Port CLGEMM to memory injecting interface

Moves the following kernels:
 - CLGEMMMatrixMultiplyKernel
 - CLGEMMMatrixMultiplyNativeKernel
 - CLGEMMMatrixMultipluReshapedKernel
 - CLGEMMMatrixMultiplyReshapedOnlyRHSKernel

 Moves the following functions
 - CLGEMM

Introduces facilities to easy handling of auxiliary temporary buffers
under then new run interface. Such are:
 - CLAuxTensorHandler: That allows wrapping of workspace buffers memory
 to CLBuffer objects
 - Ability to inject TensorInfo to allocator without transferring
 ownership. This reduce the copy overhead if needed.

Resolves: COMPMID-4188

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I7055435d831b05b749b26302082e4ac45f26dfb0
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5498
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

14 files changed, 6 insertions, 3076 deletions

diff --git a/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedKernel.h b/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedKernel.h
index 100100b1b1..06a73f173d 100644
--- a/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedKernel.h
+++ b/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedKernel.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2020 Arm Limited.
+ * Copyright (c) 2019-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -32,7 +32,9 @@ class ICLTensor;
 
 /** OpenCL kernel to multiply matrices when both the input matrices LHS (input0) and RHS (input1) have been reshaped
  *
- * @note The input matrices @p input0 and @p input1 must be reshaped through @ref CLGEMMReshapeLHSMatrixKernel and  @ref CLGEMMReshapeRHSMatrixKernel
+ * @note The input matrices @p input0 and @p input1 must be reshaped through:
+ *  - @ref opencl::kernels::ClGemmReshapeLhsMatrixKernel
+ *  - @ref opencl::kernels::ClGemmReshapeRhsMatrixKernel
  */
 class CLGEMMLowpMatrixMultiplyReshapedKernel : public ICLKernel
 {
diff --git a/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel.h b/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel.h
index 222a8615e4..e79f6dfe05 100644
--- a/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel.h
+++ b/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2020 Arm Limited.
+ * Copyright (c) 2019-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -33,7 +33,7 @@ class ICLTensor;
 
 /** OpenCL kernel to multiply matrices with QASYMM8 data type when only the input matrix RHS (input1) has been reshaped
  *
- * @note The input matrix input1 must be reshaped through @ref CLGEMMReshapeRHSMatrixKernel
+ * @note The input matrix input1 must be reshaped through @ref opencl::kernels::ClGemmReshapeRhsMatrixKernel
  * @note For fused output stage, only GEMMLowpOutputStageType::QUANTIZE_DOWN_FIXEDPOINT type is supported
  */
 class CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel : public ICLKernel
diff --git a/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.cpp b/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.cpp
deleted file mode 100644
index 479c06330d..0000000000
--- a/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.cpp
+++ /dev/null
@@ -1,540 +0,0 @@
-/*
- * 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/CL/kernels/CLGEMMMatrixMultiplyKernel.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/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "src/core/utils/helpers/float_ops.h"
-#include "support/StringSupport.h"
-
-#include <set>
-#include <string>
-
-namespace arm_compute
-{
-using namespace arm_compute::misc::shape_calculator;
-
-namespace
-{
-using ElementsProcessed = Steps;
-
-inline Status validate_arguments(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float beta,
-                                 bool is_interleaved_transposed, const GEMMReshapeInfo &reshape_info, bool fp_mixed_precision)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input0, input1, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input0);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::F16, DataType::F32);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG((fp_mixed_precision && (input0->data_type() != DataType::F16)), "Mixed precision floating point is supported only for F16 data");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input0->num_dimensions() > 4, "The number of dimensions for the matrix A must be <= 4");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input1->num_dimensions() > 3, "The number of dimensions for the matrix B must be <= 3");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(is_interleaved_transposed && reshape_info.reinterpret_input_as_3d(), "The input tensor cannot be reinterpreted as 3D if is_interleaved_transposed is true");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input1->num_dimensions() > 2 && reshape_info.reinterpret_input_as_3d(), "The input1 tensor cannot have more than 2 dimensions if input0 has to be reinterpreted as 3D");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG((reshape_info.reinterpret_input_as_3d() || reshape_info.depth_output_gemm3d() != 0) && (input2 != nullptr)
-                                    && (!reshape_info.broadcast_bias()),
-                                    "Bias addition only supported with broadcast mode in case the input or output has to be reinterpreted as 3D");
-
-    if(!is_interleaved_transposed)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(input0->dimension(0) != input1->dimension(1));
-
-        if(input2 != nullptr && !(helpers::float_ops::is_zero(beta)))
-        {
-            const unsigned int m           = reshape_info.reinterpret_input_as_3d() ? input0->dimension(1) * input0->dimension(2) : input0->dimension(1);
-            const unsigned int n           = input1->dimension(0);
-            const unsigned int input2_dim0 = input2->dimension(0);
-            const unsigned int input2_dim1 = input2->dimension(1);
-
-            ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input2, input1);
-            if(reshape_info.broadcast_bias())
-            {
-                ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim1 != 1 || input2_dim0 != n), "Incorrect dimension of bias matrix which is to be broadcasted");
-            }
-            else
-            {
-                ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim0 != n || input2_dim1 != m), "Incorrect dimension of bias matrix");
-            }
-        }
-    }
-    else
-    {
-        GEMMRHSMatrixInfo rhs_info;
-        GEMMLHSMatrixInfo lhs_info;
-        const auto        m                         = static_cast<unsigned int>(reshape_info.m());
-        const auto        n                         = static_cast<unsigned int>(reshape_info.n());
-        const int         k                         = reshape_info.k();
-        const int         mult_transpose1xW_width   = reshape_info.mult_transpose1xW_width();
-        const int         mult_interleave4x4_height = reshape_info.mult_interleave4x4_height();
-        rhs_info.n0                                 = max_cl_vector_width / input1->element_size();
-        rhs_info.k0                                 = 1;
-        rhs_info.h0                                 = mult_transpose1xW_width;
-        rhs_info.interleave                         = false;
-        rhs_info.transpose                          = false;
-        lhs_info.m0                                 = 4;
-        lhs_info.k0                                 = 4;
-        lhs_info.v0                                 = mult_interleave4x4_height;
-        lhs_info.interleave                         = true;
-        lhs_info.transpose                          = true;
-
-        TensorShape tensor_shape0{ input0->tensor_shape() };
-        tensor_shape0.set(0, k);
-        tensor_shape0.set(1, m);
-
-        TensorShape tensor_shape1{ input1->tensor_shape() };
-        tensor_shape1.set(0, n);
-        tensor_shape1.set(1, k);
-
-        const TensorInfo tensor_info0 = input0->clone()->set_tensor_shape(tensor_shape0);
-        const TensorInfo tensor_info1 = input1->clone()->set_tensor_shape(tensor_shape1);
-
-        const TensorInfo tensor_info_reshaped0 = input0->clone()->set_tensor_shape(compute_lhs_reshaped_shape(tensor_info0, lhs_info));
-        const TensorInfo tensor_info_reshaped1 = input1->clone()->set_tensor_shape(compute_rhs_reshaped_shape(tensor_info1, rhs_info));
-
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input0, &tensor_info_reshaped0);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input1, &tensor_info_reshaped1);
-
-        if(input2 != nullptr && !(helpers::float_ops::is_zero(beta)))
-        {
-            const unsigned int input2_dim0 = input2->dimension(0);
-            const unsigned int input2_dim1 = input2->dimension(1);
-
-            ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input2, input1);
-            if(reshape_info.broadcast_bias())
-            {
-                ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim1 != 1 || input2_dim0 != n), "Incorrect dimension of bias matrix which is to be broadcasted");
-            }
-            else
-            {
-                ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim0 != n || input2_dim1 != m), "Incorrect dimension of bias matrix");
-            }
-        }
-    }
-
-    if(output->total_size() != 0)
-    {
-        const TensorInfo tensor_info_output = output->clone()->set_tensor_shape(compute_mm_shape(*input0, *input1, is_interleaved_transposed, reshape_info));
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(output, &tensor_info_output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input0, output);
-    }
-
-    return Status{};
-}
-
-inline std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input0, ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output,
-                                                               float beta, bool is_interleaved_transposed, const GEMMReshapeInfo &reshape_info, GPUTarget gpu_target,
-                                                               ElementsProcessed &num_elements_processed)
-{
-    ARM_COMPUTE_UNUSED(beta);
-    bool   window_changed = false;
-    Window win{};
-    Window win_out{};
-
-    const DataType data_type                           = input0->data_type();
-    unsigned int &num_elems_processed_per_iteration_x = num_elements_processed[0];
-    unsigned int &num_elems_processed_per_iteration_y = num_elements_processed[1];
-    bool           reinterpret_input_as_3d             = reshape_info.reinterpret_input_as_3d();
-    bool           reinterpret_output_as_3d            = (reshape_info.depth_output_gemm3d() != 0);
-
-    // In case both input and output have to be reinterpreted as 3D tensors,
-    // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
-    if(reinterpret_input_as_3d == reinterpret_output_as_3d)
-    {
-        reinterpret_input_as_3d  = false;
-        reinterpret_output_as_3d = false;
-    }
-
-    // Output tensor auto inizialitation if not yet initialized
-    auto_init_if_empty(*output, input0->clone()->set_tensor_shape(compute_mm_shape(*input0, *input1, is_interleaved_transposed, reshape_info)));
-
-    TensorInfo tmp_info(*output);
-
-    if(reinterpret_output_as_3d)
-    {
-        // Since the output tensor has to be reinterpreted as 3D and the execute window is based on a 2D GEMM,
-        // the window needs to be constructed on the 2D collapsed version of the tensor
-        TensorShape tmp_shape(output->tensor_shape());
-        tmp_shape.collapse(2U, 1U);
-        tmp_info.set_tensor_shape(tmp_shape);
-    }
-
-    if(is_interleaved_transposed)
-    {
-        // reinterpret_input_as_3d is not supported if is_interleaved_transposed is set
-        ARM_COMPUTE_ERROR_ON(reshape_info.reinterpret_input_as_3d());
-
-        // Configure kernel window
-        num_elems_processed_per_iteration_x = max_cl_vector_width / data_size_from_type(data_type);
-        num_elems_processed_per_iteration_y = 4;
-
-        win = calculate_max_window(tmp_info, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-        if(input2 != nullptr)
-        {
-            const int bias_processed_per_iteration_x = num_elems_processed_per_iteration_x;
-
-            const int bias_processed_per_iteration_y = reshape_info.broadcast_bias() ? 1 : num_elems_processed_per_iteration_y;
-
-            AccessWindowStatic input2_access(input2, 0, 0,
-                                             ceil_to_multiple(input2->dimension(0), bias_processed_per_iteration_x),
-                                             ceil_to_multiple(input2->dimension(1), bias_processed_per_iteration_y));
-
-            window_changed = update_window_and_padding(win, input2_access); // window used by the execute_window_loop
-        }
-    }
-    else // The input tensors have not been reshaped
-    {
-        // Special case for 1xN, 2xN, 3xN and 4xN input0 tensor. num_elems_processed_per_iteration_x is set up for the default case.
-        num_elems_processed_per_iteration_x = max_cl_vector_width / data_size_from_type(data_type);
-        num_elems_processed_per_iteration_y = std::min(static_cast<int>(output->dimension(1)), 4);
-
-        // Create kernels according to the architecture, data type and input size.
-        GPUTarget arch_target = get_arch_from_target(gpu_target);
-        if(arch_target == GPUTarget::BIFROST && data_type == DataType::F32)
-        {
-            num_elems_processed_per_iteration_x = (input1->dimension(0) <= 1000 && input0->num_dimensions() == 1) ? 2 : 4;
-        }
-
-        // Configure window
-        win     = calculate_max_window(tmp_info, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-        win_out = calculate_max_window(*output, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-        AccessWindowStatic input0_access(input0, 0, 0, input0->dimension(0), input0->dimension(1));
-        AccessWindowStatic input1_access(input1, 0, 0, ceil_to_multiple(input1->dimension(0), num_elems_processed_per_iteration_x), input1->dimension(1));
-        AccessWindowStatic output_access(output, 0, 0,
-                                         output->dimension(0),
-                                         output->dimension(1));
-
-        if(input2 != nullptr)
-        {
-            const int bias_processed_per_iteration_x = num_elems_processed_per_iteration_x;
-
-            AccessWindowStatic input2_access(input2, 0, 0,
-                                             ceil_to_multiple(input2->dimension(0), bias_processed_per_iteration_x),
-                                             input2->dimension(1));
-
-            window_changed = update_window_and_padding(win, input0_access, input1_access, input2_access) || // window used by the execute_window_loop
-                             update_window_and_padding(win_out, output_access);                             // window used to update the padding requirements of output tensor
-        }
-        else
-        {
-            window_changed = update_window_and_padding(win, input0_access, input1_access) || // window used by the execute_window_loop
-                             update_window_and_padding(win_out, output_access);              // window used to update the padding requirements of output tensor
-        }
-    }
-
-    // Collapse along the Z direction
-    // This collapse needs to be here in order to tune the Z dimension of LWS
-    Window             collapsed             = win;
-    const unsigned int dimension_to_collapse = std::min(static_cast<unsigned int>(output->num_dimensions()), 2u);
-    collapsed                                = win.collapse(win, dimension_to_collapse);
-
-    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
-    return std::make_pair(err, collapsed);
-}
-} // namespace
-
-CLGEMMMatrixMultiplyKernel::CLGEMMMatrixMultiplyKernel()
-    : _input0(nullptr), _input1(nullptr), _input2(nullptr), _output(nullptr), _slide_matrix_b(true), _reinterpret_input_as_3d(false), _reinterpret_output_as_3d(false), _add_bias(false),
-      _broadcast_bias(false)
-{
-}
-
-void CLGEMMMatrixMultiplyKernel::configure(const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta,
-                                           bool is_interleaved_transposed, const GEMMReshapeInfo &reshape_info, bool fp_mixed_precision, const ActivationLayerInfo &activation_info)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input0, input1, input2, output, alpha, beta, is_interleaved_transposed, reshape_info, fp_mixed_precision, activation_info);
-}
-
-void CLGEMMMatrixMultiplyKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha,
-                                           float beta,
-                                           bool is_interleaved_transposed, const GEMMReshapeInfo &reshape_info, bool fp_mixed_precision, const ActivationLayerInfo &activation_info)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input0, input1, output);
-
-    // Perform validate step
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input0->info(), input1->info(), (input2 != nullptr) ? input2->info() : nullptr, output->info(), beta,
-                                                  is_interleaved_transposed, reshape_info, fp_mixed_precision));
-
-    auto padding_info = is_interleaved_transposed ? get_padding_info({ input0, input1, output }) : get_padding_info({ input0, output });
-
-    _input0                   = input0;
-    _input1                   = input1;
-    _input2                   = helpers::float_ops::is_zero(beta) ? nullptr : input2;
-    _output                   = output;
-    _reinterpret_input_as_3d  = reshape_info.reinterpret_input_as_3d();
-    _reinterpret_output_as_3d = (reshape_info.depth_output_gemm3d() != 0);
-    _add_bias                 = _input2 != nullptr;
-    _broadcast_bias           = reshape_info.broadcast_bias();
-
-    // In case both input and output have to be reinterpreted as 3D tensors,
-    // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
-    if(_reinterpret_input_as_3d == _reinterpret_output_as_3d)
-    {
-        _reinterpret_input_as_3d  = false;
-        _reinterpret_output_as_3d = false;
-    }
-
-    // Check if we need to slide the matrix B
-    const unsigned int num_dimensions_input0 = _reinterpret_input_as_3d ? _input0->info()->num_dimensions() - 1 : _input0->info()->num_dimensions();
-
-    _slide_matrix_b = (_input1->info()->num_dimensions() >= num_dimensions_input0);
-
-    const DataType data_type = input0->info()->data_type();
-
-    // Get target architecture
-    GPUTarget gpu_target = get_target();
-
-    ElementsProcessed num_elements_processed{};
-
-    // Configure kernel window
-    auto win_config = validate_and_configure_window(input0->info(), input1->info(), (input2 != nullptr) ? input2->info() : nullptr, output->info(), beta, is_interleaved_transposed, reshape_info,
-                                                    gpu_target, num_elements_processed);
-    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
-    ICLKernel::configure_internal(win_config.second);
-
-    // If _reinterpret_input_as_3d = _reinterpret_output_as_3d = true, both will be turned off (false)
-    // in which case we will dispatch a batched-GEMM to reduce the complexity of the address calculation within the OpenCL kernel.
