COMPMID-1723: CL: Implement Reverse

Change-Id: Id0d4a07af24e2331161996083b0c1bab072bd405
Reviewed-on: https://review.mlplatform.org/322
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

15 files changed, 838 insertions, 6 deletions

diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h
index 0fe1e04ccd..201679347a 100644
--- a/arm_compute/core/CL/CLKernels.h
+++ b/arm_compute/core/CL/CLKernels.h
@@ -113,6 +113,7 @@
 #include "arm_compute/core/CL/kernels/CLRemapKernel.h"
 #include "arm_compute/core/CL/kernels/CLReorgLayerKernel.h"
 #include "arm_compute/core/CL/kernels/CLReshapeLayerKernel.h"
+#include "arm_compute/core/CL/kernels/CLReverseKernel.h"
 #include "arm_compute/core/CL/kernels/CLScaleKernel.h"
 #include "arm_compute/core/CL/kernels/CLScharr3x3Kernel.h"
 #include "arm_compute/core/CL/kernels/CLSelectKernel.h"
diff --git a/arm_compute/core/CL/kernels/CLChannelShuffleLayerKernel.h b/arm_compute/core/CL/kernels/CLChannelShuffleLayerKernel.h
index f7bd205ec7..7ae77580d2 100644
--- a/arm_compute/core/CL/kernels/CLChannelShuffleLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLChannelShuffleLayerKernel.h
@@ -55,9 +55,9 @@ public:
     void configure(const ICLTensor *input, ICLTensor *output, unsigned int num_groups);
     /** Static function to check if given info will lead to a valid configuration of @ref CLChannelShuffleLayerKernel
      *
-     * @param[in]  input      Input tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
-     * @param[out] output     Output tensor. Data type supported: Same as @p input
-     * @param[in]  num_groups Number of groups. Must be greater than 1 and the number of channels of the tensors must be a multiple of the number of groups.
+     * @param[in] input      Input tensor info. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in] output     Output tensor info. Data type supported: Same as @p input
+     * @param[in] num_groups Number of groups. Must be greater than 1 and the number of channels of the tensors must be a multiple of the number of groups.
      *
      * @return a status
      */
diff --git a/arm_compute/core/CL/kernels/CLReverseKernel.h b/arm_compute/core/CL/kernels/CLReverseKernel.h
new file mode 100644
index 0000000000..e11fbf53b1
--- /dev/null
+++ b/arm_compute/core/CL/kernels/CLReverseKernel.h
@@ -0,0 +1,76 @@
+/*
+ * 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_CLREVERSEKERNEL_H__
+#define __ARM_COMPUTE_CLREVERSEKERNEL_H__
+
+#include "arm_compute/core/CL/ICLKernel.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+/** Interface for the reverse kernel */
+class CLReverseKernel : public ICLKernel
+{
+public:
+    /** Default constructor */
+    CLReverseKernel();
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLReverseKernel(const CLReverseKernel &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLReverseKernel &operator=(const CLReverseKernel &) = delete;
+    /** Allow instances of this class to be moved */
+    CLReverseKernel(CLReverseKernel &&) = default;
+    /** Allow instances of this class to be moved */
+    CLReverseKernel &operator=(CLReverseKernel &&) = default;
+    /** Default destructor */
+    ~CLReverseKernel() = default;
+    /** Initialise the kernel's inputis and output
+     *
+     * @param[in]  input  Input tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[out] output Output tensor. Data type supported: Same as @p input
+     * @param[in]  axis   Axis tensor. Contains the indices of the dimensions to reverse. Data type supported: U32
+     */
+    void configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *axis);
+
+    /** Static function to check if given info will lead to a valid configuration of @ref CLReverseKernel
+     *
+     * @param[in] input  Input tensor info. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in] output Output tensor info. Data type supported: Same as @p input
+     * @param[in] axis   Axis tensor info. Contains the indices of the dimensions to reverse. Data type supported: U32
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *axis);
+
+    // Inherited methods overridden:
+    void run(const Window &window, cl::CommandQueue &queue) override;
+
+public:
+    const ICLTensor *_input;
+    ICLTensor       *_output;
+    const ICLTensor *_axis;
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_CLREVERSEKERNEL_H__ */
diff --git a/arm_compute/runtime/CL/CLFunctions.h b/arm_compute/runtime/CL/CLFunctions.h
