Make all CL Concatenate kernels and functions state-less

Resolves COMPMID-3995

Change-Id: I84172bed20924f1d9ae3b4d14d7b321e9494296e
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4887
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

13 files changed, 1 insertions, 1336 deletions

diff --git a/src/core/CL/CLKernels.h b/src/core/CL/CLKernels.h
index f23871d4db..11f1d2d7cf 100644
--- a/src/core/CL/CLKernels.h
+++ b/src/core/CL/CLKernels.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2020 Arm Limited.
+ * Copyright (c) 2016-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -29,7 +29,6 @@
 #include "src/core/CL/kernels/CLAccumulateKernel.h"
 #include "src/core/CL/kernels/CLActivationLayerKernel.h"
 #include "src/core/CL/kernels/CLArgMinMaxLayerKernel.h"
-#include "src/core/CL/kernels/CLBatchConcatenateLayerKernel.h"
 #include "src/core/CL/kernels/CLBatchNormalizationLayerKernel.h"
 #include "src/core/CL/kernels/CLBatchToSpaceLayerKernel.h"
 #include "src/core/CL/kernels/CLBitwiseKernel.h"
@@ -48,7 +47,6 @@
 #include "src/core/CL/kernels/CLCropKernel.h"
 #include "src/core/CL/kernels/CLDeconvolutionLayerUpsampleKernel.h"
 #include "src/core/CL/kernels/CLDeconvolutionReshapeOutputKernel.h"
-#include "src/core/CL/kernels/CLDepthConcatenateLayerKernel.h"
 #include "src/core/CL/kernels/CLDepthConvertLayerKernel.h"
 #include "src/core/CL/kernels/CLDepthToSpaceLayerKernel.h"
 #include "src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.h"
@@ -92,7 +90,6 @@
 #include "src/core/CL/kernels/CLHOGDescriptorKernel.h"
 #include "src/core/CL/kernels/CLHOGDetectorKernel.h"
 #include "src/core/CL/kernels/CLHarrisCornersKernel.h"
-#include "src/core/CL/kernels/CLHeightConcatenateLayerKernel.h"
 #include "src/core/CL/kernels/CLHistogramKernel.h"
 #include "src/core/CL/kernels/CLIm2ColKernel.h"
 #include "src/core/CL/kernels/CLInstanceNormalizationLayerKernel.h"
@@ -144,9 +141,6 @@
 #include "src/core/CL/kernels/CLWarpAffineKernel.h"
 #include "src/core/CL/kernels/CLWarpPerspectiveKernel.h"
 #include "src/core/CL/kernels/CLWeightsReshapeKernel.h"
-#include "src/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h"
-#include "src/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h"
-#include "src/core/CL/kernels/CLWidthConcatenateLayerKernel.h"
 #include "src/core/CL/kernels/CLWinogradFilterTransformKernel.h"
 #include "src/core/CL/kernels/CLWinogradInputTransformKernel.h"
 #include "src/core/CL/kernels/CLWinogradOutputTransformKernel.h"
diff --git a/src/core/CL/kernels/CLBatchConcatenateLayerKernel.cpp b/src/core/CL/kernels/CLBatchConcatenateLayerKernel.cpp
deleted file mode 100644
index ccd6a5a0fc..0000000000
--- a/src/core/CL/kernels/CLBatchConcatenateLayerKernel.cpp
+++ /dev/null
@@ -1,150 +0,0 @@
-/*
- * Copyright (c) 2019-2020 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 "src/core/CL/kernels/CLBatchConcatenateLayerKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Utils.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/Cast.h"
-
-#include "support/StringSupport.h"
-
-namespace arm_compute
-{
-namespace
-{
-Status validate_arguments(const ITensorInfo *input, unsigned int batch_offset, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimX) != output->dimension(Window::DimX));
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimY) != output->dimension(Window::DimY));
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimZ) != output->dimension(Window::DimZ));
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(3) + batch_offset > output->dimension(3));
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(4, input, output);
-
-    return Status{};
-}
-} // namespace
-
-CLBatchConcatenateLayerKernel::CLBatchConcatenateLayerKernel()
-    : _batch_offset(0)
-{
-}
-
-void CLBatchConcatenateLayerKernel::configure(const CLCompileContext &compile_context, ITensorInfo *input, unsigned int batch_offset, ITensorInfo *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input, batch_offset, output));
-
-    auto padding_info = get_padding_info({ input, output });
-
-    _batch_offset = batch_offset;
-
-    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16 / input->element_size(), input->dimension(0));
-
-    // Add build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->data_type()));
-    build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
-    build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(input->dimension(0) % num_elems_processed_per_iteration));
-    if(is_data_type_quantized_asymmetric(input->data_type()) && input->quantization_info() != output->quantization_info())
-    {
-        const UniformQuantizationInfo iq_info = input->quantization_info().uniform();
-        const UniformQuantizationInfo oq_info = output->quantization_info().uniform();
-
-        build_opts.add_option("-DOFFSET_IN1=" + float_to_string_with_full_precision(iq_info.offset));
-        build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oq_info.offset));
-        build_opts.add_option("-DSCALE_IN1=" + float_to_string_with_full_precision(iq_info.scale));
-        build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oq_info.scale));
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "concatenate", build_opts.options());
-
-    // Configure kernel window
-    auto win = calculate_max_window(*output, Steps(num_elems_processed_per_iteration));
-    win.set(3, Window::Dimension(0, input->tensor_shape()[3], 1));