-    // This means that the actual m used by the kernel is given by output->info()->dimension(1)
-    const unsigned int internal_m = _reinterpret_output_as_3d ? output->info()->dimension(1) * output->info()->dimension(2) : output->info()->dimension(1);
-    const unsigned int n          = output->info()->dimension(0);
-
-    const unsigned int h_gemm_3d = _reinterpret_output_as_3d ? output->info()->dimension(1) : input0->info()->dimension(1);
-    const unsigned int d_gemm_3d = _reinterpret_output_as_3d ? output->info()->dimension(2) : input0->info()->dimension(2);
-
-    const unsigned int m0 = num_elements_processed.y();
-    const unsigned int n0 = num_elements_processed.x();
-
-    // Calculate partial (store instead of load) M0 and partial N0 for the partial blocks at the end of a row/column if any. This is to avoid padding.
-    const unsigned int partial_store_m0 = internal_m % m0;
-    const unsigned int partial_store_n0 = n % n0;
-
-    // Create build options
-    CLBuildOptions build_opts;
-
-    build_opts.add_option_if(!(helpers::float_ops::is_one(alpha)), "-DALPHA=" + float_to_string_with_full_precision(alpha));
-    build_opts.add_option_if(_input2 != nullptr, "-DBETA=" + float_to_string_with_full_precision(beta));
-    build_opts.add_option_if(helpers::float_ops::is_one(beta), "-DUNIT_BETA");
-    build_opts.add_option_if(reshape_info.broadcast_bias(), "-DBROADCAST_BIAS");
-    build_opts.add_option_if(_reinterpret_input_as_3d, "-DREINTERPRET_INPUT_AS_3D");
-    build_opts.add_option_if(_reinterpret_output_as_3d, "-DREINTERPRET_OUTPUT_AS_3D");
-    build_opts.add_option_if(_reinterpret_input_as_3d || _reinterpret_output_as_3d, "-DHEIGHT_GEMM3D=" + support::cpp11::to_string(h_gemm_3d));
-    build_opts.add_option_if(_reinterpret_input_as_3d || _reinterpret_output_as_3d, "-DDEPTH_GEMM3D=" + support::cpp11::to_string(d_gemm_3d));
-    build_opts.add_option_if(!_slide_matrix_b, "-DMATRIX_B_DEPTH=" + support::cpp11::to_string(input1->info()->dimension(2)));
-    build_opts.add_option_if(activation_info.enabled(), "-DACTIVATION_TYPE=" + lower_string(string_from_activation_func(activation_info.activation())));
-    build_opts.add_option_if(activation_info.enabled(), "-DA_VAL=" + float_to_string_with_full_precision(activation_info.a()));
-    build_opts.add_option_if(activation_info.enabled(), "-DB_VAL=" + float_to_string_with_full_precision(activation_info.b()));
-    build_opts.add_option("-DIN1_DIM_X=" + support::cpp11::to_string(input1->info()->dimension(0)));
-
-    const bool is_bifrost = get_arch_from_target(gpu_target) == GPUTarget::BIFROST;
-
-    std::string kernel_name;
-    if(is_interleaved_transposed)
-    {
-        const int mult_transpose1xW_width   = reshape_info.mult_transpose1xW_width();
-        const int mult_interleave4x4_height = reshape_info.mult_interleave4x4_height();
-
-        build_opts.add_option("-DM=" + support::cpp11::to_string(internal_m));
-        build_opts.add_option("-DN=" + support::cpp11::to_string(n));
-        build_opts.add_option("-DK=" + support::cpp11::to_string(input1->info()->dimension(0) / (n0 * mult_transpose1xW_width)));
-        build_opts.add_option("-DH0=" + support::cpp11::to_string(mult_transpose1xW_width));
-        build_opts.add_option("-DV0=" + support::cpp11::to_string(mult_interleave4x4_height));
-        build_opts.add_option("-DPARTIAL_STORE_M0=" + support::cpp11::to_string(partial_store_m0));
-        build_opts.add_option("-DPARTIAL_STORE_N0=" + support::cpp11::to_string(partial_store_n0));
-
-        if(is_data_type_float(data_type) && is_bifrost)
-        {
-            kernel_name = "gemm_mm_interleaved_transposed_" + lower_string(string_from_data_type(data_type)) + "_bifrost";
-        }
-        else
-        {
-            kernel_name = "gemm_mm_interleaved_transposed_" + lower_string(string_from_data_type(data_type));
-            if(fp_mixed_precision && data_type == DataType::F16)
-            {
-                // currently wider accumulator is only supported for fp16 kernels.
-                kernel_name += "_acc32";
-            }
-        }
-    }
-    else // The input tensors have not been reshaped
-    {
-        build_opts.add_option("-DN=" + support::cpp11::to_string(n));
-        build_opts.add_option("-DK=" + support::cpp11::to_string(input0->info()->dimension(0)));
-        build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
-        build_opts.add_option("-DM0=" + support::cpp11::to_string(m0));
-        build_opts.add_option("-DN0=" + support::cpp11::to_string(n0));
-        build_opts.add_option("-DPARTIAL_STORE_M0=" + support::cpp11::to_string(partial_store_m0));
-        build_opts.add_option("-DPARTIAL_STORE_N0=" + support::cpp11::to_string(partial_store_n0));
-
-        // Create kernels according to the architecture, data type and input size.
-        if(is_data_type_float(data_type) && is_bifrost)
-        {
-            kernel_name = "gemm_mm_floating_point";
-
-            if(input0->info()->num_dimensions() != 1)
-            {
-                kernel_name += "_" + lower_string(string_from_data_type(data_type)) + "_bifrost";
-                if(fp_mixed_precision && data_type == DataType::F16)
-                {
-                    // currently wider accumulator is only supported for fp16 kernels.
-                    kernel_name += "_acc32";
-                }
-            }
-            else if(input1->info()->dimension(0) <= 1000 && data_type == DataType::F32)
-            {
-                // The first kernel is optimized for the case of 1000 or less output elements (e.g. FC8 of AlexNet and VGG-16, and
-                // FC1 of Inception v3). The second kernel is optimized for the case of greater than 1000 output elements (e.g.
-                // FC6 and FC7 of AlexNet and VGG-16).
-                kernel_name += "_" + lower_string(string_from_data_type(data_type)) + "_bifrost_1000";
-            }
-
-            // The work-group size equal to the Bifrost quad size has been proved to be optimal for these kernels
-            // via exhaustive autotuning over a range of representative layer configurations.
-            set_lws_hint(cl::NDRange(4));
-        }
-        else // (MIDGARD and F32) or (F16)
-        {
-            kernel_name = "gemm_mm_floating_point";
-        }
-    }
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
-
-    // Set config_id for enabling LWS tuning
-    _config_id = "gemm_";
-    _config_id += (is_interleaved_transposed ? "reshaped_" : "");
-    _config_id += (_add_bias ? "add_bias_" : "");
-    _config_id += (_broadcast_bias ? "broadcast_bias_" : "");
-    _config_id += (fp_mixed_precision ? "fp_mixed_" : "");
-    _config_id += (_reinterpret_input_as_3d ? "3di_" : "");
-    _config_id += (_reinterpret_output_as_3d ? "3do_" : "");
-    _config_id += lower_string(string_from_data_type(input0->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(2));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(3));
-    _config_id += "_";
-    _config_id += (is_interleaved_transposed ? support::cpp11::to_string(input1->info()->dimension(0)) : support::cpp11::to_string(input1->info()->dimension(1)));
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-Status CLGEMMMatrixMultiplyKernel::validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta,
-                                            bool is_interleaved_transposed, const GEMMReshapeInfo &reshape_info, GPUTarget gpu_target, bool fp_mixed_precision, const ActivationLayerInfo &activation_info)
-{
-    // Note: num_elements_processed will be set in validate_and_configure_window()
-    ElementsProcessed num_elements_processed{};
-    ARM_COMPUTE_UNUSED(alpha);
-    ARM_COMPUTE_UNUSED(activation_info);
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input0, input1, input2, output, beta, is_interleaved_transposed, reshape_info, fp_mixed_precision));
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input0->clone().get(),
-                                                              input1->clone().get(),
-                                                              (input2 != nullptr) ? input2->clone().get() : nullptr,
-                                                              output->clone().get(),
-                                                              beta,
-                                                              is_interleaved_transposed,
-                                                              reshape_info,
-                                                              gpu_target,
-                                                              num_elements_processed)
-                                .first);
-
-    return Status{};
-}
-
-void CLGEMMMatrixMultiplyKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    if(_input1->info()->num_dimensions() < 3)
-    {
-        // The stride_z for matrix B must be zero if we do not slice
-        ARM_COMPUTE_ERROR_ON(_input1->info()->strides_in_bytes()[3] != 0);
-    }
-
-    Window slice          = window.first_slice_window_3D();
-    Window slice_matrix_b = slice;
-
-    slice_matrix_b.set(Window::DimX, Window::Dimension(0, 1, 1));
-    slice_matrix_b.set(Window::DimY, Window::Dimension(0, 1, 1));
-
-    const unsigned int num_arguments_bias = _add_bias ? num_arguments_per_2D_tensor() + 1 : 0;
-
-    if(_reinterpret_input_as_3d)
-    {
-        // Pass bottom paddings to the kernel if the input has to be reinterpreted as 3D tensor
-        const unsigned int idx0                  = 3 * num_arguments_per_2D_tensor() + 3 + num_arguments_bias;
-        const unsigned int total_cross_plane_pad = _input0->info()->padding().top + _input0->info()->padding().bottom;
-        _kernel.setArg<cl_uint>(idx0, static_cast<unsigned int>(total_cross_plane_pad));
-    }
-
-    if(_reinterpret_output_as_3d)
-    {
-        // Pass bottom paddings to the kernel if the output has to be reinterpreted as 3D tensor
-        const unsigned int idx0                  = 3 * num_arguments_per_2D_tensor() + 3 + (_reinterpret_input_as_3d ? 1 : 0) + num_arguments_bias;
-        const unsigned int total_cross_plane_pad = _output->info()->padding().top + _output->info()->padding().bottom;
-        _kernel.setArg<cl_uint>(idx0, static_cast<unsigned int>(total_cross_plane_pad));
-    }
-
-    do
-    {
-        Window slice_b = slice;
-        // Don't slice matrix B along the z dimension if matrix B has just 2 dimensions and matrix A more than 2
-        // This scenario can happen when the matrix multiplication is used to perform a convolution operation
-        if(!_slide_matrix_b)
-        {
-            slice_b = slice_matrix_b;
-        }
-
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input0, slice);
-        add_2D_tensor_argument(idx, _input1, slice_b);
-        if(_add_bias)
-        {
-            add_2D_tensor_argument(idx, _input2, slice);
-        }
-        add_2D_tensor_argument(idx, _output, slice);
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input0->info()->strides_in_bytes()[2]));
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input1->info()->strides_in_bytes()[2]));
-        if(_add_bias)
-        {
-            _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input2->info()->strides_in_bytes()[2]));
-        }
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_output->info()->strides_in_bytes()[2]));
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_3D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.h b/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.h
deleted file mode 100644
index 71d223b8ac..0000000000
--- a/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.h
+++ /dev/null
@@ -1,122 +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_CLGEMMMATRIXMULTIPLYKERNEL_H
-#define ARM_COMPUTE_CLGEMMMATRIXMULTIPLYKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to multiply two input matrices "A" and "B" and add a martix "C" if provided. All elements of the output matrix will be multiplied by alpha. In case matrix C is passed, it will be added to the previous result.
- *  For the matrix C, the broadcast addition is supported if the flag "broadcast_bias" is set in the GEMMReshapeInfo object
- *
- * @note If the input tensors @p input0 and @p input1 have been reshaped respectively with @ref CLGEMMReshapeLHSMatrixKernel" and @ref CLGEMMReshapeRHSMatrixKernel,
- *       the flag @p is_interleaved_transposed must be set to true
- *
- * @attention @p input1 tensor must have at least 2 dimensions (matrix)
- *
- */
-class CLGEMMMatrixMultiplyKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLGEMMMatrixMultiplyKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMMatrixMultiplyKernel(const CLGEMMMatrixMultiplyKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMMatrixMultiplyKernel &operator=(const CLGEMMMatrixMultiplyKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLGEMMMatrixMultiplyKernel(CLGEMMMatrixMultiplyKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLGEMMMatrixMultiplyKernel &operator=(CLGEMMMatrixMultiplyKernel &&) = default;
-    /** Initialise the kernel's input, output and alpha
-     *
-     * @param[in]  input0                    Input tensor containing the Matrix A. Data types supported: F16/F32
-     * @param[in]  input1                    Input tensor containing the Matrix B. Data type supported: same as @p input0
-     * @param[in]  input2                    Input tensor containing the Matrix C (bias). Can be nullptr. Data type supported: same as @p input0
-     * @param[out] output                    Output tensor to store the result of matrix multiplication. Data type supported: same as @p input0
-     * @param[in]  alpha                     Weight of the matrix product
-     * @param[in]  beta                      (Optional) Weight of vector C. Default value is 0. Only beta = 1 is currently supported.
-     * @param[in]  is_interleaved_transposed (Optional) True if input0 and input1 have been reshaped respectively using @ref CLGEMMReshapeLHSMatrixKernel and @ref CLGEMMReshapeRHSMatrixKernel
-     * @param[in]  reshape_info              (Optional) GEMM reshape info. If is_interleaved_transposed = true, this object must contain the information to understand how the matrix A and matrix B have been reshaped
-     * @param[in]  fp_mixed_precision        (Optional) Use wider accumulators (32 bit instead of 16 for FP16) to improve accuracy
-     * @param[in]  activation_info           (Optional) Activation to apply after the matrix multiplication
-     *
-     */
-    void configure(const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta = 0.f,
-                   bool is_interleaved_transposed = true, const GEMMReshapeInfo &reshape_info = GEMMReshapeInfo(), bool fp_mixed_precision = false, const ActivationLayerInfo &activation_info = ActivationLayerInfo());
-    /** Initialise the kernel's input, output and alpha
-     *
-     * @param[in]  compile_context           The compile context to be used.
-     * @param[in]  input0                    Input tensor containing the Matrix A. Data types supported: F16/F32
-     * @param[in]  input1                    Input tensor containing the Matrix B. Data type supported: same as @p input0
-     * @param[in]  input2                    Input tensor containing the Matrix C (bias). Can be nullptr. Data type supported: same as @p input0
-     * @param[out] output                    Output tensor to store the result of matrix multiplication. Data type supported: same as @p input0
-     * @param[in]  alpha                     Weight of the matrix product
-     * @param[in]  beta                      (Optional) Weight of vector C. Default value is 0. Only beta = 1 is currently supported.
-     * @param[in]  is_interleaved_transposed (Optional) True if input0 and input1 have been reshaped respectively using @ref CLGEMMReshapeLHSMatrixKernel and @ref CLGEMMReshapeRHSMatrixKernel
-     * @param[in]  reshape_info              (Optional) GEMM reshape info. If is_interleaved_transposed = true, this object must contain the information to understand how the matrix A and matrix B have been reshaped
-     * @param[in]  fp_mixed_precision        (Optional) Use wider accumulators (32 bit instead of 16 for FP16) to improve accuracy
-     * @param[in]  activation_info           (Optional) Activation to apply after the matrix multiplication
-     *
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta = 0.f,
-                   bool is_interleaved_transposed = true, const GEMMReshapeInfo &reshape_info = GEMMReshapeInfo(), bool fp_mixed_precision = false, const ActivationLayerInfo &activation_info = ActivationLayerInfo());
-    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMMatrixMultiplyKernel
-     *
-     * @param[in] input0                    Input tensor containing the Matrix A info. Data types supported: F16/F32
-     * @param[in] input1                    Input tensor containing the Matrix B info. Data type supported: same as @p input0
-     * @param[in] input2                    Input tensor containing the Matrix C (bias) info. Can be nullptr. Data type supported: same as @p input0
-     * @param[in] output                    Output tensor to store the result of matrix multiplication. Data type supported: same as @p input0
-     * @param[in] alpha                     Weight of the matrix product
-     * @param[in] beta                      Weight of vector C. Default value is 0. Only beta = 1 is currently supported.