index d340ce1e0c..3f2df0efff 100644
--- a/arm_compute/runtime/CL/CLFunctions.h
+++ b/arm_compute/runtime/CL/CLFunctions.h
@@ -115,6 +115,7 @@
 #include "arm_compute/runtime/CL/functions/CLRemap.h"
 #include "arm_compute/runtime/CL/functions/CLReorgLayer.h"
 #include "arm_compute/runtime/CL/functions/CLReshapeLayer.h"
+#include "arm_compute/runtime/CL/functions/CLReverse.h"
 #include "arm_compute/runtime/CL/functions/CLScale.h"
 #include "arm_compute/runtime/CL/functions/CLScharr3x3.h"
 #include "arm_compute/runtime/CL/functions/CLSelect.h"
diff --git a/arm_compute/runtime/CL/functions/CLChannelShuffleLayer.h b/arm_compute/runtime/CL/functions/CLChannelShuffleLayer.h
index 0bcbfd6276..813f410e44 100644
--- a/arm_compute/runtime/CL/functions/CLChannelShuffleLayer.h
+++ b/arm_compute/runtime/CL/functions/CLChannelShuffleLayer.h
@@ -48,9 +48,9 @@ public:
     void configure(const ICLTensor *input, ICLTensor *output, unsigned int num_groups);
     /** Static function to check if given info will lead to a valid configuration of @ref CLChannelShuffleLayerKernel
      *
-     * @param[in]  input      Input tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
-     * @param[out] output     Output tensor. Data type supported: Same as @p input
-     * @param[in]  num_groups Number of groups. Must be greater than 1 and the number of channels of the tensors must be a multiple of the number of groups.
+     * @param[in] input      Input tensor info. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in] output     Output tensor info. Data type supported: Same as @p input
+     * @param[in] num_groups Number of groups. Must be greater than 1 and the number of channels of the tensors must be a multiple of the number of groups.
      *
      * @return a status
      */
diff --git a/arm_compute/runtime/CL/functions/CLReverse.h b/arm_compute/runtime/CL/functions/CLReverse.h
new file mode 100644
index 0000000000..d23e2f4d14
--- /dev/null
+++ b/arm_compute/runtime/CL/functions/CLReverse.h
@@ -0,0 +1,55 @@
+/*
+ * 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_CLREVERSE_H__
+#define __ARM_COMPUTE_CLREVERSE_H__
+
+#include "arm_compute/runtime/CL/ICLSimpleFunction.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+/** Basic function to run @ref CLReverseKernel */
+class CLReverse : public ICLSimpleFunction
+{
+public:
+    /** Initialize the function
+     *
+     * @param[in]  input  Input tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[out] output Output tensor. Data type supported: Same as @p input
+     * @param[in]  axis   Axis tensor. Contains the indices of the dimensions to reverse. Data type supported: U32
+     */
+    void configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *axis);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLReverseKernel
+     *
+     * @param[in] input  Input tensor info. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in] output Output tensor info. Data type supported: Same as @p input
+     * @param[in] axis   Axis tensor info. Contains the indices of the dimensions to reverse. Data type supported: U32
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *axis);
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_CLREVERSE_H__ */
diff --git a/src/core/CL/CLKernelLibrary.cpp b/src/core/CL/CLKernelLibrary.cpp
index e48ff03e05..1361d02f74 100644
--- a/src/core/CL/CLKernelLibrary.cpp
+++ b/src/core/CL/CLKernelLibrary.cpp
@@ -387,6 +387,7 @@ const std::map<std::string, std::string> CLKernelLibrary::_kernel_program_map =
     { "reorg_layer_nhwc", "reorg_layer.cl" },
     { "reshape_layer", "reshape_layer.cl" },
     { "reshape_to_columns", "convolution_layer.cl" },
+    { "reverse", "reverse.cl" },
     { "RGB888_to_IYUV_bt709", "color_convert.cl" },
     { "RGB888_to_NV12_bt709", "color_convert.cl" },
     { "RGB888_to_RGBA8888_bt709", "color_convert.cl" },
@@ -822,6 +823,10 @@ const std::map<std::string, std::string> CLKernelLibrary::_program_source_map =
 #include "./cl_kernels/reshape_layer.clembed"
     },
     {
+        "reverse.cl",
+#include "./cl_kernels/reverse.clembed"
+    },
+    {
         "roi_align_layer.cl",
 #include "./cl_kernels/roi_align_layer.clembed"
     },
diff --git a/src/core/CL/cl_kernels/reverse.cl b/src/core/CL/cl_kernels/reverse.cl
new file mode 100644
index 0000000000..6afd382fec
--- /dev/null
+++ b/src/core/CL/cl_kernels/reverse.cl
@@ -0,0 +1,102 @@
+/*
+* 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.
