src/core/NEON/kernels/NEGEMMTranspose1xWKernel.cpp

/*
 * Copyright (c) 2016-2019 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/NEON/kernels/NEGEMMTranspose1xWKernel.h"

#include "arm_compute/core/AccessWindowStatic.h"
#include "arm_compute/core/Coordinates.h"
#include "arm_compute/core/Error.h"
#include "arm_compute/core/Helpers.h"
#include "arm_compute/core/ITensor.h"
#include "arm_compute/core/NEON/INEKernel.h"
#include "arm_compute/core/TensorInfo.h"
#include "arm_compute/core/TensorShape.h"
#include "arm_compute/core/Types.h"
#include "arm_compute/core/Validate.h"
#include "arm_compute/core/Window.h"

#include <arm_neon.h>
#include <cstddef>
#include <cstring>

using namespace arm_compute;

namespace
{
TensorShape get_output_shape(const ITensorInfo *input)
{
    TensorShape  output_shape{ input->tensor_shape() };
    const size_t transpose_w = 16 / input->element_size();
    output_shape.set(0, input->dimension(1) * transpose_w);
    output_shape.set(1, static_cast<size_t>(std::ceil((input->dimension(0) / static_cast<float>(transpose_w)))));
    return output_shape;
}

Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
{
    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input);
    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
    //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);

    if(output->total_size() != 0)
    {
        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), get_output_shape(input));
        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 unsigned int num_elems_processed_per_iteration = 16 / input->element_size();

    // Configure kernel window
    Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration));

    AccessWindowHorizontal input_access(input, 0, num_elems_processed_per_iteration);

    // Configure window in case of configured output
    if(output->total_size() != 0)
    {
        AccessWindowStatic output_access(output, 0, 0, output->dimension(0), output->dimension(1));
        output_access.set_valid_region(win, ValidRegion(Coordinates(), output->tensor_shape()));
    }

    const bool window_changed = update_window_and_padding(win, input_access);

    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
    return std::make_pair(err, win);
}
} // namespace

void NEGEMMTranspose1xWKernel::configure(const ITensor *input, ITensor *output)
{
    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);

    // Output tensor auto inizialitation if not yet initialized
    auto_init_if_empty(*output->info(), get_output_shape(input->info()), 1, input->info()->data_type());

    // Perform validate step
    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info()));

    _input  = input;
    _output = output;

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

Status NEGEMMTranspose1xWKernel::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 NEGEMMTranspose1xWKernel::run(const Window &window, const ThreadInfo &info)
{
    ARM_COMPUTE_UNUSED(info);
    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INESimpleKernel::window(), window);

    /*
     * Following an example of how the transposition1xW works when the input data type is F32
     *
     *         |a00 a01 a02 a03|
     *         |a10 a11 a12 a13|
     *         |a20 a21 a22 a23| = | a00 a01 a02 a03 || a10 a11 a12 a13 || a20 a21 a22 a23 || a30 a31 a32 a33 |
     *         |a30 a31 a32 a33|
     *
     * The output matrix will have the following shape: [ height * W, ceil(width / W) ], where W = (16 / element size of the tensor)
     */

    // Set window for output tensor. Set to 0 the X and Y dimensions in order to allow multi-threading implementation and future batched matrix multiplications
    Window win_out(window);
    win_out.set(Window::DimX, Window::Dimension(0, 0, 0));
    win_out.set(Window::DimY, Window::Dimension(0, 0, 0));

    Iterator in(_input, window);
    Iterator out(_output, win_out);

    switch(_input->info()->element_size())
    {
        case 1:
        {
            const size_t out_stride = _output->info()->strides_in_bytes()[1];
            execute_window_loop(window, [&](const Coordinates & id)
            {
                // Output address = base addr + (y * 16) + (x / 16 ) * stride
                const uint8_t *in_ptr  = in.ptr();
                uint8_t *const out_ptr = out.ptr() + (id.y() << 4) + (id.x() >> 4) * out_stride;
                vst1q_u8(out_ptr, vld1q_u8(in_ptr));
            },
            in, out);
            break;
        }
        case 2:
        {
            const size_t out_stride = _output->info()->strides_in_bytes()[1] / sizeof(int16_t);
            execute_window_loop(window, [&](const Coordinates & id)
            {
                // Output address = base addr + (y * 8) + (x / 8 ) * stride
                const auto in_ptr  = reinterpret_cast<const uint16_t *>(in.ptr());
                const auto out_ptr = reinterpret_cast<uint16_t *>(out.ptr()) + (id.y() << 3) + (id.x() >> 3) * out_stride;
                vst1q_u16(out_ptr, vld1q_u16(in_ptr));
            },
            in, out);
            break;
        }
        case 4:
        {
            const size_t out_stride = _output->info()->strides_in_bytes()[1] / sizeof(float);
            execute_window_loop(window, [&](const Coordinates & id)
            {
                // Output address = base addr + (y * 4) + (x / 4 ) * stride
                const auto in_ptr  = reinterpret_cast<const uint32_t *>(in.ptr());
                const auto out_ptr = reinterpret_cast<uint32_t *>(out.ptr()) + (id.y() << 2) + (id.x() >> 2) * out_stride;
                vst1q_u32(out_ptr, vld1q_u32(in_ptr));
            },
            in, out);
            break;
        }
        default:
        {
            ARM_COMPUTE_ERROR("Element size not supported");
            break;
        }
    }
}