COMPMID-1235: Implements StridedSlice for CL

Change-Id: If2b44da31fae528c76be742b4b3a21fb0eb06b49
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/145284
Reviewed-by: Giuseppe Rossini <giuseppe.rossini@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>

7 files changed, 278 insertions, 0 deletions

diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h
index ea942bde5d..ff16394a8a 100644
--- a/arm_compute/core/CL/CLKernels.h
+++ b/arm_compute/core/CL/CLKernels.h
@@ -109,6 +109,7 @@
 #include "arm_compute/core/CL/kernels/CLSobel5x5Kernel.h"
 #include "arm_compute/core/CL/kernels/CLSobel7x7Kernel.h"
 #include "arm_compute/core/CL/kernels/CLSoftmaxLayerKernel.h"
+#include "arm_compute/core/CL/kernels/CLStridedSliceKernel.h"
 #include "arm_compute/core/CL/kernels/CLTableLookupKernel.h"
 #include "arm_compute/core/CL/kernels/CLThresholdKernel.h"
 #include "arm_compute/core/CL/kernels/CLTransposeKernel.h"
diff --git a/arm_compute/core/CL/kernels/CLStridedSliceKernel.h b/arm_compute/core/CL/kernels/CLStridedSliceKernel.h
new file mode 100644
index 0000000000..e104dcfdd7
--- /dev/null
+++ b/arm_compute/core/CL/kernels/CLStridedSliceKernel.h
@@ -0,0 +1,97 @@
+/*
+ * 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_CL_STRIDED_SLICE_KERNEL_H__
+#define __ARM_COMPUTE_CL_STRIDED_SLICE_KERNEL_H__
+
+#include "arm_compute/core/CL/ICLKernel.h"
+#include "arm_compute/core/Types.h"
+
+#include <cstdint>
+
+namespace arm_compute
+{
+// Forward declarations
+class ICLTensor;
+
+/** Interface for the kernel to perform tensor strided slicing */
+class CLStridedSliceKernel : public ICLKernel
+{
+public:
+    /** Default constructor */
+    CLStridedSliceKernel();
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLStridedSliceKernel(const CLStridedSliceKernel &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLStridedSliceKernel &operator=(const CLStridedSliceKernel &) = delete;
+    /** Allow instances of this class to be moved */
+    CLStridedSliceKernel(CLStridedSliceKernel &&) = default;
+    /** Allow instances of this class to be moved */
+    CLStridedSliceKernel &operator=(CLStridedSliceKernel &&) = default;
+    /** Default destructor */
+    ~CLStridedSliceKernel() = default;
+    /** Configure kernel
+     *
+     * @note Supported tensor rank: up to 4
+     *
+     * @param[in]  input            Source tensor. Data type supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
+     * @param[out] output           Destination tensor. Data type supported: Same as @p input
+     * @param[in]  starts           The starts of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in]  ends             The ends of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in]  strides          The strides of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in]  begin_mask       If the ith bit of begin_mask is set, starts[i] is ignored and the fullest possible range in that dimension is used instead.
+     * @param[in]  end_mask         If the ith bit of end_mask is set, ends[i] is ignored and the fullest possible range in that dimension is used instead.
+     * @param[in]  shrink_axis_mask If the ith bit of shrink_axis_mask is set, it implies that the ith specification shrinks the dimensionality by 1.
+     *                              A slice of size 1 starting from starts[i] in the dimension must be preserved.
+     */
+    void configure(const ICLTensor *input, ICLTensor *output,
+                   const Coordinates &starts, const Coordinates &ends, const BiStrides &strides,
+                   int32_t begin_mask, int32_t end_mask, int32_t shrink_axis_mask);
+
+    /** Static function to check if given info will lead to a valid configuration of @ref CLStridedSliceKernel
+     *
+     * @note Supported tensor rank: up to 4
+     *
+     * @param[in] input            Source tensor. Data type supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
+     * @param[in] output           Destination tensor. Data type supported: Same as @p input
+     * @param[in] starts           The starts of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in] ends             The ends of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in] strides          The strides of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in] begin_mask       If the ith bit of begin_mask is set, starts[i] is ignored and the fullest possible range in that dimension is used instead.
+     * @param[in] end_mask         If the ith bit of end_mask is set, ends[i] is ignored and the fullest possible range in that dimension is used instead.
+     * @param[in] shrink_axis_mask If the ith bit of shrink_axis_mask is set, it implies that the ith specification shrinks the dimensionality by 1.
+     *                             A slice of size 1 starting from starts[i] in the dimension must be preserved.
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output,
+                           const Coordinates &starts, const Coordinates &ends, const BiStrides &strides,
+                           int32_t begin_mask, int32_t end_mask, int32_t shrink_axis_mask);
+
+    // Inherited methods overridden:
+    void run(const Window &window, cl::CommandQueue &queue) override;
+
+private:
+    const ICLTensor *_input;  /**< Source tensor */
+    ICLTensor       *_output; /**< Destination tensor */
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_CL_STRIDED_SLICE_KERNEL_H__ */
diff --git a/arm_compute/core/Types.h b/arm_compute/core/Types.h
index 37a8850237..31f2c88278 100644
--- a/arm_compute/core/Types.h
+++ b/arm_compute/core/Types.h
@@ -45,6 +45,8 @@ using half = half_float::half;
 