+*/
+#include "helpers.h"
+
+#if defined(DATA_TYPE) && defined(NUM_REVERSE_DIMS)
+
+#if NUM_REVERSE_DIMS > 4
+#error("Reversing more than 4 dimensions is not currently supported")
+#endif /* NUM_REVERSE_DIMS > 4 */
+
+/** Performs reverse along the specified axis.
+ *
+ * @note The data type must be given as a preprocessor argument using -DDATA_TYPE=num. e.g. -DDATA_TYPE=uint
+ * @note The number of dimensions to reverse must be given as a preprocessor argument using -DNUM_REVERSE_DIMS=num, e.g. -DNUM_REVERSE_DIMS=3
+ *
+ * @param[in]  src_ptr                            Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in]  src_stride_x                       Stride of the first source tensor in X dimension (in bytes)
+ * @param[in]  src_step_x                         src_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  src_stride_y                       Stride of the first source tensor in Y dimension (in bytes)
+ * @param[in]  src_step_y                         src_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in]  src_stride_z                       Stride of the first source tensor in Z dimension (in bytes)
+ * @param[in]  src_step_z                         src_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  src_stride_w                       Stride of the first source tensor in Z dimension (in bytes)
+ * @param[in]  src_step_w                         src_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  src_offset_first_element_in_bytes  The offset of the first element in the first source tensor
+ * @param[in]  axis_ptr                           Pointer to the source vector. Supported data types: U32
+ * @param[in]  axis_stride_x                      Stride of the first source tensor in X dimension (in bytes)
+ * @param[in]  axis_step_x                        src_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  axis_offset_first_element_in_bytes The offset of the first element in the first source tensor
+ * @param[out] dst_ptr                            Pointer to the destination tensor. Supported data types: same as @p src_ptr
+ * @param[in]  dst_stride_x                       Stride of the destination tensor in X dimension (in bytes)
+ * @param[in]  dst_step_x                         output_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  dst_stride_y                       Stride of the destination tensor in Y dimension (in bytes)
+ * @param[in]  dst_step_y                         output_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in]  dst_stride_z                       Stride of the destination tensor in Z dimension (in bytes)
+ * @param[in]  dst_step_z                         output_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  dst_stride_w                       Stride of the destination tensor in Z dimension (in bytes)
+ * @param[in]  dst_step_w                         output_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  dst_offset_first_element_in_bytes  The offset of the first element in the destination tensor
+ */
+__kernel void reverse(TENSOR4D_DECLARATION(src),
+                      VECTOR_DECLARATION(axis),
+                      TENSOR4D_DECLARATION(dst),
+                      const uint width,
+                      const uint height,
+                      const uint depth,
+                      const uint batches)
+{
+    Tensor4D src  = CONVERT_TO_TENSOR4D_STRUCT(src, depth);
+    Vector   axis = CONVERT_TO_VECTOR_STRUCT_NO_STEP(axis);
+    Tensor4D dst  = CONVERT_TO_TENSOR4D_STRUCT_NO_STEP(dst, depth);
+
+    const uint x_in = get_global_id(0);
+    const uint y_in = get_global_id(1);
+    const uint z_in = get_global_id(2) % depth;
+    const uint w_in = get_global_id(2) / depth;
+
+    const uint4 dims       = (uint4)(0, 1, 2, 3);
+    int4        to_reverse = (int4)(0, 0, 0, 0);
+#if NUM_REVERSE_DIMS == 1
+    const uint index = *((__global uint *)axis.ptr);
+    to_reverse       = (uint4)index == dims;
+#elif NUM_REVERSE_DIMS == 2
+    const uint2 indices = vload2(0, (__global uint *)axis.ptr);
+    to_reverse          = ((uint4)indices.s0 == dims) || ((uint4)indices.s1 == dims);
+#elif NUM_REVERSE_DIMS == 3
+    const uint2 indices01 = vload2(0, (__global uint *)axis.ptr);
+    const uint index2     = *((__global uint *)axis.ptr + 2);
+    to_reverse            = ((uint4)indices01.s0 == dims) || ((uint4)indices01.s1 == dims) || ((uint4)index2 == dims);
+#else  /* NUM_REVERSE_DIMS == 3 */
+    const uint4 indices = vload4(0, (__global uint *)axis.ptr);
+    to_reverse          = ((uint4)indices.s0 == dims) || ((uint4)indices.s1 == dims) || ((uint4)indices.s2 == dims) || ((uint4)indices.s3 == dims);
+#endif /* NUM_REVERSE_DIMS == 1 */
+    const uint x_out = to_reverse.s0 ? width - x_in - 1 : x_in;
