diff --git a/src/core/gpu/cl/ClCompileContext.h b/src/core/gpu/cl/ClCompileContext.h
new file mode 100644
index 0000000..e69cc02
--- /dev/null
+++ b/src/core/gpu/cl/ClCompileContext.h
@@ -0,0 +1,36 @@
+/*
+ * Copyright (c) 2021 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_CL_COMPILE_CONTEXT_H
+#define ARM_COMPUTE_CL_COMPILE_CONTEXT_H
+
+#include "arm_compute/core/CL/CLCompileContext.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+using ClCompileContext = arm_compute::CLCompileContext;
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_COMPILE_CONTEXT_H */
diff --git a/src/core/gpu/cl/IClKernel.h b/src/core/gpu/cl/IClKernel.h
new file mode 100644
index 0000000..52ea3c9
--- /dev/null
+++ b/src/core/gpu/cl/IClKernel.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2021 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_ICL_KERNEL_H
+#define ARM_COMPUTE_ICL_KERNEL_H
+
+#include "arm_compute/core/ITensorInfo.h"
+#include "src/core/CL/ICLKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+using IClKernel = arm_compute::ICLKernel;
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_ICL_KERNEL_H */
diff --git a/src/core/gpu/cl/kernels/ClBatchConcatenateKernel.cpp b/src/core/gpu/cl/kernels/ClBatchConcatenateKernel.cpp
new file mode 100644
index 0000000..c16ff1f
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClBatchConcatenateKernel.cpp
@@ -0,0 +1,156 @@
+/*
+ * Copyright (c) 2019-2021 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/gpu/cl/kernels/ClBatchConcatenateKernel.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 opencl
+{
+namespace kernels
+{
+namespace
+{
+Status validate_arguments(const ITensorInfo *src, unsigned int batch_offset, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(src);
+    ARM_COMPUTE_RETURN_ERROR_ON(src->data_type() == DataType::UNKNOWN);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, dst);
+
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(Window::DimX) != dst->dimension(Window::DimX));
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(Window::DimY) != dst->dimension(Window::DimY));
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(Window::DimZ) != dst->dimension(Window::DimZ));
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(3) + batch_offset > dst->dimension(3));
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(4, src, dst);
+
+    return Status{};
+}
+} // namespace
+
+ClBatchConcatenateKernel::ClBatchConcatenateKernel()
+    : _batch_offset(0)
+{
+}
+
+void ClBatchConcatenateKernel::configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int batch_offset, ITensorInfo *dst)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, batch_offset, dst));
+
+    auto padding_info = get_padding_info({ src, dst });
+
+    _batch_offset = batch_offset;
+
+    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16 / src->element_size(), src->dimension(0));
+
+    // Add build options
+    CLBuildOptions build_opts;
+    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(src->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(src->dimension(0) % num_elems_processed_per_iteration));
+    if(is_data_type_quantized_asymmetric(src->data_type()) && src->quantization_info() != dst->quantization_info())
+    {
+        const UniformQuantizationInfo iq_info = src->quantization_info().uniform();
+        const UniformQuantizationInfo oq_info = dst->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(*dst, Steps(num_elems_processed_per_iteration));
+    win.set(3, Window::Dimension(0, src->tensor_shape()[3], 1));
+    ICLKernel::configure_internal(win);
+
+    // Set dst valid region
+    dst->set_valid_region(ValidRegion(Coordinates(), dst->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(src->dimension(0));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src->dimension(1));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src->dimension(2));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src->dimension(3));
+
+    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
+}
+
+Status ClBatchConcatenateKernel::validate(const arm_compute::ITensorInfo *src,
+                                          unsigned int                    batch_offset,
+                                          const arm_compute::ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, batch_offset, dst));
+    return Status{};
+}
+
+void ClBatchConcatenateKernel::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 src and dst 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 opencl
+} // namespace kernels
+} // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/ClBatchConcatenateKernel.h b/src/core/gpu/cl/kernels/ClBatchConcatenateKernel.h
new file mode 100644
index 0000000..378a08a
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClBatchConcatenateKernel.h
@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2021 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_CL_BATCH_CONCATENATE_KERNEL_H
+#define ARM_COMPUTE_CL_BATCH_CONCATENATE_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** Interface for the batch concatenate kernel.