-    ICLKernel::configure_internal(win);
-
-    // Set output valid region
-    output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape()));
-
-    // Set config_id for enabling LWS tuning
-    _config_id = "concatenate_";
-    _config_id += support::cpp11::to_string(3);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(batch_offset);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->dimension(2));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input->dimension(3));
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-Status CLBatchConcatenateLayerKernel::validate(const arm_compute::ITensorInfo *input,
-                                               unsigned int                    batch_offset,
-                                               const arm_compute::ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, batch_offset, output));
-    return Status{};
-}
-
-void CLBatchConcatenateLayerKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    const auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
-    auto       dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
-
-    Window slice = window.first_slice_window_3D();
-
-    const int offset_to_first_elements_in_bytes = _batch_offset * dst->info()->strides_in_bytes()[3];
-
-    unsigned int idx = 2 * num_arguments_per_3D_tensor(); // Skip the input and output parameters
-    _kernel.setArg<cl_int>(idx, offset_to_first_elements_in_bytes);
-
-    do
-    {
-        unsigned int idx = 0;
-        add_3D_tensor_argument(idx, src, slice);
-        add_3D_tensor_argument(idx, dst, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_3D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLBatchConcatenateLayerKernel.h b/src/core/CL/kernels/CLBatchConcatenateLayerKernel.h
deleted file mode 100644
index 54a89eb243..0000000000
--- a/src/core/CL/kernels/CLBatchConcatenateLayerKernel.h
+++ /dev/null
@@ -1,82 +0,0 @@
-/*
- * Copyright (c) 2019-2020 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_CLBATCHCONCATENATEKERNEL_H
-#define ARM_COMPUTE_CLBATCHCONCATENATEKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the batch concatenate kernel.
- *  The input tensor will be concatenated into the output tensor.
- */
-class CLBatchConcatenateLayerKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLBatchConcatenateLayerKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLBatchConcatenateLayerKernel(const CLBatchConcatenateLayerKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLBatchConcatenateLayerKernel &operator=(const CLBatchConcatenateLayerKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLBatchConcatenateLayerKernel(CLBatchConcatenateLayerKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLBatchConcatenateLayerKernel &operator=(CLBatchConcatenateLayerKernel &&) = default;
-    /** Default destructor */
-    ~CLBatchConcatenateLayerKernel() = default;
-    /** Initialise the kernel's inputs and output
-     *
-     * @param[in]     compile_context The compile context to be used.
-     * @param[in]     input           Input tensor. Data types supported: All.
-     * @param[in]     batch_offset    The offset on axis # 3.
-     * @param[in,out] output          Output tensor. Data types supported: Same as @p input.
-     *
-     * @note: The output tensor's low two dimensions can't be smaller than the input one's.
-     * @note: The gaps between the two lowest dimensions of input and output need to be divisible by 2.
-     *
-     */
-    void configure(const CLCompileContext &compile_context, ITensorInfo *input, unsigned int batch_offset, ITensorInfo *output);
-    /**  Static function to check if given info will lead to a valid configuration of @ref CLBatchConcatenateLayerKernel
-     *
-     * @param[in] input        Input tensor info. Data types supported: All.
-     * @param[in] batch_offset The offset on axis # 3.
-     * @param[in] output       Output tensor info. Data types supported: Same as @p input.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, unsigned int batch_offset, const ITensorInfo *output);
-
-    // Inherited methods overridden:
-    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    unsigned int _batch_offset;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLBATCHCONCATENATEKERNEL_H */
diff --git a/src/core/CL/kernels/CLDepthConcatenateLayerKernel.cpp b/src/core/CL/kernels/CLDepthConcatenateLayerKernel.cpp
deleted file mode 100644
index eb5bfc2d86..0000000000
--- a/src/core/CL/kernels/CLDepthConcatenateLayerKernel.cpp
+++ /dev/null
@@ -1,135 +0,0 @@
-/*
- * Copyright (c) 2017-2020 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 "src/core/CL/kernels/CLDepthConcatenateLayerKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Utils.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/Cast.h"
-
-#include "support/StringSupport.h"
-
-namespace arm_compute
-{
-namespace
-{
-Status validate_arguments(const ITensorInfo *input, unsigned int depth_offset, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::F16, DataType::F32);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimX) != output->dimension(Window::DimX));
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimY) != output->dimension(Window::DimY));
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(2) + depth_offset > output->dimension(2));
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(3, input, output);
-