+    const uint y_out = to_reverse.s1 ? height - y_in - 1 : y_in;
+    const uint z_out = to_reverse.s2 ? depth - z_in - 1 : z_in;
+    const uint w_out = to_reverse.s3 ? batches - w_in - 1 : w_in;
+
+    *((__global DATA_TYPE *)tensor4D_offset(&dst, x_out, y_out, z_out, w_out)) = *((__global DATA_TYPE *)src.ptr);
+}
+#endif // defined(DATA_TYPE) && defined(NUM_REVERSE_DIMS)
diff --git a/src/core/CL/kernels/CLReverseKernel.cpp b/src/core/CL/kernels/CLReverseKernel.cpp
new file mode 100644
index 0000000000..2859a51ce1
--- /dev/null
+++ b/src/core/CL/kernels/CLReverseKernel.cpp
@@ -0,0 +1,135 @@
+/*
+ * 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.
+ */
+#include "arm_compute/core/CL/kernels/CLReverseKernel.h"
+
+#include "arm_compute/core/CL/CLHelpers.h"
+#include "arm_compute/core/CL/CLKernelLibrary.h"
+#include "arm_compute/core/CL/CLValidate.h"
+#include "arm_compute/core/CL/ICLTensor.h"
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/TensorInfo.h"
+#include "arm_compute/core/Window.h"
+
+namespace arm_compute
+{
+namespace
+{
+Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *axis)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output, axis);
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QASYMM8,
+                                                         DataType::U16, DataType::S16,
+                                                         DataType::U32, DataType::S32,
+                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(axis, 1, DataType::U32);
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(axis->num_dimensions() > 1, "Axis must be a 1D tensor");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(axis->dimension(0) > 4, "Only up to 4 dimensions can be reversed");
+
+    // Checks performed when output is configured
+    if(output->total_size() != 0)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+    }
+
+    return Status{};
+}
+} // namespace
+
+CLReverseKernel::CLReverseKernel()
+    : _input(nullptr), _output(nullptr), _axis(nullptr)
+{
+}
+
+void CLReverseKernel::configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *axis)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output, axis);
+
+    _input  = input;
+    _output = output;
+    _axis   = axis;
+
+    // Output tensor auto initialization if not yet initialized
+    auto_init_if_empty(*output->info(), *input->info()->clone());
+
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), axis->info()));
+
+    // Set kernel build options
+    CLBuildOptions build_opts;
+    build_opts.add_option("-DNUM_REVERSE_DIMS=" + support::cpp11::to_string(axis->info()->dimension(0)));
+    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+
+    // Create kernel
+    _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("reverse", build_opts.options()));
+
+    // Set static kernel arguments
+    unsigned int idx = 2 * num_arguments_per_4D_tensor() + num_arguments_per_1D_tensor();
+    add_argument<cl_uint>(idx, input->info()->dimension(0));
+    add_argument<cl_uint>(idx, input->info()->dimension(1));
+    add_argument<cl_uint>(idx, input->info()->dimension(2));
+    add_argument<cl_uint>(idx, input->info()->dimension(3));
+
+    // Configure kernel window
+    Window win = calculate_max_window(*output->info(), Steps());
+    ICLKernel::configure_internal(win);
+
+    // Set config_id for enabling LWS tuning
+    _config_id += "reverse_";
+    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(input->info()->dimension(0));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(input->info()->dimension(1));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(input->info()->dimension(2));
+}
+
+Status CLReverseKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *axis)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, axis));
+    return Status{};
+}
+
+void CLReverseKernel::run(const Window &window, cl::CommandQueue &queue)
+{
+    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
+
+    Window collapsed  = window.collapse(ICLKernel::window(), Window::DimZ);
+    Window slice      = collapsed.first_slice_window_4D();
+    Window axis_slice = collapsed.first_slice_window_1D();
+
+    do
+    {
+        unsigned int idx = 0;
+        add_4D_tensor_argument(idx, _input, slice);
+        add_1D_tensor_argument(idx, _axis, axis_slice);
+        add_4D_tensor_argument(idx, _output, slice);
+        enqueue(queue, *this, slice, lws_hint());
+    }
+    while(collapsed.slide_window_slice_4D(slice));
+}
+} // namespace arm_compute
diff --git a/src/runtime/CL/functions/CLReverse.cpp b/src/runtime/CL/functions/CLReverse.cpp
new file mode 100644
index 0000000000..0f86b9f326
--- /dev/null
+++ b/src/runtime/CL/functions/CLReverse.cpp
@@ -0,0 +1,43 @@
+/*
+ * 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.