+ *  The src tensor will be concatenated into the destination tensor.
+ */
+class ClBatchConcatenateKernel : public IClKernel
+{
+public:
+    /** Default constructor */
+    ClBatchConcatenateKernel();
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClBatchConcatenateKernel);
+    /** Initialise the kernel's source and destination
+     *
+     * @param[in]     compile_context The compile context to be used.
+     * @param[in]     src             Source tensor. Data types supported: All.
+     * @param[in]     batch_offset    The offset on axis # 3.
+     * @param[in,out] dst             Destination tensor. Data types supported: Same as @p src.
+     *
+     * @note: The dst tensor's low two dimensions can't be smaller than the src one's.
+     * @note: The gaps between the two lowest dimensions of src and dst need to be divisible by 2.
+     *
+     */
+    void configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int batch_offset, ITensorInfo *dst);
+    /**  Static function to check if given info will lead to a valid configuration of @ref ClBatchConcatenateKernel
+     *
+     * @param[in] src          Input tensor info. Data types supported: All.
+     * @param[in] batch_offset The offset on axis # 3.
+     * @param[in] dst          Destination tensor info. Data types supported: Same as @p src.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src, unsigned int batch_offset, const ITensorInfo *dst);
+
+    // Inherited methods overridden:
+    void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override;
+
+private:
+    unsigned int _batch_offset;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_BATCH_CONCATENATE_KERNEL_H */
diff --git a/src/core/gpu/cl/kernels/ClDepthConcatenateKernel.cpp b/src/core/gpu/cl/kernels/ClDepthConcatenateKernel.cpp
new file mode 100644
index 0000000..e8893d7
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClDepthConcatenateKernel.cpp
@@ -0,0 +1,141 @@
+/*
+ * Copyright (c) 2017-2021 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/gpu/cl/kernels/ClDepthConcatenateKernel.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 opencl
+{
+namespace kernels
+{
+namespace
+{
+Status validate_arguments(const ITensorInfo *src, unsigned int depth_offset, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(src);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(src, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, dst);
+
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(Window::DimX) != dst->dimension(Window::DimX));
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(Window::DimY) != dst->dimension(Window::DimY));
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(2) + depth_offset > dst->dimension(2));
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(3, src, dst);
+
+    return Status{};
+}
+} // namespace
+
+ClDepthConcatenateKernel::ClDepthConcatenateKernel()
+    : _depth_offset(0)
+{
+}
+
+void ClDepthConcatenateKernel::configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int depth_offset, ITensorInfo *dst)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, depth_offset, dst));
+
+    auto padding_info = get_padding_info({ src, dst });
+
+    _depth_offset = depth_offset;
+
+    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16 / src->element_size(), src->dimension(0));
+
+    // Add build options
+    CLBuildOptions build_opts;
+    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(src->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(src->dimension(0) % num_elems_processed_per_iteration));
+    if(is_data_type_quantized_asymmetric(src->data_type()) && src->quantization_info() != dst->quantization_info())
+    {
+        const UniformQuantizationInfo iq_info = src->quantization_info().uniform();
+        const UniformQuantizationInfo oq_info = dst->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(*dst, Steps(num_elems_processed_per_iteration));
+    win.set(Window::DimZ, Window::Dimension(0, src->tensor_shape().z(), 1));
+    ICLKernel::configure_internal(win);
+
+    // Set dst valid region
+    dst->set_valid_region(ValidRegion(Coordinates(), dst->tensor_shape()));
+
+    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
+}
+
+Status ClDepthConcatenateKernel::validate(const arm_compute::ITensorInfo *src,
+                                          unsigned int                    depth_offset,
+                                          const arm_compute::ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, depth_offset, dst));
+    return Status{};
+}
+
+void ClDepthConcatenateKernel::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 src and dst 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 kernels
+} // namespace opencl
+} // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/ClDepthConcatenateKernel.h b/src/core/gpu/cl/kernels/ClDepthConcatenateKernel.h
new file mode 100644
index 0000000..144d7d4
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClDepthConcatenateKernel.h
@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2017-2021 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_CL_DEPTH_CONCATENATE_KERNEL_H
+#define ARM_COMPUTE_CL_DEPTH_CONCATENATE_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** Interface for the depth concatenate kernel.
