Remove CLArithmeticSubtractionEx from ArmComputeEx (#4275)

This commit removes CLArithmeticSubtractionEx from ArmComputeEx.
  - remove CLArithmeticSubtractionEx
  - replace CLArithmeticSubtractionEx with CLArithmeticSubtraction since CLArithmeticSubtraction supports the broadcast.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

6 files changed, 0 insertions, 484 deletions

diff --git a/libs/ARMComputeEx/arm_compute/core/CL/kernels/CLArithmeticSubtractionExKernel.h b/libs/ARMComputeEx/arm_compute/core/CL/kernels/CLArithmeticSubtractionExKernel.h
deleted file mode 100644
index 9a765f310..000000000
--- a/libs/ARMComputeEx/arm_compute/core/CL/kernels/CLArithmeticSubtractionExKernel.h
+++ /dev/null
@@ -1,81 +0,0 @@
-/*
- * Copyright (c) 2018 Samsung Electronics Co., Ltd. All Rights Reserved
- * Copyright (c) 2016-2018 ARM Limited.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#ifndef __ARM_COMPUTE_CLARITHMETICSUBTRACTIONEXKERNEL_H__
-#define __ARM_COMPUTE_CLARITHMETICSUBTRACTIONEXKERNEL_H__
-
-#include "arm_compute/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the arithmetic subtraction kernel (support broadcasting)
- *
- * Arithmetic subtraction is computed by:
- * @f[ output(x,y) = input1(x,y) - input2(x,y) @f]
- */
-class CLArithmeticSubtractionExKernel : public ICLKernel
-{
-public:
-  /** Default constructor */
-  CLArithmeticSubtractionExKernel();
-  /** Prevent instances of this class from being copied (As this class contains pointers) */
-  CLArithmeticSubtractionExKernel(const CLArithmeticSubtractionExKernel &) = delete;
-  /** Prevent instances of this class from being copied (As this class contains pointers) */
-  CLArithmeticSubtractionExKernel &operator=(const CLArithmeticSubtractionExKernel &) = delete;
-  /** Allow instances of this class to be moved */
-  CLArithmeticSubtractionExKernel(CLArithmeticSubtractionExKernel &&) = default;
-  /** Allow instances of this class to be moved */
-  CLArithmeticSubtractionExKernel &operator=(CLArithmeticSubtractionExKernel &&) = default;
-  /** Default destructor */
-  ~CLArithmeticSubtractionExKernel() = default;
-
-  /** Initialise the kernel's inputs, output and convertion policy.
-   *
-   * @param[in]  input1 First tensor input. Data types supported: U8/S16/F16/F32.
-   * @param[in]  input2 Second tensor input. Data types supported: U8/S16/F16/F32.
-   * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8),
-   * S16/F16/F32.
-   * @param[in]  policy Policy to use to handle overflow.
-   */
-  void configure(const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output,
-                 ConvertPolicy policy);
-  /** Static function to check if given info will lead to a valid configuration of @ref
-   * CLArithmeticSubtractionExKernel
-   *
-   * @param[in] input1 First tensor input info. Data types supported: U8/S16/F16/F32.
-   * @param[in] input2 Second tensor input info. Data types supported: U8/S16/F16/F32.
-   * @param[in] output Output tensor info. Data types supported: U8 (Only if both inputs are U8),
-   * S16/F16/F32.
-   * @param[in] policy Policy to use to handle overflow.