+ */
+#include "arm_compute/runtime/CL/functions/CLReverse.h"
+
+#include "arm_compute/core/CL/kernels/CLReverseKernel.h"
+#include "arm_compute/core/Types.h"
+#include "support/ToolchainSupport.h"
+
+namespace arm_compute
+{
+void CLReverse::configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *axis)
+{
+    auto k = arm_compute::support::cpp14::make_unique<CLReverseKernel>();
+    k->configure(input, output, axis);
+    _kernel = std::move(k);
+}
+
+Status CLReverse::validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *axis)
+{
+    return CLReverseKernel::validate(input, output, axis);
+}
+} // namespace arm_compute
diff --git a/tests/datasets/ShapeDatasets.h b/tests/datasets/ShapeDatasets.h
index ffacf34620..f7e7ae26b9 100644
--- a/tests/datasets/ShapeDatasets.h
+++ b/tests/datasets/ShapeDatasets.h
@@ -38,6 +38,20 @@ namespace datasets
 /** Parent type for all for shape datasets. */
 using ShapeDataset = framework::dataset::ContainerDataset<std::vector<TensorShape>>;
 
+/** Data set containing tiny 1D tensor shapes. */
+class Tiny1DShapes final : public ShapeDataset
+{
+public:
+    Tiny1DShapes()
+        : ShapeDataset("Shape",
+    {
+        TensorShape{ 2U },
+                     TensorShape{ 3U },
+    })
+    {
+    }
+};
+
 /** Data set containing small 1D tensor shapes. */
 class Small1DShapes final : public ShapeDataset
 {
diff --git a/tests/validation/CL/Reverse.cpp b/tests/validation/CL/Reverse.cpp
new file mode 100644
index 0000000000..d18c490950
--- /dev/null
+++ b/tests/validation/CL/Reverse.cpp
@@ -0,0 +1,159 @@
+/*
+ * 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.
+ */
+
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/CLTensorAllocator.h"
+#include "arm_compute/runtime/CL/functions/CLReverse.h"
+#include "tests/CL/CLAccessor.h"
+#include "tests/PaddingCalculator.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "tests/validation/Validation.h"
+#include "tests/validation/fixtures/ReverseFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+auto run_small_dataset = combine(datasets::SmallShapes(), datasets::Tiny1DShapes());
+auto run_large_dataset = combine(datasets::LargeShapes(), datasets::Tiny1DShapes());
+
+} // namespace
+TEST_SUITE(CL)
+TEST_SUITE(Reverse)
+
+// *INDENT-OFF*
+// clang-format off
+DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
+        framework::dataset::make("InputInfo", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S8), // Invalid axis datatype
+                                            TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8), // Invalid axis shape
+                                            TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8), // Invalid axis length (> 4)
+                                            TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8), // Mismatching shapes
+                                            TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                            TensorInfo(TensorShape(2U), 1, DataType::U8),
+        }),
+        framework::dataset::make("OutputInfo", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S8),
+                                            TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                            TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                            TensorInfo(TensorShape(2U, 13U, 2U), 1, DataType::U8),
+                                            TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                            TensorInfo(TensorShape(2U), 1, DataType::U8),
+        })),
+        framework::dataset::make("AxisInfo",{ TensorInfo(TensorShape(3U), 1, DataType::U8),
+                                           TensorInfo(TensorShape(2U, 10U), 1, DataType::U32),
+                                           TensorInfo(TensorShape(8U), 1, DataType::U32),
+                                           TensorInfo(TensorShape(2U), 1, DataType::U32),
+                                           TensorInfo(TensorShape(2U), 1, DataType::U32),
+                                           TensorInfo(TensorShape(2U), 1, DataType::U32),
+        })),
+        framework::dataset::make("Expected", { false, false, false, false, true, true})),