 /** Permutation vector */
 using PermutationVector = Strides;
+/** Bidirectional strides */
+using BiStrides = Coordinates;
 
 /** Image colour formats */
 enum class Format
diff --git a/arm_compute/core/utils/helpers/tensor_transform.h b/arm_compute/core/utils/helpers/tensor_transform.h
new file mode 100644
index 0000000000..e5b32a5372
--- /dev/null
+++ b/arm_compute/core/utils/helpers/tensor_transform.h
@@ -0,0 +1,84 @@
+/*
+ * 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_UTILS_HELPERS_TENSOR_TRANSFORM_H__
+#define __ARM_COMPUTE_UTILS_HELPERS_TENSOR_TRANSFORM_H__
+
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+namespace helpers
+{
+namespace tensor_transform
+{
+/** Returns the absolute start coordinates of strided slice
+ *
+ * @param[in] input_shape Input tensor shape
+ * @param[in] starts      Start coordinates
+ * @param[in] strides     Slice strides
+ * @param[in] begin_mask  (Optional) If the ith bit of begin_mask is set, begin[i] is ignored and
+ *                        the fullest possible range in that dimension is used instead.
+ *
+ * @return Absolute start coordinates
+ */
+Coordinates strided_slice_absolute_start_coords(TensorShape input_shape, Coordinates starts, Coordinates strides, int32_t begin_mask = 0);
+
+/** Returns the absolute ends coordinates of strided slice
+ *
+ * @param[in] input_shape      Input tensor shape
+ * @param[in] starts_abs       Absolute start coordinates
+ * @param[in] ends             End coordinates
+ * @param[in] strides          Slice strides
+ * @param[in] end_mask         (Optional) If the ith bit of end_mask is set, end[i] is ignored and
+ *                             the fullest possible range in that dimension is used instead.
+ * @param[in] shrink_axis_mask (Optional) If the ith bit of shrink_axis_mask is set, it implies that the ith specification shrinks the dimensionality by 1.
+ *                             A slice of size 1 starting from begin[i] in the dimension must be preserved.
+ *
+ * @return Absolute end coordinates
+ */
+Coordinates strided_slice_absolute_end_coords(TensorShape input_shape, Coordinates starts_abs, Coordinates ends, Coordinates strides,
+                                              int32_t end_mask = 0, int32_t shrink_axis_mask = 0);
+/** Returns the final strides of strided slice
+ *
+ * @param[in] input_shape Input tensor shape
+ * @param[in] strides     Slice strides
+ *
+ * @return The final strides need by strided slice
+ */
+Coordinates strided_slice_strides(TensorShape input_shape, Coordinates strides);
+
+/** Computes output shape of a strided slice
+ *
+ * @param[in] input_shape   Input tensor shape
+ * @param[in] starts_abs    Absolute start coordinates
+ * @param[in] ends_abs      Absolute end coordinates
+ * @param[in] final_strides Slice strides
+ *
+ * @return The output tensor shape
+ */
+TensorShape compute_strided_slice_output_shape(TensorShape input_shape, Coordinates starts_abs, Coordinates ends_abs, Coordinates final_strides);
+} // namespace tensor_tranform
+} // namespace helpers
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_UTILS_HELPERS_TENSOR_TRANSFORM_H__ */
diff --git a/arm_compute/core/utils/misc/ShapeCalculator.h b/arm_compute/core/utils/misc/ShapeCalculator.h
index 09f558d8b0..e0b6ce639f 100644
--- a/arm_compute/core/utils/misc/ShapeCalculator.h
+++ b/arm_compute/core/utils/misc/ShapeCalculator.h
@@ -28,6 +28,8 @@
 #include "arm_compute/core/ITensorInfo.h"
 #include "arm_compute/core/Utils.h"
 
+#include "arm_compute/core/utils/helpers/tensor_transform.h"
+
 #include <cmath>
 
 namespace arm_compute
@@ -432,6 +434,22 @@ inline TensorShape compute_mm_shape(const ITensorInfo &input0, const ITensorInfo
     return output_shape;
 }
 