+ *  The src tensor will be concatenated into the dst tensor.
+ */
+class ClDepthConcatenateKernel : public ICLKernel
+{
+public:
+    /** Default constructor */
+    ClDepthConcatenateKernel();
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClDepthConcatenateKernel);
+    /** Initialise the kernel's source and destination
+     *
+     * @param[in]     compile_context The compile context to be used.
+     * @param[in]     src             Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
+     * @param[in]     depth_offset    The offset on the Z axis.
+     * @param[in,out] dst             Destination tensor. Data types supported: Same as @p src.
+     *
+     * @note: The dst tensor's low two dimensions can't be smaller than the src one's.
+     * @note: The gaps between the two lowest dimensions of src and dst need to be divisible by 2.
+     *
+     */
+    void configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int depth_offset, ITensorInfo *dst);
+    /**  Static function to check if given info will lead to a valid configuration of @ref ClDepthConcatenateKernel
+     *
+     * @param[in] src          Source tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32
+     * @param[in] depth_offset The offset on the Z axis.
+     * @param[in] dst          Destination tensor info. Data types supported: Same as @p src.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src, unsigned int depth_offset, const ITensorInfo *dst);
+
+    // Inherited methods overridden:
+    void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override;
+
+private:
+    unsigned int _depth_offset;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_DEPTH_CONCATENATE_KERNEL_H */
diff --git a/src/core/gpu/cl/kernels/ClHeightConcatenateKernel.cpp b/src/core/gpu/cl/kernels/ClHeightConcatenateKernel.cpp
new file mode 100644
index 0000000..83e976e
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClHeightConcatenateKernel.cpp
@@ -0,0 +1,134 @@
+/*
+ * Copyright (c) 2019-2021 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/gpu/cl/kernels/ClHeightConcatenateKernel.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 opencl
+{
+namespace kernels
+{
+namespace
+{
+Status validate_arguments(const ITensorInfo *src, unsigned int height_offset, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON(src->data_type() == DataType::UNKNOWN);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(Window::DimY) + height_offset > dst->dimension(Window::DimY));
+
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(0) != dst->dimension(0));
+    for(size_t i = 2; i < Coordinates::num_max_dimensions; ++i)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(i) != dst->dimension(i));
+    }
+    ARM_COMPUTE_RETURN_ERROR_ON(src->num_dimensions() > 4);
+
+    return Status{};
+}
+} // namespace
+
+ClHeightConcatenateKernel::ClHeightConcatenateKernel()
+    : _height_offset(0)
+{
+}
+
+Status ClHeightConcatenateKernel::validate(const ITensorInfo *src, unsigned int height_offset, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, height_offset, dst));
+    return Status{};
+}
+
+void ClHeightConcatenateKernel::configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int height_offset, ITensorInfo *dst)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, height_offset, dst));
+
+    auto padding_info = get_padding_info({ src, dst });
+
+    _height_offset = height_offset;
+
+    // Add build options
+    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(4, src->dimension(0));
+
+    CLBuildOptions build_opts;
+    build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(src->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(src->dimension(2)));
+    build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(src->dimension(0) % num_elems_processed_per_iteration));
+
+    if(is_data_type_quantized_asymmetric(src->data_type()) && src->quantization_info() != dst->quantization_info())
+    {
+        const UniformQuantizationInfo iq_info = src->quantization_info().uniform();
+        const UniformQuantizationInfo oq_info = dst->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 src as we copy all the heights of src
+    Window win = calculate_max_window(*src, Steps(num_elems_processed_per_iteration));
+    ICLKernel::configure_internal(win.collapse(win, Window::DimZ));
+
+    // Set dst valid region
+    dst->set_valid_region(ValidRegion(Coordinates(), dst->tensor_shape()));
+
+    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
+}
+
+void ClHeightConcatenateKernel::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 kernels
+} // namespace opencl
+} // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/ClHeightConcatenateKernel.h b/src/core/gpu/cl/kernels/ClHeightConcatenateKernel.h
new file mode 100644
index 0000000..88cd4c4
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClHeightConcatenateKernel.h
@@ -0,0 +1,74 @@
+/*
+ * Copyright (c) 2021 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_CL_HEIGHT_CONCATENATE_LAYER_KERNEL_H
+#define ARM_COMPUTE_CL_HEIGHT_CONCATENATE_LAYER_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** Interface for the height concatenate kernel.