-     * @param[in] is_interleaved_transposed True if input0 and input1 have been reshaped respectively using @ref CLGEMMReshapeLHSMatrixKernel and @ref CLGEMMReshapeRHSMatrixKernel
-     * @param[in] reshape_info              GEMM reshape info. If is_interleaved_transposed = true, this object must contain the information to understand how the matrix A and matrix B have been reshaped
-     * @param[in] gpu_target                GPU Target
-     * @param[in] fp_mixed_precision        (Optional) Use wider accumulators (32 bit instead of 16 for FP16) to improve accuracy
-     * @param[in] activation_info           (Optional) Activation to apply after the matrix multiplication
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta,
-                           bool is_interleaved_transposed, const GEMMReshapeInfo &reshape_info, GPUTarget gpu_target, bool fp_mixed_precision = false, const ActivationLayerInfo &activation_info = ActivationLayerInfo());
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-public:
-    const ICLTensor *_input0;
-    const ICLTensor *_input1;
-    const ICLTensor *_input2;
-    ICLTensor       *_output;
-    bool             _slide_matrix_b;
-    bool             _reinterpret_input_as_3d;
-    bool             _reinterpret_output_as_3d;
-    bool             _add_bias;
-    bool             _broadcast_bias;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLGEMMMATRIXMULTIPLYKERNEL_H */
diff --git a/src/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.cpp b/src/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.cpp
deleted file mode 100644
index 1fe298c0a1..0000000000
--- a/src/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.cpp
+++ /dev/null
@@ -1,420 +0,0 @@
-/*
- * 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/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.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/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "src/core/utils/helpers/float_ops.h"
-#include "support/StringSupport.h"
-
-#include <cstddef>
-#include <cstdint>
-#include <tuple>
-
-using namespace arm_compute::misc::shape_calculator;
-
-namespace arm_compute
-{
-namespace
-{
-using ElementsProcessed = Steps;
-
-Status validate_arguments(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info,
-                          const GEMMRHSMatrixInfo &rhs_info,
-                          const GEMMKernelInfo    &gemm_info)
-{
-    ARM_COMPUTE_UNUSED(alpha);
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input0, input1, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::F32);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input0->num_dimensions() > 4, "The number of dimensions for the LHS matrix must be <= 4");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input1->num_dimensions() > 3, "The number of dimensions for the RHS matrix must be <= 3");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((rhs_info.k0 & (rhs_info.k0 - 1)) && rhs_info.k0 != 3), "Only 2,3,4,8,16 are supported for k0");
-    ARM_COMPUTE_RETURN_ERROR_ON(rhs_info.k0 > 16);
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.m0 < 1 || lhs_info.m0 > 8);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((rhs_info.n0 & (rhs_info.n0 - 1)) && rhs_info.n0 != 3), "Only 2,3,4,8,16 are supported for n0");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG((gemm_info.reinterpret_input_as_3d || gemm_info.depth_output_gemm3d != 0) && (input2 != nullptr)
-                                    && (!gemm_info.broadcast_bias),
-                                    "Bias addition only supported with broadcast mode in case the input or output has to be reinterpreted as 3D");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(gemm_info.fp_mixed_precision, "Mixed precision not supported");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(rhs_info.export_to_cl_image, "Export to CLImage not supported for GEMM native");
-
-    const unsigned int m = gemm_info.m;
-    const unsigned int n = gemm_info.n;
-    const unsigned int k = gemm_info.k;
-
-    ARM_COMPUTE_UNUSED(m);
-    ARM_COMPUTE_UNUSED(n);
-    ARM_COMPUTE_UNUSED(k);
-
-    ARM_COMPUTE_RETURN_ERROR_ON(input0->dimension(0) != k);
-    ARM_COMPUTE_RETURN_ERROR_ON(input1->dimension(0) != n);
-    ARM_COMPUTE_RETURN_ERROR_ON(input1->dimension(1) != k);
-    if(gemm_info.reinterpret_input_as_3d)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(input0->dimension(1) * input0->dimension(2) != m);
-    }
-    else
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(input0->dimension(1) != m);
-    }
-
-    if(input2 != nullptr && !(helpers::float_ops::is_zero(beta)))
-    {
-        const unsigned int input2_dim0 = input2->dimension(0);
-        const unsigned int input2_dim1 = input2->dimension(1);
-
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input2, input1);
-        if(gemm_info.broadcast_bias)
-        {
-            ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim1 != 1 || input2_dim0 != n), "Incorrect dimension of bias matrix which is to be broadcasted");
-        }
-        else
-        {
-            ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim0 != n || input2_dim1 != m), "Incorrect dimension of bias matrix");
-        }
-    }
-
-    if(output->total_size() != 0)
-    {
-        const TensorInfo tensor_info_output = output->clone()->set_tensor_shape(compute_mm_shape(*input0, *input1, gemm_info));
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(output, &tensor_info_output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input0, output);
-    }
-
-    return Status{};
-}
-
-std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input0, ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output, const GEMMLHSMatrixInfo &lhs_info,
-                                                        const GEMMRHSMatrixInfo &rhs_info,
-                                                        const GEMMKernelInfo &gemm_info, ElementsProcessed &num_elements_processed)
-{
-    unsigned int &num_elems_processed_per_iteration_x = num_elements_processed[0];
-    unsigned int &num_elems_processed_per_iteration_y = num_elements_processed[1];
-    bool          reinterpret_input_as_3d             = gemm_info.reinterpret_input_as_3d;
-    bool          reinterpret_output_as_3d            = gemm_info.depth_output_gemm3d != 0;
-
-    Window win{};
-    Window win_out{};
-    bool   window_changed = false;
-
-    // In case both input and output have to be reinterpreted as 3D tensors,
-    // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
-    if(reinterpret_input_as_3d == reinterpret_output_as_3d)
-    {
-        reinterpret_output_as_3d = false;
-    }
-
-    // Output tensor auto initialization if not yet initialized
-    auto_init_if_empty(*output, input0->clone()->set_tensor_shape(compute_mm_shape(*input0, *input1, gemm_info)));
-
-    TensorInfo tmp_info(*output);
-
-    if(reinterpret_output_as_3d)
-    {
-        // Since the output tensor has to be reinterpreted as 3D and the execute window is based on a 2D GEMM,
-        // the window needs to be constructed on the 2D collapsed version of the tensor
-        TensorShape tmp_shape(output->tensor_shape());
-        tmp_shape.collapse(2U, 1U);
-        tmp_info.set_tensor_shape(tmp_shape);
-    }
-
-    // Configure kernel window
-    num_elems_processed_per_iteration_x = rhs_info.n0;
-    num_elems_processed_per_iteration_y = lhs_info.m0;
-
-    win     = calculate_max_window(tmp_info, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-    win_out = calculate_max_window(*output, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-
-    AccessWindowStatic input0_access(input0, 0, 0,
-                                     input0->dimension(0),
-                                     input0->dimension(1));
-    AccessWindowStatic input1_access(input1, 0, 0,
-                                     ceil_to_multiple(input1->dimension(0), num_elems_processed_per_iteration_x),
-                                     input1->dimension(1));
-    AccessWindowStatic output_access(output, 0, 0,
-                                     output->dimension(0),
-                                     output->dimension(1));
-
-    if(input2 != nullptr)
-    {
-        const int bias_processed_per_iteration_x = num_elems_processed_per_iteration_x;
-
-        AccessWindowStatic input2_access(input2, 0, 0,
-                                         ceil_to_multiple(input2->dimension(0), bias_processed_per_iteration_x),
-                                         input2->dimension(1));
-
-        window_changed = update_window_and_padding(win, input0_access, input1_access, input2_access) || // window used by the execute_window_loop
-                         update_window_and_padding(win_out, output_access);                             // window used to update the padding requirements of output tensor
-    }
-    else
-    {
-        window_changed = update_window_and_padding(win, input0_access, input1_access) || // window used by the execute_window_loop
-                         update_window_and_padding(win_out, output_access);              // window used to update the padding requirements of output tensor
-    }
-
-    // Collapse along the Z direction
-    // This collapse needs to be here in order to tune the Z dimension of LWS
-    Window             collapsed             = win;
-    const unsigned int dimension_to_collapse = std::min(static_cast<unsigned int>(output->num_dimensions()), 2u);
-    collapsed                                = win.collapse(win, dimension_to_collapse);
-
-    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
-    return std::make_pair(err, collapsed);
-}
-} // namespace
-
-CLGEMMMatrixMultiplyNativeKernel::CLGEMMMatrixMultiplyNativeKernel()
-    : _input0(nullptr), _input1(nullptr), _input2(nullptr), _output(nullptr), _slide_matrix_b(true), _reinterpret_input_as_3d(false), _reinterpret_output_as_3d(false), _use_dummy_work_items(false),
-      _add_bias(false), _broadcast_bias(false)
-{
-}
-
-void CLGEMMMatrixMultiplyNativeKernel::configure(const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta,
-                                                 const GEMMLHSMatrixInfo &lhs_info,
-                                                 const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input0, input1, input2, output, alpha, beta, lhs_info, rhs_info, gemm_info);
-}
-
-void CLGEMMMatrixMultiplyNativeKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha,
-                                                 float                    beta,
-                                                 const GEMMLHSMatrixInfo &lhs_info,
-                                                 const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input0, input1, output);
-
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input0->info(), input1->info(), (input2 != nullptr ? input2->info() : nullptr), output->info(), alpha, beta, lhs_info, rhs_info, gemm_info));
-
-    auto padding_info         = get_padding_info({ input0, output });
-    _input0                   = input0;
-    _input1                   = input1;
-    _input2                   = helpers::float_ops::is_zero(beta) ? nullptr : input2;
-    _output                   = output;
-    _reinterpret_input_as_3d  = gemm_info.reinterpret_input_as_3d;
-    _reinterpret_output_as_3d = gemm_info.depth_output_gemm3d != 0;
-    _use_dummy_work_items     = preferred_dummy_work_items_support(CLKernelLibrary::get().get_device());
-    _add_bias                 = _input2 != nullptr;
-    _broadcast_bias           = gemm_info.broadcast_bias;
-
-    // In case both input and output have to be reinterpreted as 3D tensors,
-    // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
-    if(_reinterpret_input_as_3d == _reinterpret_output_as_3d)
-    {
-        _reinterpret_input_as_3d  = false;
-        _reinterpret_output_as_3d = false;
-    }
-
-    // Check if we need to slide the matrix B
-    const unsigned int num_dimensions_input0 = _input0->info()->num_dimensions();
-    _slide_matrix_b                          = (_input1->info()->num_dimensions() >= num_dimensions_input0);
-
-    ElementsProcessed num_elements_processed{};
-
-    // Configure kernel window
-    auto win_config = validate_and_configure_window(input0->info(), input1->info(), input2 != nullptr ? input2->info() : nullptr, output->info(), lhs_info, rhs_info, gemm_info, num_elements_processed);
-    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
-    ICLKernel::configure_internal(win_config.second);
-
-    // If _reinterpret_input_as_3d = _reinterpret_output_as_3d = true,
-    // we will dispatch a batched-GEMM to reduce the complexity of the address calculation within the OpenCL kernel.
-    // This means that the actual m used by the kernel is given by output->info()->dimension(1) and not by gemm_info.m
-    const unsigned int internal_m = _reinterpret_output_as_3d ? gemm_info.m : output->info()->dimension(1);
-
-    const unsigned int h_gemm_3d = _reinterpret_output_as_3d ? output->info()->dimension(1) : input0->info()->dimension(1);
-    const unsigned int d_gemm_3d = _reinterpret_output_as_3d ? output->info()->dimension(2) : input0->info()->dimension(2);
-
-    // Calculate partial (store instead of load) M0 and partial N0 for the partial blocks at the end of a row/column if any. This is to avoid padding.
-    const unsigned int partial_store_m0 = internal_m % lhs_info.m0;
-    const unsigned int partial_store_n0 = gemm_info.n % rhs_info.n0;
-
-    // Shrink M0 to be always <= M (internal_m) to prevent out-of-bounds reads.
-    // NOTE: This might have implications on heuristics and performance
-    const unsigned int internal_m0 = std::min(internal_m, lhs_info.m0);
-
-    // Create build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input0->info()->data_type()));
-    build_opts.add_option_if(!(helpers::float_ops::is_one(alpha)), "-DALPHA=" + float_to_string_with_full_precision(alpha));
-    build_opts.add_option_if(_input2 != nullptr, "-DBETA=" + float_to_string_with_full_precision(beta));
-    build_opts.add_option_if(helpers::float_ops::is_one(beta), "-DUNIT_BETA");
-    build_opts.add_option_if(gemm_info.broadcast_bias, "-DBROADCAST_BIAS");
-    build_opts.add_option_if(_reinterpret_input_as_3d, "-DREINTERPRET_INPUT_AS_3D");
-    build_opts.add_option_if(_reinterpret_output_as_3d, "-DREINTERPRET_OUTPUT_AS_3D");
-    build_opts.add_option_if(_reinterpret_input_as_3d || _reinterpret_output_as_3d, "-DHEIGHT_GEMM3D=" + support::cpp11::to_string(h_gemm_3d));
-    build_opts.add_option_if(_reinterpret_input_as_3d || _reinterpret_output_as_3d, "-DDEPTH_GEMM3D=" + support::cpp11::to_string(d_gemm_3d));
-    build_opts.add_option_if(!_slide_matrix_b, "-DMATRIX_B_DEPTH=" + support::cpp11::to_string(input1->info()->dimension(2)));
-    build_opts.add_option_if(_use_dummy_work_items, "-DDUMMY_WORK_ITEMS");
-    build_opts.add_option("-DM=" + support::cpp11::to_string(internal_m));
-    build_opts.add_option("-DN=" + support::cpp11::to_string(gemm_info.n));
-    build_opts.add_option("-DK=" + support::cpp11::to_string(gemm_info.k));
-    build_opts.add_option("-DM0=" + support::cpp11::to_string(internal_m0));
-    build_opts.add_option("-DN0=" + support::cpp11::to_string(rhs_info.n0));
-    build_opts.add_option("-DK0=" + support::cpp11::to_string(rhs_info.k0));
-    build_opts.add_option("-DPARTIAL_STORE_M0=" + support::cpp11::to_string(partial_store_m0));
-    build_opts.add_option("-DPARTIAL_STORE_N0=" + support::cpp11::to_string(partial_store_n0));
-    build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DACTIVATION_TYPE=" + lower_string(string_from_activation_func(gemm_info.activation_info.activation())));
-    build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DA_VAL=" + float_to_string_with_full_precision(gemm_info.activation_info.a()));
-    build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DB_VAL=" + float_to_string_with_full_precision(gemm_info.activation_info.b()));
-
-    std::string kernel_name("gemm_mm_native");
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += (_add_bias ? "add_bias_" : "");
-    _config_id += (_broadcast_bias ? "broadcast_bias_" : "");
-    _config_id += (_reinterpret_input_as_3d ? "3di_" : "");
-    _config_id += (_reinterpret_output_as_3d ? "3do_" : "");
-    _config_id += (gemm_info.activation_info.enabled() ? "fused_activation_" : "");
-    _config_id += lower_string(string_from_data_type(input0->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(gemm_info.k);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(2));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.m0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(rhs_info.n0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(rhs_info.k0);
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-Status CLGEMMMatrixMultiplyNativeKernel::validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta,
-                                                  const GEMMLHSMatrixInfo &lhs_info,
-                                                  const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
-{
-    ElementsProcessed num_elements_processed{};
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input0, input1, input2, output, alpha, beta, lhs_info, rhs_info, gemm_info));
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input0->clone().get(),
-                                                              input1->clone().get(),
-                                                              input2 != nullptr ? input2->clone().get() : nullptr,
-                                                              output->clone().get(),
-                                                              lhs_info,
-                                                              rhs_info,
-                                                              gemm_info,
-                                                              num_elements_processed)
-                                .first);
-
-    return Status{};
-}
-
-void CLGEMMMatrixMultiplyNativeKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    if(_input1->info()->num_dimensions() < 3)
-    {
-        // The stride_z for matrix B must be zero if we do not slice
-        ARM_COMPUTE_ERROR_ON(_input1->info()->strides_in_bytes()[3] != 0);
-    }
-
-    Window slice          = window.first_slice_window_3D();
-    Window slice_matrix_b = slice;
-
-    slice_matrix_b.set(Window::DimX, Window::Dimension(0, 1, 1));
-    slice_matrix_b.set(Window::DimY, Window::Dimension(0, 1, 1));
-
-    if(_reinterpret_input_as_3d)
-    {
-        // Pass bottom paddings to the kernel if the input has to be reinterpreted as 3D tensor
-        unsigned int idx0;
-        if(_add_bias)
-        {
-            idx0 = 4 * num_arguments_per_2D_tensor() + 4;
-        }
-        else
-        {
-            idx0 = 3 * num_arguments_per_2D_tensor() + 3;
-        }
-        const unsigned int total_cross_plane_pad = _input0->info()->padding().top + _input0->info()->padding().bottom;
-        _kernel.setArg<cl_uint>(idx0, static_cast<unsigned int>(total_cross_plane_pad));
-    }
-
-    if(_reinterpret_output_as_3d)
-    {
-        // Pass bottom paddings to the kernel if the output has to be reinterpreted as 3D tensor
-        unsigned int idx0;
-        if(_add_bias)
-        {
-            idx0 = 4 * num_arguments_per_2D_tensor() + 4 + (_reinterpret_input_as_3d ? 1 : 0);
-        }
-        else
-        {
-            idx0 = 3 * num_arguments_per_2D_tensor() + 3 + (_reinterpret_input_as_3d ? 1 : 0);
-        }
-        const unsigned int total_cross_plane_pad = _output->info()->padding().top + _output->info()->padding().bottom;
-        _kernel.setArg<cl_uint>(idx0, static_cast<unsigned int>(total_cross_plane_pad));
-    }
-
-    do
-    {
-        Window slice_b = slice;
-        // Don't slice matrix B along the z dimension if matrix B has just 2 dimensions and matrix A more than 2
-        // This scenario can happen when the matrix multiplication is used to perform a convolution operation
-        if(!_slide_matrix_b)
-        {
-            slice_b = slice_matrix_b;
-        }
-
-        unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _input0, slice);
-        add_2D_tensor_argument(idx, _input1, slice_b);
-        if(_add_bias)
-        {
-            add_2D_tensor_argument(idx, _input2, slice);
-        }
-        add_2D_tensor_argument(idx, _output, slice);
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input0->info()->strides_in_bytes()[2]));
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input1->info()->strides_in_bytes()[2]));
-        if(_add_bias)
-        {
-            _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input2->info()->strides_in_bytes()[2]));
-        }
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_output->info()->strides_in_bytes()[2]));
-        enqueue(queue, *this, slice, lws_hint(), _use_dummy_work_items);
-    }
-    while(window.slide_window_slice_3D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.h b/src/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.h
deleted file mode 100644
index 6b6004b464..0000000000
--- a/src/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.h
+++ /dev/null
@@ -1,127 +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_CLGEMMMATRIXMULTIPLYNATIVEKERNEL_H
-#define ARM_COMPUTE_CLGEMMMATRIXMULTIPLYNATIVEKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-#include "arm_compute/core/KernelDescriptors.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to multiply matrices when neither of the input matrices have been reshaped */
-class CLGEMMMatrixMultiplyNativeKernel : public ICLKernel
-{
-public:
-    /** Default Constructor */
-    CLGEMMMatrixMultiplyNativeKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMMatrixMultiplyNativeKernel(const CLGEMMMatrixMultiplyNativeKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMMatrixMultiplyNativeKernel &operator=(const CLGEMMMatrixMultiplyNativeKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLGEMMMatrixMultiplyNativeKernel(CLGEMMMatrixMultiplyNativeKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLGEMMMatrixMultiplyNativeKernel &operator=(CLGEMMMatrixMultiplyNativeKernel &&) = default;
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  input0    Input tensor for the LHS matrix. Data type supported: F32. The number of dimensions for the LHS matrix must be less or equal than 4.