-    return Status{};
-}
-} // namespace
-
-CLDepthConcatenateLayerKernel::CLDepthConcatenateLayerKernel()
-    : _depth_offset(0)
-{
-}
-
-void CLDepthConcatenateLayerKernel::configure(const CLCompileContext &compile_context, ITensorInfo *input, unsigned int depth_offset, ITensorInfo *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input, depth_offset, output));
-
-    auto padding_info = get_padding_info({ input, output });
-
-    _depth_offset = depth_offset;
-
-    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16 / input->element_size(), input->dimension(0));
-
-    // Add build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->data_type()));
-    build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
-    build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(input->dimension(0) % num_elems_processed_per_iteration));
-    if(is_data_type_quantized_asymmetric(input->data_type()) && input->quantization_info() != output->quantization_info())
-    {
-        const UniformQuantizationInfo iq_info = input->quantization_info().uniform();
-        const UniformQuantizationInfo oq_info = output->quantization_info().uniform();
-
-        build_opts.add_option("-DOFFSET_IN1=" + float_to_string_with_full_precision(iq_info.offset));
-        build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oq_info.offset));
-        build_opts.add_option("-DSCALE_IN1=" + float_to_string_with_full_precision(iq_info.scale));
-        build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oq_info.scale));
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "concatenate", build_opts.options());
-
-    // Configure kernel window
-    auto win = calculate_max_window(*output, Steps(num_elems_processed_per_iteration));
-    win.set(Window::DimZ, Window::Dimension(0, input->tensor_shape().z(), 1));
-    ICLKernel::configure_internal(win);
-
-    // Set output valid region
-    output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape()));
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-Status CLDepthConcatenateLayerKernel::validate(const arm_compute::ITensorInfo *input,
-                                               unsigned int                    depth_offset,
-                                               const arm_compute::ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, depth_offset, output));
-    return Status{};
-}
-
-void CLDepthConcatenateLayerKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    const auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
-    auto       dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
-
-    Window slice = window.first_slice_window_3D();
-
-    const int offset_to_first_elements_in_bytes = _depth_offset * dst->info()->strides_in_bytes()[2];
-
-    unsigned int idx = 2 * num_arguments_per_3D_tensor(); // Skip the input and output parameters
-    _kernel.setArg<cl_int>(idx, offset_to_first_elements_in_bytes);
-
-    do
-    {
-        unsigned int idx = 0;
-        add_3D_tensor_argument(idx, src, slice);
-        add_3D_tensor_argument(idx, dst, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_3D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLDepthConcatenateLayerKernel.h b/src/core/CL/kernels/CLDepthConcatenateLayerKernel.h
deleted file mode 100644
index 6c73bd4bf4..0000000000
--- a/src/core/CL/kernels/CLDepthConcatenateLayerKernel.h
+++ /dev/null
@@ -1,80 +0,0 @@
-/*
- * Copyright (c) 2017-2020 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_CLDEPTHCONCATENATEKERNEL_H
-#define ARM_COMPUTE_CLDEPTHCONCATENATEKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-/** Interface for the depth concatenate kernel.
- *  The input tensor will be concatenated into the output tensor.
- */
-class CLDepthConcatenateLayerKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLDepthConcatenateLayerKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLDepthConcatenateLayerKernel(const CLDepthConcatenateLayerKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLDepthConcatenateLayerKernel &operator=(const CLDepthConcatenateLayerKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLDepthConcatenateLayerKernel(CLDepthConcatenateLayerKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLDepthConcatenateLayerKernel &operator=(CLDepthConcatenateLayerKernel &&) = default;
-    /** Default destructor */
-    ~CLDepthConcatenateLayerKernel() = default;
-    /** Initialise the kernel's inputs and output
-     *
-     * @param[in]     compile_context The compile context to be used.
-     * @param[in]     input           Input tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
-     * @param[in]     depth_offset    The offset on the Z axis.
-     * @param[in,out] output          Output tensor. Data types supported: Same as @p input.
-     *
-     * @note: The output tensor's low two dimensions can't be smaller than the input one's.
-     * @note: The gaps between the two lowest dimensions of input and output need to be divisible by 2.
-     *
-     */
-    void configure(const CLCompileContext &compile_context, ITensorInfo *input, unsigned int depth_offset, ITensorInfo *output);
-    /**  Static function to check if given info will lead to a valid configuration of @ref CLDepthConcatenateLayerKernel
-     *
-     * @param[in] input        Input tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32
-     * @param[in] depth_offset The offset on the Z axis.