-   *
-   * @return a status
-   */
-  static Status validate(const ITensorInfo *input1, const ITensorInfo *input2,
-                         const ITensorInfo *output, ConvertPolicy policy);
-
-  // Inherited methods overridden:
-  void run(const Window &window, cl::CommandQueue &queue) override;
-  BorderSize border_size() const override;
-
-private:
-  const ICLTensor *_input1; /**< Source tensor 1 */
-  const ICLTensor *_input2; /**< Source tensor 2 */
-  ICLTensor *_output;       /**< Destination tensor */
-};
-} // namespace arm_compute
-#endif /* __ARM_COMPUTE_CLARITHMETICSUBTRACTIONEXKERNEL_H__ */
diff --git a/libs/ARMComputeEx/arm_compute/runtime/CL/functions/CLArithmeticSubtractionEx.h b/libs/ARMComputeEx/arm_compute/runtime/CL/functions/CLArithmeticSubtractionEx.h
deleted file mode 100644
index 34e6c6334..000000000
--- a/libs/ARMComputeEx/arm_compute/runtime/CL/functions/CLArithmeticSubtractionEx.h
+++ /dev/null
@@ -1,62 +0,0 @@
-/*
- * Copyright (c) 2018 Samsung Electronics Co., Ltd. All Rights Reserved
- * Copyright (c) 2016-2018 ARM Limited.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#ifndef __ARM_COMPUTE_CLARITHMETICSUBTRACTIONEX_H__
-#define __ARM_COMPUTE_CLARITHMETICSUBTRACTIONEX_H__
-
-#include "arm_compute/runtime/CL/ICLSimpleFunction.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Basic function to run @ref CLArithmeticSubtractionExKernel
- *
- * @note The tensor data type for the inputs must be U8/S16/F16/F32.
- * @note The function performs an arithmetic subtraction between two tensors.
- */
-class CLArithmeticSubtractionEx : public ICLSimpleFunction
-{
-public:
-  /** Initialise the kernel's inputs, output and convertion policy.
-   *
-   * @param[in, out] input1 An input tensor. Data types supported: U8/S16/F16/F32.
-   *                        The input tensor is [in, out] because its TensorInfo might be modified
-   * inside the kernel in case of broadcasting of dimension 0.
-   * @param[in, out] input2 An input tensor. Data types supported: same as @p input1.
-   *                        The input tensor is [in, out] because its TensorInfo might be modified
-   * inside the kernel in case of broadcasting of dimension 0.
-   * @param[out]     output Output tensor. Data types supported: U8 (Only if both inputs are U8),
-   * S16/F16/F32.
-   * @param[in]      policy Policy to use to handle overflow.
-   */
-  void configure(ICLTensor *input1, ICLTensor *input2, ICLTensor *output, ConvertPolicy policy);
-  /** Static function to check if given info will lead to a valid configuration of @ref
-   * CLArithmeticSubtractionEx
-   *
-   * @param[in] input1 First tensor input info. Data types supported: U8/S16/F16/F32.
-   * @param[in] input2 Second tensor input info. Data types supported: U8/S16/F16/F32.
-   * @param[in] output Output tensor info. Data types supported: U8 (Only if both inputs are U8),
-   * S16/F16/F32.
-   * @param[in] policy Policy to use to handle overflow.
-   *
-   * @return a status
-   */
-  static Status validate(const ITensorInfo *input1, const ITensorInfo *input2,
-                         const ITensorInfo *output, ConvertPolicy policy);
-};
-}
-#endif /* __ARM_COMPUTE_CLARITHMETICSUBTRACTIONEX_H__ */
diff --git a/libs/ARMComputeEx/src/core/CL/CLKernelLibrary.cpp b/libs/ARMComputeEx/src/core/CL/CLKernelLibrary.cpp
index b67023870..4cbcb651f 100644
--- a/libs/ARMComputeEx/src/core/CL/CLKernelLibrary.cpp
+++ b/libs/ARMComputeEx/src/core/CL/CLKernelLibrary.cpp
@@ -41,7 +41,6 @@ const std::map<std::string, std::string> CLKernelLibraryEx::_kernel_program_map
     // ARMComputeEx kernels
     {"activation_layer_ex", "activation_layer_ex.cl"},
     {"arg_op", "arg_operation.cl"},
-    {"arithmetic_sub_ex", "arithmetic_op_ex.cl"},
     {"arithmetic_add_qasymm8", "arithmetic_op_quantized.cl"},
