COMPMID-2240 Implement DEPTH_TO_SPACE for NEON
Change-Id: I705aa0f804093c3628c691e46cca475f2819dc65
Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1198
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Giuseppe Rossini <giuseppe.rossini@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
diff --git a/arm_compute/core/NEON/NEKernels.h b/arm_compute/core/NEON/NEKernels.h
index b9716b1..d19b0c7 100644
--- a/arm_compute/core/NEON/NEKernels.h
+++ b/arm_compute/core/NEON/NEKernels.h
@@ -50,6 +50,7 @@
#include "arm_compute/core/NEON/kernels/NECumulativeDistributionKernel.h"
#include "arm_compute/core/NEON/kernels/NEDepthConcatenateLayerKernel.h"
#include "arm_compute/core/NEON/kernels/NEDepthConvertLayerKernel.h"
+#include "arm_compute/core/NEON/kernels/NEDepthToSpaceLayerKernel.h"
#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayer3x3Kernel.h"
#include "arm_compute/core/NEON/kernels/NEDepthwiseIm2ColKernel.h"
#include "arm_compute/core/NEON/kernels/NEDepthwiseVectorToTensorKernel.h"
diff --git a/arm_compute/core/NEON/kernels/NEDepthToSpaceLayerKernel.h b/arm_compute/core/NEON/kernels/NEDepthToSpaceLayerKernel.h
new file mode 100644
index 0000000..a62da04
--- /dev/null
+++ b/arm_compute/core/NEON/kernels/NEDepthToSpaceLayerKernel.h
@@ -0,0 +1,79 @@
+/*
+ * 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_NEDEPTHTOSPACELAYERKERNEL_H__
+#define __ARM_COMPUTE_NEDEPTHTOSPACELAYERKERNEL_H__
+
+#include "arm_compute/core/NEON/INEKernel.h"
+
+namespace arm_compute
+{
+class ITensor;
+
+/** Interface for the depth to space kernel */
+class NEDepthToSpaceLayerKernel : public INEKernel
+{
+public:
+ const char *name() const override
+ {
+ return "NEDepthToSpaceLayerKernel";
+ }
+ /** Default constructor */
+ NEDepthToSpaceLayerKernel();
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ NEDepthToSpaceLayerKernel(const NEDepthToSpaceLayerKernel &) = delete;
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ NEDepthToSpaceLayerKernel &operator=(const NEDepthToSpaceLayerKernel &) = delete;
+ /** Allow instances of this class to be moved */
+ NEDepthToSpaceLayerKernel(NEDepthToSpaceLayerKernel &&) = default;
+ /** Allow instances of this class to be moved */
+ NEDepthToSpaceLayerKernel &operator=(NEDepthToSpaceLayerKernel &&) = default;
+ /** Default destructor */
+ ~NEDepthToSpaceLayerKernel() = default;
+ /** Initialise the kernel's inputs and output.
+ *
+ * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32.
+ * @param[out] output Tensor output. Data types supported: same as @p input
+ * @param[in] block_shape Block shape x value.
+ */
+ void configure(const ITensor *input, ITensor *output, int32_t block_shape);
+ /** Static function to check if given info will lead to a valid configuration of @ref NEDepthToSpaceLayerKernel.
+ *
+ * @param[in] input Tensor input info. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32.
+ * @param[in] output Tensor output info. Data types supported: same as @p input
+ * @param[in] block_shape Block shape value.