-     * @param[in] output       Output tensor info. Data types supported: Same as @p input.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, unsigned int depth_offset, const ITensorInfo *output);
-
-    // Inherited methods overridden:
-    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    unsigned int _depth_offset;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLDEPTHCONCATENATEKERNEL_H */
diff --git a/src/core/CL/kernels/CLHeightConcatenateLayerKernel.cpp b/src/core/CL/kernels/CLHeightConcatenateLayerKernel.cpp
deleted file mode 100644
index 8aa7366d50..0000000000
--- a/src/core/CL/kernels/CLHeightConcatenateLayerKernel.cpp
+++ /dev/null
@@ -1,128 +0,0 @@
-/*
- * Copyright (c) 2019-2020 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 "src/core/CL/kernels/CLHeightConcatenateLayerKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/Cast.h"
-
-#include "support/StringSupport.h"
-
-namespace arm_compute
-{
-namespace
-{
-Status validate_arguments(const ITensorInfo *input, unsigned int height_offset, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimY) + height_offset > output->dimension(Window::DimY));
-
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(0) != output->dimension(0));
-    for(size_t i = 2; i < Coordinates::num_max_dimensions; ++i)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(i) != output->dimension(i));
-    }
-    ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
-
-    return Status{};
-}
-} // namespace
-
-CLHeightConcatenateLayerKernel::CLHeightConcatenateLayerKernel()
-    : _height_offset(0)
-{
-}
-
-Status CLHeightConcatenateLayerKernel::validate(const ITensorInfo *input, unsigned int height_offset, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, height_offset, output));
-    return Status{};
-}
-
-void CLHeightConcatenateLayerKernel::configure(const CLCompileContext &compile_context, ITensorInfo *input, unsigned int height_offset, ITensorInfo *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input, height_offset, output));
-
-    auto padding_info = get_padding_info({ input, output });
-
-    _height_offset = height_offset;
-
-    // Add build options
-    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(4, input->dimension(0));
-
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(input->element_size()));
-    build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
-    build_opts.add_option("-DHEIGHT_OFFSET=" + support::cpp11::to_string(_height_offset));
-    build_opts.add_option("-DDEPTH=" + support::cpp11::to_string(input->dimension(2)));
-    build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(input->dimension(0) % num_elems_processed_per_iteration));
-
-    if(is_data_type_quantized_asymmetric(input->data_type()) && input->quantization_info() != output->quantization_info())
-    {
-        const UniformQuantizationInfo iq_info = input->quantization_info().uniform();
-        const UniformQuantizationInfo oq_info = output->quantization_info().uniform();
-
-        build_opts.add_option("-DOFFSET_IN1=" + float_to_string_with_full_precision(iq_info.offset));
-        build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oq_info.offset));
-        build_opts.add_option("-DSCALE_IN1=" + float_to_string_with_full_precision(iq_info.scale));
-        build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oq_info.scale));
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "concatenate_height", build_opts.options());
-    // Configure kernel window
-
-    // The window needs to be based on input as we copy all the heights of input
-    Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration));
-    ICLKernel::configure_internal(win.collapse(win, Window::DimZ));
-
-    // Set output valid region
-    output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape()));
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-void CLHeightConcatenateLayerKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    const auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
-    auto       dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
-
-    unsigned int idx = 0;
-    add_4D_tensor_argument(idx, src, window);
-    add_4D_tensor_argument(idx, dst, window);
-    enqueue(queue, *this, window, lws_hint());
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLHeightConcatenateLayerKernel.h b/src/core/CL/kernels/CLHeightConcatenateLayerKernel.h
deleted file mode 100644
index f4cb627052..0000000000
--- a/src/core/CL/kernels/CLHeightConcatenateLayerKernel.h
+++ /dev/null
@@ -1,77 +0,0 @@
-/*
- * Copyright (c) 2019-2020 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_CLHEIGHTCONCATENATELAYERKERNEL_H
-#define ARM_COMPUTE_CLHEIGHTCONCATENATELAYERKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-/** Interface for the height concatenate kernel.
- *  The input tensor will be concatenated into the output tensor.
- */
-class CLHeightConcatenateLayerKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLHeightConcatenateLayerKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHeightConcatenateLayerKernel(const CLHeightConcatenateLayerKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLHeightConcatenateLayerKernel &operator=(const CLHeightConcatenateLayerKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLHeightConcatenateLayerKernel(CLHeightConcatenateLayerKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLHeightConcatenateLayerKernel &operator=(CLHeightConcatenateLayerKernel &&) = default;
-    /** Default destructor */
-    ~CLHeightConcatenateLayerKernel() = default;
-    /** Initialise the kernel's inputs and output
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Input tensor. Data types supported: All.
-     * @param[in]  height_offset   The starting offset on the Y axis for the output tensor.
-     * @param[out] output          Output tensor. Data types supported: Same as @p input.
-     *
-     */
-    void configure(const CLCompileContext &compile_context, ITensorInfo *input, unsigned int height_offset, ITensorInfo *output);
-    /**  Static function to check if given info will lead to a valid configuration of @ref CLHeightConcatenateLayerKernel
-     *
-     * @param[in] input         Input tensor info. Data types supported: All.
-     * @param[in] height_offset The starting offset on the Y axis for the output tensor.
