diff options
Diffstat (limited to 'src/core/CL/kernels/CLFFTDigitReverseKernel.cpp')
-rw-r--r-- | src/core/CL/kernels/CLFFTDigitReverseKernel.cpp | 51 |
1 files changed, 36 insertions, 15 deletions
diff --git a/src/core/CL/kernels/CLFFTDigitReverseKernel.cpp b/src/core/CL/kernels/CLFFTDigitReverseKernel.cpp index 922e50aa73..3d8f875ef7 100644 --- a/src/core/CL/kernels/CLFFTDigitReverseKernel.cpp +++ b/src/core/CL/kernels/CLFFTDigitReverseKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2020 Arm Limited. + * Copyright (c) 2019-2021, 2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -26,6 +26,9 @@ #include "arm_compute/core/CL/CLKernelLibrary.h" #include "arm_compute/core/CL/ICLTensor.h" #include "arm_compute/core/TensorInfo.h" +#include "arm_compute/core/Utils.h" +#include "arm_compute/core/utils/StringUtils.h" + #include "src/core/CL/CLValidate.h" #include "src/core/helpers/AutoConfiguration.h" #include "src/core/helpers/WindowHelpers.h" @@ -35,17 +38,20 @@ namespace arm_compute { namespace { -Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *idx, const FFTDigitReverseKernelInfo &config) +Status validate_arguments(const ITensorInfo *input, + const ITensorInfo *output, + const ITensorInfo *idx, + const FFTDigitReverseKernelInfo &config) { ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input); - ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() != DataType::F32); + ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_NOT_IN(input, DataType::F16, DataType::F32); ARM_COMPUTE_RETURN_ERROR_ON(input->num_channels() != 1 && input->num_channels() != 2); ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(idx, 1, DataType::U32); - ARM_COMPUTE_RETURN_ERROR_ON(std::set<unsigned int>({ 0, 1 }).count(config.axis) == 0); + ARM_COMPUTE_RETURN_ERROR_ON(std::set<unsigned int>({0, 1}).count(config.axis) == 0); ARM_COMPUTE_RETURN_ERROR_ON(input->tensor_shape()[config.axis] != idx->tensor_shape().x()); // Checks performed when output is configured - if((output != nullptr) && (output->total_size() != 0)) + if ((output != nullptr) && (output->total_size() != 0)) { ARM_COMPUTE_RETURN_ERROR_ON(output->num_channels() != 2); ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output); @@ -55,32 +61,42 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, c return Status{}; } -std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, ITensorInfo *idx, const FFTDigitReverseKernelInfo &config) +std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, + ITensorInfo *output, + ITensorInfo *idx, + const FFTDigitReverseKernelInfo &config) { ARM_COMPUTE_UNUSED(idx, config); auto_init_if_empty(*output, input->clone()->set_num_channels(2)); Window win = calculate_max_window(*output, Steps()); - output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape())); return std::make_pair(Status{}, win); } } // namespace -CLFFTDigitReverseKernel::CLFFTDigitReverseKernel() - : _input(nullptr), _output(nullptr), _idx(nullptr) +CLFFTDigitReverseKernel::CLFFTDigitReverseKernel() : _input(nullptr), _output(nullptr), _idx(nullptr) { + _type = CLKernelType::ELEMENTWISE; } -void CLFFTDigitReverseKernel::configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *idx, const FFTDigitReverseKernelInfo &config) +void CLFFTDigitReverseKernel::configure(const ICLTensor *input, + ICLTensor *output, + const ICLTensor *idx, + const FFTDigitReverseKernelInfo &config) { configure(CLKernelLibrary::get().get_compile_context(), input, output, idx, config); } -void CLFFTDigitReverseKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const ICLTensor *idx, const FFTDigitReverseKernelInfo &config) +void CLFFTDigitReverseKernel::configure(const CLCompileContext &compile_context, + const ICLTensor *input, + ICLTensor *output, + const ICLTensor *idx, + const FFTDigitReverseKernelInfo &config) { ARM_COMPUTE_ERROR_ON_NULLPTR(input, output, idx); + auto padding_info = get_padding_info({input, output, idx}); ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), idx->info(), config)); _input = input; @@ -90,6 +106,7 @@ void CLFFTDigitReverseKernel::configure(const CLCompileContext &compile_context, // Create kernel CLBuildOptions build_opts; build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(input->info()->num_channels())); + build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type())); build_opts.add_option_if(config.conjugate, "-DCONJ"); std::string kernel_name = "fft_digit_reverse_axis_" + support::cpp11::to_string(config.axis); _kernel = create_kernel(compile_context, kernel_name, build_opts.options()); @@ -107,12 +124,17 @@ void CLFFTDigitReverseKernel::configure(const CLCompileContext &compile_context, _config_id += support::cpp11::to_string(input->info()->dimension(0)); _config_id += "_"; _config_id += support::cpp11::to_string(input->info()->dimension(1)); + ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info)); } -Status CLFFTDigitReverseKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *idx, const FFTDigitReverseKernelInfo &config) +Status CLFFTDigitReverseKernel::validate(const ITensorInfo *input, + const ITensorInfo *output, + const ITensorInfo *idx, + const FFTDigitReverseKernelInfo &config) { ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, idx, config)); - ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get(), idx->clone().get(), config).first); + ARM_COMPUTE_RETURN_ON_ERROR( + validate_and_configure_window(input->clone().get(), output->clone().get(), idx->clone().get(), config).first); return Status{}; } @@ -132,7 +154,6 @@ void CLFFTDigitReverseKernel::run(const Window &window, cl::CommandQueue &queue) add_3D_tensor_argument(idx, _output, slice); add_1D_tensor_argument(idx, _idx, slice); enqueue(queue, *this, slice, lws_hint()); - } - while(collapsed.slide_window_slice_3D(slice)); + } while (collapsed.slide_window_slice_3D(slice)); } } // namespace arm_compute |