+        src_info, dst_info, axis_info, expected)
+{
+    Status s = CLReverse::validate(&src_info.clone()->set_is_resizable(false),
+                                  &dst_info.clone()->set_is_resizable(false),
+                                  &axis_info.clone()->set_is_resizable(false));
+    ARM_COMPUTE_EXPECT(bool(s) == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
+template <typename T>
+using CLReverseFixture = ReverseValidationFixture<CLTensor, CLAccessor, CLReverse, T>;
+
+TEST_SUITE(Float)
+TEST_SUITE(F16)
+FIXTURE_DATA_TEST_CASE(RunSmall,
+                       CLReverseFixture<half>,
+                       framework::DatasetMode::PRECOMMIT,
+                       combine(run_small_dataset, framework::dataset::make("DataType", DataType::F16)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge,
+                       CLReverseFixture<half>,
+                       framework::DatasetMode::NIGHTLY,
+                       combine(run_large_dataset, framework::dataset::make("DataType", DataType::F16)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END() // F16
+
+TEST_SUITE(FP32)
+FIXTURE_DATA_TEST_CASE(RunSmall,
+                       CLReverseFixture<float>,
+                       framework::DatasetMode::PRECOMMIT,
+                       combine(run_small_dataset, framework::dataset::make("DataType", DataType::F32)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge,
+                       CLReverseFixture<float>,
+                       framework::DatasetMode::NIGHTLY,
+                       combine(run_large_dataset, framework::dataset::make("DataType", DataType::F32)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END() // F32
+TEST_SUITE_END() // Float
+
+TEST_SUITE(Quantized)
+TEST_SUITE(QASYMM8)
+FIXTURE_DATA_TEST_CASE(RunSmall,
+                       CLReverseFixture<uint8_t>,
+                       framework::DatasetMode::PRECOMMIT,
+                       combine(run_small_dataset, framework::dataset::make("DataType", DataType::QASYMM8)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge,
+                       CLReverseFixture<uint8_t>,
+                       framework::DatasetMode::NIGHTLY,
+                       combine(run_large_dataset, framework::dataset::make("DataType", DataType::QASYMM8)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END() // QASYMM8
+TEST_SUITE_END() // Quantized
+
+TEST_SUITE_END() // Reverse
+TEST_SUITE_END() // CL
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/fixtures/ReverseFixture.h b/tests/validation/fixtures/ReverseFixture.h
new file mode 100644
index 0000000000..172357c4b7
--- /dev/null
+++ b/tests/validation/fixtures/ReverseFixture.h
@@ -0,0 +1,121 @@
+/*
+ * 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_TEST_REVERSE_FIXTURE
+#define ARM_COMPUTE_TEST_REVERSE_FIXTURE
+
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/TensorShape.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/Tensor.h"
+#include "tests/AssetsLibrary.h"
+#include "tests/Globals.h"
+#include "tests/IAccessor.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Fixture.h"
+#include "tests/validation/reference/Reverse.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ReverseValidationFixture : public framework::Fixture
+{
+public:
+    template <typename...>
+    void setup(TensorShape shape, TensorShape axis_shape, DataType data_type)
+    {
+        _target    = compute_target(shape, axis_shape, data_type);
+        _reference = compute_reference(shape, axis_shape, data_type);
+    }
+
+protected:
+    template <typename U>
+    void fill(U &&tensor)
+    {
+        library->fill_tensor_uniform(tensor, 0);
+    }
+    template <typename U>
+    void fill_axis(U &&tensor)
+    {
+        std::uniform_int_distribution<> distribution(0, 3);
+        library->fill(tensor, distribution, 0);
+    }
+
+    TensorType compute_target(const TensorShape &shape, const TensorShape &axis_shape, DataType data_type)
+    {
+        // Create tensors
+        TensorType src  = create_tensor<TensorType>(shape, data_type, 1);
+        TensorType axis = create_tensor<TensorType>(axis_shape, DataType::U32, 1);
+        TensorType dst;
+
+        // Create and configure function