-     * @param[in]  input1    Input tensor for the RHS matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3.
-     * @param[in]  input2    Input tensor containing the bias matrix. Data type supported: same as @p input0.
-     * @param[out] output    Output tensor info. Data type supported: same as @p input0
-     * @param[in]  alpha     Weight of the matrix product
-     * @param[in]  beta      Weight of the matrix bias
-     * @param[in]  lhs_info  LHS matrix information used to retrieve the number of rows and accumulations to be processed by each thread. Only the following values are supported:
-     *                       lhs_info.m0: 1,2,3,4,5,6,7,8
-     *                       lhs_info.k0: 2,3,4,8,16
-     * @param[in]  rhs_info  RHS matrix information used to retrieve the number of columns and accumulations to be processed by each thread. Only the following values are supported:
-     *                       rhs_info.n0: 2,3,4,8,16
-     *                       rhs_info.k0: same of lhs_info.k0
-     * @param[in]  gemm_info GEMM information used to retrieve the original dimensions of the input matrices
-     */
-    void configure(const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info,
-                   const GEMMRHSMatrixInfo &rhs_info,
-                   const GEMMKernelInfo    &gemm_info);
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input0          Input tensor for the LHS matrix. Data type supported: F32. The number of dimensions for the LHS matrix must be less or equal than 4.
-     * @param[in]  input1          Input tensor for the RHS matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3.
-     * @param[in]  input2          Input tensor containing the bias matrix. Data type supported: same as @p input0.
-     * @param[out] output          Output tensor info. Data type supported: same as @p input0
-     * @param[in]  alpha           Weight of the matrix product
-     * @param[in]  beta            Weight of the matrix bias
-     * @param[in]  lhs_info        LHS matrix information used to retrieve the number of rows and accumulations to be processed by each thread. Only the following values are supported:
-     *                             lhs_info.m0: 1,2,3,4,5,6,7,8
-     *                             lhs_info.k0: 2,3,4,8,16
-     * @param[in]  rhs_info        RHS matrix information used to retrieve the number of columns and accumulations to be processed by each thread. Only the following values are supported:
-     *                             rhs_info.n0: 2,3,4,8,16
-     *                             rhs_info.k0: same of lhs_info.k0
-     * @param[in]  gemm_info       GEMM information used to retrieve the original dimensions of the input matrices
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta,
-                   const GEMMLHSMatrixInfo &lhs_info,
-                   const GEMMRHSMatrixInfo &rhs_info,
-                   const GEMMKernelInfo    &gemm_info);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMMatrixMultiplyNativeKernel
-     *
-     * @param[in] input0    Input tensor info for the LHS matrix. Data type supported: F32. The number of dimensions for the LHS matrix must be less or equal than 4.
-     * @param[in] input1    Input tensor info for the RHS matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3.
-     * @param[in] input2    Input tensor info containing the bias matrix. Data type supported: same as @p input0.
-     * @param[in] output    Output tensor info. Data type supported: same as @p input0
-     * @param[in] alpha     Weight of the matrix product
-     * @param[in] beta      Weight of the matrix bias
-     * @param[in] lhs_info  LHS matrix information used to retrieve the number of rows and accumulations to be processed by each thread. Only the following values are supported:
-     *                      lhs_info.m0: 1,2,3,4,5,6,7,8
-     *                      lhs_info.k0: 2,3,4,8,16
-     * @param[in] rhs_info  RHS matrix information used to retrieve the number of columns and accumulations to be processed by each thread. Only the following values are supported:
-     *                      rhs_info.n0: 2,3,4,8,16
-     *                      rhs_info.k0: same of lhs_info.k0
-     * @param[in] gemm_info GEMM information used to retrieve the original dimensions of the input matrices
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info,
-                           const GEMMRHSMatrixInfo &rhs_info,
-                           const GEMMKernelInfo    &gemm_info);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input0;
-    const ICLTensor *_input1;
-    const ICLTensor *_input2;
-    ICLTensor       *_output;
-    bool             _slide_matrix_b;
-    bool             _reinterpret_input_as_3d;
-    bool             _reinterpret_output_as_3d;
-    bool             _use_dummy_work_items;
-    bool             _add_bias;
-    bool             _broadcast_bias;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLGEMMMATRIXMULTIPLYNATIVEKERNEL_H*/
diff --git a/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.cpp b/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.cpp
deleted file mode 100644
index d270f92615..0000000000
--- a/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.cpp
+++ /dev/null
@@ -1,425 +0,0 @@
-/*
- * 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/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.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/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/CL/CLUtils.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/CL/gemm/CLGEMMHelpers.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "src/core/utils/helpers/float_ops.h"
-#include "support/StringSupport.h"
-
-#include <cstddef>
-#include <cstdint>
-#include <tuple>
-
-using namespace arm_compute;
-using namespace arm_compute::misc::shape_calculator;
-
-namespace arm_compute
-{
-class Coordinates;
-} // namespace arm_compute
-
-namespace
-{
-using ElementsProcessed = Steps;
-
-Status validate_arguments(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info,
-                          const GEMMRHSMatrixInfo &rhs_info,
-                          const GEMMKernelInfo    &gemm_info)
-{
-    ARM_COMPUTE_UNUSED(alpha);
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input0, input1, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input0);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::F16, DataType::F32);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input0->num_dimensions() > 4, "The number of dimensions for the LHS matrix must be <= 4");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input1->num_dimensions() > 3, "The number of dimensions for the RHS matrix must be <= 3");
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.k0 != rhs_info.k0);
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.transpose == rhs_info.transpose);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((lhs_info.k0 & (lhs_info.k0 - 1)) && lhs_info.k0 != 3), "Only 2,3,4,8,16 are supported for k0");
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.k0 > 16);
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.m0 < 2 || lhs_info.m0 > 8);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG((lhs_info.transpose) && ((lhs_info.m0 & (lhs_info.m0 - 1)) && lhs_info.m0 != 3), "Only 2,3,4,8,16 are supported for m0");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG((rhs_info.transpose) && ((rhs_info.n0 & (rhs_info.n0 - 1)) && rhs_info.n0 != 3), "Only 2,3,4,8,16 are supported for n0");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG((gemm_info.reinterpret_input_as_3d || gemm_info.depth_output_gemm3d != 0) && (input2 != nullptr)
-                                    && (!gemm_info.broadcast_bias),
-                                    "Bias addition only supported with broadcast mode in case the input or output has to be reinterpreted as 3D");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(gemm_info.fp_mixed_precision && (input0->data_type() == DataType::F32), "Mixed precision only supported for F16 data type");
-    ARM_COMPUTE_RETURN_ON_ERROR(cl_gemm::validate_image2d_support_on_rhs(*input1, rhs_info));
-
-    const unsigned int m = gemm_info.m;
-    const unsigned int n = gemm_info.n;
-    const unsigned int k = gemm_info.k;
-
-    TensorShape tensor_shape0{ input0->tensor_shape() };
-    tensor_shape0.set(0, k);
-    tensor_shape0.set(1, m);
-
-    TensorShape tensor_shape1{ input1->tensor_shape() };
-    tensor_shape1.set(0, n);
-    tensor_shape1.set(1, k);
-
-    if(input2 != nullptr && !(helpers::float_ops::is_zero(beta)))
-    {
-        const unsigned int input2_dim0 = input2->dimension(0);
-        const unsigned int input2_dim1 = input2->dimension(1);
-
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input2, input1);
-        if(gemm_info.broadcast_bias)
-        {
-            ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim1 != 1 || input2_dim0 != n), "Incorrect dimension of bias matrix which is to be broadcasted");
-        }
-        else
-        {
-            ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim0 != n || input2_dim1 != m), "Incorrect dimension of bias matrix");
-        }
-    }
-
-    const TensorInfo tensor_info0 = input0->clone()->set_tensor_shape(tensor_shape0);
-    const TensorInfo tensor_info1 = input1->clone()->set_tensor_shape(tensor_shape1);
-
-    const TensorInfo tensor_info_reshaped0 = input0->clone()->set_tensor_shape(compute_lhs_reshaped_shape(tensor_info0, lhs_info));
-    const TensorInfo tensor_info_reshaped1 = input1->clone()->set_tensor_shape(compute_rhs_reshaped_shape(tensor_info1, rhs_info));
-
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input0, &tensor_info_reshaped0);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input1, &tensor_info_reshaped1);
-
-    if(output->total_size() != 0)
-    {
-        const TensorInfo tensor_info_output = output->clone()->set_tensor_shape(compute_mm_shape(*input0, *input1, gemm_info));
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(output, &tensor_info_output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input0, output);
-    }
-
-    return Status{};
-}
-
-std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input0, ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output, const GEMMLHSMatrixInfo &lhs_info,
-                                                        const GEMMRHSMatrixInfo &rhs_info,
-                                                        const GEMMKernelInfo &gemm_info, ElementsProcessed &num_elements_processed)
-{
-    unsigned int &num_elems_processed_per_iteration_x = num_elements_processed[0];
-    unsigned int &num_elems_processed_per_iteration_y = num_elements_processed[1];
-    bool          reinterpret_output_as_3d            = gemm_info.depth_output_gemm3d != 0;
-
-    Window win{};
-    Window win_out{};
-    bool   window_changed = false;
-
-    // Output tensor auto initialization if not yet initialized
-    auto_init_if_empty(*output, input0->clone()->set_tensor_shape(compute_mm_shape(*input0, *input1, gemm_info)));
-
-    TensorInfo tmp_info(*output);
-
-    if(reinterpret_output_as_3d)
-    {
-        // Since the output tensor has to be reinterpreted as 3D and the execute window is based on a 2D GEMM,
-        // the window needs to be constructed on the 2D collapsed version of the tensor
-        TensorShape tmp_shape(output->tensor_shape());
-        tmp_shape.collapse(2U, 1U);
-        tmp_info.set_tensor_shape(tmp_shape);
-    }
-
-    // Configure kernel window
-    num_elems_processed_per_iteration_x = rhs_info.n0;
-    num_elems_processed_per_iteration_y = lhs_info.m0;
-
-    win     = calculate_max_window(tmp_info, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-    win_out = calculate_max_window(*output, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-
-    if(input2 != nullptr)
-    {
-        const int bias_processed_per_iteration_x = num_elems_processed_per_iteration_x;
-
-        const int bias_processed_per_iteration_y = gemm_info.broadcast_bias ? 1 : num_elems_processed_per_iteration_y;
-
-        AccessWindowStatic input2_access(input2, 0, 0,
-                                         ceil_to_multiple(input2->dimension(0), bias_processed_per_iteration_x),
-                                         ceil_to_multiple(input2->dimension(1), bias_processed_per_iteration_y));
-
-        window_changed = update_window_and_padding(win, input2_access); // window used by the execute_window_loop
-    }
-
-    // Collapse along the Z direction
-    // This collapse needs to be here in order to tune the Z dimension of LWS
-    Window             collapsed             = win;
-    const unsigned int dimension_to_collapse = std::min(static_cast<unsigned int>(output->num_dimensions()), 2u);
-    collapsed                                = win.collapse(win, dimension_to_collapse);
-
-    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
-    return std::make_pair(err, collapsed);
-}
-} // namespace
-
-CLGEMMMatrixMultiplyReshapedKernel::CLGEMMMatrixMultiplyReshapedKernel()
-    : _input0(nullptr), _input1(nullptr), _input2(nullptr), _output(nullptr), _slide_matrix_b(true), _reinterpret_output_as_3d(false), _use_dummy_work_items(false), _add_bias(false),
-      _broadcast_bias(false), _export_to_cl_image(false), _k(1)
-{
-}
-
-void CLGEMMMatrixMultiplyReshapedKernel::configure(const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta,
-                                                   const GEMMLHSMatrixInfo &lhs_info,
-                                                   const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input0, input1, input2, output, alpha, beta, lhs_info, rhs_info, gemm_info);
-}
-
-void CLGEMMMatrixMultiplyReshapedKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha,
-                                                   float                    beta,
-                                                   const GEMMLHSMatrixInfo &lhs_info,
-                                                   const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input0, input1, output);
-
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input0->info(), input1->info(), (input2 != nullptr ? input2->info() : nullptr), output->info(), alpha, beta, lhs_info, rhs_info, gemm_info));
-
-    auto padding_info         = get_padding_info({ input0, output });
-    _input0                   = input0;
-    _input1                   = input1;
-    _input2                   = helpers::float_ops::is_zero(beta) ? nullptr : input2;
-    _output                   = output;
-    _reinterpret_output_as_3d = gemm_info.depth_output_gemm3d != 0;
-    _use_dummy_work_items     = preferred_dummy_work_items_support(CLKernelLibrary::get().get_device());
-    _add_bias                 = _input2 != nullptr;
-    _broadcast_bias           = gemm_info.broadcast_bias;
-    _export_to_cl_image       = rhs_info.export_to_cl_image;
-    _k                        = gemm_info.k;
-
-    // Check if we need to slide the matrix B
-    const unsigned int num_dimensions_input0 = _input0->info()->num_dimensions();
-    _slide_matrix_b                          = (_input1->info()->num_dimensions() >= num_dimensions_input0);
-
-    ElementsProcessed num_elements_processed{};
-
-    // Configure kernel window
-    auto win_config = validate_and_configure_window(input0->info(), input1->info(), input2 != nullptr ? input2->info() : nullptr, output->info(), lhs_info, rhs_info, gemm_info, num_elements_processed);
-    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
-    ICLKernel::configure_internal(win_config.second);
-
-    const bool     enable_mixed_precision = gemm_info.fp_mixed_precision;
-    const DataType data_type              = input0->info()->data_type();
-
-    // Calculate partial (store instead of load) M0 and partial N0 for the partial blocks at the end of a row/column if any. This is to avoid padding.
