COMPMID-700: Adds CPPPermute function

Change-Id: I8f6adc67f6e6157710fca540f217d70e6f297217
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/111721
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

20 files changed, 944 insertions, 30 deletions

diff --git a/SConscript b/SConscript
index 91dc60166a..3637858626 100644
--- a/SConscript
+++ b/SConscript
@@ -142,6 +142,9 @@ core_files += Glob('src/core/CPP/kernels/*.cpp')
 core_files += Glob('src/core/utils/*/*.cpp')
 
 runtime_files = Glob('src/runtime/*.cpp')
+runtime_files += Glob('src/runtime/CPP/ICPPSimpleFunction.cpp')
+runtime_files += Glob('src/runtime/CPP/functions/*.cpp')
+
 # CLHarrisCorners uses the Scheduler to run CPP kernels
 runtime_files += Glob('src/runtime/CPP/SingleThreadScheduler.cpp')
 
diff --git a/arm_compute/core/CPP/CPPKernels.h b/arm_compute/core/CPP/CPPKernels.h
index 1eabfa9437..f55f41b0eb 100644
--- a/arm_compute/core/CPP/CPPKernels.h
+++ b/arm_compute/core/CPP/CPPKernels.h
@@ -27,6 +27,7 @@
 /* Header regrouping all the CPP kernels */
 #include "arm_compute/core/CPP/kernels/CPPCornerCandidatesKernel.h"
 #include "arm_compute/core/CPP/kernels/CPPDetectionWindowNonMaximaSuppressionKernel.h"
+#include "arm_compute/core/CPP/kernels/CPPPermuteKernel.h"
 #include "arm_compute/core/CPP/kernels/CPPSortEuclideanDistanceKernel.h"
 
 #endif /* __ARM_COMPUTE_CPPKERNELS_H__ */
diff --git a/arm_compute/core/CPP/kernels/CPPPermuteKernel.h b/arm_compute/core/CPP/kernels/CPPPermuteKernel.h
new file mode 100644
index 0000000000..31011e4a5c
--- /dev/null
+++ b/arm_compute/core/CPP/kernels/CPPPermuteKernel.h
@@ -0,0 +1,93 @@
+/*
+ * 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_CPPPERMUTEKERNEL_H__
+#define __ARM_COMPUTE_CPPPERMUTEKERNEL_H__
+
+#include "arm_compute/core/CPP/ICPPKernel.h"
+
+namespace arm_compute
+{
+class ITensor;
+
+/** CPP kernel to perform tensor permutation.
+ *
+ * Permutes given a permutation vector
+ */
+class CPPPermuteKernel : public ICPPKernel
+{
+public:
+    /** Default constructor */
+    CPPPermuteKernel();
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CPPPermuteKernel(const CPPPermuteKernel &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CPPPermuteKernel &operator=(const CPPPermuteKernel &) = delete;
+    /** Allow instances of this class to be moved */
+    CPPPermuteKernel(CPPPermuteKernel &&) = default;
+    /** Allow instances of this class to be moved */
+    CPPPermuteKernel &operator=(CPPPermuteKernel &&) = default;
+    /** Default destructor */
+    ~CPPPermuteKernel() = default;
+
+    /** Set the input and output of the kernel.
+     *
+     * @param[in]  input  The input tensor to permute. Data types supported: U8/S8/QS8/QASYMM8/U16/S16/QS16/F16/U32/S32/F32
+     * @param[out] output The output tensor. Data types supported: Same as @p input
+     * @param[in]  perm   Permutation vector
+     */
+    void configure(const ITensor *input, ITensor *output, const PermutationVector &perm);
+    /** Static function to check if given info will lead to a valid configuration of @ref CPPPermuteKernel
+     *
+     * @param[in] input  The input tensor to permute. Data types supported: U8/S8/QS8/QASYMM8/U16/S16/QS16/F16/U32/S32/F32
+     * @param[in] output The output tensor. Data types supported: Same as @p input
+     * @param[in] perm   Permutation vector
+     *
+     * @return an error status
+     */
+    static Error validate(const ITensorInfo *input, const ITensorInfo *output, const PermutationVector &perm);
+
+    // Inherited methods overridden:
+    void run(const Window &window, const ThreadInfo &info) override;
+
+private:
+    /** Template function to run the permute
+     *
+     * @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_permute(const Window &window);
+
+    /** Common signature for all the specialised permute functions
+     *
+     * @param[in] window Region on which to execute the kernel.
+     */
+    using PermuteFunctionPtr = void (CPPPermuteKernel::*)(const Window &window);
+
+    PermuteFunctionPtr _func;
+    const ITensor     *_input;
+    ITensor           *_output;
+    PermutationVector  _perm;
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_CPPPERMUTEKERNEL_H__ */
diff --git a/arm_compute/core/Dimensions.h b/arm_compute/core/Dimensions.h
index 912b9d57d7..ae8d6c3503 100644
--- a/arm_compute/core/Dimensions.h
+++ b/arm_compute/core/Dimensions.h
@@ -100,7 +100,20 @@ public:
      *
      * @return The size of the requested dimension.
      */
-    T operator[](size_t dimension) const
+    const T &operator[](size_t dimension) const
+    {
+        ARM_COMPUTE_ERROR_ON(dimension >= num_max_dimensions);
+        return _id[dimension];
+    }
+    /** Generic accessor to get the size of any dimension
+     *
+     * @note Precondition: dimension < Dimensions::num_max_dimensions
+     *
+     * @param[in] dimension Dimension of the wanted size
+     *
+     * @return The size of the requested dimension.
+     */
+    T &operator[](size_t dimension)
     {
         ARM_COMPUTE_ERROR_ON(dimension >= num_max_dimensions);
         return _id[dimension];
diff --git a/arm_compute/core/Helpers.h b/arm_compute/core/Helpers.h
index 13d1f6c99f..1be24e1841 100644
--- a/arm_compute/core/Helpers.h
+++ b/arm_compute/core/Helpers.h
@@ -33,6 +33,7 @@
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
 #include "arm_compute/core/Window.h"
+#include "arm_compute/core/utils/misc/utility.h"
 