+ *  The source tensor will be concatenated into the destination tensor.
+ */
+class ClHeightConcatenateKernel : public IClKernel
+{
+public:
+    /** Default constructor */
+    ClHeightConcatenateKernel();
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClHeightConcatenateKernel);
+    /** Initialise the kernel's source and destination
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src             Source tensor. Data types supported: All.
+     * @param[in]  height_offset   The starting offset on the Y axis for the dst tensor.
+     * @param[out] dst             Destination tensor. Data types supported: same as @p src.
+     *
+     */
+    void configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int height_offset, ITensorInfo *dst);
+    /**  Static function to check if given info will lead to a valid configuration of @ref ClHeightConcatenateKernel
+     *
+     * @param[in] src           Source tensor info. Data types supported: All.
+     * @param[in] height_offset The starting offset on the Y axis for the dst tensor.
+     * @param[in] dst           Destination tensor info. Data types supported: same as @p src.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src, unsigned int height_offset, const ITensorInfo *dst);
+
+    // Inherited methods overridden:
+    void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override;
+
+private:
+    unsigned int _height_offset;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_HEIGHT_CONCATENATE_LAYER_KERNEL_H */
diff --git a/src/core/gpu/cl/kernels/ClWidthConcatenate2TensorsKernel.cpp b/src/core/gpu/cl/kernels/ClWidthConcatenate2TensorsKernel.cpp
new file mode 100644
index 0000000..6a2ab3b
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClWidthConcatenate2TensorsKernel.cpp
@@ -0,0 +1,156 @@
+/*
+ * Copyright (c) 2018-2021 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/gpu/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 opencl
+{
+namespace kernels
+{
+namespace
+{
+Status validate_arguments(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src1, src2, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(src1);
+    ARM_COMPUTE_RETURN_ERROR_ON(src1->data_type() == DataType::UNKNOWN);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src1, src2, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON(src1->dimension(0) + src2->dimension(0) > dst->dimension(0));
+
+    for(size_t i = 1; i < Coordinates::num_max_dimensions; ++i)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON(src1->dimension(i) != dst->dimension(i));
+        ARM_COMPUTE_RETURN_ERROR_ON(src2->dimension(i) != dst->dimension(i));
+    }
+    ARM_COMPUTE_RETURN_ERROR_ON(src1->num_dimensions() > 4);
+
+    return Status{};
+}
+} // namespace
+
+Status ClWidthConcatenate2TensorsKernel::validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src1, src2, dst));
+    return Status{};
+}
+
+void ClWidthConcatenate2TensorsKernel::configure(const CLCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(src1, src2, dst);
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src1, src2, dst));
+
+    auto padding_info = get_padding_info({ src1, src2, dst });
+
+    const unsigned int min_dimension                     = std::min(src1->dimension(0), src2->dimension(0));
+    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(8, min_dimension);
+    const unsigned int vec_size_leftover                 = dst->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(src1->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(src1->dimension(2)));
+    build_opts.add_option("-DINPUT1_WIDTH=" + support::cpp11::to_string(src1->dimension(0)));
+    build_opts.add_option("-DINPUT2_WIDTH=" + support::cpp11::to_string(src2->dimension(0)));
+    build_opts.add_option("-DELEMENT_SIZE=" + support::cpp11::to_string(src1->element_size()));
+    build_opts.add_option("-DINPUT1_ROTATE_N=" + support::cpp11::to_string((src1->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(dst, src1, src2);
+    if(is_data_type_quantized_asymmetric(src1->data_type()) && have_different_qinfo)
+    {
+        const UniformQuantizationInfo iq1_info = src1->quantization_info().uniform();
+        const UniformQuantizationInfo iq2_info = src2->quantization_info().uniform();
+        const UniformQuantizationInfo oq_info  = dst->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(*dst, Steps(num_elems_processed_per_iteration));
+    ICLKernel::configure_internal(win.collapse(win, Window::DimZ));
+
+    // Set dst valid region
+    dst->set_valid_region(ValidRegion(Coordinates(), dst->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(src1->data_type()));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src1->dimension(0));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src1->dimension(1));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src2->dimension(0));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src2->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 kernels
+} // namespace opencl
+} // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/ClWidthConcatenate2TensorsKernel.h b/src/core/gpu/cl/kernels/ClWidthConcatenate2TensorsKernel.h
new file mode 100644
index 0000000..9271500
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClWidthConcatenate2TensorsKernel.h
@@ -0,0 +1,70 @@
+/*
+ * Copyright (c) 2021 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_CL_WIDTHCONCATENATE_2TENSORS_KERNEL_H
+#define ARM_COMPUTE_CL_WIDTHCONCATENATE_2TENSORS_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** Interface for the width concatenate kernel of 2 tensors.