-    const unsigned int internal_m = _reinterpret_output_as_3d ? gemm_info.m : output->info()->dimension(1);
-
-    const unsigned int partial_store_m0 = internal_m % lhs_info.m0;
-    const unsigned int partial_store_n0 = gemm_info.n % rhs_info.n0;
-
-    // Create build options
-    CLBuildOptions build_opts;
-    build_opts.add_option_if(!(helpers::float_ops::is_one(alpha)), "-DALPHA=" + float_to_string_with_full_precision(alpha));
-    build_opts.add_option_if(_input2 != nullptr, "-DBETA=" + float_to_string_with_full_precision(beta));
-    build_opts.add_option_if(helpers::float_ops::is_one(beta), "-DUNIT_BETA");
-    build_opts.add_option_if(_reinterpret_output_as_3d, "-DREINTERPRET_OUTPUT_AS_3D");
-    build_opts.add_option_if(_reinterpret_output_as_3d, "-DHEIGHT_GEMM3D=" + support::cpp11::to_string(output->info()->dimension(1)));
-    build_opts.add_option_if(_reinterpret_output_as_3d, "-DDEPTH_GEMM3D=" + support::cpp11::to_string(output->info()->dimension(2)));
-    build_opts.add_option_if(gemm_info.broadcast_bias, "-DBROADCAST_BIAS");
-    build_opts.add_option_if(!_slide_matrix_b, "-DMATRIX_B_DEPTH=" + support::cpp11::to_string(input1->info()->dimension(2)));
-    build_opts.add_option_if(lhs_info.interleave, "-DLHS_INTERLEAVE");
-    build_opts.add_option_if(rhs_info.interleave, "-DRHS_INTERLEAVE");
-    build_opts.add_option_if(lhs_info.transpose, "-DLHS_TRANSPOSE");
-    build_opts.add_option_if(_use_dummy_work_items, "-DDUMMY_WORK_ITEMS");
-    build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DACTIVATION_TYPE=" + lower_string(string_from_activation_func(gemm_info.activation_info.activation())));
-    build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DA_VAL=" + float_to_string_with_full_precision(gemm_info.activation_info.a()));
-    build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DB_VAL=" + float_to_string_with_full_precision(gemm_info.activation_info.b()));
-    build_opts.add_option_if(enable_mixed_precision, "-DMIXED_PRECISION");
-    build_opts.add_option_if(rhs_info.export_to_cl_image, "-DOPENCL_IMAGE_SUPPORT");
-    build_opts.add_option("-DRHS_HEIGHT=" + support::cpp11::to_string(input1->info()->dimension(1)));
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
-    build_opts.add_option("-DDATA_TYPE_ACCUMULATOR=" + (enable_mixed_precision ? get_cl_type_from_data_type(DataType::F32) : get_cl_type_from_data_type(data_type)));
-    build_opts.add_option("-DM=" + support::cpp11::to_string(gemm_info.m));
-    build_opts.add_option("-DN=" + support::cpp11::to_string(gemm_info.n));
-    build_opts.add_option("-DK=" + support::cpp11::to_string(gemm_info.k));
-    build_opts.add_option("-DM0=" + support::cpp11::to_string(lhs_info.m0));
-    build_opts.add_option("-DN0=" + support::cpp11::to_string(rhs_info.n0));
-    build_opts.add_option("-DK0=" + support::cpp11::to_string(lhs_info.k0));
-    build_opts.add_option("-DV0=" + support::cpp11::to_string(lhs_info.v0));
-    build_opts.add_option("-DH0=" + support::cpp11::to_string(rhs_info.h0));
-    build_opts.add_option("-DPARTIAL_STORE_M0=" + support::cpp11::to_string(partial_store_m0));
-    build_opts.add_option("-DPARTIAL_STORE_N0=" + support::cpp11::to_string(partial_store_n0));
-
-    std::string kernel_name("gemm_mm_reshaped_");
-    kernel_name += lhs_info.transpose ? "lhs_t_" : "lhs_nt_";
-    kernel_name += rhs_info.transpose ? "rhs_t" : "rhs_nt";
-    kernel_name += rhs_info.export_to_cl_image ? "_texture" : "";
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += (_add_bias ? "add_bias_" : "");
-    _config_id += (_broadcast_bias ? "broadcast_bias_" : "");
-    _config_id += (_reinterpret_output_as_3d ? "3do_" : "");
-    _config_id += (gemm_info.activation_info.enabled() ? "fused_activation_" : "");
-    _config_id += lower_string(string_from_data_type(input0->info()->data_type()));
-    _config_id += "_";
-    _config_id += (enable_mixed_precision ? "mixed_precision_" : "");
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(gemm_info.k);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(2));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.m0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(rhs_info.n0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.k0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.v0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(rhs_info.h0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.interleave);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(rhs_info.interleave);
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-Status CLGEMMMatrixMultiplyReshapedKernel::validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta,
-                                                    const GEMMLHSMatrixInfo &lhs_info,
-                                                    const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
-{
-    ElementsProcessed num_elements_processed{};
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input0, input1, input2, output, alpha, beta, lhs_info, rhs_info, gemm_info));
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input0->clone().get(),
-                                                              input1->clone().get(),
-                                                              input2 != nullptr ? input2->clone().get() : nullptr,
-                                                              output->clone().get(),
-                                                              lhs_info,
-                                                              rhs_info,
-                                                              gemm_info,
-                                                              num_elements_processed)
-                                .first);
-
-    return Status{};
-}
-
-void CLGEMMMatrixMultiplyReshapedKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    if(_input1->info()->num_dimensions() < 3)
-    {
-        // The stride_z for matrix B must be zero if we do not slice
-        ARM_COMPUTE_ERROR_ON(_input1->info()->strides_in_bytes()[3] != 0);
-    }
-
-    Window slice          = window.first_slice_window_3D();
-    Window slice_matrix_b = slice;
-
-    slice_matrix_b.set(Window::DimX, Window::Dimension(0, 1, 1));
-    slice_matrix_b.set(Window::DimY, Window::Dimension(0, 1, 1));
-
-    const unsigned int total_cross_plane_pad = _output->info()->padding().top + _output->info()->padding().bottom;
-
-    cl::Image2D input1_image2d;
-
-    if(_export_to_cl_image)
-    {
-        const TensorShape shape2d(_input1->info()->dimension(0) / 4, _input1->info()->dimension(1) * _input1->info()->dimension(2));
-        const size_t      image_row_pitch = _input1->info()->strides_in_bytes()[1];
-
-        input1_image2d = create_image2d_from_buffer(CLKernelLibrary::get().context(), _input1->cl_buffer(), shape2d, _input1->info()->data_type(), image_row_pitch);
-    }
-
-    do
-    {
-        Window slice_b = slice;
-        // Don't slice matrix B along the z dimension if matrix B has just 2 dimensions and matrix A more than 2
-        // This scenario can happen when the matrix multiplication is used to perform a convolution operation
-        if(!_slide_matrix_b)
-        {
-            slice_b = slice_matrix_b;
-        }
-
-        unsigned int idx = 0;
-
-        // LHS buffer
-        add_2D_tensor_argument(idx, _input0, slice);
-
-        // RHS buffer or RHS OpenCL image (_export_to_cl_image == true)
-        if(_export_to_cl_image)
-        {
-            _kernel.setArg(idx++, input1_image2d);
-        }
-        else
-        {
-            add_2D_tensor_argument(idx, _input1, slice_b);
-        }
-
-        // Bias buffer (_add_bias == true)
-        add_2D_tensor_argument_if(_add_bias, idx, _input2, slice);
-
-        // Output buffer
-        add_2D_tensor_argument(idx, _output, slice);
-
-        // K dimension (not used if _export_to_cl_image == true)
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_k));
-
-        // LHS stride_z
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input0->info()->strides_in_bytes()[2]));
-
-        // RHS stride_z (not used if _export_to_cl_image == true)
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input1->info()->strides_in_bytes()[2]));
-
-        // Bias stride_z (if _add_bias == true)
-        if(_add_bias)
-        {
-            _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input2->info()->strides_in_bytes()[2]));
-        }
-
-        // Output stride_z
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_output->info()->strides_in_bytes()[2]));
-
-        // Cross-plan padding (if _reinterpret_output_as_3d = true)
-        if(_reinterpret_output_as_3d)
-        {
-            _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(total_cross_plane_pad));
-        }
-
-        // Dispatch kernel
-        enqueue(queue, *this, slice, lws_hint(), _use_dummy_work_items);
-    }
-    while(window.slide_window_slice_3D(slice));
-}
diff --git a/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.h b/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.h
deleted file mode 100644
index 2ffc322def..0000000000
--- a/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.h
+++ /dev/null
@@ -1,188 +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_CLGEMMMATRIXMULTIPLYRESHAPEDKERNEL_H
-#define ARM_COMPUTE_CLGEMMMATRIXMULTIPLYRESHAPEDKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-#include "arm_compute/core/KernelDescriptors.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to multiply matrices when both the input matrices LHS (input0) and RHS (input1) have been reshaped
- *
- * @note The input matrices @p input0 and @p input1 must be reshaped through @ref CLGEMMReshapeLHSMatrixKernel and  @ref CLGEMMReshapeRHSMatrixKernel
- */
-class CLGEMMMatrixMultiplyReshapedKernel : public ICLKernel
-{
-public:
-    /** Default Constructor */
-    CLGEMMMatrixMultiplyReshapedKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMMatrixMultiplyReshapedKernel(const CLGEMMMatrixMultiplyReshapedKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMMatrixMultiplyReshapedKernel &operator=(const CLGEMMMatrixMultiplyReshapedKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLGEMMMatrixMultiplyReshapedKernel(CLGEMMMatrixMultiplyReshapedKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLGEMMMatrixMultiplyReshapedKernel &operator=(CLGEMMMatrixMultiplyReshapedKernel &&) = default;
-    /** Initialise the kernel's input and output.
-     *
-     * @note The F16 computation also supports mixed precision through the gemm_info.fp_mixed_precision flag.
-     *       Mixed precision combines different floating precisions during the computation, in particular, F32 for the accumulations and F16 for the
-     *       multiplications. i.e. float c = (half)a * (half)b
-     *
-     * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will fetch the RHS data using the OpenCL read_image built-in function.
-     *       Reading from the OpenCL image object can increase the performance. However, since the OpenCL image object is created importing the OpenCL buffer,
-     *       the following conditions are required:
-     *       -# rhs_info.n0 can only be 4, 8 and 16
-     *       -# rhs_info.k0 can only be 4, 8 and 16
-     *       -# Data type can only be F32
-     *       -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension
-     *       -# The stride Y for the input1 should satisfy the OpenCL pitch alignment requirement
-     *       -# input1 width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
-     *       -# input1 (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
-     *
-     * @param[in]  input0    Input tensor containing the LHS reshaped matrix. Data type supported: F16/F32 (only F32 if rhs_info.export_to_cl_image = true). The number of dimensions for the LHS matrix must be less or equal than 4
-     * @param[in]  input1    Input tensor containing the RHS reshaped matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3
-     * @param[in]  input2    Input tensor containing the bias matrix. Data type supported: same as @p input0.
-     * @param[out] output    Output tensor to store the result of matrix multiplication. Data type supported: same as @p input0
-     * @param[in]  alpha     Weight of the matrix product
-     * @param[in]  beta      Weight of the matrix bias
-     * @param[in]  lhs_info  LHS matrix information used for reshaping the input0 tensor. Only the following values are supported:
-     *                       lhs_info.m0: 2,3,4,5,6,7,8
-     *                       lhs_info.k0: 2,3,4,8,16
-     *                       lhs_info.transpose: false
-     * @param[in]  rhs_info  RHS matrix information used for reshaping the input1 tensor.  Only the following values are supported:
-     *                       rhs_info.n0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image = true)
-     *                       rhs_info.k0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image = true)
-     *                       rhs_info.transpose: true
-     * @param[in]  gemm_info GEMM information used to retrieve the original dimensions of the input matrices
-     *
-     * @note lhs_info.k0 must be equal to rhs_info.k0
-     */
-    void configure(const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info,
-                   const GEMMRHSMatrixInfo &rhs_info,
-                   const GEMMKernelInfo    &gemm_info);
-    /** Initialise the kernel's input and output.
-     *
-     * @note The F16 computation also supports mixed precision through the gemm_info.fp_mixed_precision flag.
-     *       Mixed precision combines different floating precisions during the computation, in particular, F32 for the accumulations and F16 for the
-     *       multiplications. i.e. float c = (half)a * (half)b
-     *
-     * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will fetch the RHS data using the OpenCL read_image built-in function.
-     *       Reading from the OpenCL image object can increase the performance. However, since the OpenCL image object is created importing the OpenCL buffer,
-     *       the following conditions are required:
-     *       -# rhs_info.n0 can only be 4, 8 and 16
-     *       -# rhs_info.k0 can only be 4, 8 and 16
-     *       -# Data type can only be F32
-     *       -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension
-     *       -# The stride Y for the input1 should satisfy the OpenCL pitch alignment requirement
-     *       -# input1 width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
-     *       -# input1 (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input0          Input tensor containing the LHS reshaped matrix. Data type supported: F16/F32  (only F32 if rhs_info.export_to_cl_image = true). The number of dimensions for the LHS matrix must be less or equal than 4
-     * @param[in]  input1          Input tensor containing the RHS reshaped matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3
-     * @param[in]  input2          Input tensor containing the bias matrix. Data type supported: same as @p input0.
-     * @param[out] output          Output tensor to store the result of matrix multiplication. Data type supported: same as @p input0
-     * @param[in]  alpha           Weight of the matrix product
-     * @param[in]  beta            Weight of the matrix bias
-     * @param[in]  lhs_info        LHS matrix information used for reshaping the input0 tensor.  Only the following values are supported:
-     *                             lhs_info.m0: 2,3,4,5,6,7,8
-     *                             lhs_info.k0: 2,3,4,8,16
-     *                             lhs_info.transpose: false
-     * @param[in]  rhs_info        RHS matrix information used for reshaping the input1 tensor.  Only the following values are supported:
-     *                             rhs_info.n0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image = true)
-     *                             rhs_info.k0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image = true)
-     *                             rhs_info.transpose: true
-     * @param[in]  gemm_info       GEMM information used to retrieve the original dimensions of the input matrices
-     *
-     * @note lhs_info.k0 must be equal to rhs_info.k0
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta,
-                   const GEMMLHSMatrixInfo &lhs_info,
-                   const GEMMRHSMatrixInfo &rhs_info,
-                   const GEMMKernelInfo    &gemm_info);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMMatrixMultiplyReshapedKernel
-     *
-     * @note The F16 computation also supports mixed precision through the gemm_info.fp_mixed_precision flag.
-     *       Mixed precision combines different floating precisions during the computation, in particular, F32 for the accumulations and F16 for the
-     *       multiplications. i.e. float c = (half)a * (half)b
-     *
-     * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will fetch the RHS data using the OpenCL read_image built-in function.
-     *       Reading from the OpenCL image object can increase the performance. However, since the OpenCL image object is created importing the OpenCL buffer,
-     *       the following conditions are required:
-     *       -# rhs_info.n0 can only be 4, 8 and 16
-     *       -# rhs_info.k0 can only be 4, 8 and 16
-     *       -# Data type can only be F32
-     *       -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension
-     *       -# The stride Y for the input1 should satisfy the OpenCL pitch alignment requirement
-     *       -# input1 width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
-     *       -# input1 (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
-     *
-     * @param[in] input0    Input tensor containing the LHS reshaped matrix. Data type supported: F16/F32  (only F32 if rhs_info.export_to_cl_image = true). The number of dimensions for the LHS matrix must be less or equal than 4
-     * @param[in] input1    Input tensor containing the RHS reshaped matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3
-     * @param[in] input2    Input tensor info containing the bias matrix. Data type supported: same as @p input0.
