COMPMID-1761: NEON: Implement Pack

Change-Id: Icc3392494b1e3361e8fd925da200827c494351b3
Reviewed-on: https://review.mlplatform.org/430
Reviewed-by: Manuel Bottini <manuel.bottini@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Giuseppe Rossini <giuseppe.rossini@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

6 files changed, 203 insertions, 5 deletions

diff --git a/arm_compute/core/CL/kernels/CLStackLayerKernel.h b/arm_compute/core/CL/kernels/CLStackLayerKernel.h
index 4d377daf8b..1511a4ed66 100644
--- a/arm_compute/core/CL/kernels/CLStackLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLStackLayerKernel.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -50,6 +50,8 @@ public:
     ~CLStackLayerKernel() = default;
     /** Initialise the kernel's inputs and output
      *
+     * @note Supported input tensor rank: up to 4
+     *
      * @param[in]  input       Input tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
      * @param[in]  axis        The dimension to stack the tensors along. It must be smaller than the number of input dimensions.
      * @param[in]  idx_input   Index of the input tensor in the list of tensors to stack.
@@ -59,7 +61,9 @@ public:
      *
      */
     void configure(const ICLTensor *input, unsigned int axis, unsigned int idx_input, unsigned int num_tensors, ICLTensor *output);
-    /**  Static function to check if given info will lead to a valid configuration of @ref CLStackLayerKernel
+    /** Static function to check if given info will lead to a valid configuration of @ref CLStackLayerKernel
+     *
+     * @note Supported input tensor rank: up to 4
      *
      * @param[in] input       Input tensor info. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
      * @param[in] axis        The dimension to stack the tensors along. It must be smaller than the number of input dimensions.
diff --git a/arm_compute/core/NEON/NEKernels.h b/arm_compute/core/NEON/NEKernels.h
index 26d2acaf5c..a32c507266 100644
--- a/arm_compute/core/NEON/NEKernels.h
+++ b/arm_compute/core/NEON/NEKernels.h
@@ -117,6 +117,7 @@
 #include "arm_compute/core/NEON/kernels/NESobel5x5Kernel.h"
 #include "arm_compute/core/NEON/kernels/NESobel7x7Kernel.h"
 #include "arm_compute/core/NEON/kernels/NESoftmaxLayerKernel.h"
+#include "arm_compute/core/NEON/kernels/NEStackLayerKernel.h"
 #include "arm_compute/core/NEON/kernels/NEStridedSliceKernel.h"
 #include "arm_compute/core/NEON/kernels/NETableLookupKernel.h"
 #include "arm_compute/core/NEON/kernels/NEThresholdKernel.h"
diff --git a/arm_compute/core/NEON/kernels/NEStackLayerKernel.h b/arm_compute/core/NEON/kernels/NEStackLayerKernel.h
new file mode 100644
index 0000000000..3a9e81fa94
--- /dev/null
+++ b/arm_compute/core/NEON/kernels/NEStackLayerKernel.h
@@ -0,0 +1,107 @@
+/*
+ * Copyright (c) 2018-2019 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_NESTACKLAYERKERNEL_H__
+#define __ARM_COMPUTE_NESTACKLAYERKERNEL_H__
+
+#include "arm_compute/core/NEON/INEKernel.h"
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+class ITensor;
+
+/** NEON kernel to stacks a rank-R tensor into one with rank-(R+1) along the axis dimension.*/
+class NEStackLayerKernel : public INEKernel
+{
+public:
+    const char *name() const override
+    {
+        return "NEStackLayerKernel";
+    }
+    /** Default constructor */
+    NEStackLayerKernel();
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    NEStackLayerKernel(const NEStackLayerKernel &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    NEStackLayerKernel &operator=(const NEStackLayerKernel &) = delete;
+    /** Allow instances of this class to be moved */
+    NEStackLayerKernel(NEStackLayerKernel &&) = default;
+    /** Allow instances of this class to be moved */
+    NEStackLayerKernel &operator=(NEStackLayerKernel &&) = default;
+    /** Default destructor */
+    ~NEStackLayerKernel() = default;
+    /** Initialise the kernel's inputs and output
+     *
+     * @note Supported input tensor rank: up to 4
+     *
+     * @param[in]  input       Input tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in]  axis        The dimension to stack the tensors along. It must be smaller than the number of input dimensions.
+     * @param[in]  idx_input   Index of the input tensor in the list of tensors to stack.
+     *                         All tensors in the list must have the same shape
+     * @param[in]  num_tensors Number of tensors to stack
+     * @param[out] output      Output tensor. Data types supported: Same as @p input.
+     *
+     */
+    void configure(const ITensor *input, unsigned int axis, unsigned int idx_input, unsigned int num_tensors, ITensor *output);
+    /**  Static function to check if given info will lead to a valid configuration of @ref NEStackLayerKernel
+     *
+     * @note Supported input tensor rank: up to 4
+     *
+     * @param[in] input       Input tensor info. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in] axis        The dimension to stack the tensors along. It must be smaller than the number of input dimensions.
+     * @param[in] idx_input   Index of the input tensor in the list of tensors to stack
+     *                        All tensors in the list must have the same shape
+     * @param[in] num_tensors Number of tensors to stack
+     * @param[in] output      Output tensor info. Data types supported: Same as @p input.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, unsigned int axis, unsigned int idx_input, unsigned int num_tensors, const ITensorInfo *output);
+
+    // Inherited methods overridden