+ *  The src1 and src2 tensors will be concatenated into the dst tensor.
+ */
+class ClWidthConcatenate2TensorsKernel : public IClKernel
+{
+public:
+    /** Default constructor */
+    ClWidthConcatenate2TensorsKernel() = default;
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClWidthConcatenate2TensorsKernel);
+    /** Initialise the kernel's sources and destination
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src1            First source tensor. Data types supported: All.
+     * @param[in]  src2            Second source tensor. Data types supported: same as @p src1
+     * @param[out] dst             Destination tensor. Data types supported: Same as @p src1.
+     */
+    void configure(const CLCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst);
+    /**  Static function to check if given info will lead to a valid configuration of @ref ClWidthConcatenate2TensorsKernel
+     *
+     * @param[in] src1 First tensor info. Data types supported: All.
+     * @param[in] src2 Second tensor info. Data types supported: same as @p src1
+     * @param[in] dst  Destination tensor info. Data types supported: Same as @p src1.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst);
+
+    // Inherited methods overridden:
+    void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_WIDTH_CONCATENATE_2TENSORS_KERNEL_H */
diff --git a/src/core/gpu/cl/kernels/ClWidthConcatenate4TensorsKernel.cpp b/src/core/gpu/cl/kernels/ClWidthConcatenate4TensorsKernel.cpp
new file mode 100644
index 0000000..4b49652
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClWidthConcatenate4TensorsKernel.cpp
@@ -0,0 +1,186 @@
+/*
+ * Copyright (c) 2018-2021 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/gpu/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 opencl
+{
+namespace kernels
+{
+namespace
+{
+Status validate_arguments(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *src3, const ITensorInfo *src4, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src1, src2, src3, src4, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(src1);
+    ARM_COMPUTE_RETURN_ERROR_ON(src1->data_type() == DataType::UNKNOWN);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src1, src2, src3, src4, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON(src1->dimension(0) + src2->dimension(0) + src3->dimension(0) + src4->dimension(0) > dst->dimension(0));
+
+    for(size_t i = 1; i < Coordinates::num_max_dimensions; ++i)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON(src1->dimension(i) != dst->dimension(i));
+        ARM_COMPUTE_RETURN_ERROR_ON(src2->dimension(i) != dst->dimension(i));
+        ARM_COMPUTE_RETURN_ERROR_ON(src3->dimension(i) != dst->dimension(i));
+        ARM_COMPUTE_RETURN_ERROR_ON(src4->dimension(i) != dst->dimension(i));
+    }
+    ARM_COMPUTE_RETURN_ERROR_ON(src1->num_dimensions() > 4);
+
+    return Status{};
+}
+} // namespace
+
+ClWidthConcatenate4TensorsKernel::ClWidthConcatenate4TensorsKernel()
+{
+}
+
+Status ClWidthConcatenate4TensorsKernel::validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *src3, const ITensorInfo *src4, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src1, src2, src3, src4, dst));
+    return Status{};
+}
+
+void ClWidthConcatenate4TensorsKernel::configure(const CLCompileContext &compile_context,
+                                                 ITensorInfo *src1, ITensorInfo *src2,
+                                                 ITensorInfo *src3, ITensorInfo *src4,
+                                                 ITensorInfo *dst)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(src1, src2, src3, src4, dst);
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src1, src2, src3, src4, dst));
+
+    auto               padding_info                      = get_padding_info({ src1, src2, src3, src4, dst });
+    const unsigned int min_dimension                     = std::min(std::min(src1->dimension(0), src2->dimension(0)), std::min(src3->dimension(0), src4->dimension(0)));
+    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(8, min_dimension);
+    const unsigned int vec_size_leftover                 = dst->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(src1->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(src1->dimension(2)));
+    build_opts.add_option("-DINPUT1_WIDTH=" + support::cpp11::to_string(src1->dimension(0)));