 #include <array>
 #include <cstddef>
@@ -459,6 +460,23 @@ inline Strides compute_strides(const ITensorInfo &info)
     return compute_strides(info, info.element_size());
 }
 
+/** Permutes given Dimensions according to a permutation vector
+ *
+ * @warning Validity of permutation is not checked
+ *
+ * @param[in, out] dimensions Dimensions to permute
+ * @param[in]      perm       Permutation vector
+ */
+template <typename T>
+inline void permute(Dimensions<T> &dimensions, const PermutationVector &perm)
+{
+    auto copy_dimensions = utility::make_array<Dimensions<T>::num_max_dimensions>(dimensions.begin(), dimensions.end());
+    for(unsigned int i = 0; i < perm.num_dimensions(); ++i)
+    {
+        dimensions[i] = copy_dimensions[perm[i]];
+    }
+}
+
 /* Auto initialize the tensor info (shape, number of channels, data type and fixed point position) if the current assignment is empty.
  *
  * @param[in,out] info                 Tensor info used to check and assign.
diff --git a/arm_compute/core/Strides.h b/arm_compute/core/Strides.h
index 329fafb5f8..105fdfde4e 100644
--- a/arm_compute/core/Strides.h
+++ b/arm_compute/core/Strides.h
@@ -26,7 +26,6 @@
 
 #include "arm_compute/core/Dimensions.h"
 #include "arm_compute/core/Error.h"
-#include "arm_compute/core/Types.h"
 
 #include <algorithm>
 #include <array>
@@ -58,5 +57,5 @@ public:
     /** Default destructor */
     ~Strides() = default;
 };
-}
+} // namespace arm_compute
 #endif /*__ARM_COMPUTE_STRIDES_H__*/
diff --git a/arm_compute/core/Types.h b/arm_compute/core/Types.h
index 4ecaec1eb9..36ec38ff68 100644
--- a/arm_compute/core/Types.h
+++ b/arm_compute/core/Types.h
@@ -25,6 +25,7 @@
 #define __ARM_COMPUTE_TYPES_H__
 