+    void run(const Window &window, const ThreadInfo &info) override;
+
+private:
+    /** Template function to run the stack
+     *
+     * @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_stack(const Window &window);
+
+    /** Common signature for all the specialised stack functions
+     *
+     * @param[in] window Region on which to execute the kernel.
+     */
+    using StackFunctionPtr = void (NEStackLayerKernel::*)(const Window &window);
+
+    const ITensor *_input;
+    ITensor       *_output;
+    unsigned int   _axis;
+    unsigned int   _idx_input;
+    StackFunctionPtr _func;
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_NESTACKLAYERKERNEL_H__ */
diff --git a/arm_compute/runtime/CL/functions/CLStackLayer.h b/arm_compute/runtime/CL/functions/CLStackLayer.h
index 9794014889..5b821b863a 100644
--- a/arm_compute/runtime/CL/functions/CLStackLayer.h
+++ b/arm_compute/runtime/CL/functions/CLStackLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -48,13 +48,17 @@ public:
     CLStackLayer();
     /** Initialise the kernel's inputs vector and output.
      *
+     * @note Supported input tensor rank: up to 4
+     *
      * @param[in]  input  The vectors containing all the tensors with the same shape to stack. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
      * @param[in]  axis   The dimension to stack the tensors along. It must be smaller than the number of input dimensions.
      *                    Negative values wrap around
      * @param[out] output Output tensor. Data types supported: Same as @p input.
      */
     void configure(const std::vector<ICLTensor *> &input, int axis, ICLTensor *output);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLDepthConcatenateLayer
+    /** Static function to check if given info will lead to a valid configuration of @ref CLStackLayerKernel
+     *
+     * @note Supported input tensor rank: up to 4
      *
      * @param[in] input  The vectors containing all the tensors info with the same shape to stack. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
      * @param[in] axis   The dimension to stack the tensors along. It must be smaller than the number of input dimensions.
@@ -73,5 +77,5 @@ private:
     std::unique_ptr<CLStackLayerKernel[]> _stack_kernels;
     unsigned int                          _num_inputs;
 };
-}
+} // namespace arm_compute
 #endif /* __ARM_COMPUTE_CLSTACKLAYER_H__ */
diff --git a/arm_compute/runtime/NEON/NEFunctions.h b/arm_compute/runtime/NEON/NEFunctions.h
index 2daef70cef..da61853785 100644
--- a/arm_compute/runtime/NEON/NEFunctions.h
+++ b/arm_compute/runtime/NEON/NEFunctions.h
@@ -123,6 +123,7 @@
 #include "arm_compute/runtime/NEON/functions/NESobel7x7.h"
 #include "arm_compute/runtime/NEON/functions/NESoftmaxLayer.h"
 #include "arm_compute/runtime/NEON/functions/NESplit.h"
+#include "arm_compute/runtime/NEON/functions/NEStackLayer.h"
 #include "arm_compute/runtime/NEON/functions/NEStridedSlice.h"
 #include "arm_compute/runtime/NEON/functions/NETableLookup.h"
 #include "arm_compute/runtime/NEON/functions/NEThreshold.h"
diff --git a/arm_compute/runtime/NEON/functions/NEStackLayer.h b/arm_compute/runtime/NEON/functions/NEStackLayer.h
new file mode 100644
index 0000000000..6032dae0cb
--- /dev/null
+++ b/arm_compute/runtime/NEON/functions/NEStackLayer.h
@@ -0,0 +1,81 @@
+/*
+ * Copyright (c) 2018-2019 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_NESTACKLAYER_H__
+#define __ARM_COMPUTE_NESTACKLAYER_H__
+
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/IFunction.h"
+
+#include "arm_compute/core/NEON/kernels/NEStackLayerKernel.h"
+
+#include <memory>
+#include <vector>
+
+namespace arm_compute
+{
+class ITensor;
+
+/** Basic function to stack tensors along an axis. This function calls the following kernel:
+ *
+ * -# @ref NEStackLayerKernel
+ *
+ */
+class NEStackLayer : public IFunction
+{
+public:
+    /** Default constructor */
+    NEStackLayer();
+    /** Initialise the kernel's inputs vector and output.
+     *
+     * @note Supported input tensor rank: up to 4
+     *
+     * @param[in]  input  The vectors containing all the tensors with the same shape to stack. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in]  axis   The dimension to stack the tensors along. It must be smaller than the number of input dimensions.
+     *                    Negative values wrap around
+     * @param[out] output Output tensor. Data types supported: Same as @p input.
+     */
+    void configure(const std::vector<ITensor *> &input, int axis, ITensor *output);
+    /** Static function to check if given info will lead to a valid configuration of @ref NEStackLayerKernel
+     *
+     * @note Supported input tensor rank: up to 4
+     *
+     * @param[in] input  The vectors containing all the tensors info with the same shape to stack. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in] axis   The dimension to stack the tensors along. It must be smaller than the number of input dimensions.
+     *                   Negative values wrap around
+     * @param[in] output Output tensor info. Data types supported: Same as @p input.
+     *
+     * @return a status
+     */
+    static Status validate(const std::vector<ITensorInfo *> &input, int axis, const ITensorInfo *output);
+
+    // Inherited methods overridden:
+    void run() override;
+
+private:
+    std::vector<ITensor *>                _input;
+    std::unique_ptr<NEStackLayerKernel[]> _stack_kernels;
+    unsigned int                          _num_inputs;
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_NESTACKLAYER_H__ */