+    build_opts.add_option("-DINPUT2_WIDTH=" + support::cpp11::to_string(src2->dimension(0)));
+    build_opts.add_option("-DINPUT3_WIDTH=" + support::cpp11::to_string(src3->dimension(0)));
+    build_opts.add_option("-DINPUT4_WIDTH=" + support::cpp11::to_string(src4->dimension(0)));
+    build_opts.add_option("-DELEMENT_SIZE=" + support::cpp11::to_string(src1->element_size()));
+    build_opts.add_option("-DINPUT1_ROTATE_N=" + support::cpp11::to_string((src1->dimension(0) - vec_size_leftover) % num_elems_processed_per_iteration));
+    build_opts.add_option("-DINPUT2_ROTATE_N=" + support::cpp11::to_string((src1->dimension(0) + src2->dimension(0) - vec_size_leftover) % num_elems_processed_per_iteration));
+    build_opts.add_option("-DINPUT3_ROTATE_N=" + support::cpp11::to_string((src1->dimension(0) + src2->dimension(0) + src3->dimension(0) - vec_size_leftover) % num_elems_processed_per_iteration));
+
+    // If soources 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(dst, src1, src2, src3, src4);
+    if(is_data_type_quantized_asymmetric(src1->data_type()) && have_different_qinfo)
+    {
+        const UniformQuantizationInfo iq1_info = src1->quantization_info().uniform();
+        const UniformQuantizationInfo iq2_info = src2->quantization_info().uniform();
+        const UniformQuantizationInfo iq3_info = src3->quantization_info().uniform();
+        const UniformQuantizationInfo iq4_info = src4->quantization_info().uniform();
+        const UniformQuantizationInfo oq_info  = dst->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(*dst, Steps(num_elems_processed_per_iteration));
+    ICLKernel::configure_internal(win.collapse(win, Window::DimZ));
+
+    // Set dst valid region
+    dst->set_valid_region(ValidRegion(Coordinates(), dst->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(src1->data_type()));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src1->dimension(0));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src1->dimension(1));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src2->dimension(0));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src2->dimension(1));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src3->dimension(0));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src3->dimension(1));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src4->dimension(0));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(src4->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 kernels
+} // namespace opencl
+} // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/ClWidthConcatenate4TensorsKernel.h b/src/core/gpu/cl/kernels/ClWidthConcatenate4TensorsKernel.h
new file mode 100644
index 0000000..06d6c03
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClWidthConcatenate4TensorsKernel.h
@@ -0,0 +1,75 @@
+/*
+ * Copyright (c) 2021 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef ARM_COMPUTE_CL_WIDTH_CONCATENATE_4TENSORS_KERNEL_H
+#define ARM_COMPUTE_CL_WIDTH_CONCATENATE_4TENSORS_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** Interface for the width concatenate kernel of 4 tensors.
+ *  All source tensors will be concatenated into the destination tensor.
+ */
+class ClWidthConcatenate4TensorsKernel : public IClKernel
+{
+public:
+    /** Default constructor */
+    ClWidthConcatenate4TensorsKernel();
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClWidthConcatenate4TensorsKernel);
+    /** Initialise the kernel's sources and destination
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src1            First source tensor. Data types supported: All.
+     * @param[in]  src2            Second source tensor. Data types supported: same as @p src1
+     * @param[in]  src3            Third source tensor. Data types supported: same as @p src1
+     * @param[in]  src4            Fourth source tensor. Data types supported: same as @p src1
+     * @param[out] dst             Destination tensor. Data types supported: same as @p src1.
+     */
+    void configure(const CLCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *src3, ITensorInfo *src4, ITensorInfo *dst);
+    /**  Static function to check if given info will lead to a valid configuration of @ref ClWidthConcatenate4TensorsKernel
+     *
+     * @param[in] src1 First tensor info. Data types supported: All.
