFunctionPtr = void (NEReorgLayerKernel::*)(const Window &window);
+
+private:
+    ReorgFunctionPtr _func;
+    const ITensor   *_input;
+    ITensor         *_output;
+    int32_t          _stride;
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_NEREORGLAYERKERNEL_H__ */
diff --git a/arm_compute/core/utils/misc/ShapeCalculator.h b/arm_compute/core/utils/misc/ShapeCalculator.h
index 1a86d27727..9c7cfecd4c 100644
--- a/arm_compute/core/utils/misc/ShapeCalculator.h
+++ b/arm_compute/core/utils/misc/ShapeCalculator.h
@@ -57,6 +57,22 @@ inline TensorShape compute_permutation_output_shape(const ITensorInfo &input, co
     permute(output_shape, perm);
     return output_shape;
 }
+inline TensorShape compute_reorg_output_shape(const ITensorInfo &input, int32_t stride)
+{
+    ARM_COMPUTE_ERROR_ON(stride <= 0);
+
+    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);
+
+    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);
+
+    return output_shape;
+}
 inline TensorShape compute_weights_reshaped_shape(const ITensorInfo &weights, bool has_bias = false, unsigned int num_groups = 1)
 {
     // Number of groups greater than one are only supported for NCHW data layout, and the number of weights must be a multiple of it.
diff --git a/arm_compute/runtime/NEON/NEFunctions.h b/arm_compute/runtime/NEON/NEFunctions.h
index dabd78c256..89a5c4aaaf 100644
--- a/arm_compute/runtime/NEON/NEFunctions.h
+++ b/arm_compute/runtime/NEON/NEFunctions.h
@@ -103,6 +103,7 @@
 #include "arm_compute/runtime/NEON/functions/NEROIPoolingLayer.h"
 #include "arm_compute/runtime/NEON/functions/NEReductionOperation.h"
 #include "arm_compute/runtime/NEON/functions/NERemap.h"
+#include "arm_compute/runtime/NEON/functions/NEReorgLayer.h"
 #include "arm_compute/runtime/NEON/functions/NEReshapeLayer.h"
 #include "arm_compute/runtime/NEON/functions/NEScale.h"
 #include "arm_compute/runtime/NEON/functions/NEScharr3x3.h"
diff --git a/arm_compute/runtime/NEON/functions/NEReorgLayer.h b/arm_compute/runtime/NEON/functions/NEReorgLayer.h
new file mode 100644
index 0000000000..f29b2f1964
--- /dev/null
+++ b/arm_compute/runtime/NEON/functions/NEReorgLayer.h
@@ -0,0 +1,58 @@
+/*
+ * 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_NEREORGLAYER_H__
+#define __ARM_COMPUTE_NEREORGLAYER_H__
+
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/NEON/INESimpleFunction.h"
+
+namespace arm_compute
+{
+// Forward declarations
+class ITensor;
+
+/** Basic function to run @ref NEReorgLayerKernel */
+class NEReorgLayer : public INESimpleFunction
+{
+public:
+    /** Initialise the kernel's inputs and outputs
+     *
+     * @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
+     */
+    void configure(const ITensor *input, ITensor *output, int32_t stride);
+
+    /** Static function to check if given info will lead to a valid configuration of @ref NEReorgLayer
+     *
+     * @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
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, int32_t stride);
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_NEREORGLAYER_H__ */