-     * @param[in] output        Output tensor info. Data types supported: Same as @p input.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, unsigned int height_offset, const ITensorInfo *output);
-
-    // Inherited methods overridden:
-    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    unsigned int _height_offset;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLHEIGHTCONCATENATELAYERKERNEL_H */
diff --git a/src/core/CL/kernels/CLWidthConcatenate2TensorsKernel.cpp b/src/core/CL/kernels/CLWidthConcatenate2TensorsKernel.cpp
deleted file mode 100644
index d6697ba46b..0000000000
--- a/src/core/CL/kernels/CLWidthConcatenate2TensorsKernel.cpp
+++ /dev/null
@@ -1,150 +0,0 @@
-/*
- * Copyright (c) 2018-2020 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 "src/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "src/core/utils/helpers/tensor_info.h"
-#include "support/Cast.h"
-
-#include "support/StringSupport.h"
-
-namespace arm_compute
-{
-namespace
-{
-Status validate_arguments(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input1, input2, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input1);
-    ARM_COMPUTE_RETURN_ERROR_ON(input1->data_type() == DataType::UNKNOWN);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input1, input2, output);
-    ARM_COMPUTE_RETURN_ERROR_ON(input1->dimension(0) + input2->dimension(0) > output->dimension(0));
-
-    for(size_t i = 1; i < Coordinates::num_max_dimensions; ++i)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(input1->dimension(i) != output->dimension(i));
-        ARM_COMPUTE_RETURN_ERROR_ON(input2->dimension(i) != output->dimension(i));
-    }
-    ARM_COMPUTE_RETURN_ERROR_ON(input1->num_dimensions() > 4);
-
-    return Status{};
-}
-} // namespace
-
-Status CLWidthConcatenate2TensorsKernel::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input1, input2, output));
-    return Status{};
-}
-
-void CLWidthConcatenate2TensorsKernel::configure(const CLCompileContext &compile_context, ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input1, input2, output);
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input1, input2, output));
-
-    auto padding_info = get_padding_info({ input1, input2, output });
-
-    const unsigned int min_dimension                     = std::min(input1->dimension(0), input2->dimension(0));
-    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(8, min_dimension);
-    const unsigned int vec_size_leftover                 = output->dimension(0) % num_elems_processed_per_iteration;
-
-    // Add build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input1->data_type()));
-    build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
-    build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(vec_size_leftover));
-    build_opts.add_option("-DDEPTH=" + support::cpp11::to_string(input1->dimension(2)));
-    build_opts.add_option("-DINPUT1_WIDTH=" + support::cpp11::to_string(input1->dimension(0)));
-    build_opts.add_option("-DINPUT2_WIDTH=" + support::cpp11::to_string(input2->dimension(0)));
-    build_opts.add_option("-DELEMENT_SIZE=" + support::cpp11::to_string(input1->element_size()));
-    build_opts.add_option("-DINPUT1_ROTATE_N=" + support::cpp11::to_string((input1->dimension(0) - vec_size_leftover) % num_elems_processed_per_iteration));
-
-    // If input have different quantization info set quantization parameters needed for the re-quantization process
-    const bool have_different_qinfo = helpers::tensor_info::tensors_have_different_quantization_info(output, input1, input2);
-    if(is_data_type_quantized_asymmetric(input1->data_type()) && have_different_qinfo)
-    {
-        const UniformQuantizationInfo iq1_info = input1->quantization_info().uniform();
-        const UniformQuantizationInfo iq2_info = input2->quantization_info().uniform();
-        const UniformQuantizationInfo oq_info  = output->quantization_info().uniform();
-
-        build_opts.add_option("-DOFFSET_IN1=" + float_to_string_with_full_precision(iq1_info.offset));
-        build_opts.add_option("-DSCALE_IN1=" + float_to_string_with_full_precision(iq1_info.scale));
-        build_opts.add_option("-DOFFSET_IN2=" + float_to_string_with_full_precision(iq2_info.offset));
-        build_opts.add_option("-DSCALE_IN2=" + float_to_string_with_full_precision(iq2_info.scale));
-        build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oq_info.offset));
-        build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oq_info.scale));
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "concatenate_width_x2", build_opts.options());
-
-    // Configure kernel window
-    Window win = calculate_max_window(*output, Steps(num_elems_processed_per_iteration));
-    ICLKernel::configure_internal(win.collapse(win, Window::DimZ));
-
-    // Set output valid region
-    output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape()));
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-
-    // Set config_id for enabling LWS tuning
-    _config_id = "concatenate_width_x2_";
-    _config_id += lower_string(string_from_data_type(input1->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input1->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input1->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input2->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input2->dimension(1));
-}
-
-void CLWidthConcatenate2TensorsKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    Window slice = window.first_slice_window_4D();
-
-    const auto src0 = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC_VEC));
-    const auto src1 = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC_VEC + 1));
-    auto       dst  = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
-
-    do
-    {
-        unsigned int idx = 0;
-        add_4D_tensor_argument(idx, src0, slice);
-        add_4D_tensor_argument(idx, src1, slice);
-        add_4D_tensor_argument(idx, dst, slice);
-        enqueue(queue, *this, window, lws_hint());
-    }
-    while(window.slide_window_slice_4D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h b/src/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h
deleted file mode 100644
index 2af89e12eb..0000000000
--- a/src/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h
+++ /dev/null
@@ -1,73 +0,0 @@
-/*
- * Copyright (c) 2018-2020 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_CLWIDTHCONCATENATE_2TENSORS_KERNEL_H
-#define ARM_COMPUTE_CLWIDTHCONCATENATE_2TENSORS_KERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-/** Interface for the width concatenate kernel of 2 tensors.
- *  The input1 and input2 tensors will be concatenated into the output tensor.
- */
-class CLWidthConcatenate2TensorsKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLWidthConcatenate2TensorsKernel() = default;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLWidthConcatenate2TensorsKernel(const CLWidthConcatenate2TensorsKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLWidthConcatenate2TensorsKernel &operator=(const CLWidthConcatenate2TensorsKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLWidthConcatenate2TensorsKernel(CLWidthConcatenate2TensorsKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLWidthConcatenate2TensorsKernel &operator=(CLWidthConcatenate2TensorsKernel &&) = default;
-    /** Default destructor */
-    ~CLWidthConcatenate2TensorsKernel() = default;
-    /** Initialise the kernel's input1s and output
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input1          First input tensor. Data types supported: All.
