src/core/CL/kernels/CLFlattenLayerKernel.cpp

/*
 * Copyright (c) 2018 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/CLFlattenLayerKernel.h"

#include "arm_compute/core/CL/CLHelpers.h"
#include "arm_compute/core/CL/CLKernelLibrary.h"
#include "arm_compute/core/CL/CLValidate.h"
#include "arm_compute/core/CL/ICLTensor.h"
#include "arm_compute/core/CL/OpenCL.h"
#include "arm_compute/core/Error.h"
#include "arm_compute/core/Helpers.h"
#include "arm_compute/core/IAccessWindow.h"
#include "arm_compute/core/TensorInfo.h"
#include "arm_compute/core/Types.h"
#include "arm_compute/core/Validate.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "support/ToolchainSupport.h"

using namespace arm_compute::misc::shape_calculator;

namespace arm_compute
{
namespace
{
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
{
    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
    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);
    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(output);

    // Checks performed when output is configured
    if(output->total_size() != 0)
    {
        const TensorInfo tensor_info_output = input->clone()->set_tensor_shape(compute_flatten_shape(input));

        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(output, &tensor_info_output);
        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
    }

    return Status{};
}

std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output)
{
    // Output tensor auto initialization if not yet initialized
    auto_init_if_empty(*output, input->clone()->set_tensor_shape(compute_flatten_shape(input)));

    Window win = calculate_max_window(*input, Steps()); // Flatten does not need paddings

    output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape()));

    return std::make_pair(Status{}, win);
}
} // namespace

CLFlattenLayerKernel::CLFlattenLayerKernel()
    : _input(nullptr), _output(nullptr)
{
}

void CLFlattenLayerKernel::configure(const ICLTensor *input, ICLTensor *output)
{
    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info()));

    _input  = input;
    _output = output;

    CLBuildOptions build_opts;
    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
    build_opts.add_option("-DSRC_WIDTH=" + support::cpp11::to_string(input->info()->dimension(0)));
    build_opts.add_option("-DSRC_HEIGHT=" + support::cpp11::to_string(input->info()->dimension(1)));

    // Create kernel
    _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("flatten", build_opts.options()));

    // Configure kernel window
    auto win_config = validate_and_configure_window(input->info(), output->info());
    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
    ICLKernel::configure_internal(win_config.second);

    // Set config_id for enabling LWS tuning
    _config_id = "flatten";
    _config_id += "_";
    _config_id += lower_string(string_from_data_type(input->info()->data_type()));
    _config_id += "_";
    _config_id += support::cpp11::to_string(input->info()->dimension(0));
    _config_id += "_";
    _config_id += support::cpp11::to_string(input->info()->dimension(1));
    _config_id += "_";
    _config_id += support::cpp11::to_string(input->info()->dimension(2));
    _config_id += "_";
    _config_id += support::cpp11::to_string(output->info()->dimension(0));
    _config_id += "_";
    _config_id += support::cpp11::to_string(output->info()->dimension(1));
}

Status CLFlattenLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *output)
{
    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output));
    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get()).first);
    return Status{};
}

void CLFlattenLayerKernel::run(const Window &window, cl::CommandQueue &queue)
{
    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
    ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);

    Window out_window;
    out_window.use_tensor_dimensions(_output->info()->tensor_shape());

    Window out_slice = out_window.first_slice_window_1D();
    Window in_slice  = window.first_slice_window_3D();

    // Run kernel
    do
    {
        // Set arguments
        unsigned int idx = 0;
        add_3D_tensor_argument(idx, _input, in_slice);
        add_1D_tensor_argument(idx, _output, out_slice);
        enqueue(queue, *this, in_slice, lws_hint());
    }
    while(window.slide_window_slice_3D(in_slice) && out_window.slide_window_slice_1D(out_slice));
}
} // namespace arm_compute