     {"batch_to_space_nd", "batch_to_space_nd.cl"},
     {"binary_logical_op", "binary_logical_op.cl"},
@@ -95,10 +94,6 @@ const std::map<std::string, std::string> CLKernelLibraryEx::_program_source_map
 #include "./cl_kernels/arg_operation.clembed"
     },
     {
-        "arithmetic_op_ex.cl",
-#include "./cl_kernels/arithmetic_op_ex.clembed"
-    },
-    {
         "batch_to_space_nd.cl",
 #include "./cl_kernels/batch_to_space_nd.clembed"
     },
diff --git a/libs/ARMComputeEx/src/core/CL/cl_kernels/arithmetic_op_ex.cl b/libs/ARMComputeEx/src/core/CL/cl_kernels/arithmetic_op_ex.cl
deleted file mode 100644
index 2ed698951..000000000
--- a/libs/ARMComputeEx/src/core/CL/cl_kernels/arithmetic_op_ex.cl
+++ /dev/null
@@ -1,74 +0,0 @@
-/*
- * Copyright (c) 2018 Samsung Electronics Co., Ltd. All Rights Reserved
- * Copyright (c) 2017 ARM Limited.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "helpers.h"
-
-#ifdef SATURATE
-#define SUB(x, y) sub_sat((x), (y))
-#else /* SATURATE */
-#define SUB(x, y) (x) - (y)
-#endif /* SATURATE */
-
-/** This function subtracts one tensors from another.
- *
- * @attention The input and output data_types need to be passed at compile time using -DDATA_TYPE_IN1, -DDATA_TYPE_IN2 and -DDATA_TYPE_OUT:
- * e.g. -DDATA_TYPE_IN1=uchar -DDATA_TYPE_IN2=uchar -DDATA_TYPE_OUT=short
- * @attention To perform saturating operation -DSATURATE has to be passed to the compiler otherwise wrapping policy will be used.
- *
- * @param[in]  in1_ptr                           Pointer to the source tensor. Supported data types: U8, S16
- * @param[in]  in1_stride_x                      Stride of the source tensor in X dimension (in bytes)
- * @param[in]  in1_step_x                        in1_stride_x * number of elements along X processed per workitem(in bytes)
- * @param[in]  in1_stride_y                      Stride of the source tensor in Y dimension (in bytes)
- * @param[in]  in1_step_y                        in1_stride_y * number of elements along Y processed per workitem(in bytes)
- * @param[in]  in1_stride_z                      Stride of the source tensor in Z dimension (in bytes)
- * @param[in]  in1_step_z                        in1_stride_z * number of elements along Z processed per workitem(in bytes)
- * @param[in]  in1_offset_first_element_in_bytes The offset of the first element in the source tensor
- * @param[in]  in2_ptr                           Pointer to the source tensor. Supported data types: U8, S16
- * @param[in]  in2_stride_x                      Stride of the source tensor in X dimension (in bytes)
- * @param[in]  in2_step_x                        in2_stride_x * number of elements along X processed per workitem(in bytes)
- * @param[in]  in2_stride_y                      Stride of the source tensor in Y dimension (in bytes)
- * @param[in]  in2_step_y                        in2_stride_y * number of elements along Y processed per workitem(in bytes)
- * @param[in]  in2_stride_z                      Stride of the source tensor in Z dimension (in bytes)
- * @param[in]  in2_step_z                        in2_stride_z * number of elements along Z processed per workitem(in bytes)
- * @param[in]  in2_offset_first_element_in_bytes The offset of the first element in the source tensor
- * @param[out] out_ptr                           Pointer to the destination tensor. Supported data types: U8, S16
- * @param[in]  out_stride_x                      Stride of the destination tensor in X dimension (in bytes)
- * @param[in]  out_step_x                        out_stride_x * number of elements along X processed per workitem(in bytes)
- * @param[in]  out_stride_y                      Stride of the destination tensor in Y dimension (in bytes)
- * @param[in]  out_step_y                        out_stride_y * number of elements along Y processed per workitem(in bytes)