-     * @param[in]  input2          Second input tensor. Data types supported: same as @p input1
-     * @param[out] output          Output tensor. Data types supported: Same as @p input1.
-     */
-    void configure(const CLCompileContext &compile_context, ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output);
-    /**  Static function to check if given info will lead to a valid configuration of @ref CLWidthConcatenate2TensorsKernel
-     *
-     * @param[in] input1 First tensor info. Data types supported: All.
-     * @param[in] input2 Second tensor info. Data types supported: same as @p input1
-     * @param[in] output Output tensor info. Data types supported: Same as @p input1.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
-
-    // Inherited methods overridden:
-    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLWIDTHCONCATENATE_2TENSORS_KERNEL_H */
diff --git a/src/core/CL/kernels/CLWidthConcatenate4TensorsKernel.cpp b/src/core/CL/kernels/CLWidthConcatenate4TensorsKernel.cpp
deleted file mode 100644
index 7ecdd30224..0000000000
--- a/src/core/CL/kernels/CLWidthConcatenate4TensorsKernel.cpp
+++ /dev/null
@@ -1,180 +0,0 @@
-/*
- * Copyright (c) 2018-2020 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 "src/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Utils.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "src/core/utils/helpers/tensor_info.h"
-#include "support/Cast.h"
-
-#include "support/StringSupport.h"
-
-namespace arm_compute
-{
-namespace
-{
-Status validate_arguments(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *input3, const ITensorInfo *input4, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input1, input2, input3, input4, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input1);
-    ARM_COMPUTE_RETURN_ERROR_ON(input1->data_type() == DataType::UNKNOWN);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input1, input2, input3, input4, output);
-    ARM_COMPUTE_RETURN_ERROR_ON(input1->dimension(0) + input2->dimension(0) + input3->dimension(0) + input4->dimension(0) > output->dimension(0));
-
-    for(size_t i = 1; i < Coordinates::num_max_dimensions; ++i)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(input1->dimension(i) != output->dimension(i));
-        ARM_COMPUTE_RETURN_ERROR_ON(input2->dimension(i) != output->dimension(i));
-        ARM_COMPUTE_RETURN_ERROR_ON(input3->dimension(i) != output->dimension(i));
-        ARM_COMPUTE_RETURN_ERROR_ON(input4->dimension(i) != output->dimension(i));
-    }
-    ARM_COMPUTE_RETURN_ERROR_ON(input1->num_dimensions() > 4);
-
-    return Status{};
-}
-} // namespace
-
-CLWidthConcatenate4TensorsKernel::CLWidthConcatenate4TensorsKernel()
-{
-}
-
-Status CLWidthConcatenate4TensorsKernel::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *input3, const ITensorInfo *input4, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input1, input2, input3, input4, output));
-    return Status{};
-}
-
-void CLWidthConcatenate4TensorsKernel::configure(const CLCompileContext &compile_context,
-                                                 ITensorInfo *input1, ITensorInfo *input2,
-                                                 ITensorInfo *input3, ITensorInfo *input4,
-                                                 ITensorInfo *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input1, input2, input3, input4, output);
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input1, input2, input3, input4, output));
-
-    auto               padding_info                      = get_padding_info({ input1, input2, input3, input4, output });
-    const unsigned int min_dimension                     = std::min(std::min(input1->dimension(0), input2->dimension(0)), std::min(input3->dimension(0), input4->dimension(0)));
-    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(8, min_dimension);
-    const unsigned int vec_size_leftover                 = output->dimension(0) % num_elems_processed_per_iteration;
-
-    // Add build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input1->data_type()));
-    build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
-    build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(vec_size_leftover));
-    build_opts.add_option("-DDEPTH=" + support::cpp11::to_string(input1->dimension(2)));
-    build_opts.add_option("-DINPUT1_WIDTH=" + support::cpp11::to_string(input1->dimension(0)));
-    build_opts.add_option("-DINPUT2_WIDTH=" + support::cpp11::to_string(input2->dimension(0)));
-    build_opts.add_option("-DINPUT3_WIDTH=" + support::cpp11::to_string(input3->dimension(0)));
-    build_opts.add_option("-DINPUT4_WIDTH=" + support::cpp11::to_string(input4->dimension(0)));
-    build_opts.add_option("-DELEMENT_SIZE=" + support::cpp11::to_string(input1->element_size()));
-    build_opts.add_option("-DINPUT1_ROTATE_N=" + support::cpp11::to_string((input1->dimension(0) - vec_size_leftover) % num_elems_processed_per_iteration));
-    build_opts.add_option("-DINPUT2_ROTATE_N=" + support::cpp11::to_string((input1->dimension(0) + input2->dimension(0) - vec_size_leftover) % num_elems_processed_per_iteration));
-    build_opts.add_option("-DINPUT3_ROTATE_N=" + support::cpp11::to_string((input1->dimension(0) + input2->dimension(0) + input3->dimension(0) - vec_size_leftover) % num_elems_processed_per_iteration));
-
-    // If input have different quantization info set quantization parameters needed for the re-quantization process
-    const bool have_different_qinfo = helpers::tensor_info::tensors_have_different_quantization_info(output, input1, input2, input3, input4);
-    if(is_data_type_quantized_asymmetric(input1->data_type()) && have_different_qinfo)
-    {