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *input, const ITensorInfo *output, int32_t block_shape);
+
+ // Inherited methods overridden:
+ void run(const Window &window, const ThreadInfo &info) override;
+
+private:
+ const ITensor *_input; /**< Source tensor */
+ ITensor *_output; /**< Destination tensor */
+ int32_t _block_shape; /**< Block shape */
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_NEDEPTHTOSPACELAYERKERNEL_H__ */
diff --git a/arm_compute/core/utils/misc/ShapeCalculator.h b/arm_compute/core/utils/misc/ShapeCalculator.h
index d66c87f..8394bba 100644
--- a/arm_compute/core/utils/misc/ShapeCalculator.h
+++ b/arm_compute/core/utils/misc/ShapeCalculator.h
@@ -975,6 +975,30 @@
return output_shape;
}
+/** Calculate the depth to space output shape of a tensor
+ *
+ * @param[in] input Input tensor info
+ * @param[in] block Block shape value
+ *
+ * @return the calculated shape
+ */
+inline TensorShape compute_depth_to_space_shape(const ITensorInfo *input, int block)
+{
+ ARM_COMPUTE_ERROR_ON(block < 2);
+
+ const DataLayout data_layout = input->data_layout();
+ const int idx_width = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
+ const int idx_height = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
+ const int idx_channel = get_data_layout_dimension_index(data_layout, DataLayoutDimension::CHANNEL);
+
+ TensorShape output_shape{ input->tensor_shape() };
+ output_shape.set(idx_width, input->dimension(idx_width) * block);
+ output_shape.set(idx_height, input->dimension(idx_height) * block);
+ output_shape.set(idx_channel, input->dimension(idx_channel) / (block * block));
+
+ return output_shape;
+}
+
/** Calculate the split output shape of a tensor
*
* @param[in] input Input tensor info
diff --git a/arm_compute/runtime/NEON/NEFunctions.h b/arm_compute/runtime/NEON/NEFunctions.h
index 9094ee4..971ddfa 100644
--- a/arm_compute/runtime/NEON/NEFunctions.h
+++ b/arm_compute/runtime/NEON/NEFunctions.h
@@ -53,6 +53,7 @@
#include "arm_compute/runtime/NEON/functions/NECropResize.h"
#include "arm_compute/runtime/NEON/functions/NEDeconvolutionLayer.h"
#include "arm_compute/runtime/NEON/functions/NEDepthConvertLayer.h"
+#include "arm_compute/runtime/NEON/functions/NEDepthToSpaceLayer.h"
#include "arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h"
#include "arm_compute/runtime/NEON/functions/NEDepthwiseSeparableConvolutionLayer.h"
#include "arm_compute/runtime/NEON/functions/NEDequantizationLayer.h"
diff --git a/arm_compute/runtime/NEON/functions/NEDepthToSpaceLayer.h b/arm_compute/runtime/NEON/functions/NEDepthToSpaceLayer.h
new file mode 100644
index 0000000..54a6dda
--- /dev/null
+++ b/arm_compute/runtime/NEON/functions/NEDepthToSpaceLayer.h
@@ -0,0 +1,59 @@
+/*
+ * 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_NEDEPTHTOSPACELAYER_H__
+#define __ARM_COMPUTE_NEDEPTHTOSPACELAYER_H__
+
+#include "arm_compute/runtime/IFunction.h"
+
+#include "arm_compute/core/NEON/kernels/NEDepthToSpaceLayerKernel.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/NEON/INESimpleFunctionNoBorder.h"
+
+namespace arm_compute
+{
+class ITensor;
+
+/** Basic function to run @ref NEDepthToSpaceLayerKernel. */
+class NEDepthToSpaceLayer : public INESimpleFunctionNoBorder
+{
+public:
+ /** Set the input and output tensors.
+ *
+ * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32.
+ * @param[out] output Tensor output. Data types supported: same as @p input
+ * @param[in] block_shape Block shape value.
+ */
+ void configure(const ITensor *input, ITensor *output, int32_t block_shape);
+ /** Static function to check if given info will lead to a valid configuration of @ref NEDepthToSpaceLayer.
+ *
+ * @param[in] input Tensor input info. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32.
+ * @param[in] output Tensor output info. Data types supported: same as @p input
+ * @param[in] block_shape Block shape x value.
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *input, const ITensorInfo *output, int32_t block_shape);
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_NEDEPTHTOSPACELAYER_H__ */