 #include "arm_compute/core/Coordinates.h"
+#include "arm_compute/core/Strides.h"
 #include "arm_compute/core/TensorShape.h"
 #include "support/Half.h"
 
@@ -38,6 +39,9 @@ namespace arm_compute
 /** 16-bit floating point type */
 using half = half_float::half;
 
+/** Permutation vector */
+using PermutationVector = Strides;
+
 /** Image colour formats */
 enum class Format
 {
diff --git a/arm_compute/core/utils/misc/utility.h b/arm_compute/core/utils/misc/utility.h
new file mode 100644
index 0000000000..898d0cdea8
--- /dev/null
+++ b/arm_compute/core/utils/misc/utility.h
@@ -0,0 +1,72 @@
+/*
+ * 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_MISC_UTILITY_H__
+#define __ARM_COMPUTE_MISC_UTILITY_H__
+
+#include <array>
+
+namespace arm_compute
+{
+namespace utility
+{
+/** @cond */
+template <std::size_t...>
+struct index_sequence
+{
+};
+
+template <std::size_t N, std::size_t... S>
+struct index_sequence_generator : index_sequence_generator < N - 1, N - 1, S... >
+{
+};
+
+template <std::size_t... S>
+struct index_sequence_generator<0u, S...> : index_sequence<S...>
+{
+    using type = index_sequence<S...>;
+};
+
+template <std::size_t N>
+using index_sequence_t = typename index_sequence_generator<N>::type;
+/** @endcond */
+
+namespace detail
+{
+template <std::size_t... S,
+          typename Iterator,
+          typename T = std::array<typename std::iterator_traits<Iterator>::value_type, sizeof...(S)>>
+T make_array(Iterator first, index_sequence<S...>)
+{
+    return T{ { first[S]... } };
+}
+} // namespace detail
+
+template <std::size_t N, typename Iterator>
+std::array<typename std::iterator_traits<Iterator>::value_type, N> make_array(Iterator first, Iterator last)
+{
+    return detail::make_array(first, index_sequence_t<N> {});
+}
+} // namespace misc
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_MISC_UTILITY_H__ */
diff --git a/arm_compute/runtime/CPP/CPPFunctions.h b/arm_compute/runtime/CPP/CPPFunctions.h
new file mode 100644
index 0000000000..1f01ffac82
--- /dev/null
+++ b/arm_compute/runtime/CPP/CPPFunctions.h
@@ -0,0 +1,30 @@
+/*
+ * 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_CPPFUNCTIONS_H__
+#define __ARM_COMPUTE_CPPFUNCTIONS_H__
+
+/* Header regrouping all the CPP functions */
+#include "arm_compute/runtime/CPP/functions/CPPPermute.h"
+
+#endif /* __ARM_COMPUTE_CPPFUNCTIONS_H__ */
diff --git a/arm_compute/runtime/CPP/ICPPSimpleFunction.h b/arm_compute/runtime/CPP/ICPPSimpleFunction.h
new file mode 100644
index 0000000000..d1bd232588
--- /dev/null
+++ b/arm_compute/runtime/CPP/ICPPSimpleFunction.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_ICPPSIMPLEFUNCTION_H__
+#define __ARM_COMPUTE_ICPPSIMPLEFUNCTION_H__
+
+#include "arm_compute/core/CPP/ICPPKernel.h"
+#include "arm_compute/runtime/IFunction.h"
+
+#include <memory>
+
+namespace arm_compute
+{
+/** Basic interface for functions which have a single CPP kernel */
+class ICPPSimpleFunction : public IFunction
+{
+public:
+    /** Constructor */
+    ICPPSimpleFunction();
+
+    // Inherited methods overridden:
+    void run() override final;
+
+protected:
+    std::unique_ptr<ICPPKernel> _kernel; /**< Kernel to run */
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_ICPPSIMPLEFUNCTION_H__ */
diff --git a/arm_compute/runtime/CPP/functions/CPPPermute.h b/arm_compute/runtime/CPP/functions/CPPPermute.h
new file mode 100644
index 0000000000..b7643608ab
--- /dev/null
+++ b/arm_compute/runtime/CPP/functions/CPPPermute.h
@@ -0,0 +1,57 @@
+/*
