diff --git a/src/core/CL/cl_kernels/copy_tensor.cl b/src/core/CL/cl_kernels/copy_tensor.cl
index 4b37dec..930a676 100644
--- a/src/core/CL/cl_kernels/copy_tensor.cl
+++ b/src/core/CL/cl_kernels/copy_tensor.cl
@@ -25,24 +25,35 @@
 
 /** Performs a copy of input tensor to the output tensor.
  *
- * @param[in]  in_ptr                            Pointer to the source image. Supported data types: U8.
- * @param[in]  in_stride_x                       Stride of the source image in X dimension (in bytes)
- * @param[in]  in_step_x                         in_stride_x * number of elements along X processed per work item (in bytes)
- * @param[in]  in_offset_first_element_in_bytes  Offset of the first element in the source image
- * @param[out] out_ptr                           Pointer to the destination image. Supported data types: U8.
- * @param[in]  out_stride_x                      Stride of the destination image in X dimension (in bytes)
- * @param[in]  out_step_x                        out_stride_x * number of elements along X processed per work item (in bytes)
- * @param[in]  out_offset_first_element_in_bytes Offset of the first element in the destination image
+ * @param[in]  in_ptr                            Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in]  in_stride_x                       Stride of the source tensor in X dimension (in bytes)
+ * @param[in]  in_step_x                         input_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  in_stride_y                       Stride of the source tensor in Y dimension (in bytes)
+ * @param[in]  in_step_y                         input_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in]  in_stride_z                       Stride of the source tensor in Z dimension (in bytes)
+ * @param[in]  in_step_z                         input_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  in_offset_first_element_in_bytes  The offset of the first element in the source tensor
+ * @param[out] out_ptr                           Pointer to the destination tensor. Supported data types: same as @p in_ptr
+ * @param[in]  out_stride_x                      Stride of the destination tensor in X dimension (in bytes)
+ * @param[in]  out_step_x                        output_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  out_stride_y                      Stride of the destination tensor in Y dimension (in bytes)
+ * @param[in]  out_step_y                        output_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in]  out_stride_z                      Stride of the source tensor in Z dimension (in bytes)
+ * @param[in]  out_step_z                        output_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  out_offset_first_element_in_bytes The offset of the first element in the destination tensor
  */
 __kernel void copy_tensor(
-    VECTOR_DECLARATION(in),
-    VECTOR_DECLARATION(out))
+    TENSOR3D_DECLARATION(in),
+    TENSOR3D_DECLARATION(out))
 {
-    Vector in  = CONVERT_TO_VECTOR_STRUCT(in);
-    Vector out = CONVERT_TO_VECTOR_STRUCT(out);
+    Tensor3D in  = CONVERT_TO_TENSOR3D_STRUCT(in);
+    Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(out);
 
-    VEC_DATA_TYPE(DATA_TYPE, 16)
-    data = vload16(0, (__global DATA_TYPE *)in.ptr);
+    // Load data
+    VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
+    data = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)in.ptr);
 
-    vstore16(data, 0, (__global DATA_TYPE *)out.ptr);
+    // Store result
+    VSTORE(VEC_SIZE)
+    (data, 0, (__global DATA_TYPE *)out.ptr);
 }
\ No newline at end of file
diff --git a/src/core/CL/kernels/CLCopyKernel.cpp b/src/core/CL/kernels/CLCopyKernel.cpp
index 4f00ef9..1fc8b5b 100644
--- a/src/core/CL/kernels/CLCopyKernel.cpp
+++ b/src/core/CL/kernels/CLCopyKernel.cpp
@@ -33,10 +33,44 @@
 #include "arm_compute/core/Validate.h"
 #include "arm_compute/core/Window.h"
 
-#include <algorithm>
-
 using namespace arm_compute;
 
+namespace
+{
+Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
+
+    // Validate output if initialized
+    if(output->total_size() != 0)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(input->tensor_shape(), output->tensor_shape());
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+    }
+
+    return Status{};
+}
+
+std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output)
+{
+    // Output auto inizialitation if not yet initialized
+    auto_init_if_empty(*output, *input);
+
+    // Configure window
+    const unsigned int num_elems_processed_per_iteration = 16 / input->element_size();
+
+    Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration));
+
+    AccessWindowHorizontal input_access(input, 0, num_elems_processed_per_iteration);
+    AccessWindowHorizontal output_access(output, 0, num_elems_processed_per_iteration);
+
+    bool window_changed = update_window_and_padding(win, input_access, output_access);
+
+    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
+    return std::make_pair(err, win);
+}
+} // namespace
+
 CLCopyKernel::CLCopyKernel()
     : _input(nullptr), _output(nullptr)
 {
@@ -44,28 +78,32 @@
 
 void CLCopyKernel::configure(const ICLTensor *input, ICLTensor *output)
 {
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(input->info()->tensor_shape(), output->info()->tensor_shape());
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info()));
 
     _input  = input;
     _output = output;
 
+    const unsigned int num_elems_processed_per_iteration = 16 / input->info()->element_size();
+
     // Create kernel
     CLBuildOptions build_opts;
     build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+    build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
     _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("copy_tensor", build_opts.options()));
 
-    // Configure window
-    constexpr unsigned int num_elems_processed_per_iteration = 16;
+    // Configure kernel window
+    auto win_config = validate_and_configure_window(input->info(), output->info());
+    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
+    ICLKernel::configure(win_config.second);
+}
 
-    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
+Status CLCopyKernel::validate(const arm_compute::ITensorInfo *input, const arm_compute::ITensorInfo *output)
+{
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output));
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get()).first);
 
-    AccessWindowHorizontal input_access(input->info(), 0, num_elems_processed_per_iteration);
-    AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
-
-    update_window_and_padding(win, input_access, output_access);
-
-    ICLKernel::configure(win);
+    return Status{};
 }
 
 void CLCopyKernel::run(const Window &window, cl::CommandQueue &queue)
@@ -73,15 +111,15 @@
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
     ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
 
-    Window collapsed = window.collapse_if_possible(ICLKernel::window(), Window::DimX);
-    Window slice     = collapsed.first_slice_window_1D();
+    Window collapsed = window.collapse_if_possible(ICLKernel::window(), Window::DimZ);
+    Window slice     = collapsed.first_slice_window_3D();
 
     do
     {
         unsigned int idx = 0;
-        add_1D_tensor_argument(idx, _input, slice);
-        add_1D_tensor_argument(idx, _output, slice);
+        add_3D_tensor_argument(idx, _input, slice);
+        add_3D_tensor_argument(idx, _output, slice);
         enqueue(queue, *this, slice);
     }
-    while(collapsed.slide_window_slice_1D(slice));
+    while(collapsed.slide_window_slice_3D(slice));
 }