+        FunctionType reverse_func;
+        reverse_func.configure(&src, &dst, &axis);
+
+        ARM_COMPUTE_EXPECT(src.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(axis.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(dst.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+        // Allocate tensors
+        src.allocator()->allocate();
+        axis.allocator()->allocate();
+        dst.allocator()->allocate();
+
+        ARM_COMPUTE_EXPECT(!src.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!axis.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!dst.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+        // Fill tensors
+        fill(AccessorType(src));
+        fill_axis(AccessorType(axis));
+
+        // Compute function
+        reverse_func.run();
+
+        return dst;
+    }
+
+    SimpleTensor<T> compute_reference(const TensorShape &shape, const TensorShape &axis_shape, DataType data_type)
+    {
+        // Create reference
+        SimpleTensor<T>        src{ shape, data_type };
+        SimpleTensor<uint32_t> axis{ axis_shape, DataType::U32 };
+
+        // Fill reference
+        fill(src);
+        fill_axis(axis);
+
+        return reference::reverse<T>(src, axis);
+    }
+
+    TensorType      _target{};
+    SimpleTensor<T> _reference{};
+};
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_TEST_REVERSE_FIXTURE */
diff --git a/tests/validation/reference/Reverse.cpp b/tests/validation/reference/Reverse.cpp
new file mode 100644
index 0000000000..1662dc24a0
--- /dev/null
+++ b/tests/validation/reference/Reverse.cpp
@@ -0,0 +1,77 @@
+/*
+ * 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.
+ */
+#include "Reverse.h"
+
+#include "arm_compute/core/Types.h"
+#include "tests/validation/Helpers.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace reference
+{
+template <typename T>
+SimpleTensor<T> reverse(const SimpleTensor<T> &src, const SimpleTensor<uint32_t> &axis)
+{
+    ARM_COMPUTE_ERROR_ON(axis.shape().num_dimensions() > 1);
+    ARM_COMPUTE_ERROR_ON(axis.shape().x() > 4);
+
+    // Create reference
+    SimpleTensor<T> dst{ src.shape(), src.data_type(), src.num_channels(), src.quantization_info() };
+
+    const unsigned int width   = src.shape()[0];
+    const unsigned int height  = src.shape()[1];
+    const unsigned int depth   = src.shape()[2];
+    const unsigned int batches = src.shape()[3];
+
+    std::array<bool, 4> to_reverse = { { false, false, false, false } };
+    for(int i = 0; i < axis.num_elements(); ++i)
+    {
+        to_reverse[axis[i]] = true;
+    }
+
+    for(int i = 0; i < src.num_elements(); ++i)
+    {
+        const Coordinates  src_coord = index2coord(src.shape(), i);
+        const unsigned int dst_x     = to_reverse[0] ? width - src_coord[0] - 1 : src_coord[0];
+        const unsigned int dst_y     = to_reverse[1] ? height - src_coord[1] - 1 : src_coord[1];
+        const unsigned int dst_z     = to_reverse[2] ? depth - src_coord[2] - 1 : src_coord[2];
+        const unsigned int dst_w     = to_reverse[3] ? batches - src_coord[3] - 1 : src_coord[3];
+
+        dst[coord2index(src.shape(), Coordinates(dst_x, dst_y, dst_z, dst_w))] = src[i];
+    }
+
+    return dst;
+}
+
+template SimpleTensor<uint8_t> reverse(const SimpleTensor<uint8_t> &src, const SimpleTensor<uint32_t> &axis);
+template SimpleTensor<half> reverse(const SimpleTensor<half> &src, const SimpleTensor<uint32_t> &axis);
+template SimpleTensor<float> reverse(const SimpleTensor<float> &src, const SimpleTensor<uint32_t> &axis);
+} // namespace reference
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/reference/Reverse.h b/tests/validation/reference/Reverse.h
new file mode 100644
index 0000000000..a5ff7b066f
--- /dev/null
+++ b/tests/validation/reference/Reverse.h
@@ -0,0 +1,43 @@
+/*
+ * 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_TEST_REVERSE_H__
+#define __ARM_COMPUTE_TEST_REVERSE_H__
+
+#include "tests/SimpleTensor.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace reference
+{
+template <typename T>
+SimpleTensor<T> reverse(const SimpleTensor<T> &src, const SimpleTensor<uint32_t> &axis);
+} // namespace reference
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_TEST_REVERSE_H__ */