+ * 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_CPPPERMUTE_H__
+#define __ARM_COMPUTE_CPPPERMUTE_H__
+
+#include "arm_compute/runtime/CPP/ICPPSimpleFunction.h"
+
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+class ITensor;
+
+/** Basic function to run @ref CPPPermuteKernel */
+class CPPPermute : public ICPPSimpleFunction
+{
+public:
+    /** Configure the permute CPP kernel
+     *
+     * @param[in]  input  The input tensor to permute. Data types supported: U8/S8/QS8/QASYMM8/U16/S16/QS16/F16/U32/S32/F32
+     * @param[out] output The output tensor. Data types supported: Same as @p input
+     * @param[in]  perm   Permutation vector
+     */
+    void configure(const ITensor *input, ITensor *output, const PermutationVector &perm);
+    /** Static function to check if given info will lead to a valid configuration of @ref CPPPermute
+     *
+     * @param[in] input  The input tensor to permute. Data types supported: U8/S8/QS8/QASYMM8/U16/S16/QS16/F16/U32/S32/F32
+     * @param[in] output The output tensor. Data types supported: Same as @p input
+     * @param[in] perm   Permutation vector
+     *
+     * @return an error status
+     */
+    static Error validate(const ITensorInfo *input, const ITensorInfo *output, const PermutationVector &perm);
+};
+}
+#endif /* __ARM_COMPUTE_CPPPERMUTE_H__ */
diff --git a/src/core/CPP/kernels/CPPPermuteKernel.cpp b/src/core/CPP/kernels/CPPPermuteKernel.cpp
new file mode 100644
index 0000000000..55cc00ffbf
--- /dev/null
+++ b/src/core/CPP/kernels/CPPPermuteKernel.cpp
@@ -0,0 +1,169 @@
+/*
+ * 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.
+ */
+#include "arm_compute/core/CPP/kernels/CPPPermuteKernel.h"
+
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/ITensor.h"
+#include "arm_compute/core/TensorInfo.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/core/Validate.h"
+
+#include <cstddef>
+#include <cstdint>
+
+using namespace arm_compute;
+
+namespace
+{
+TensorShape get_output_shape(const ITensorInfo *input, const PermutationVector &perm)
+{
+    TensorShape output_shape = input->tensor_shape();
+    permute(output_shape, perm);
+    return output_shape;
+}
+
+Error validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const PermutationVector &perm)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::QASYMM8,
+                                                         DataType::U16, DataType::S16, DataType::QS16,
+                                                         DataType::U32, DataType::S32,
+                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input->num_dimensions() < 3, "Invalid input size!");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(perm.num_dimensions() != 3 && ((perm[0] != 2 && perm[1] != 0 && perm[2] != 1) || (perm[0] != 1 && perm[1] != 2 && perm[2] != 0)),
+                                    "Only [2, 0, 1] and [1, 2, 0] permutation is supported");
+
+    // Validate configured output
+    if(output->total_size() != 0)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), get_output_shape(input, perm));
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_FIXED_POINT(input, output);
+    }
+
+    return Error{};
+}
+} // namespace
+
+template <typename T>
+void CPPPermuteKernel::run_permute(const Window &window)
+{
+    const int output_stride_x = _output->info()->strides_in_bytes().x();
+    const int output_stride_y = _output->info()->strides_in_bytes().y();
+    const int output_stride_z = _output->info()->strides_in_bytes().z();
+