- * @param[in]  out_stride_z                      Stride of the source tensor in Z dimension (in bytes)
- * @param[in]  out_step_z                        out_stride_z * number of elements along Z processed per workitem(in bytes)
- * @param[in]  out_offset_first_element_in_bytes The offset of the first element in the destination tensor
- */
-__kernel void arithmetic_sub_ex(
-    TENSOR3D_DECLARATION(in1),
-    TENSOR3D_DECLARATION(in2),
-    TENSOR3D_DECLARATION(out))
-{
-    // Get pixels pointer
-    Tensor3D in1 = CONVERT_TO_TENSOR3D_STRUCT(in1);
-    Tensor3D in2 = CONVERT_TO_TENSOR3D_STRUCT(in2);
-    Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(out);
-
-    // Load values
-    VEC_DATA_TYPE(DATA_TYPE_OUT, 16)
-    in_a = CONVERT(vload16(0, (__global DATA_TYPE_IN1 *)in1.ptr), VEC_DATA_TYPE(DATA_TYPE_OUT, 16));
-    VEC_DATA_TYPE(DATA_TYPE_OUT, 16)
-    in_b = CONVERT(vload16(0, (__global DATA_TYPE_IN2 *)in2.ptr), VEC_DATA_TYPE(DATA_TYPE_OUT, 16));
-
-    // Calculate and store result
-    vstore16(SUB(in_a, in_b), 0, (__global DATA_TYPE_OUT *)out.ptr);
-}
diff --git a/libs/ARMComputeEx/src/core/CL/kernels/CLArithmeticSubtractionExKernel.cpp b/libs/ARMComputeEx/src/core/CL/kernels/CLArithmeticSubtractionExKernel.cpp
deleted file mode 100644
index 1c505b4d5..000000000
--- a/libs/ARMComputeEx/src/core/CL/kernels/CLArithmeticSubtractionExKernel.cpp
+++ /dev/null
@@ -1,216 +0,0 @@
-/*
- * Copyright (c) 2018 Samsung Electronics Co., Ltd. All Rights Reserved
- * Copyright (c) 2016-2018 ARM Limited.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "arm_compute/core/CL/kernels/CLArithmeticSubtractionExKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibraryEx.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-
-using namespace arm_compute;
-
-namespace
-{
-constexpr unsigned int num_elems_processed_per_iteration = 16;
-
-Status validate_arguments(const ITensorInfo *input1, const ITensorInfo *input2,
-                          const ITensorInfo *output, ConvertPolicy policy)
-{
-  ARM_COMPUTE_UNUSED(policy);
-
-  ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input1, 1, DataType::U8, DataType::S16,
-                                                       DataType::F16, DataType::F32);
-  ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input2, 1, DataType::U8, DataType::S16,
-                                                       DataType::F16, DataType::F32);
-
-  const TensorShape &out_shape =
-      TensorShape::broadcast_shape(input1->tensor_shape(), input2->tensor_shape());
-
-  ARM_COMPUTE_RETURN_ERROR_ON_MSG(out_shape.total_size() == 0,
-                                  "Inputs are not broadcast compatible");
-
-  // Validate in case of configured output
-  if (output->total_size() > 0)
-  {
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8, DataType::S16,
-                                                         DataType::F16, DataType::F32);
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(
-        output->data_type() == DataType::U8 &&
-            (input1->data_type() != DataType::U8 || input2->data_type() != DataType::U8),
-        "Output can only be U8 if both inputs are U8");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(
-        detail::have_different_dimensions(out_shape, output->tensor_shape(), 0),
-        "Wrong shape for output");
-  }
-
-  return Status{};
-}
-
-std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input1, ITensorInfo *input2,
-                                                        ITensorInfo *output)
-{
-  const std::pair<TensorShape, ValidRegion> broadcast_pair =
-      ITensorInfo::broadcast_shape_and_valid_region(*input1, *input2);
-  const TensorShape &out_shape = broadcast_pair.first;
-  const ValidRegion &valid_region = broadcast_pair.second;