-     * @param[in] output    Output tensor to store the result of matrix multiplication. Data type supported: same as @p input0
-     * @param[in] alpha     Weight of the matrix product
-     * @param[in] beta      Weight of the matrix bias
-     * @param[in] lhs_info  LHS matrix information used for reshaping the input0 tensor.  Only the following values are supported:
-     *                      lhs_info.m0: 2,3,4,5,6,7,8
-     *                      lhs_info.k0: 2,3,4,8,16
-     *                      lhs_info.transpose: false
-     * @param[in] rhs_info  RHS matrix information used for reshaping the input1 tensor.  Only the following values are supported:
-     *                      rhs_info.n0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image = true)
-     *                      rhs_info.k0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image = true)
-     *                      rhs_info.transpose: true
-     * @param[in] gemm_info GEMM information used to retrieve the original dimensions of the input matrices
-     *
-     * @note lhs_info.k0 must be equal to rhs_info.k0
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info,
-                           const GEMMRHSMatrixInfo &rhs_info,
-                           const GEMMKernelInfo    &gemm_info);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input0;
-    const ICLTensor *_input1;
-    const ICLTensor *_input2;
-    ICLTensor       *_output;
-    bool             _slide_matrix_b;
-    bool             _reinterpret_output_as_3d;
-    bool             _use_dummy_work_items;
-    bool             _add_bias;
-    bool             _broadcast_bias;
-    bool             _export_to_cl_image;
-    unsigned int     _k;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLGEMMMATRIXMULTIPLYRESHAPEDKERNEL_H*/
\ No newline at end of file
diff --git a/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.cpp b/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.cpp
deleted file mode 100644
index 3dee4f24cd..0000000000
--- a/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.cpp
+++ /dev/null
@@ -1,449 +0,0 @@
-/*
- * 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/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.h"
-
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/CL/CLUtils.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/CL/gemm/CLGEMMHelpers.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "src/core/utils/helpers/float_ops.h"
-#include "support/StringSupport.h"
-
-#include <tuple>
-
-using namespace arm_compute::misc::shape_calculator;
-
-namespace arm_compute
-{
-namespace
-{
-using ElementsProcessed = Steps;
-
-Status validate_arguments(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info,
-                          const GEMMRHSMatrixInfo &rhs_info,
-                          const GEMMKernelInfo    &gemm_info)
-{
-    ARM_COMPUTE_UNUSED(alpha);
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input0, input1, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input0);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::F16, DataType::F32);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input0->num_dimensions() > 4, "The number of dimensions for the LHS matrix must be <= 4");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(input1->num_dimensions() > 3, "The number of dimensions for the RHS matrix must be <= 3");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(lhs_info.m0 < 1 || lhs_info.m0 > 8, "Only 1,2,3,4,5,6,7,8 are supported for m0");
-    ARM_COMPUTE_RETURN_ERROR_ON(rhs_info.k0 > 16 || rhs_info.k0 < 2);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((rhs_info.k0 & (rhs_info.k0 - 1)) && rhs_info.k0 != 3), "Only 2,3,4,8,16 are supported for k0");
-    ARM_COMPUTE_RETURN_ERROR_ON(rhs_info.n0 > 16 || rhs_info.n0 < 2);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((rhs_info.n0 & (rhs_info.n0 - 1)) && rhs_info.n0 != 3), "Only 2,3,4,8,16 are supported for n0");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG((gemm_info.reinterpret_input_as_3d || gemm_info.depth_output_gemm3d != 0) && (input2 != nullptr)
-                                    && (!gemm_info.broadcast_bias),
-                                    "Bias addition only supported with broadcast mode in case the input or output has to be reinterpreted as 3D");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(gemm_info.fp_mixed_precision, "Mixed precision not supported");
-    ARM_COMPUTE_RETURN_ON_ERROR(cl_gemm::validate_image2d_support_on_rhs(*input1, rhs_info));
-
-    const unsigned int m = gemm_info.m;
-    const unsigned int n = gemm_info.n;
-    const unsigned int k = gemm_info.k;
-
-    TensorShape tensor_shape1{ input1->tensor_shape() };
-    tensor_shape1.set(0, n);
-    tensor_shape1.set(1, k);
-
-    if(input2 != nullptr && !(helpers::float_ops::is_zero(beta)))
-    {
-        const unsigned int input2_dim0 = input2->dimension(0);
-        const unsigned int input2_dim1 = input2->dimension(1);
-
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input2, input0);
-        if(gemm_info.broadcast_bias)
-        {
-            ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim1 != 1 || input2_dim0 != n), "Incorrect dimension of bias matrix which is to be broadcasted");
-        }
-        else
-        {
-            ARM_COMPUTE_RETURN_ERROR_ON_MSG((input2_dim0 != n || input2_dim1 != m), "Incorrect dimension of bias matrix");
-        }
-    }
-
-    const TensorInfo tensor_info1 = input1->clone()->set_tensor_shape(tensor_shape1);
-
-    const TensorInfo tensor_info_reshaped1 = input1->clone()->set_tensor_shape(compute_rhs_reshaped_shape(tensor_info1, rhs_info));
-
-    ARM_COMPUTE_RETURN_ERROR_ON(input0->dimension(0) != k);
-    if(gemm_info.reinterpret_input_as_3d)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(input0->dimension(1) * input0->dimension(2) != m);
-    }
-    else
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(input0->dimension(1) != m);
-    }
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input1, &tensor_info_reshaped1);
-
-    if(output->total_size() != 0)
-    {
-        const TensorInfo tensor_info_output = output->clone()->set_tensor_shape(compute_mm_shape(*input0, *input1, gemm_info));
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(output, &tensor_info_output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input0, output);
-    }
-
-    return Status{};
-}
-
-std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input0, ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output, const GEMMLHSMatrixInfo &lhs_info,
-                                                        const GEMMRHSMatrixInfo &rhs_info,
-                                                        const GEMMKernelInfo &gemm_info, ElementsProcessed &num_elements_processed)
-{
-    unsigned int &num_elems_processed_per_iteration_x = num_elements_processed[0];
-    unsigned int &num_elems_processed_per_iteration_y = num_elements_processed[1];
-    bool          reinterpret_input_as_3d             = gemm_info.reinterpret_input_as_3d;
-    bool          reinterpret_output_as_3d            = gemm_info.depth_output_gemm3d != 0;
-
-    Window win{};
-    Window win_out{};
-    bool   window_changed = false;
-
-    // In case both input and output have to be reinterpreted as 3D tensors,
-    // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
-    // This approach should only be used when the input/output tensors have pad on the y direction
-    if((reinterpret_input_as_3d == reinterpret_output_as_3d) && gemm_info.has_pad_y)
-    {
-        reinterpret_output_as_3d = false;
-    }
-
-    // Output tensor auto initialization if not yet initialized
-    auto_init_if_empty(*output, input0->clone()->set_tensor_shape(compute_mm_shape(*input0, *input1, gemm_info)));
-
-    TensorInfo tmp_info(*output);
-
-    if(reinterpret_output_as_3d)
-    {
-        // Since the output tensor has to be reinterpreted as 3D and the execute window is based on a 2D GEMM,
-        // the window needs to be constructed on the 2D collapsed version of the tensor
-        TensorShape tmp_shape(output->tensor_shape());
-        tmp_shape.collapse(2U, 1U);
-        tmp_info.set_tensor_shape(tmp_shape);
-    }
-
-    // Configure kernel window
-    num_elems_processed_per_iteration_x = rhs_info.n0;
-    num_elems_processed_per_iteration_y = lhs_info.m0;
-
-    win     = calculate_max_window(tmp_info, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-    win_out = calculate_max_window(*output, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-
-    if(input2 != nullptr)
-    {
-        const int bias_processed_per_iteration_x = num_elems_processed_per_iteration_x;
-
-        AccessWindowStatic input2_access(input2, 0, 0,
-                                         ceil_to_multiple(input2->dimension(0), bias_processed_per_iteration_x),
-                                         input2->dimension(1));
-
-        window_changed = update_window_and_padding(win, input2_access);
-    }
-
-    // Collapse along the Z direction
-    // This collapse needs to be here in order to tune the Z dimension of LWS
-    Window             collapsed             = win;
-    const unsigned int dimension_to_collapse = std::min(static_cast<unsigned int>(output->num_dimensions()), 2u);
-    collapsed                                = win.collapse(win, dimension_to_collapse);
-
-    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
-    return std::make_pair(err, collapsed);
-}
-} // namespace
-
-CLGEMMMatrixMultiplyReshapedOnlyRHSKernel::CLGEMMMatrixMultiplyReshapedOnlyRHSKernel()
-    : _input0(nullptr), _input1(nullptr), _input2(nullptr), _output(nullptr), _slide_matrix_b(true), _reinterpret_input_as_3d(false), _reinterpret_output_as_3d(false), _use_dummy_work_items(false),
-      _add_bias(false), _broadcast_bias(false), _export_to_cl_image(false), _has_pad_y(false)
-{
-}
-
-void CLGEMMMatrixMultiplyReshapedOnlyRHSKernel::configure(const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta,
-                                                          const GEMMLHSMatrixInfo &lhs_info,
-                                                          const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input0, input1, input2, output, alpha, beta, lhs_info, rhs_info, gemm_info);
-}
-
-void CLGEMMMatrixMultiplyReshapedOnlyRHSKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output,
-                                                          float                    alpha,
-                                                          float                    beta,
-                                                          const GEMMLHSMatrixInfo &lhs_info,
-                                                          const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input0, input1, output);
-
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input0->info(), input1->info(), (input2 != nullptr ? input2->info() : nullptr), output->info(), alpha, beta, lhs_info, rhs_info, gemm_info));
-
-    _input0                   = input0;
-    _input1                   = input1;
-    _input2                   = helpers::float_ops::is_zero(beta) ? nullptr : input2;
-    _output                   = output;
-    _reinterpret_input_as_3d  = gemm_info.reinterpret_input_as_3d;
-    _reinterpret_output_as_3d = gemm_info.depth_output_gemm3d != 0;
-    _use_dummy_work_items     = preferred_dummy_work_items_support(CLKernelLibrary::get().get_device());
-    _add_bias                 = _input2 != nullptr;
-    _broadcast_bias           = gemm_info.broadcast_bias;
-    _export_to_cl_image       = rhs_info.export_to_cl_image;
-    _has_pad_y                = gemm_info.has_pad_y;
-
-    auto padding_info = get_padding_info({ input0, input1, output });
-
-    // In case both input and output have to be reinterpreted as 3D tensors,
-    // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
-    if((_reinterpret_input_as_3d == _reinterpret_output_as_3d) && _has_pad_y)
-    {
-        _reinterpret_input_as_3d  = false;
-        _reinterpret_output_as_3d = false;
-    }
-
-    // Check if we need to slide the matrix B
-    const unsigned int num_dimensions_input0 = _input0->info()->num_dimensions();
-    _slide_matrix_b                          = (_input1->info()->num_dimensions() >= num_dimensions_input0);
-
-    ElementsProcessed num_elements_processed{};
-
-    // Configure kernel window
-    auto win_config = validate_and_configure_window(input0->info(), input1->info(), input2 != nullptr ? input2->info() : nullptr, output->info(), lhs_info, rhs_info, gemm_info, num_elements_processed);
-    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
-    ICLKernel::configure_internal(win_config.second);
-
-    // If _reinterpret_input_as_3d = _reinterpret_output_as_3d = true,
-    // we will dispatch a batched-GEMM to reduce the complexity of the address calculation within the OpenCL kernel.
-    // This means that the actual m used by the kernel is given by output->info()->dimension(1) and not by gemm_info.m
-    const unsigned int internal_m = _reinterpret_output_as_3d ? gemm_info.m : output->info()->dimension(1);
-
-    // These variables are used only if gemm_info.has_pad_y == true
-    const unsigned int h_gemm_3d = _reinterpret_output_as_3d ? output->info()->dimension(1) : input0->info()->dimension(1);
-    const unsigned int d_gemm_3d = _reinterpret_output_as_3d ? output->info()->dimension(2) : input0->info()->dimension(2);
-
-    // Shrink M0 to be always <= M (internal_m) to prevent out-of-bounds reads.
-    // NOTE: This might have implications on heuristics and performance
-    const unsigned int internal_m0 = std::min(internal_m, lhs_info.m0);
-
-    // Calculate partial (store instead of load) M0 and partial N0 for the partial blocks at the end of a row/column if any. This is to avoid padding.
-    const unsigned int partial_store_m0 = internal_m % internal_m0;
-    const unsigned int partial_store_n0 = gemm_info.n % rhs_info.n0;
-
-    // Create build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input0->info()->data_type()));
-    build_opts.add_option_if(!(helpers::float_ops::is_one(alpha)), "-DALPHA=" + float_to_string_with_full_precision(alpha));
-    build_opts.add_option_if(_input2 != nullptr, "-DBETA=" + float_to_string_with_full_precision(beta));
-    build_opts.add_option_if(helpers::float_ops::is_one(beta), "-DUNIT_BETA");
-    build_opts.add_option_if(gemm_info.broadcast_bias, "-DBROADCAST_BIAS");
-    build_opts.add_option_if(!_slide_matrix_b, "-DMATRIX_B_DEPTH=" + support::cpp11::to_string(input1->info()->dimension(2)));
-    build_opts.add_option_if(rhs_info.interleave, "-DRHS_INTERLEAVE");
-    build_opts.add_option_if(_use_dummy_work_items, "-DDUMMY_WORK_ITEMS");
-    build_opts.add_option_if(rhs_info.export_to_cl_image, "-DOPENCL_IMAGE_SUPPORT");
-    build_opts.add_option("-DRHS_HEIGHT=" + support::cpp11::to_string(input1->info()->dimension(1)));
-    build_opts.add_option("-DM=" + support::cpp11::to_string(internal_m));
-    build_opts.add_option("-DN=" + support::cpp11::to_string(gemm_info.n));
-    build_opts.add_option("-DK=" + support::cpp11::to_string(gemm_info.k));
-    build_opts.add_option("-DM0=" + support::cpp11::to_string(internal_m0));
-    build_opts.add_option("-DN0=" + support::cpp11::to_string(rhs_info.n0));
-    build_opts.add_option("-DK0=" + support::cpp11::to_string(rhs_info.k0));
-    build_opts.add_option("-DH0=" + support::cpp11::to_string(rhs_info.h0));
-    build_opts.add_option("-DPARTIAL_STORE_M0=" + support::cpp11::to_string(partial_store_m0));
-    build_opts.add_option("-DPARTIAL_STORE_N0=" + support::cpp11::to_string(partial_store_n0));
-    build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DACTIVATION_TYPE=" + lower_string(string_from_activation_func(gemm_info.activation_info.activation())));
-    build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DA_VAL=" + float_to_string_with_full_precision(gemm_info.activation_info.a()));
-    build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DB_VAL=" + float_to_string_with_full_precision(gemm_info.activation_info.b()));
-    if(_has_pad_y)
-    {
-        build_opts.add_option_if(_reinterpret_input_as_3d, "-DREINTERPRET_INPUT_AS_3D");
-        build_opts.add_option_if(_reinterpret_output_as_3d, "-DREINTERPRET_OUTPUT_AS_3D");
-        build_opts.add_option_if(_reinterpret_input_as_3d || _reinterpret_output_as_3d, "-DHEIGHT_GEMM3D=" + support::cpp11::to_string(h_gemm_3d));
-        build_opts.add_option_if(_reinterpret_input_as_3d || _reinterpret_output_as_3d, "-DDEPTH_GEMM3D=" + support::cpp11::to_string(d_gemm_3d));
-    }
-
-    std::string kernel_name("gemm_mm_reshaped_only_rhs_");
-    kernel_name += rhs_info.transpose ? "t" : "nt";
-    kernel_name += rhs_info.export_to_cl_image ? "_texture" : "";
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
-
-    // Set config_id for enabling LWS tuning
-    _config_id = kernel_name;
-    _config_id += "_";
-    _config_id += (_has_pad_y ? "" : "no_pad_y_");
-    _config_id += (_add_bias ? "add_bias_" : "");
-    _config_id += (_broadcast_bias ? "broadcast_bias_" : "");
-    _config_id += (_reinterpret_input_as_3d ? "3di_" : "");
-    _config_id += (_reinterpret_output_as_3d ? "3do_" : "");
-    _config_id += (gemm_info.activation_info.enabled() ? "fused_activation_" : "");
-    _config_id += lower_string(string_from_data_type(input0->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(gemm_info.k);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(2));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.m0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(rhs_info.n0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(rhs_info.k0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(rhs_info.h0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(rhs_info.interleave);
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-Status CLGEMMMatrixMultiplyReshapedOnlyRHSKernel::validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta,
-                                                           const GEMMLHSMatrixInfo &lhs_info,
-                                                           const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
-{
-    ElementsProcessed num_elements_processed{};
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input0, input1, input2, output, alpha, beta, lhs_info, rhs_info, gemm_info));
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input0->clone().get(),
-                                                              input1->clone().get(),
-                                                              input2 != nullptr ? input2->clone().get() : nullptr,
-                                                              output->clone().get(),
-                                                              lhs_info,
-                                                              rhs_info,
-                                                              gemm_info,
-                                                              num_elements_processed)
-                                .first);
-
-    return Status{};
-}
-
-void CLGEMMMatrixMultiplyReshapedOnlyRHSKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-    if(_input1->info()->num_dimensions() < 3)
-    {
-        // The stride_z for matrix B must be zero if we do not slice
-        ARM_COMPUTE_ERROR_ON(_input1->info()->strides_in_bytes()[3] != 0);
-    }
-
-    const size_t lhs_idx_batch_size = _reinterpret_input_as_3d && !_has_pad_y ? 3u : 2u;
-    const size_t rhs_idx_batch_size = 2u;
-    const size_t bia_idx_batch_size = 2u;
-    const size_t out_idx_batch_size = _reinterpret_output_as_3d && !_has_pad_y ? 3u : 2u;
-
-    Window slice          = window.first_slice_window_3D();
-    Window slice_matrix_b = slice;
-
-    slice_matrix_b.set(Window::DimX, Window::Dimension(0, 1, 1));
-    slice_matrix_b.set(Window::DimY, Window::Dimension(0, 1, 1));
-
-    // Get cross plane pads
-    const unsigned int total_cross_plane_pad_lhs = _input0->info()->padding().top + _input0->info()->padding().bottom;
-    const unsigned int total_cross_plane_pad_out = _output->info()->padding().top + _output->info()->padding().bottom;
-
-    // The execution should fail if we try to run with has_pad_y = false but we have padding in either the LHS or DST tensor
-    ARM_COMPUTE_ERROR_ON(!_has_pad_y && ((total_cross_plane_pad_lhs != 0) || (total_cross_plane_pad_out != 0)));
-
-    cl::Image2D input1_image2d;
-
-    if(_export_to_cl_image)
-    {
-        const TensorShape shape2d(_input1->info()->dimension(0) / 4, _input1->info()->dimension(1) * _input1->info()->dimension(2));
-        const size_t      image_row_pitch = _input1->info()->strides_in_bytes()[1];
-
-        input1_image2d = create_image2d_from_buffer(CLKernelLibrary::get().context(), _input1->cl_buffer(), shape2d, _input1->info()->data_type(), image_row_pitch);
-    }
-
-    do
-    {
-        Window slice_b = slice;
-        // Don't slice matrix B along the z dimension if matrix B has just 2 dimensions and matrix A more than 2
-        // This scenario can happen when the matrix multiplication is used to perform a convolution operation
-        if(!_slide_matrix_b)
-        {
-            slice_b = slice_matrix_b;
-        }
-
-        unsigned int idx = 0;
-
-        // LHS buffer
-        add_2D_tensor_argument(idx, _input0, slice);
-
-        // RHS buffer or RHS OpenCL image (_export_to_cl_image == true)
-        if(_export_to_cl_image)
-        {
-            _kernel.setArg(idx++, input1_image2d);
-        }
-        else
-        {
-            add_2D_tensor_argument(idx, _input1, slice_b);
-        }
-
-        // Bias buffer (_add_bias == true)
-        add_2D_tensor_argument_if(_add_bias, idx, _input2, slice);
-
-        // Output buffer
-        add_2D_tensor_argument(idx, _output, slice);
-
-        // LHS stride_z
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input0->info()->strides_in_bytes()[lhs_idx_batch_size]));
-
-        // RHS stride_z (not used if _export_to_cl_image == true)
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input1->info()->strides_in_bytes()[rhs_idx_batch_size]));
-
-        // Bias stride_z (if _add_bias == true)
-        if(_add_bias)
-        {
-            _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_input2->info()->strides_in_bytes()[bia_idx_batch_size]));
-        }
-
-        // Output stride_z
-        _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(_output->info()->strides_in_bytes()[out_idx_batch_size]));
-
-        // Cross-plan padding (if _reinterpret_input_as_3d = true)
-        if(_reinterpret_input_as_3d && _has_pad_y)
-        {
-            _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(total_cross_plane_pad_lhs));
-        }
-
-        // Cross-plan padding (if _reinterpret_output_as_3d = true)
-        if(_reinterpret_output_as_3d && _has_pad_y)
-        {
-            _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(total_cross_plane_pad_out));
-        }
-
-        enqueue(queue, *this, slice, lws_hint(), _use_dummy_work_items);
-    }
-    while(window.slide_window_slice_3D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.h b/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.h
deleted file mode 100644
index 5b96679a46..0000000000
--- a/src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.h
+++ /dev/null
@@ -1,168 +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_CLGEMMMATRIXMULTIPLYRESHAPEDONLYRHSKERNEL_H
-#define ARM_COMPUTE_CLGEMMMATRIXMULTIPLYRESHAPEDONLYRHSKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-#include "arm_compute/core/KernelDescriptors.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to multiply matrices when only the input matrix RHS (input1) has been reshaped
- *
- * @note The input matrix input1 must be reshaped through @ref CLGEMMReshapeRHSMatrixKernel
- */
-class CLGEMMMatrixMultiplyReshapedOnlyRHSKernel : public ICLKernel
-{
-public:
-    /** Default Constructor */
-    CLGEMMMatrixMultiplyReshapedOnlyRHSKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMMatrixMultiplyReshapedOnlyRHSKernel(const CLGEMMMatrixMultiplyReshapedOnlyRHSKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMMatrixMultiplyReshapedOnlyRHSKernel &operator=(const CLGEMMMatrixMultiplyReshapedOnlyRHSKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLGEMMMatrixMultiplyReshapedOnlyRHSKernel(CLGEMMMatrixMultiplyReshapedOnlyRHSKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLGEMMMatrixMultiplyReshapedOnlyRHSKernel &operator=(CLGEMMMatrixMultiplyReshapedOnlyRHSKernel &&) = default;
-    /** Initialise the kernel's input and output.