+    Window window_out(window);
+    window_out.set(Window::DimX, Window::Dimension(0, 0, 0));
+    window_out.set(Window::DimY, Window::Dimension(0, 0, 0));
+    window_out.set(Window::DimZ, Window::Dimension(0, 0, 0));
+
+    // Create iterators
+    Iterator in(_input, window);
+    Iterator out(_output, window_out);
+
+    // Run [2, 0, 1] permute
+    if(_perm[0] == 2 && _perm[1] == 0 && _perm[2] == 1)
+    {
+        execute_window_loop(window, [&](const Coordinates & id)
+        {
+            const int idx                             = id.y() * output_stride_z + id.x() * output_stride_y + id.z() * output_stride_x;
+            *(reinterpret_cast<T *>(out.ptr() + idx)) = *(reinterpret_cast<const T *>(in.ptr()));
+        },
+        in, out);
+    }
+    // Run [1, 2, 0] permute
+    else
+    {
+        execute_window_loop(window, [&](const Coordinates & id)
+        {
+            const int idx                             = id.x() * output_stride_z + id.z() * output_stride_y + id.y() * output_stride_x;
+            *(reinterpret_cast<T *>(out.ptr() + idx)) = *(reinterpret_cast<const T *>(in.ptr()));
+        },
+        in, out);
+    }
+}
+
+CPPPermuteKernel::CPPPermuteKernel()
+    : _func(), _input(nullptr), _output(nullptr), _perm()
+{
+}
+
+void CPPPermuteKernel::configure(const ITensor *input, ITensor *output, const PermutationVector &perm)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
+
+    // Output auto inizialitation if not yet initialized
+    auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(get_output_shape(input->info(), perm)));
+
+    // Perform validation step
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), perm));
+
+    _input  = input;
+    _output = output;
+    _perm   = perm;
+
+    switch(input->info()->element_size())
+    {
+        case 1:
+            _func = &CPPPermuteKernel::run_permute<uint8_t>;
+            break;
+        case 2:
+            _func = &CPPPermuteKernel::run_permute<uint16_t>;
+            break;
+        case 4:
+            _func = &CPPPermuteKernel::run_permute<uint32_t>;
+            break;
+        default:
+            ARM_COMPUTE_ERROR("Element size not supported");
+            break;
+    }
+
+    // Configure kernel window
+    Window win = calculate_max_window(*input->info(), Steps());
+
+    // The CPPPermute doesn't need padding so update_window_and_padding() can be skipped
+    Coordinates coord;
+    coord.set_num_dimensions(output->info()->num_dimensions());
+    output->info()->set_valid_region(ValidRegion(coord, output->info()->tensor_shape()));
+
+    ICPPKernel::configure(win);
+}
+
+Error CPPPermuteKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const PermutationVector &perm)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, perm));
+    return Error{};
+}
+
+void CPPPermuteKernel::run(const Window &window, const ThreadInfo &info)
+{
+    ARM_COMPUTE_UNUSED(info);
+    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICPPKernel::window(), window);
+
+    if(_func != nullptr)
+    {
+        (this->*_func)(window);
+    }
+}
diff --git a/src/core/NEON/kernels/NEMinMaxLocationKernel.cpp b/src/core/NEON/kernels/NEMinMaxLocationKernel.cpp
index c7dc03c8c8..ad66acd8fa 100644
--- a/src/core/NEON/kernels/NEMinMaxLocationKernel.cpp
+++ b/src/core/NEON/kernels/NEMinMaxLocationKernel.cpp
@@ -32,6 +32,7 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/core/Validate.h"
 #include "arm_compute/core/Window.h"
+#include "arm_compute/core/utils/misc/utility.h"
 