+inline TensorShape compute_strided_slice_shape(const ITensorInfo &input,
+                                               const Coordinates &starts, const Coordinates &ends, const Coordinates &strides,
+                                               int32_t begin_mask, int32_t end_mask, int32_t shrink_axis_mask)
+{
+    using namespace arm_compute::helpers::tensor_transform;
+
+    const TensorShape &input_shape = input.tensor_shape();
+
+    // Get actual start, end coordinates and strides
+    const Coordinates final_strides = strided_slice_strides(input_shape, strides);
+    const Coordinates starts_abs    = strided_slice_absolute_start_coords(input_shape, starts, final_strides, begin_mask);
+    const Coordinates ends_abs      = strided_slice_absolute_end_coords(input_shape, starts_abs, ends, final_strides, end_mask, shrink_axis_mask);
+
+    return compute_strided_slice_output_shape(input_shape, starts_abs, ends_abs, final_strides);
+}
+
 template <typename T>
 inline TensorShape extract_shape(T *data)
 {
diff --git a/arm_compute/runtime/CL/CLFunctions.h b/arm_compute/runtime/CL/CLFunctions.h
index 5e42715c2f..f81a1babf9 100644
--- a/arm_compute/runtime/CL/CLFunctions.h
+++ b/arm_compute/runtime/CL/CLFunctions.h
@@ -109,6 +109,7 @@
 #include "arm_compute/runtime/CL/functions/CLSobel5x5.h"
 #include "arm_compute/runtime/CL/functions/CLSobel7x7.h"
 #include "arm_compute/runtime/CL/functions/CLSoftmaxLayer.h"
+#include "arm_compute/runtime/CL/functions/CLStridedSlice.h"
 #include "arm_compute/runtime/CL/functions/CLTableLookup.h"
 #include "arm_compute/runtime/CL/functions/CLThreshold.h"
 #include "arm_compute/runtime/CL/functions/CLTranspose.h"
diff --git a/arm_compute/runtime/CL/functions/CLStridedSlice.h b/arm_compute/runtime/CL/functions/CLStridedSlice.h
new file mode 100644
index 0000000000..4a336f6fdc
--- /dev/null
+++ b/arm_compute/runtime/CL/functions/CLStridedSlice.h
@@ -0,0 +1,75 @@
+/*
+ * 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_CL_STRIDED_SLICE_H__
+#define __ARM_COMPUTE_CL_STRIDED_SLICE_H__
+
+#include "arm_compute/runtime/CL/ICLSimpleFunction.h"
+
+namespace arm_compute
+{
+// Forward Declarations
+class ICLTensor;
+
+/** Basic function to run @ref CLStridedSliceKernel */
+class CLStridedSlice : public ICLSimpleFunction
+{
+public:
+    /** Configure kernel
+     *
+     * @note Supported tensor rank: up to 4
+     *
+     * @param[in]  input            Source tensor. Data type supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
+     * @param[out] output           Destination tensor. Data type supported: Same as @p input
+     * @param[in]  starts           The starts of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in]  ends             The ends of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in]  strides          The strides of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in]  begin_mask       (Optional) If the ith bit of begin_mask is set, starts[i] is ignored and the fullest possible range in that dimension is used instead.
+     * @param[in]  end_mask         (Optional) If the ith bit of end_mask is set, ends[i] is ignored and the fullest possible range in that dimension is used instead.
+     * @param[in]  shrink_axis_mask (Optional) If the ith bit of shrink_axis_mask is set, it implies that the ith specification shrinks the dimensionality by 1.
+     *                              A slice of size 1 starting from starts[i] in the dimension must be preserved.
+     */
+    void configure(const ICLTensor *input, ICLTensor *output,
+                   const Coordinates &starts, const Coordinates &ends, const BiStrides &strides,
+                   int32_t begin_mask = 0, int32_t end_mask = 0, int32_t shrink_axis_mask = 0);
+
+    /** Static function to check if given info will lead to a valid configuration of @ref CLStridedSlice
+     *
+     * @note Supported tensor rank: up to 4
+     *
+     * @param[in] input            Source tensor. Data type supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
+     * @param[in] output           Destination tensor. Data type supported: Same as @p input
+     * @param[in] starts           The starts of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in] ends             The ends of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in] strides          The strides of the dimensions of the input tensor to be sliced. The length must be of rank(input).
+     * @param[in] begin_mask       (Optional) If the ith bit of begin_mask is set, starts[i] is ignored and the fullest possible range in that dimension is used instead.
+     * @param[in] end_mask         (Optional) If the ith bit of end_mask is set, ends[i] is ignored and the fullest possible range in that dimension is used instead.
+     * @param[in] shrink_axis_mask (Optional) If the ith bit of shrink_axis_mask is set, it implies that the ith specification shrinks the dimensionality by 1.
+     *                             A slice of size 1 starting from starts[i] in the dimension must be preserved.
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output,
+                           const Coordinates &starts, const Coordinates &ends, const BiStrides &strides,
+                           int32_t begin_mask = 0, int32_t end_mask = 0, int32_t shrink_axis_mask = 0);
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_CL_STRIDED_SLICE_H__ */