-
-  // Auto initialize output if not initialized
-  {
-    set_shape_if_empty(*output, out_shape);
-
-    if (input1->data_type() == DataType::S16 || input2->data_type() == DataType::S16)
-    {
-      set_format_if_unknown(*output, Format::S16);
-    }
-    else if (input1->data_type() == DataType::F16 && input2->data_type() == DataType::F16)
-    {
-      set_format_if_unknown(*output, Format::F16);
-    }
-    else if (input1->data_type() == DataType::F32 || input2->data_type() == DataType::F32)
-    {
-      set_format_if_unknown(*output, Format::F32);
-    }
-  }
-
-  Window win = calculate_max_window(valid_region, Steps(num_elems_processed_per_iteration));
-  Window win_input1 = win.broadcast_if_dimension_le_one(*input1);
-  Window win_input2 = win.broadcast_if_dimension_le_one(*input2);
-
-  AccessWindowHorizontal input1_access(input1, 0, num_elems_processed_per_iteration);
-  AccessWindowHorizontal input2_access(input2, 0, num_elems_processed_per_iteration);
-  AccessWindowHorizontal output_access(output, 0, num_elems_processed_per_iteration);
-
-  bool window_changed = update_window_and_padding(win_input1, input1_access) ||
-                        update_window_and_padding(win_input2, input2_access) ||
-                        update_window_and_padding(win, output_access);
-
-  output_access.set_valid_region(win, valid_region);
-
-  Status err = (window_changed)
-                   ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!")
-                   : Status{};
-  return std::make_pair(err, win);
-}
-} // namespace
-
-CLArithmeticSubtractionExKernel::CLArithmeticSubtractionExKernel()
-    : _input1(nullptr), _input2(nullptr), _output(nullptr)
-{
-}
-
-void CLArithmeticSubtractionExKernel::configure(const ICLTensor *input1, const ICLTensor *input2,
-                                                ICLTensor *output, ConvertPolicy policy)
-{
-  ARM_COMPUTE_ERROR_ON_NULLPTR(input1, input2, output);
-  ARM_COMPUTE_ERROR_THROW_ON(
-      validate_arguments(input1->info(), input2->info(), output->info(), policy));
-
-  // Configure kernel window
-  auto win_config = validate_and_configure_window(input1->info(), input2->info(), output->info());
-  ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
-
-  _input1 = input1;
-  _input2 = input2;
-  _output = output;
-
-  const bool has_float_out = is_data_type_float(output->info()->data_type());
-
-  // Set kernel build options
-  std::set<std::string> build_opts;
-  build_opts.emplace((policy == ConvertPolicy::WRAP || has_float_out) ? "-DWRAP" : "-DSATURATE");
-  build_opts.emplace("-DDATA_TYPE_IN1=" + get_cl_type_from_data_type(input1->info()->data_type()));
-  build_opts.emplace("-DDATA_TYPE_IN2=" + get_cl_type_from_data_type(input2->info()->data_type()));
-  build_opts.emplace("-DDATA_TYPE_OUT=" + get_cl_type_from_data_type(output->info()->data_type()));
-
-  // Create kernel
-  _kernel = static_cast<cl::Kernel>(
-      CLKernelLibraryEx::get().create_kernel("arithmetic_sub_ex", build_opts));
-
-  ICLKernel::configure_internal(win_config.second);
-}
-
-Status CLArithmeticSubtractionExKernel::validate(const ITensorInfo *input1,
-                                                 const ITensorInfo *input2,
-                                                 const ITensorInfo *output, ConvertPolicy policy)
-{
-  ARM_COMPUTE_ERROR_ON_NULLPTR(input1, input2, output);
-  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input1, input2, output, policy));
-  ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input1->clone().get(),
-                                                            input2->clone().get(),
-                                                            output->clone().get())
-                                  .first);
-
-  return Status{};
-}
-
-void CLArithmeticSubtractionExKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-  ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-  ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
-
-  const TensorShape &in_shape1 = _input1->info()->tensor_shape();
-  const TensorShape &in_shape2 = _input2->info()->tensor_shape();
-  const TensorShape &out_shape = _output->info()->tensor_shape();
-
-  bool can_collapse = true;
-  if (std::min(in_shape1.total_size(), in_shape2.total_size()) > 1)
-  {
-    can_collapse =
-        (std::min(in_shape1.num_dimensions(), in_shape2.num_dimensions()) > Window::DimZ);
-    for (size_t d = Window::DimZ; can_collapse && (d < out_shape.num_dimensions()); d++)
-    {
-      can_collapse = (in_shape1[d] == in_shape2[d]);
-    }
-  }
-
-  bool has_collapsed = false;
-  Window collapsed =
-      can_collapse ? window.collapse_if_possible(ICLKernel::window(), Window::DimZ, &has_collapsed)
-                   : window;
-
-  const TensorShape &in_shape1_collapsed =
-      has_collapsed ? in_shape1.collapsed_from(Window::DimZ) : in_shape1;
-  const TensorShape &in_shape2_collapsed =
-      has_collapsed ? in_shape2.collapsed_from(Window::DimZ) : in_shape2;
-
-  Window slice = collapsed.first_slice_window_3D();
-  Window slice_input1 = slice.broadcast_if_dimension_le_one(in_shape1_collapsed);
-  Window slice_input2 = slice.broadcast_if_dimension_le_one(in_shape2_collapsed);
-
-  do
-  {
-    unsigned int idx = 0;
-
-    add_3D_tensor_argument(idx, _input1, slice_input1);
-    add_3D_tensor_argument(idx, _input2, slice_input2);
-    add_3D_tensor_argument(idx, _output, slice);
-
-    enqueue(queue, *this, slice);
-
-    collapsed.slide_window_slice_3D(slice_input1);
-    collapsed.slide_window_slice_3D(slice_input2);
-  } while (collapsed.slide_window_slice_3D(slice));
-}
-
-BorderSize CLArithmeticSubtractionExKernel::border_size() const
-{
-  const unsigned int replicateSize =
-      _output->info()->dimension(0) -
-      std::min(_input1->info()->dimension(0), _input2->info()->dimension(0));
-  const unsigned int border =
-      std::min<unsigned int>(num_elems_processed_per_iteration - 1U, replicateSize);
-  return BorderSize(0, border, 0, 0);
-}
diff --git a/libs/ARMComputeEx/src/runtime/CL/functions/CLArithmeticSubtractionEx.cpp b/libs/ARMComputeEx/src/runtime/CL/functions/CLArithmeticSubtractionEx.cpp
deleted file mode 100644
index 3f403c80a..000000000
--- a/libs/ARMComputeEx/src/runtime/CL/functions/CLArithmeticSubtractionEx.cpp
+++ /dev/null
@@ -1,46 +0,0 @@
-/*
- * Copyright (c) 2018 Samsung Electronics Co., Ltd. All Rights Reserved
- * Copyright (c) 2016-2018 ARM Limited.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "arm_compute/runtime/CL/functions/CLArithmeticSubtractionEx.h"
-
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/kernels/CLArithmeticSubtractionExKernel.h"
-
-using namespace arm_compute;
-
-void CLArithmeticSubtractionEx::configure(ICLTensor *input1, ICLTensor *input2, ICLTensor *output,
-                                          ConvertPolicy policy)
-{
-  auto k = arm_compute::support::cpp14::make_unique<CLArithmeticSubtractionExKernel>();
-  k->configure(input1, input2, output, policy);
-  _kernel = std::move(k);
-
-  if (output->info()->dimension(0) > 1)
-  {
-    ICLTensor *broadcasted_info = (input1->info()->dimension(0) == 1) ? input1 : input2;
-
-    if (broadcasted_info->info()->dimension(0) == 1)
-    {
-      _border_handler.configure(broadcasted_info, _kernel->border_size(), BorderMode::REPLICATE);
-    }
-  }
-}
-
-Status CLArithmeticSubtractionEx::validate(const ITensorInfo *input1, const ITensorInfo *input2,
-                                           const ITensorInfo *output, ConvertPolicy policy)
-{
-  return CLArithmeticSubtractionExKernel::validate(input1, input2, output, policy);
-}