 #include <algorithm>
 #include <arm_neon.h>
@@ -319,34 +320,14 @@ bool NEMinMaxLocationKernel::is_parallelisable() const
     return false;
 }
 
-template <unsigned int...>
-struct index_seq
-{
-    index_seq()                  = default;
-    index_seq(const index_seq &) = default;
-    index_seq &operator=(const index_seq &) = default;
-    index_seq(index_seq &&) noexcept        = default;
-    index_seq &operator=(index_seq &&) noexcept = default;
-    virtual ~index_seq()                        = default;
-};
-template <unsigned int N, unsigned int... S>
-struct gen_index_seq : gen_index_seq < N - 1, N - 1, S... >
-{
-};
-template <unsigned int... S>
-struct gen_index_seq<0u, S...> : index_seq<S...>
-{
-    using type = index_seq<S...>;
-};
-
-template <class T, unsigned int... N>
-struct NEMinMaxLocationKernel::create_func_table<T, index_seq<N...>>
+template <class T, std::size_t... N>
+struct NEMinMaxLocationKernel::create_func_table<T, utility::index_sequence<N...>>
 {
     static const NEMinMaxLocationKernel::MinMaxLocFunction func_table[sizeof...(N)];
 };
 
-template <class T, unsigned int... N>
-const NEMinMaxLocationKernel::MinMaxLocFunction NEMinMaxLocationKernel::create_func_table<T, index_seq<N...>>::func_table[sizeof...(N)] =
+template <class T, std::size_t... N>
+const NEMinMaxLocationKernel::MinMaxLocFunction NEMinMaxLocationKernel::create_func_table<T, utility::index_sequence<N...>>::func_table[sizeof...(N)] =
 {
     &NEMinMaxLocationKernel::minmax_loc<T, bool(N & 8), bool(N & 4), bool(N & 2), bool(N & 1)>...
 };
@@ -378,13 +359,13 @@ void NEMinMaxLocationKernel::configure(const IImage *input, void *min, void *max
     switch(input->info()->data_type())
     {
         case DataType::U8:
-            _func = create_func_table<uint8_t, gen_index_seq<16>::type>::func_table[table_idx];
+            _func = create_func_table<uint8_t, utility::index_sequence_t<16>>::func_table[table_idx];
             break;
         case DataType::S16:
-            _func = create_func_table<int16_t, gen_index_seq<16>::type>::func_table[table_idx];
+            _func = create_func_table<int16_t, utility::index_sequence_t<16>>::func_table[table_idx];
             break;
         case DataType::F32:
-            _func = create_func_table<float, gen_index_seq<16>::type>::func_table[table_idx];
+            _func = create_func_table<float, utility::index_sequence_t<16>>::func_table[table_idx];
             break;
         default:
             ARM_COMPUTE_ERROR("Unsupported data type");
diff --git a/src/runtime/CPP/ICPPSimpleFunction.cpp b/src/runtime/CPP/ICPPSimpleFunction.cpp
new file mode 100644
index 0000000000..42a2d2228c
--- /dev/null
+++ b/src/runtime/CPP/ICPPSimpleFunction.cpp
@@ -0,0 +1,38 @@
+/*
+ * 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.
+ */
+#include "arm_compute/runtime/CPP/ICPPSimpleFunction.h"
+
+#include "arm_compute/runtime/Scheduler.h"
+
+using namespace arm_compute;
+
+ICPPSimpleFunction::ICPPSimpleFunction() // NOLINT
+    : _kernel()
+{
+}
+
+void ICPPSimpleFunction::run()
+{
+    Scheduler::get().schedule(_kernel.get(), Window::DimY);
+}
diff --git a/src/runtime/CPP/functions/CPPPermute.cpp b/src/runtime/CPP/functions/CPPPermute.cpp
new file mode 100644
index 0000000000..5b92718542
--- /dev/null
+++ b/src/runtime/CPP/functions/CPPPermute.cpp
@@ -0,0 +1,41 @@
+/*
+ * 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.
+ */
+#include "arm_compute/runtime/CPP/functions/CPPPermute.h"
+
+#include "arm_compute/core/CPP/kernels/CPPPermuteKernel.h"
+#include "support/ToolchainSupport.h"
+
+using namespace arm_compute;
+
+void CPPPermute::configure(const ITensor *input, ITensor *output, const PermutationVector &perm)
+{
+    auto k = arm_compute::support::cpp14::make_unique<CPPPermuteKernel>();
+    k->configure(input, output, perm);
+    _kernel = std::move(k);
+}
+
+Error CPPPermute::validate(const ITensorInfo *input, const ITensorInfo *output, const PermutationVector &perm)
+{
+    return CPPPermuteKernel::validate(input, output, perm);
+}
diff --git a/tests/SConscript b/tests/SConscript
index 37ec95f589..240a8fba45 100644
--- a/tests/SConscript
+++ b/tests/SConscript
@@ -87,6 +87,10 @@ files_validation += Glob('validation/reference/*.cpp')
 files_validation += Glob('validation/UNIT/*/*.cpp')
 files_validation += Glob('validation/UNIT/*.cpp')
 
