Georgios Pinitas | 8be9148 | 2019-03-26 17:23:28 +0000 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2019 ARM Limited. |
| 3 | * |
| 4 | * SPDX-License-Identifier: MIT |
| 5 | * |
| 6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | * of this software and associated documentation files (the "Software"), to |
| 8 | * deal in the Software without restriction, including without limitation the |
| 9 | * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or |
| 10 | * sell copies of the Software, and to permit persons to whom the Software is |
| 11 | * furnished to do so, subject to the following conditions: |
| 12 | * |
| 13 | * The above copyright notice and this permission notice shall be included in all |
| 14 | * copies or substantial portions of the Software. |
| 15 | * |
| 16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 22 | * SOFTWARE. |
| 23 | */ |
| 24 | #include "arm_compute/core/CL/kernels/CLFFTScaleKernel.h" |
| 25 | |
| 26 | #include "arm_compute/core/CL/CLKernelLibrary.h" |
| 27 | #include "arm_compute/core/CL/CLValidate.h" |
| 28 | #include "arm_compute/core/CL/ICLTensor.h" |
| 29 | #include "arm_compute/core/TensorInfo.h" |
| 30 | #include "arm_compute/core/Types.h" |
| 31 | #include "arm_compute/core/Window.h" |
| 32 | |
| 33 | namespace arm_compute |
| 34 | { |
| 35 | namespace |
| 36 | { |
| 37 | Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output) |
| 38 | { |
| 39 | ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input); |
| 40 | ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 2, DataType::F32); |
| 41 | |
| 42 | // Checks performed when output is configured |
| 43 | if((output != nullptr) && (output->total_size() != 0)) |
| 44 | { |
| 45 | ARM_COMPUTE_RETURN_ERROR_ON(output->num_channels() != 1 && output->num_channels() != 2); |
| 46 | ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output); |
| 47 | ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); |
| 48 | } |
| 49 | |
| 50 | return Status{}; |
| 51 | } |
| 52 | |
| 53 | std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output) |
| 54 | { |
| 55 | // Configure kernel window |
| 56 | Window win = calculate_max_window(*input, Steps()); |
| 57 | |
| 58 | if(output != nullptr) |
| 59 | { |
| 60 | // Output auto inizialitation if not yet initialized |
| 61 | auto_init_if_empty(*output, *input->clone()); |
| 62 | |
| 63 | // CLFFTScaleKernel doesn't need padding so update_window_and_padding() can be skipped |
| 64 | Coordinates coord; |
| 65 | coord.set_num_dimensions(output->num_dimensions()); |
| 66 | output->set_valid_region(ValidRegion(coord, output->tensor_shape())); |
| 67 | } |
| 68 | |
| 69 | return std::make_pair(Status{}, win); |
| 70 | } |
| 71 | } // namespace |
| 72 | |
| 73 | CLFFTScaleKernel::CLFFTScaleKernel() |
| 74 | : _input(nullptr), _output(nullptr), _run_in_place(false) |
| 75 | { |
| 76 | } |
| 77 | |
| 78 | void CLFFTScaleKernel::configure(ICLTensor *input, ICLTensor *output, const FFTScaleKernelInfo &config) |
| 79 | { |
| 80 | ARM_COMPUTE_ERROR_ON_NULLPTR(input); |
| 81 | ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), (output != nullptr) ? output->info() : nullptr)); |
| 82 | |
| 83 | _input = input; |
| 84 | _output = output; |
| 85 | _run_in_place = (output == nullptr) || (output == input); |
| 86 | |
| 87 | // Create kernel |
| 88 | CLBuildOptions build_opts; |
| 89 | build_opts.add_option_if(_run_in_place, "-DIN_PLACE"); |
| 90 | build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(output != nullptr ? output->info()->num_channels() : input->info()->num_channels())); |
| 91 | build_opts.add_option_if(config.conjugate, "-DCONJ"); |
| 92 | std::string kernel_name = "fft_scale_conj"; |
| 93 | _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel(kernel_name, build_opts.options())); |
| 94 | |
| 95 | // Set static arguments |
| 96 | unsigned int idx = (1 + (_run_in_place ? 0 : 1)) * num_arguments_per_3D_tensor(); // Skip the input and output parameters |
| 97 | _kernel.setArg<cl_float>(idx, config.scale); |
| 98 | |
| 99 | // Configure kernel window |
| 100 | auto win_config = validate_and_configure_window(input->info(), _run_in_place ? nullptr : output->info()); |
| 101 | ARM_COMPUTE_ERROR_THROW_ON(win_config.first); |
| 102 | ICLKernel::configure_internal(win_config.second); |
| 103 | |
| 104 | // Set config_id for enabling LWS tuning |
| 105 | _config_id = kernel_name; |
| 106 | _config_id += "_"; |
| 107 | _config_id += lower_string(string_from_data_type(input->info()->data_type())); |
| 108 | _config_id += "_"; |
| 109 | _config_id += support::cpp11::to_string(input->info()->dimension(0)); |
| 110 | _config_id += "_"; |
| 111 | _config_id += support::cpp11::to_string(input->info()->dimension(1)); |
| 112 | } |
| 113 | |
| 114 | Status CLFFTScaleKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const FFTScaleKernelInfo &config) |
| 115 | { |
| 116 | ARM_COMPUTE_UNUSED(config); |
| 117 | ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output)); |
| 118 | ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get()).first); |
| 119 | |
| 120 | return Status{}; |
| 121 | } |
| 122 | |
| 123 | void CLFFTScaleKernel::run(const Window &window, cl::CommandQueue &queue) |
| 124 | { |
| 125 | ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); |
| 126 | ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window); |
| 127 | |
| 128 | Window collapsed = window.collapse_if_possible(ICLKernel::window(), Window::DimZ); |
| 129 | Window slice = collapsed.first_slice_window_3D(); |
| 130 | |
| 131 | do |
| 132 | { |
| 133 | unsigned int idx = 0; |
| 134 | add_3D_tensor_argument(idx, _input, slice); |
| 135 | if(!_run_in_place) |
| 136 | { |
| 137 | add_3D_tensor_argument(idx, _output, slice); |
| 138 | } |
| 139 | enqueue(queue, *this, slice, lws_hint()); |
| 140 | } |
| 141 | while(collapsed.slide_window_slice_3D(slice)); |
| 142 | } |
| 143 | } // namespace arm_compute |