aboutsummaryrefslogtreecommitdiff
path: root/arm_compute/core
diff options
context:
space:
mode:
authorGeorgios Pinitas <georgios.pinitas@arm.com>2017-09-07 17:29:16 +0100
committerAnthony Barbier <anthony.barbier@arm.com>2018-11-02 16:35:24 +0000
commit5ee66eae5dff60f615dfe9427cf1c4116d95f478 (patch)
treef2922f584deb20e92404edcd6fba14f6df59c0d8 /arm_compute/core
parentce54b56e0d91a8e73f3ecfede6a2b2aa323aa1fd (diff)
downloadComputeLibrary-5ee66eae5dff60f615dfe9427cf1c4116d95f478.tar.gz
COMPMID-462: Implement TensorReshape for NEON and CL.
Change-Id: I11b39c2ceca26ade73822e29a384ef866ae05729 Reviewed-on: http://mpd-gerrit.cambridge.arm.com/87707 Reviewed-by: Pablo Tello <pablo.tello@arm.com> Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Diffstat (limited to 'arm_compute/core')
-rw-r--r--arm_compute/core/CL/CLKernels.h1
-rw-r--r--arm_compute/core/CL/kernels/CLReshapeLayerKernel.h65
-rw-r--r--arm_compute/core/Helpers.h16
-rw-r--r--arm_compute/core/Helpers.inl37
-rw-r--r--arm_compute/core/NEON/NEKernels.h1
-rw-r--r--arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h48
6 files changed, 168 insertions, 0 deletions
diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h
index 04ea04f04f..f8aa5f8968 100644
--- a/arm_compute/core/CL/CLKernels.h
+++ b/arm_compute/core/CL/CLKernels.h
@@ -82,6 +82,7 @@
#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/CLReshapeLayerKernel.h"
#include "arm_compute/core/CL/kernels/CLScaleKernel.h"
#include "arm_compute/core/CL/kernels/CLScharr3x3Kernel.h"
#include "arm_compute/core/CL/kernels/CLSobel3x3Kernel.h"
diff --git a/arm_compute/core/CL/kernels/CLReshapeLayerKernel.h b/arm_compute/core/CL/kernels/CLReshapeLayerKernel.h
new file mode 100644
index 0000000000..b6efc3db1e
--- /dev/null
+++ b/arm_compute/core/CL/kernels/CLReshapeLayerKernel.h
@@ -0,0 +1,65 @@
+/*
+ * Copyright (c) 2017 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_CLRESHAPELAYERKERNEL_H__
+#define __ARM_COMPUTE_CLRESHAPELAYERKERNEL_H__
+
+#include "arm_compute/core/CL/ICLKernel.h"
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+/** Interface for the kernel to perform tensor reshaping */
+class CLReshapeLayerKernel : public ICLKernel
+{
+public:
+ /** Default constructor */
+ CLReshapeLayerKernel();
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ CLReshapeLayerKernel(const CLReshapeLayerKernel &) = delete;
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ CLReshapeLayerKernel &operator=(const CLReshapeLayerKernel &) = delete;
+ /** Allow instances of this class to be moved */
+ CLReshapeLayerKernel(CLReshapeLayerKernel &&) = default;
+ /** Allow instances of this class to be moved */
+ CLReshapeLayerKernel &operator=(CLReshapeLayerKernel &&) = default;
+ /** Default destructor */
+ ~CLReshapeLayerKernel() = default;
+ /** Set the input and output of the kernel
+ *
+ * @param[in] input Source tensor. Data type supported: U8/S8/QS8/U16/S16/QS16/U32/S32/F16/F32
+ * @param[out] output Destination tensor. Data type supported: Same as @p input
+ */
+ void configure(const ICLTensor *input, ICLTensor *output);
+
+ // Inherited methods overridden:
+ void run(const Window &window, cl::CommandQueue &queue) override;
+
+private:
+ const ICLTensor *_input; /**< Source tensor */
+ ICLTensor *_output; /**< Destination tensor */
+};
+}
+#endif /*__ARM_COMPUTE_CLRESHAPELAYERKERNEL_H__ */
diff --git a/arm_compute/core/Helpers.h b/arm_compute/core/Helpers.h
index f3702e7c93..b6461bc47a 100644
--- a/arm_compute/core/Helpers.h
+++ b/arm_compute/core/Helpers.h
@@ -472,6 +472,22 @@ bool set_fixed_point_position_if_zero(ITensorInfo &info, int fixed_point_positio
* @return The corrispondent valid region
*/
ValidRegion calculate_valid_region_scale(const ITensorInfo &src_info, const TensorShape &dst_shape, InterpolationPolicy policy, BorderSize border_size, bool border_undefined);
+/** Convert a linear index into n-dimensional coordinates.
