/* * Copyright (c) 2017-2020 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/NEFloorKernel.h" #include "arm_compute/core/CPP/Validate.h" #include "arm_compute/core/Coordinates.h" #include "arm_compute/core/Helpers.h" #include "arm_compute/core/IAccessWindow.h" #include "arm_compute/core/ITensor.h" #include "arm_compute/core/NEON/INEKernel.h" #include "arm_compute/core/NEON/NEMath.h" #include "arm_compute/core/Validate.h" #include namespace arm_compute { namespace { Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output) { ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output); ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input); ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32); // Validate in case of configured output if(output->total_size() > 0) { ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output); } return Status{}; } } // namespace void NEFloorKernel::configure(const ITensor *input, ITensor *output) { ARM_COMPUTE_ERROR_ON_NULLPTR(input, output); // Auto initialize output auto_init_if_empty(*output->info(), input->info()->tensor_shape(), 1, input->info()->data_type()); // Validate ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info())); _input = input; _output = output; // Configure kernel window Window win = calculate_max_window(*input->info(), Steps()); Coordinates coord; coord.set_num_dimensions(output->info()->num_dimensions()); output->info()->set_valid_region(ValidRegion(coord, output->info()->tensor_shape())); INEKernel::configure(win); } Status NEFloorKernel::validate(const ITensorInfo *input, const ITensorInfo *output) { ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output)); return Status{}; } void NEFloorKernel::run(const Window &window, const ThreadInfo &info) { ARM_COMPUTE_UNUSED(info); ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window); const DataType data_type = _input->info()->data_type(); const auto window_start_x = static_cast(window.x().start()); const auto window_end_x = static_cast(window.x().end()); const int window_step_x = 16 / _input->info()->element_size(); Window win{ window }; win.set(Window::DimX, Window::Dimension(0, 1, 1)); Iterator input(_input, win); Iterator output(_output, win); if(data_type == DataType::F32) { execute_window_loop(win, [&](const Coordinates &) { const auto input_ptr = reinterpret_cast(input.ptr()); const auto output_ptr = reinterpret_cast(output.ptr()); int x = window_start_x; for(; x <= (window_end_x - window_step_x); x += window_step_x) { const float32x4_t res = vfloorq_f32(vld1q_f32(input_ptr + x)); vst1q_f32(output_ptr + x, res); } // Compute left-over elements for(; x < window_end_x; ++x) { *(output_ptr + x) = std::floor(*(input_ptr + x)); } }, input, output); } #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC else if(data_type == DataType::F16) { execute_window_loop(win, [&](const Coordinates &) { const auto input_ptr = reinterpret_cast(input.ptr()); const auto output_ptr = reinterpret_cast(output.ptr()); int x = window_start_x; for(; x <= (window_end_x - window_step_x); x += window_step_x) { const float16x8_t res = vfloorq_f16(vld1q_f16(input_ptr + x)); vst1q_f16(output_ptr + x, res); } // Compute left-over elements for(; x < window_end_x; ++x) { *(output_ptr + x) = std::floor(*(input_ptr + x)); } }, input, output); } #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC else { ARM_COMPUTE_ERROR("Invalid data type!"); } } } // namespace arm_compute