-        const UniformQuantizationInfo iq1_info = input1->quantization_info().uniform();
-        const UniformQuantizationInfo iq2_info = input2->quantization_info().uniform();
-        const UniformQuantizationInfo iq3_info = input3->quantization_info().uniform();
-        const UniformQuantizationInfo iq4_info = input4->quantization_info().uniform();
-        const UniformQuantizationInfo oq_info  = output->quantization_info().uniform();
-
-        build_opts.add_option("-DOFFSET_IN1=" + float_to_string_with_full_precision(iq1_info.offset));
-        build_opts.add_option("-DSCALE_IN1=" + float_to_string_with_full_precision(iq1_info.scale));
-        build_opts.add_option("-DOFFSET_IN2=" + float_to_string_with_full_precision(iq2_info.offset));
-        build_opts.add_option("-DSCALE_IN2=" + float_to_string_with_full_precision(iq2_info.scale));
-        build_opts.add_option("-DOFFSET_IN3=" + float_to_string_with_full_precision(iq3_info.offset));
-        build_opts.add_option("-DSCALE_IN3=" + float_to_string_with_full_precision(iq3_info.scale));
-        build_opts.add_option("-DOFFSET_IN4=" + float_to_string_with_full_precision(iq4_info.offset));
-        build_opts.add_option("-DSCALE_IN4=" + float_to_string_with_full_precision(iq4_info.scale));
-        build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oq_info.offset));
-        build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oq_info.scale));
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "concatenate_width_x4", build_opts.options());
-
-    // Configure kernel window
-    Window win = calculate_max_window(*output, Steps(num_elems_processed_per_iteration));
-    ICLKernel::configure_internal(win.collapse(win, Window::DimZ));
-
-    // Set output valid region
-    output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape()));
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-
-    // Set config_id for enabling LWS tuning
-    _config_id = "concatenate_width_x4_";
-    _config_id += lower_string(string_from_data_type(input1->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input1->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input1->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input2->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input2->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input3->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input3->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input4->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(input4->dimension(1));
-}
-
-void CLWidthConcatenate4TensorsKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    const auto src0 = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC_VEC));
-    const auto src1 = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC_VEC + 1));
-    const auto src2 = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC_VEC + 2));
-    const auto src3 = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC_VEC + 3));
-    auto       dst  = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
-
-    Window slice = window.first_slice_window_4D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_4D_tensor_argument(idx, src0, slice);
-        add_4D_tensor_argument(idx, src1, slice);
-        add_4D_tensor_argument(idx, src2, slice);
-        add_4D_tensor_argument(idx, src3, slice);
-        add_4D_tensor_argument(idx, dst, slice);
-        enqueue(queue, *this, window, lws_hint());
-    }
-    while(window.slide_window_slice_4D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h b/src/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h
deleted file mode 100644
index 0caf87114d..0000000000
--- a/src/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h
+++ /dev/null
@@ -1,77 +0,0 @@
-/*
- * Copyright (c) 2018-2020 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_CLWIDTHCONCATENATE_4TENSORS_KERNEL_H
-#define ARM_COMPUTE_CLWIDTHCONCATENATE_4TENSORS_KERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-/** Interface for the width concatenate kernel of 4 tensors.
- *  All input tensors will be concatenated into the output tensor.
- */
-class CLWidthConcatenate4TensorsKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLWidthConcatenate4TensorsKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLWidthConcatenate4TensorsKernel(const CLWidthConcatenate4TensorsKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLWidthConcatenate4TensorsKernel &operator=(const CLWidthConcatenate4TensorsKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLWidthConcatenate4TensorsKernel(CLWidthConcatenate4TensorsKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLWidthConcatenate4TensorsKernel &operator=(CLWidthConcatenate4TensorsKernel &&) = default;
-    /** Default destructor */
-    ~CLWidthConcatenate4TensorsKernel() = default;
-    /** Initialise the kernel's input1s and output
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input1          First input tensor. Data types supported: All.
-     * @param[in]  input2          Second input tensor. Data types supported: same as @p input1
-     * @param[in]  input3          Third input tensor. Data types supported: same as @p input1
-     * @param[in]  input4          Fourth input tensor. Data types supported: same as @p input1
-     * @param[out] output          Output tensor. Data types supported: Same as @p input1.
-     */
-    void configure(const CLCompileContext &compile_context, ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *input3, ITensorInfo *input4, ITensorInfo *output);
-    /**  Static function to check if given info will lead to a valid configuration of @ref CLWidthConcatenate4TensorsKernel
-     *
-     * @param[in] input1 First tensor info. Data types supported: All.