+ *
+ * @param[in] shape Shape of the n-dimensional tensor.
+ * @param[in] index Linear index specifying the i-th element.
+ *
+ * @return n-dimensional coordinates.
+ */
+inline Coordinates index2coords(const TensorShape &shape, int index);
+/** Convert n-dimensional coordinates into a linear index.
+ *
+ * @param[in] shape Shape of the n-dimensional tensor.
+ * @param[in] coord N-dimensional coordinates.
+ *
+ * @return linead index
+ */
+inline int coords2index(const TensorShape &shape, const Coordinates &coord);
} // namespace arm_compute
#include "arm_compute/core/Helpers.inl"
diff --git a/arm_compute/core/Helpers.inl b/arm_compute/core/Helpers.inl
index e20bdb58a1..c2ca3b44b3 100644
--- a/arm_compute/core/Helpers.inl
+++ b/arm_compute/core/Helpers.inl
@@ -333,4 +333,41 @@ inline ValidRegion calculate_valid_region_scale(const ITensorInfo &src_info, con
return ValidRegion(std::move(anchor), std::move(new_dst_shape));
}
+
+inline Coordinates index2coords(const TensorShape &shape, int index)
+{
+ int num_elements = shape.total_size();
+
+ ARM_COMPUTE_ERROR_ON_MSG(index < 0 || index >= num_elements, "Index has to be in [0, num_elements]!");
+ ARM_COMPUTE_ERROR_ON_MSG(num_elements == 0, "Cannot create coordinate from empty shape!");
+
+ Coordinates coord{ 0 };
+
+ for(int d = shape.num_dimensions() - 1; d >= 0; --d)
+ {
+ num_elements /= shape[d];
+ coord.set(d, index / num_elements);
+ index %= num_elements;
+ }
+
+ return coord;
+}
+
+inline int coords2index(const TensorShape &shape, const Coordinates &coord)
+{
+ int num_elements = shape.total_size();
+ ARM_COMPUTE_UNUSED(num_elements);
+ ARM_COMPUTE_ERROR_ON_MSG(num_elements == 0, "Cannot create linear index from empty shape!");
+
+ int index = 0;
+ int stride = 1;
+
+ for(unsigned int d = 0; d < coord.num_dimensions(); ++d)
+ {
+ index += coord[d] * stride;
+ stride *= shape[d];
+ }
+
+ return index;
+}
} // namespace arm_compute
diff --git a/arm_compute/core/NEON/NEKernels.h b/arm_compute/core/NEON/NEKernels.h
index a38282c289..6fa5b5d0a4 100644
--- a/arm_compute/core/NEON/NEKernels.h
+++ b/arm_compute/core/NEON/NEKernels.h
@@ -88,6 +88,7 @@
#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/NEReshapeLayerKernel.h"
#include "arm_compute/core/NEON/kernels/NEScaleKernel.h"
#include "arm_compute/core/NEON/kernels/NEScharr3x3Kernel.h"
#include "arm_compute/core/NEON/kernels/NESobel3x3Kernel.h"
diff --git a/arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h b/arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h
new file mode 100644
index 0000000000..6507610b1d
--- /dev/null
+++ b/arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h
@@ -0,0 +1,48 @@
+/*
+ * Copyright (c) 2017 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_NERESHAPELAYERKERNEL_H__
+#define __ARM_COMPUTE_NERESHAPELAYERKERNEL_H__
+
+#include "arm_compute/core/NEON/INESimpleKernel.h"
+
+namespace arm_compute
+{
+class ITensor;
+
+/** Interface for the kernel to perform tensor reshaping */
+class NEReshapeLayerKernel : public INESimpleKernel
+{
+public:
+ /** Set the input and output of the kernel
+ *
+ * @param[in] input Source tensor. Data type supported: U8/S8/QS8/U16/S16/QS16/U32/S32/F16/F32
+ * @param[out] output Destination tensor. Data type supported: Same as @p input
+ */
+ void configure(const ITensor *input, ITensor *output);
+
+ // Inherited methods overridden:
+ void run(const Window &window, const ThreadInfo &info) override;
+};
+}
+#endif /*__ARM_COMPUTE_NERESHAPELAYERKERNEL_H__ */