+     * @param[in] src2 Second tensor info. Data types supported: same as @p src1
+     * @param[in] src3 Third tensor info. Data types supported: same as @p src1
+     * @param[in] src4 Fourth tensor info. Data types supported: same as @p src1
+     * @param[in] dst  Destination tensor info. Data types supported: same as @p src1.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *src3, const ITensorInfo *src4, const ITensorInfo *dst);
+
+    // Inherited methods overridden:
+    void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_WIDTH_CONCATENATE_4TENSORS_KERNEL_H */
diff --git a/src/core/gpu/cl/kernels/ClWidthConcatenateKernel.cpp b/src/core/gpu/cl/kernels/ClWidthConcatenateKernel.cpp
new file mode 100644
index 0000000..8cbbc27
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClWidthConcatenateKernel.cpp
@@ -0,0 +1,129 @@
+/*
+ * Copyright (c) 2018-2021 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/gpu/cl/kernels/ClWidthConcatenateKernel.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 opencl
+{
+namespace kernels
+{
+namespace
+{
+Status validate_arguments(const ITensorInfo *src, unsigned int width_offset, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(src);
+    ARM_COMPUTE_RETURN_ERROR_ON(src->data_type() == DataType::UNKNOWN);
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(0) + width_offset > dst->dimension(0));
+
+    for(size_t i = 1; i < Coordinates::num_max_dimensions; ++i)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(i) != dst->dimension(i));
+    }
+    ARM_COMPUTE_RETURN_ERROR_ON(src->num_dimensions() > 4);
+
+    return Status{};
+}
+} // namespace
+
+ClWidthConcatenateKernel::ClWidthConcatenateKernel()
+{
+}
+
+Status ClWidthConcatenateKernel::validate(const ITensorInfo *src, unsigned int width_offset, const ITensorInfo *dst)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, width_offset, dst));
+    return Status{};
+}
+
+void ClWidthConcatenateKernel::configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int width_offset, ITensorInfo *dst)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, width_offset, dst));
+
+    auto padding_info = get_padding_info({ src, dst });
+
+    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16, src->dimension(0));
+
+    // Add build options
+    CLBuildOptions build_opts;
+    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(src->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(src->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(src->dimension(2)));
+
+    if(is_data_type_quantized_asymmetric(src->data_type()) && src->quantization_info() != dst->quantization_info())
+    {
+        const UniformQuantizationInfo iqinfo = src->quantization_info().uniform();
+        const UniformQuantizationInfo oqinfo = dst->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(*src, Steps(num_elems_processed_per_iteration));
+    ICLKernel::configure_internal(win.collapse(win, Window::DimZ));
+
+    // Set dst valid region
+    dst->set_valid_region(ValidRegion(Coordinates(), dst->tensor_shape()));
+
+    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
+}
+
+void ClWidthConcatenateKernel::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 kernels
+} // namespace opencl
+} // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/ClWidthConcatenateKernel.h b/src/core/gpu/cl/kernels/ClWidthConcatenateKernel.h
new file mode 100644
index 0000000..3bffe52
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClWidthConcatenateKernel.h
@@ -0,0 +1,71 @@
+/*
+ * Copyright (c) 2021 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_CL_WIDTH_CONCATENATE_LAYER_KERNEL_H
+#define ARM_COMPUTE_CL_WIDTH_CONCATENATE_LAYER_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** Interface for the width concatenate kernel.
+ *  The source tensor will be concatenated into the destination tensor.
+ */
+class ClWidthConcatenateKernel : public IClKernel
+{
+public:
+    /** Default constructor */
+    ClWidthConcatenateKernel();
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClWidthConcatenateKernel);
+    /** Initialise the kernel's source and destination
+     *
+     * @param[in]     compile_context The compile context to be used.
+     * @param[in]     src             Source tensor. Data types supported: All.
+     * @param[in]     width_offset    The offset on the X axis.
+     * @param[in,out] dst             Destination tensor. Data types supported: same as @p src.
+     *
+     */
+    void configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int width_offset, ITensorInfo *dst);
+    /**  Static function to check if given info will lead to a valid configuration of @ref ClWidthConcatenateKernel
+     *
+     * @param[in] src          Source tensor info. Data types supported: All.
+     * @param[in] width_offset The offset on the X axis.
+     * @param[in] dst          Destination tensor info. Data types supported: same as @p src.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src, unsigned int width_offset, const ITensorInfo *dst);
+
+    // Inherited methods overridden:
+    void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_WIDTH_CONCATENATE_LAYER_KERNEL_H */