1 files changed, 126 insertions, 0 deletions
diff --git a/libs/ARMComputeEx/src/core/CL/kernels/CLPermuteExKernel.cpp b/libs/ARMComputeEx/src/core/CL/kernels/CLPermuteExKernel.cpp
new file mode 100644
index 000000000..aa094761c
--- /dev/null
+++ b/libs/ARMComputeEx/src/core/CL/kernels/CLPermuteExKernel.cpp
@@ -0,0 +1,126 @@
+/*
+ * 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/CLPermuteExKernel.h"
+
+#include "arm_compute/core/CL/CLHelpers.h"
+#include "arm_compute/core/CL/CLKernelLibraryEx.h"
+#include "arm_compute/core/CL/ICLTensor.h"
+#include "arm_compute/core/utils/misc/ShapeCalculator.h"
+
+using namespace arm_compute;
+
+namespace
+{
+TensorShape get_output_shape(const ITensorInfo *input, const PermutationVector &perm)
+{
+  TensorShape output_shape = input->tensor_shape();
+  permute(output_shape, perm);
+  return output_shape;
+}
+
+Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output,
+                          const PermutationVector &perm)
+{
+  ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(
+      input, 1, DataType::U8, DataType::S8, DataType::QASYMM8, DataType::U16, DataType::S16,
+      DataType::U32, DataType::S32, DataType::F16, DataType::F32);
+
+  const TensorShape output_shape =
+      misc::shape_calculator::compute_permutation_output_shape(*input, perm);
+
+  // Validate configured output
+  if (output->total_size() != 0)
+  {
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), output_shape);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+  }
+  return Status{};
+}
+} // namespace
+
+CLPermuteExKernel::CLPermuteExKernel() : _input(nullptr), _output(nullptr), _perm() {}
+
+void CLPermuteExKernel::configure(const ICLTensor *input, ICLTensor *output,
+                                  const PermutationVector &perm)
+{
+  ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
+  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), perm));
+
+  _input = input;
+  _output = output;
+  _perm = perm;
+
+  const TensorShape output_shape = get_output_shape(input->info(), perm);
+  // Output auto inizialitation if not yet initialized
+  auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(output_shape));
+
+  // Create kernel
+  std::set<std::string> build_opts;
+
+  build_opts.emplace("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+  build_opts.emplace("-DDEPTH_IN=" + support::cpp11::to_string(input->info()->dimension(2)));
+
+  // New positions of batch(D), height(H), width(w) and channel(C) based on permutation vector
+  build_opts.emplace("-DP1=" + support::cpp11::to_string(perm[0]));
+  build_opts.emplace("-DP2=" + support::cpp11::to_string(perm[1]));
+  build_opts.emplace("-DP3=" + support::cpp11::to_string(perm[2]));
+  build_opts.emplace("-DP4=" + support::cpp11::to_string(perm[3]));
+  _kernel = static_cast<cl::Kernel>(
+      CLKernelLibraryEx::get().create_kernel("permute_generic", build_opts));
+
+  // Configure  kernel window
+  Window win = calculate_max_window(*input->info(), Steps());
+
+  // The CLPermute doesn't need padding so update_window_and_padding() can be skipped
+  Coordinates coord;
+  coord.set_num_dimensions(output->info()->num_dimensions());
+  output->info()->set_valid_region(ValidRegion(coord, output->info()->tensor_shape()));
+
+  ICLKernel::configure_internal(win);
+}
+
+Status CLPermuteExKernel::validate(const ITensorInfo *input, const ITensorInfo *output,
+                                   const PermutationVector &perm)
+{
+  ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
+  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, perm));
+
+  return Status{};
+}
+
+void CLPermuteExKernel::run(const Window &window, cl::CommandQueue &queue)
+{
+  ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+  ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);
+
+  Window slice_in = window.first_slice_window_4D().collapse(ICLKernel::window(), 2, 4);
+
+  // Setup output slice
+  Window slice_out(slice_in);
+  slice_out.set(Window::DimX, Window::Dimension(0, 0, 0));
+  slice_out.set(Window::DimY, Window::Dimension(0, 0, 0));
+  slice_out.set(Window::DimZ, Window::Dimension(0, 0, 0));
+  slice_out.set(3, Window::Dimension(0, 0, 0));
+
+  do
+  {
+    unsigned int idx = 0;
+    add_4D_tensor_argument(idx, _input, slice_in);
+    add_4D_tensor_argument(idx, _output, slice_out);
+    enqueue(queue, *this, slice_in);
+  } while (window.slide_window_slice_4D(slice_in) && window.slide_window_slice_4D(slice_out));
+}