-     *
-     * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will fetch the RHS data using the OpenCL read_image built-in function.
-     *       Reading from the OpenCL image object can increase the performance. However, since the OpenCL image object is created importing the OpenCL buffer,
-     *       the following conditions are required:
-     *       -# rhs_info.n0 can only be 4, 8 and 16
-     *       -# rhs_info.k0 can only be 4, 8 and 16
-     *       -# Data type can only be F32
-     *       -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension
-     *       -# The stride Y for the input1 should satisfy the OpenCL pitch alignment requirement
-     *       -# input1 width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
-     *       -# input1 (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
-     *
-     * @param[in]  input0    Input tensor containing the LHS matrix. Data type supported: F16/F32 (only F32 if rhs_info.export_to_cl_image = true).
-     *                       The number of dimensions for the LHS matrix must be less or equal than 4.
-     * @param[in]  input1    Input tensor containing the RHS reshaped matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3.
-     * @param[in]  input2    Input tensor containing the bias matrix. Data type supported: same as @p input0.
-     * @param[out] output    Output tensor to store the result of matrix multiplication. Data type supported: same as @p input0
-     * @param[in]  alpha     Weight of the matrix product
-     * @param[in]  beta      Weight of the matrix bias
-     * @param[in]  lhs_info  LHS matrix information used to retrieve the number of rows to be processed by each thread. Only the following values are supported:
-     *                       lhs_info.m0: 1,2,3,4,5,6,7,8
-     * @param[in]  rhs_info  RHS matrix information used for reshaping the input1 tensor.  Only the following values are supported:
-     *                       rhs_info.k0: 2,3,4,8,16
-     *                       rhs_info.n0: 2,3,4,8,16
-     *                       rhs_info.transpose: true,false
-     * @param[in]  gemm_info GEMM information used to retrieve the original dimensions of the input matrices
-     */
-    void configure(const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info,
-                   const GEMMRHSMatrixInfo &rhs_info,
-                   const GEMMKernelInfo    &gemm_info);
-    /** Initialise the kernel's input and output.
-     *
-     * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will fetch the RHS data using the OpenCL read_image built-in function.
-     *       Reading from the OpenCL image object can increase the performance. However, since the OpenCL image object is created importing the OpenCL buffer,
-     *       the following conditions are required:
-     *       -# rhs_info.n0 can only be 4, 8 and 16
-     *       -# rhs_info.k0 can only be 4, 8 and 16
-     *       -# Data type can only be F32
-     *       -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension
-     *       -# The stride Y for the input1 should satisfy the OpenCL pitch alignment requirement
-     *       -# input1 width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
-     *       -# input1 (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input0          Input tensor containing the LHS matrix. Data type supported: F16/F32 (only F32 if rhs_info.export_to_cl_image = true).
-     *                             The number of dimensions for the LHS matrix must be less or equal than 4.
-     * @param[in]  input1          Input tensor containing the RHS reshaped matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3.
-     * @param[in]  input2          Input tensor containing the bias matrix. Data type supported: same as @p input0.
-     * @param[out] output          Output tensor to store the result of matrix multiplication. Data type supported: same as @p input0
-     * @param[in]  alpha           Weight of the matrix product
-     * @param[in]  beta            Weight of the matrix bias
-     * @param[in]  lhs_info        LHS matrix information used to retrieve the number of rows to be processed by each thread. Only the following values are supported:
-     *                             lhs_info.m0: 1,2,3,4,5,6,7,8
-     * @param[in]  rhs_info        RHS matrix information used for reshaping the input1 tensor.  Only the following values are supported:
-     *                             rhs_info.k0: 2,3,4,8,16
-     *                             rhs_info.n0: 2,3,4,8,16
-     *                             rhs_info.transpose: true,false
-     * @param[in]  gemm_info       GEMM information used to retrieve the original dimensions of the input matrices
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta,
-                   const GEMMLHSMatrixInfo &lhs_info,
-                   const GEMMRHSMatrixInfo &rhs_info,
-                   const GEMMKernelInfo    &gemm_info);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMMatrixMultiplyReshapedOnlyRHSKernel
-     *
-     * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will fetch the RHS data using the OpenCL read_image built-in function.
-     *       Reading from the OpenCL image object can increase the performance. However, since the OpenCL image object is created importing the OpenCL buffer,
-     *       the following conditions are required:
-     *       -# rhs_info.n0 can only be 4, 8 and 16
-     *       -# rhs_info.k0 can only be 4, 8 and 16
-     *       -# Data type can only be F32
-     *       -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension
-     *       -# The stride Y for the input1 should satisfy the OpenCL pitch alignment requirement
-     *       -# input1 width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
-     *       -# input1 (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
-     *
-     * @param[in] input0    Input tensor info for the LHS matrix. Data type supported: F16/F32 (only F32 if rhs_info.export_to_cl_image = true).
-     *                      The number of dimensions for the LHS matrix must be less or equal than 4.
-     * @param[in] input1    Input tensor info for the RHS reshaped matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3.
-     * @param[in] input2    Input tensor info containing the bias matrix. Data type supported: same as @p input0.
-     * @param[in] output    Output tensor info. Data type supported: same as @p input0
-     * @param[in] alpha     Weight of the matrix product
-     * @param[in] beta      Weight of the matrix bias
-     * @param[in] lhs_info  LHS matrix information used to retrieve the number of rows to be processed by each thread. Only the following values are supported:
-     *                      lhs_info.m0: 1,2,3,4,5,6,7,8
-     * @param[in] rhs_info  RHS matrix information used for reshaping the input1 tensor.  Only the following values are supported:
-     *                      rhs_info.k0: 2,3,4,8,16
-     *                      rhs_info.n0: 2,3,4,8,16
-     *                      rhs_info.transpose: true,false
-     * @param[in] gemm_info GEMM information used to retrieve the original dimensions of the input matrices
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info,
-                           const GEMMRHSMatrixInfo &rhs_info,
-                           const GEMMKernelInfo    &gemm_info);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input0;
-    const ICLTensor *_input1;
-    const ICLTensor *_input2;
-    ICLTensor       *_output;
-    bool             _slide_matrix_b;
-    bool             _reinterpret_input_as_3d;
-    bool             _reinterpret_output_as_3d;
-    bool             _use_dummy_work_items;
-    bool             _add_bias;
-    bool             _broadcast_bias;
-    bool             _export_to_cl_image;
-    bool             _has_pad_y;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLGEMMMATRIXMULTIPLYRESHAPEDONLYRHSKERNEL_H*/
diff --git a/src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.cpp b/src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.cpp
deleted file mode 100644
index cc95315894..0000000000
--- a/src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.cpp
+++ /dev/null
@@ -1,220 +0,0 @@
-/*
- * 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/CL/kernels/CLGEMMReshapeLHSMatrixKernel.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/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-namespace arm_compute
-{
-using namespace arm_compute::misc::shape_calculator;
-
-namespace
-{
-Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const GEMMLHSMatrixInfo &lhs_info, bool reinterpret_input_as_3d)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.m0 == 0);
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.k0 == 0);
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.v0 == 0);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((lhs_info.k0 & (lhs_info.k0 - 1)) && lhs_info.k0 != 3), "Only 2,3,4,8,16 are supported for k0");
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.k0 > 16);
-    ARM_COMPUTE_RETURN_ERROR_ON(lhs_info.m0 < 2 || lhs_info.m0 > 8);
-
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
-
-    if(output->total_size() != 0)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), compute_lhs_reshaped_shape(*input, lhs_info, reinterpret_input_as_3d));
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(input, output);
-    }
-
-    return Status{};
-}
-
-std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, const GEMMLHSMatrixInfo &lhs_info, bool reinterpret_input_as_3d)
-{
-    const unsigned int num_elems_processed_per_iteration_x = lhs_info.k0;
-    const unsigned int num_elems_processed_per_iteration_y = lhs_info.m0;
-    bool               window_changed                      = false;
-
-    TensorInfo tmp_info(*input);
-
-    if(reinterpret_input_as_3d)
-    {
-        // Since the input tensor has to be reinterpreted as 3D and the execute window is based on a 2D interleave,
-        // the window needs to be constructed on the 2D collapsed version of the tensor
-        TensorShape tmp_shape(input->tensor_shape());
-        tmp_shape.collapse(2U, 1U);
-        tmp_info.set_tensor_shape(tmp_shape);
-    }
-
-    // Output auto inizialitation if not yet initialized
-    auto_init_if_empty(*output, input->clone()->set_tensor_shape(compute_lhs_reshaped_shape(*input, lhs_info, reinterpret_input_as_3d)));
-
-    // Configure window
-    Window win    = calculate_max_window(tmp_info, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-    Window win_in = calculate_max_window(*input, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-
-    AccessWindowStatic input_access(input, 0, 0,
-                                    input->dimension(0),
-                                    input->dimension(1));
-    AccessWindowStatic output_access(output, 0, 0, output->dimension(0), output->dimension(1));
-
-    window_changed = update_window_and_padding(win_in, input_access) || // window used by the execute_window_loop
-                     update_window_and_padding(win, output_access);     // window used to update the padding requirements of output tensor
-
-    // Collapse along the Z direction
-    // This collapse needs to be here in order to tune the Z dimension of LWS
-    Window collapsed = win.collapse(win, Window::DimZ);
-
-    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
-    return std::make_pair(err, collapsed);
-}
-} // namespace
-
-CLGEMMReshapeLHSMatrixKernel::CLGEMMReshapeLHSMatrixKernel()
-    : _input(nullptr), _output(nullptr), _reinterpret_input_as_3d(false)
-{
-}
-
-void CLGEMMReshapeLHSMatrixKernel::configure(const ICLTensor *input, ICLTensor *output, const GEMMLHSMatrixInfo &lhs_info, bool reinterpret_input_as_3d)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, lhs_info, reinterpret_input_as_3d);
-}
-
-void CLGEMMReshapeLHSMatrixKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const GEMMLHSMatrixInfo &lhs_info, bool reinterpret_input_as_3d)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-
-    // Perform validate step
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), lhs_info, reinterpret_input_as_3d));
-
-    auto padding_info = get_padding_info({ input });
-
-    _input                   = input;
-    _output                  = output;
-    _reinterpret_input_as_3d = reinterpret_input_as_3d;
-
-    const unsigned int src_w           = input->info()->dimension(0);
-    const unsigned int src_h           = _reinterpret_input_as_3d ? input->info()->dimension(1) * input->info()->dimension(2) : input->info()->dimension(1);
-    const unsigned int partial_load_m0 = src_h % lhs_info.m0;
-    const unsigned int partial_load_k0 = src_w % lhs_info.k0;
-
-    // Create build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DM0=" + support::cpp11::to_string(lhs_info.m0));
-    build_opts.add_option("-DK0=" + support::cpp11::to_string(lhs_info.k0));
-    build_opts.add_option("-DV0=" + support::cpp11::to_string(lhs_info.v0));
-    build_opts.add_option("-DSRC_WIDTH=" + support::cpp11::to_string(src_w));
-    build_opts.add_option("-DSRC_HEIGHT=" + support::cpp11::to_string(src_h));
-    build_opts.add_option_if(lhs_info.interleave, "-DINTERLEAVE");
-    build_opts.add_option_if(_reinterpret_input_as_3d, "-DREINTERPRET_INPUT_AS_3D");
-    build_opts.add_option_if(_reinterpret_input_as_3d, "-DHEIGHT_GEMM3D=" + support::cpp11::to_string(input->info()->dimension(1)));
-    build_opts.add_option_if(_reinterpret_input_as_3d, "-DDEPTH_GEMM3D=" + support::cpp11::to_string(input->info()->dimension(2)));
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(input->info()->element_size()));
-    build_opts.add_option("-DPARTIAL_LOAD_M0=" + support::cpp11::to_string(partial_load_m0));
-    build_opts.add_option("-DPARTIAL_LOAD_K0=" + support::cpp11::to_string(partial_load_k0));
-
-    std::string kernel_name("gemm_reshape_lhs_matrix_");
-    kernel_name += lhs_info.transpose ? "t" : "nt";
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
-
-    // Configure kernel window
-    auto win_config = validate_and_configure_window(input->info(), output->info(), lhs_info, reinterpret_input_as_3d);
-    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
-    ICLKernel::configure_internal(win_config.second);
-
-    // Set config_id for enabling LWS tuning
-    _config_id = "gemm_reshape_lhs_matrix_";
-    _config_id += (_reinterpret_input_as_3d ? "3d_" : "");
-    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(0));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(1));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(output->info()->dimension(2));
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.m0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.k0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.v0);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.interleave);
-    _config_id += "_";
-    _config_id += support::cpp11::to_string(lhs_info.transpose);
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-Status CLGEMMReshapeLHSMatrixKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const GEMMLHSMatrixInfo &lhs_info, bool reinterpret_input_as_3d)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, lhs_info, reinterpret_input_as_3d));
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get(), lhs_info, reinterpret_input_as_3d).first);
-
-    return Status{};
-}
-
-void CLGEMMReshapeLHSMatrixKernel::run(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_3D();
-
-    if(_reinterpret_input_as_3d)
-    {
-        // Pass bottom paddings to the kernel if the input has to be reinterpreted as 3D tensor
-        const unsigned int idx0                  = 2 * num_arguments_per_3D_tensor();
-        const unsigned int total_cross_plane_pad = _input->info()->padding().top + _input->info()->padding().bottom;
-        _kernel.setArg<cl_uint>(idx0, static_cast<unsigned int>(total_cross_plane_pad));
-    }
-
-    do
-    {
-        unsigned int idx = 0;
-        add_3D_tensor_argument(idx, _input, slice);
-        add_3D_tensor_argument(idx, _output, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_3D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h b/src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h
deleted file mode 100644
index 92202a26fc..0000000000
--- a/src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h
+++ /dev/null
@@ -1,105 +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_CLGEMMRESHAPELHSMATRIXKERNEL_H
-#define ARM_COMPUTE_CLGEMMRESHAPELHSMATRIXKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to reshape the LHS matrix when performing the matrix multiplication.