-     * @param[in] input2 Second tensor info. Data types supported: same as @p input1
-     * @param[in] input3 Third tensor info. Data types supported: same as @p input1
-     * @param[in] input4 Fourth tensor info. Data types supported: same as @p input1
-     * @param[in] output Output tensor info. Data types supported: Same as @p input1.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *input3, const ITensorInfo *input4, const ITensorInfo *output);
-
-    // Inherited methods overridden:
-    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLWIDTHCONCATENATE_4TENSORS_KERNEL_H */
diff --git a/src/core/CL/kernels/CLWidthConcatenateLayerKernel.cpp b/src/core/CL/kernels/CLWidthConcatenateLayerKernel.cpp
deleted file mode 100644
index 30d0a481bd..0000000000
--- a/src/core/CL/kernels/CLWidthConcatenateLayerKernel.cpp
+++ /dev/null
@@ -1,123 +0,0 @@
-/*
- * Copyright (c) 2018-2020 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 "src/core/CL/kernels/CLWidthConcatenateLayerKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Utils.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/Cast.h"
-
-#include "support/StringSupport.h"
-
-namespace arm_compute
-{
-namespace
-{
-Status validate_arguments(const ITensorInfo *input, unsigned int width_offset, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
-
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(0) + width_offset > output->dimension(0));
-
-    for(size_t i = 1; i < Coordinates::num_max_dimensions; ++i)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(i) != output->dimension(i));
-    }
-    ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
-
-    return Status{};
-}
-} // namespace
-
-CLWidthConcatenateLayerKernel::CLWidthConcatenateLayerKernel()
-{
-}
-
-Status CLWidthConcatenateLayerKernel::validate(const ITensorInfo *input, unsigned int width_offset, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, width_offset, output));
-    return Status{};
-}
-
-void CLWidthConcatenateLayerKernel::configure(const CLCompileContext &compile_context, ITensorInfo *input, unsigned int width_offset, ITensorInfo *output)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input, width_offset, output));
-
-    auto padding_info = get_padding_info({ input, output });
-
-    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16, input->dimension(0));
-
-    // Add build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->data_type()));
-    build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
-    build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(input->dimension(0) % num_elems_processed_per_iteration));
-    build_opts.add_option("-DWIDTH_OFFSET=" + support::cpp11::to_string(width_offset));
-    build_opts.add_option("-DDEPTH=" + support::cpp11::to_string(input->dimension(2)));
-
-    if(is_data_type_quantized_asymmetric(input->data_type()) && input->quantization_info() != output->quantization_info())
-    {
-        const UniformQuantizationInfo iqinfo = input->quantization_info().uniform();
-        const UniformQuantizationInfo oqinfo = output->quantization_info().uniform();
-
-        build_opts.add_option("-DOFFSET_IN1=" + float_to_string_with_full_precision(iqinfo.offset));
-        build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oqinfo.offset));
-        build_opts.add_option("-DSCALE_IN1=" + float_to_string_with_full_precision(iqinfo.scale));
-        build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oqinfo.scale));
-    }
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "concatenate_width", build_opts.options());
-    // Configure kernel window
-    Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration));
-    ICLKernel::configure_internal(win.collapse(win, Window::DimZ));
-
-    // Set output valid region
-    output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape()));
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-void CLWidthConcatenateLayerKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    const auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
-    auto       dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
-
-    unsigned int idx = 0;
-    add_4D_tensor_argument(idx, src, window);
-    add_4D_tensor_argument(idx, dst, window);
-    enqueue(queue, *this, window, lws_hint());
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLWidthConcatenateLayerKernel.h b/src/core/CL/kernels/CLWidthConcatenateLayerKernel.h
deleted file mode 100644
index 09c3f4455d..0000000000
--- a/src/core/CL/kernels/CLWidthConcatenateLayerKernel.h
+++ /dev/null
@@ -1,74 +0,0 @@
-/*
- * Copyright (c) 2018-2020 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_CLWIDTHCONCATENATELAYERKERNEL_H
-#define ARM_COMPUTE_CLWIDTHCONCATENATELAYERKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-/** Interface for the width concatenate kernel.
- *  The input tensor will be concatenated into the output tensor.
- */
-class CLWidthConcatenateLayerKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLWidthConcatenateLayerKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLWidthConcatenateLayerKernel(const CLWidthConcatenateLayerKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLWidthConcatenateLayerKernel &operator=(const CLWidthConcatenateLayerKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLWidthConcatenateLayerKernel(CLWidthConcatenateLayerKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLWidthConcatenateLayerKernel &operator=(CLWidthConcatenateLayerKernel &&) = default;
-    /** Default destructor */
-    ~CLWidthConcatenateLayerKernel() = default;
-    /** Initialise the kernel's inputs and output
-     *
-     * @param[in]     compile_context The compile context to be used.
-     * @param[in]     input           Input tensor. Data types supported: All.
-     * @param[in]     width_offset    The offset on the X axis.
-     * @param[in,out] output          Output tensor. Data types supported: Same as @p input.
-     *
-     */
-    void configure(const CLCompileContext &compile_context, ITensorInfo *input, unsigned int width_offset, ITensorInfo *output);
-    /**  Static function to check if given info will lead to a valid configuration of @ref CLWidthConcatenateLayerKernel
-     *
-     * @param[in] input        Input tensor info. Data types supported: All.
-     * @param[in] width_offset The offset on the X axis.
-     * @param[in] output       Output tensor info. Data types supported: Same as @p input.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, unsigned int width_offset, const ITensorInfo *output);
-
-    // Inherited methods overridden:
-    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLWIDTHCONCATENATELAYERKERNEL_H */