+# Add CPP tests
+filter_pattern = test_env['test_filter']
+files_validation += Glob('validation/CPP/' + filter_pattern)
+
 if env['opencl']:
     Import('opencl')
     filter_pattern = test_env['test_filter']
diff --git a/tests/validation/CPP/Permute.cpp b/tests/validation/CPP/Permute.cpp
new file mode 100644
index 0000000000..7f27f3cfff
--- /dev/null
+++ b/tests/validation/CPP/Permute.cpp
@@ -0,0 +1,122 @@
+/*
+ * 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.
+ */
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CPP/functions/CPPPermute.h"
+#include "arm_compute/runtime/Tensor.h"
+#include "arm_compute/runtime/TensorAllocator.h"
+#include "tests/NEON/Accessor.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/PermuteFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+const auto PermuteParametersSmall = combine(concat(datasets::Small3DShapes(), datasets::Small4DShapes()),
+                                            framework::dataset::make("PermutationVector", { PermutationVector(2U, 0U, 1U), PermutationVector(1U, 2U, 0U) }));
+const auto PermuteParametersLarge = combine(concat(datasets::Large3DShapes(), datasets::Large4DShapes()),
+                                            framework::dataset::make("PermutationVector", { PermutationVector(2U, 0U, 1U), PermutationVector(1U, 2U, 0U) }));
+} // namespace
+TEST_SUITE(CPP)
+TEST_SUITE(Permute)
+
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::Small4DShapes(), framework::dataset::make("DataType", { DataType::S8, DataType::U8, DataType::S16, DataType::U16, DataType::U32, DataType::S32, DataType::F16, DataType::F32 })),
+               shape, data_type)
+{
+    // Define permutation vector
+    const PermutationVector perm(2U, 0U, 1U);
+
+    // Permute shapes
+    TensorShape output_shape = shape;
+    permute(output_shape, perm);
+
+    // Create tensors
+    Tensor ref_src = create_tensor<Tensor>(shape, data_type);
+    Tensor dst     = create_tensor<Tensor>(output_shape, data_type);
+
+    // Create and Configure function
+    CPPPermute perm_func;
+    perm_func.configure(&ref_src, &dst, perm);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(output_shape);
+    validate(dst.info()->valid_region(), valid_region);
+}
+
+template <typename T>
+using CPPPermuteFixture = PermuteValidationFixture<Tensor, Accessor, CPPPermute, T>;
+
+TEST_SUITE(U8)
+FIXTURE_DATA_TEST_CASE(RunSmall, CPPPermuteFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(PermuteParametersSmall, framework::dataset::make("DataType", DataType::U8)))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, CPPPermuteFixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(PermuteParametersLarge, framework::dataset::make("DataType", DataType::U8)))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+TEST_SUITE(U16)
+FIXTURE_DATA_TEST_CASE(RunSmall, CPPPermuteFixture<uint16_t>, framework::DatasetMode::PRECOMMIT, combine(PermuteParametersSmall, framework::dataset::make("DataType", DataType::U16)))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, CPPPermuteFixture<uint16_t>, framework::DatasetMode::NIGHTLY, combine(PermuteParametersLarge, framework::dataset::make("DataType", DataType::U16)))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+TEST_SUITE(U32)
+FIXTURE_DATA_TEST_CASE(RunSmall, CPPPermuteFixture<uint32_t>, framework::DatasetMode::PRECOMMIT, combine(PermuteParametersSmall, framework::dataset::make("DataType", DataType::U32)))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, CPPPermuteFixture<uint32_t>, framework::DatasetMode::NIGHTLY, combine(PermuteParametersLarge, framework::dataset::make("DataType", DataType::U32)))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+TEST_SUITE_END()
+TEST_SUITE_END()
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/fixtures/PermuteFixture.h b/tests/validation/fixtures/PermuteFixture.h
new file mode 100644
index 0000000000..3aae384706
--- /dev/null
+++ b/tests/validation/fixtures/PermuteFixture.h
@@ -0,0 +1,112 @@
+/*
+ * 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_TEST_PERMUTE_FIXTURE