- *  In particular, this function splits the input matrix in blocks of size M0xK0 (defined through GEMMLHSInfo) and
- *  stores each one in the output matrix unrolling the values
- */
-class CLGEMMReshapeLHSMatrixKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLGEMMReshapeLHSMatrixKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMReshapeLHSMatrixKernel(const CLGEMMReshapeLHSMatrixKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMReshapeLHSMatrixKernel &operator=(const CLGEMMReshapeLHSMatrixKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLGEMMReshapeLHSMatrixKernel(CLGEMMReshapeLHSMatrixKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLGEMMReshapeLHSMatrixKernel &operator=(CLGEMMReshapeLHSMatrixKernel &&) = default;
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  input                   Input tensor. Data types supported: All
-     * @param[out] output                  Output tensor. Data type supported: same as @p input
-     * @param[in]  lhs_info                LHS matrix information to be used for reshaping. This object contains all the necessary
-     *                                     information to reshape the input tensor. Only the following values are supported:
-     *                                     lhs_info.m0: 2,3,4,5,6,7,8
-     *                                     lhs_info.k0: 2,3,4,8,16
-     *                                     lhs_info.v0: greater than 0
-     *                                     lhs_info.transpose: true, false
-     *                                     lhs_info.interleave: true, false
-     * @param[in]  reinterpret_input_as_3d (Optional) True if the input has to be reinterpreted as 3D tensor
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const GEMMLHSMatrixInfo &lhs_info, bool reinterpret_input_as_3d = false);
-    /** Initialise the kernel's input and output.
-     *
-     * @param[in]  compile_context         The compile context to be used.
-     * @param[in]  input                   Input tensor. Data types supported: All
-     * @param[out] output                  Output tensor. Data type supported: same as @p input
-     * @param[in]  lhs_info                LHS matrix information to be used for reshaping. This object contains all the necessary
-     *                                     information to reshape the input tensor. Only the following values are supported:
-     *                                     lhs_info.m0: 2,3,4,5,6,7,8
-     *                                     lhs_info.k0: 2,3,4,8,16
-     *                                     lhs_info.v0: greater than 0
-     *                                     lhs_info.transpose: true, false
-     *                                     lhs_info.interleave: true, false
-     * @param[in]  reinterpret_input_as_3d (Optional) True if the input has to be reinterpreted as 3D tensor
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const GEMMLHSMatrixInfo &lhs_info, bool reinterpret_input_as_3d = false);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMReshapeLHSMatrixKernel
-     *
-     * @param[in] input                   Input tensor info. Data types supported: All
-     * @param[in] output                  Output tensor info which stores the interleaved matrix. Data type supported: same as @p input.
-     * @param[in] lhs_info                LHS matrix information to be used for reshaping. This object contains all the necessary
-     *                                    information to reshape the input tensor. Only the following values are supported:
-     *                                    lhs_info.m0: 2,3,4,5,6,7,8
-     *                                    lhs_info.k0: 2,3,4,8,16
-     *                                    lhs_info.v0: greater than 0
-     *                                    lhs_info.transpose: true, false
-     *                                    lhs_info.interleave: true, false
-     * @param[in] reinterpret_input_as_3d True if the input has to be reinterpreted as 3D tensor
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const GEMMLHSMatrixInfo &lhs_info, bool reinterpret_input_as_3d);
-
-    // Inherited methods overridden
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input;
-    ICLTensor       *_output;
-    bool             _reinterpret_input_as_3d;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLGEMMRESHAPELHSMATRIXKERNEL_H */
\ No newline at end of file
diff --git a/src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.cpp b/src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.cpp
deleted file mode 100644
index 1c4092c0e5..0000000000
--- a/src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.cpp
+++ /dev/null
@@ -1,173 +0,0 @@
-/*
- * 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/CL/kernels/CLGEMMReshapeRHSMatrixKernel.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/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/CL/CLValidate.h"
-#include "src/core/CL/gemm/CLGEMMHelpers.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-namespace arm_compute
-{
-using namespace arm_compute::misc::shape_calculator;
-
-namespace
-{
-Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const GEMMRHSMatrixInfo &rhs_info)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON(rhs_info.n0 == 0);
-    ARM_COMPUTE_RETURN_ERROR_ON(rhs_info.k0 == 0);
-    ARM_COMPUTE_RETURN_ERROR_ON(rhs_info.h0 == 0);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((rhs_info.n0 & (rhs_info.n0 - 1)) && rhs_info.n0 != 3), "Only 2,3,4,8,16 are supported for n0");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((rhs_info.k0 & (rhs_info.k0 - 1)) && (rhs_info.k0 != 1) && (rhs_info.k0 != 3)), "Only 1,2,3,4,8,16 are supported for k0");
-    ARM_COMPUTE_RETURN_ERROR_ON(rhs_info.n0 > 16);
-    ARM_COMPUTE_RETURN_ERROR_ON(rhs_info.k0 > 16);
-    ARM_COMPUTE_RETURN_ERROR_ON((rhs_info.k0 == 1) && (rhs_info.transpose));
-
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
-
-    if(rhs_info.export_to_cl_image)
-    {
-        const TensorInfo tensor_reshaped_info(compute_rhs_reshaped_shape(*input, rhs_info), 1, input->data_type());
-        ARM_COMPUTE_RETURN_ON_ERROR(cl_gemm::validate_image2d_support_on_rhs(tensor_reshaped_info, rhs_info));
-    }
-
-    if(output->total_size() != 0)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), compute_rhs_reshaped_shape(*input, rhs_info));
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(input, output);
-    }
-
-    return Status{};
-}
-
-std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, const GEMMRHSMatrixInfo &rhs_info)
-{
-    const unsigned int num_elems_processed_per_iteration_x = rhs_info.n0;
-    const unsigned int num_elems_processed_per_iteration_y = rhs_info.k0;
-    bool               window_changed                      = false;
-
-    // Output auto initialization if not yet initialized
-    auto_init_if_empty(*output, input->clone()->set_tensor_shape(compute_rhs_reshaped_shape(*input, rhs_info)));
-
-    // Configure window
-    Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
-
-    AccessWindowRectangle input_access(input, 0, 0, num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y);
-
-    window_changed = update_window_and_padding(win, input_access);
-
-    if(rhs_info.export_to_cl_image)
-    {
-        arm_compute::cl_gemm::update_padding_for_cl_image(output);
-    }
-
-    // Collapse along the Z direction
-    // This collapse needs to be here in order to tune the Z dimension of LWS
-    Window collapsed = win.collapse(win, Window::DimZ);
-
-    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
-    return std::make_pair(err, collapsed);
-}
-} // namespace
-
-CLGEMMReshapeRHSMatrixKernel::CLGEMMReshapeRHSMatrixKernel()
-    : _input(nullptr), _output(nullptr)
-{
-}
-
-void CLGEMMReshapeRHSMatrixKernel::configure(const ICLTensor *input, ICLTensor *output, const GEMMRHSMatrixInfo &rhs_info)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, output, rhs_info);
-}
-
-void CLGEMMReshapeRHSMatrixKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const GEMMRHSMatrixInfo &rhs_info)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-
-    // Perform validate step
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), rhs_info));
-
-    _input  = input;
-    _output = output;
-
-    // Create build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DN0=" + support::cpp11::to_string(rhs_info.n0));
-    build_opts.add_option("-DK0=" + support::cpp11::to_string(rhs_info.k0));
-    build_opts.add_option("-DH0=" + support::cpp11::to_string(rhs_info.h0));
-    build_opts.add_option_if(rhs_info.transpose, "-DTRANSPOSE");
-    build_opts.add_option_if(rhs_info.interleave, "-DINTERLEAVE");
-    build_opts.add_option("-DSRC_HEIGHT=" + support::cpp11::to_string(input->info()->dimension(1)));
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(input->info()->element_size()));
-
-    std::string kernel_name("gemm_reshape_rhs_matrix_");
-    kernel_name += rhs_info.transpose ? "t" : "nt";
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
-
-    // Configure kernel window
-    auto win_config = validate_and_configure_window(input->info(), output->info(), rhs_info);
-    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
-    ICLKernel::configure_internal(win_config.second);
-}
-
-Status CLGEMMReshapeRHSMatrixKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const GEMMRHSMatrixInfo &rhs_info)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, rhs_info));
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get(), rhs_info).first);
-
-    return Status{};
-}
-
-void CLGEMMReshapeRHSMatrixKernel::run(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_3D();
-
-    do
-    {
-        unsigned int idx = 0;
-        add_3D_tensor_argument(idx, _input, slice);
-        add_3D_tensor_argument(idx, _output, slice);
-        enqueue(queue, *this, slice, lws_hint());
-    }
-    while(window.slide_window_slice_3D(slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h b/src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h
deleted file mode 100644
index 911484ea76..0000000000
--- a/src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h
+++ /dev/null
@@ -1,135 +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_CLGEMMRESHAPERHSMATRIXKERNEL_H
-#define ARM_COMPUTE_CLGEMMRESHAPERHSMATRIXKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to reshape the RHS matrix when performing the matrix multiplication
- *  In particular, this kernel splits the input matrix in blocks of size K0xN0 and stores each one in
- *  the output matrix unrolling the values */
-class CLGEMMReshapeRHSMatrixKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLGEMMReshapeRHSMatrixKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMReshapeRHSMatrixKernel(const CLGEMMReshapeRHSMatrixKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLGEMMReshapeRHSMatrixKernel &operator=(const CLGEMMReshapeRHSMatrixKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLGEMMReshapeRHSMatrixKernel(CLGEMMReshapeRHSMatrixKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLGEMMReshapeRHSMatrixKernel &operator=(CLGEMMReshapeRHSMatrixKernel &&) = default;
-    /** Default destructor */
-    ~CLGEMMReshapeRHSMatrixKernel() = default;
-    /** Initialise the kernel's input and output.
-     *
-     * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will guarantee the OpenCL pitch alignment for the output tensor,
-     *       required to create a OpenCL image object from buffer in @ref CLGEMMMatrixMultiplyReshapedKernel and in @ref CLGEMMMatrixMultiplyReshapedOnlyRHSKernel
-     *       Since the OpenCL image object is created importing the OpenCL buffer, the following conditions are required:
-     *       -# rhs_info.n0 can only be 4, 8 and 16
-     *       -# rhs_info.k0 can only be 4, 8 and 16
-     *       -# Data type can only be F32, F16
-     *       -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension
-     *       -# output width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
-     *       -# output (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
-     *       -# The output tensor should be only consumed by @ref CLGEMMMatrixMultiplyReshapedKernel or @ref CLGEMMMatrixMultiplyReshapedOnlyRHSKernel
-     *
-     * @param[in]  input    Input tensor. Data types supported: All
-     * @param[out] output   Output tensor. Data type supported: same as @p input
-     * @param[in]  rhs_info RHS matrix information to be used for reshaping. This object contains all the necessary
-     *                      information to reshape the input tensor. Only the following values are supported:
-     *                      rhs_info.n0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image == true)
-     *                      rhs_info.k0: 1,2,3,4,8,16 (k0 = 1 only if rhs_info.transpose = false), (only 4, 8 and 16 if rhs_info.export_to_cl_image == true)
-     *                      rhs_info.h0: greater than 0
-     *                      rhs_info.transpose: true, false
-     *                      rhs_info.interleave: true, false
-     */
-    void configure(const ICLTensor *input, ICLTensor *output, const GEMMRHSMatrixInfo &rhs_info);
-    /** Initialise the kernel's input and output.
-     *
-     * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will guarantee the OpenCL pitch alignment for the output tensor,
-     *       required to create a OpenCL image object from buffer in @ref CLGEMMMatrixMultiplyReshapedKernel and in @ref CLGEMMMatrixMultiplyReshapedOnlyRHSKernel
-     *       Since the OpenCL image object is created importing the OpenCL buffer, the following conditions are required:
-     *       -# rhs_info.n0 can only be 4, 8 and 16
-     *       -# rhs_info.k0 can only be 4, 8 and 16
-     *       -# Data type can only be F32, F16
-     *       -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension
-     *       -# output width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
-     *       -# output (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
-     *       -# The output tensor should be only consumed by @ref CLGEMMMatrixMultiplyReshapedKernel or @ref CLGEMMMatrixMultiplyReshapedOnlyRHSKernel
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Input tensor. Data types supported: All
-     * @param[out] output          Output tensor. Data type supported: same as @p input
-     * @param[in]  rhs_info        RHS matrix information to be used for reshaping. This object contains all the necessary
-     *                             information to reshape the input tensor. Only the following values are supported:
-     *                             rhs_info.n0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image == true)
-     *                             rhs_info.k0: 1,2,3,4,8,16 (k0 = 1 only if rhs_info.transpose = false), (only 4, 8 and 16 if rhs_info.export_to_cl_image == true)
-     *                             rhs_info.h0: greater than 0
-     *                             rhs_info.transpose: true, false
-     *                             rhs_info.interleave: true, false
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const GEMMRHSMatrixInfo &rhs_info);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMReshapeRHSMatrixKernel
-     *
-     * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will guarantee the OpenCL pitch alignment for the output tensor,
-     *       required to create a OpenCL image object from buffer in @ref CLGEMMMatrixMultiplyReshapedKernel and in @ref CLGEMMMatrixMultiplyReshapedOnlyRHSKernel
-     *       Since the OpenCL image object is created importing the OpenCL buffer, the following conditions are required:
-     *       -# rhs_info.n0 can only be 4, 8 and 16
-     *       -# rhs_info.k0 can only be 4, 8 and 16
-     *       -# Data type can only be F32, F16
-     *       -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension
-     *       -# output width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
-     *       -# output (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
-     *       -# The output tensor should be only consumed by @ref CLGEMMMatrixMultiplyReshapedKernel or @ref CLGEMMMatrixMultiplyReshapedOnlyRHSKernel
-     *
-     * @param[in] input    Input tensor info. Data types supported: All
-     * @param[in] output   Output tensor info which stores the interleaved matrix. Data type supported: same as @p input.
-     * @param[in] rhs_info RHS matrix information to be used for reshaping. This object contains all the necessary
-     *                     information to reshape the input tensor. Only the following values are supported:
-     *                     rhs_info.n0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image == true)
-     *                     rhs_info.k0: 1,2,3,4,8,16 (k0 = 1 only if rhs_info.transpose = false),(only 4, 8 and 16 if rhs_info.export_to_cl_image == true)
-     *                     rhs_info.h0: greater than 0
-     *                     rhs_info.transpose: true, false
-     *                     rhs_info.interleave: true, false
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const GEMMRHSMatrixInfo &rhs_info);
-
-    // Inherited methods overridden
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input;
-    ICLTensor       *_output;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLGEMMRESHAPERHSMATRIXKERNEL_H */
\ No newline at end of file