COMPMID-462: Implement TensorReshape for NEON and CL.

Change-Id: I11b39c2ceca26ade73822e29a384ef866ae05729
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/87707
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>

1 files changed, 109 insertions, 0 deletions

diff --git a/src/core/CL/kernels/CLReshapeLayerKernel.cpp b/src/core/CL/kernels/CLReshapeLayerKernel.cpp
new file mode 100644
index 0000000000..0131bd3856
--- /dev/null
+++ b/src/core/CL/kernels/CLReshapeLayerKernel.cpp
@@ -0,0 +1,109 @@
+/*
+ * Copyright (c) 2017 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 "arm_compute/core/CL/kernels/CLReshapeLayerKernel.h"
+
+#include "arm_compute/core/AccessWindowStatic.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/IAccessWindow.h"
+#include "arm_compute/core/TensorInfo.h"
+#include "arm_compute/core/Utils.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/core/Window.h"
+
+#include <string>
+
+using namespace arm_compute;
+
+CLReshapeLayerKernel::CLReshapeLayerKernel()
+    : _input(nullptr), _output(nullptr)
+{
+}
+
+void CLReshapeLayerKernel::configure(const ICLTensor *input, ICLTensor *output)
+{
+    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::U16, DataType::S16, DataType::QS16,
+                                                  DataType::U32, DataType::S32, DataType::F16, DataType::F32);
+    ARM_COMPUTE_ERROR_ON_NULLPTR(output);
+    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+    ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input, output);
+    ARM_COMPUTE_ERROR_ON(input->info()->tensor_shape().total_size() != output->info()->tensor_shape().total_size());
+
+    _input  = input;
+    _output = output;
+
+    constexpr unsigned int num_elems_processed_per_iteration = 1;
+
+    // Create kernel
+    std::set<std::string> build_opts = { "-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()) };
+    _kernel                          = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("reshape_layer", build_opts));
+
+    // Add static arguments
+    const cl_int2 input_shape =
+    {
+        {
+            static_cast<cl_int>(_input->info()->tensor_shape()[0]),
+            static_cast<cl_int>(_input->info()->tensor_shape()[1])
+        }
+    };
+    const cl_int2 output_shape =
+    {
+        {
+            static_cast<cl_int>(_output->info()->tensor_shape()[0]),
+            static_cast<cl_int>(_output->info()->tensor_shape()[1])
+        }
+    };
+    unsigned int idx = 2 * num_arguments_per_3D_tensor(); // Skip the input and output parameters
+    _kernel.setArg<cl_int2>(idx++, input_shape);
+    _kernel.setArg<cl_int2>(idx++, output_shape);
+
+    // Configure kernel window
+    Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
+
+    AccessWindowHorizontal input_access(input->info(), 0, num_elems_processed_per_iteration);
+    AccessWindowStatic     output_access(output->info(), 0, 0, output->info()->tensor_shape().x(), output->info()->tensor_shape().y());
+    update_window_and_padding(win, input_access, output_access);
+
+    output_access.set_valid_region(win, ValidRegion(Coordinates(), output->info()->tensor_shape()));
+
+    ICLKernel::configure(win);
+}
+
+void CLReshapeLayerKernel::run(const Window &window, cl::CommandQueue &queue)
+{
+    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
+
+    Window window_collapsed = window.collapse_if_possible(ICLKernel::window(), Window::DimZ);
+    Window slice            = window_collapsed.first_slice_window_3D();
+
+    // Set inputs
+    unsigned int idx = 0;
+    add_3D_tensor_argument(idx, _input, window_collapsed);
+    add_3D_tensor_argument(idx, _output, window_collapsed);
+    enqueue(queue, *this, slice);
+}