+#define ARM_COMPUTE_TEST_PERMUTE_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/Permute.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class PermuteValidationFixture : public framework::Fixture
+{
+public:
+    template <typename...>
+    void setup(TensorShape shape, PermutationVector perm, DataType data_type)
+    {
+        _target    = compute_target(shape, data_type, perm);
+        _reference = compute_reference(shape, data_type, perm);
+    }
+
+protected:
+    template <typename U>
+    void fill(U &&tensor)
+    {
+        library->fill_tensor_uniform(tensor, 0);
+    }
+
+    TensorType compute_target(const TensorShape &shape, DataType data_type, PermutationVector perm)
+    {
+        // Permute shapes
+        TensorShape output_shape = shape;
+        permute(output_shape, perm);
+
+        // Create tensors
+        TensorType src = create_tensor<TensorType>(shape, data_type);
+        TensorType dst = create_tensor<TensorType>(output_shape, data_type);
+
+        // Create and configure function
+        FunctionType perm_func;
+        perm_func.configure(&src, &dst, perm);
+
+        ARM_COMPUTE_EXPECT(src.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(dst.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+        // Allocate tensors
+        src.allocator()->allocate();
+        dst.allocator()->allocate();
+
+        ARM_COMPUTE_EXPECT(!src.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!dst.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+        // Fill tensors
+        fill(AccessorType(src));
+
+        // Compute function
+        perm_func.run();
+
+        return dst;
+    }
+
+    SimpleTensor<T> compute_reference(const TensorShape &shape, DataType data_type, PermutationVector perm)
+    {
+        // Create reference
+        SimpleTensor<T> src{ shape, data_type };
+
+        // Fill reference
+        fill(src);
+
+        return reference::permute<T>(src, perm);
+    }
+
+    TensorType      _target{};
+    SimpleTensor<T> _reference{};
+};
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_TEST_PERMUTE_FIXTURE */
diff --git a/tests/validation/reference/Permute.cpp b/tests/validation/reference/Permute.cpp
new file mode 100644
index 0000000000..4a12ca6959
--- /dev/null
+++ b/tests/validation/reference/Permute.cpp
@@ -0,0 +1,66 @@
+/*
+ * 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.
+ */
+#include "Permute.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> permute(const SimpleTensor<T> &src, PermutationVector perm)
+{
+    // Permute shapes
+    TensorShape dst_shape = src.shape();
+    permute(dst_shape, perm);
+
+    // Create reference
+    SimpleTensor<T> dst{ dst_shape, src.data_type() };
+
+    // Compute reference
+    for(int i = 0; i < src.num_elements(); ++i)
+    {
+        Coordinates coord = index2coord(src.shape(), i);
+        permute(coord, perm);
+        const size_t dst_index = coord2index(dst.shape(), coord);
+
+        dst[dst_index] = src[i];
+    }
+
+    return dst;
+}
+
+template SimpleTensor<uint8_t> permute(const SimpleTensor<uint8_t> &src, PermutationVector perm);
+template SimpleTensor<uint16_t> permute(const SimpleTensor<uint16_t> &src, PermutationVector perm);
+template SimpleTensor<uint32_t> permute(const SimpleTensor<uint32_t> &src, PermutationVector perm);
+} // namespace reference
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/reference/Permute.h b/tests/validation/reference/Permute.h
new file mode 100644
index 0000000000..295ad2b2ef
--- /dev/null
+++ b/tests/validation/reference/Permute.h
@@ -0,0 +1,43 @@
+/*
+ * 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_TEST_PERMUTE_H__
+#define __ARM_COMPUTE_TEST_PERMUTE_H__
+
+#include "tests/SimpleTensor.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace reference
+{
+template <typename T>
+SimpleTensor<T> permute(const SimpleTensor<T> &src, PermutationVector perm);
+} // namespace reference
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_TEST_PERMUTE_H__ */