convolution.cu

#include <iostream>
#include <vector>

#include <cuda_runtime_api.h>
#include <cufftdx.hpp>

#include "block_io.hpp"
#include "common.hpp"

template<class FFT, class IFFT>
__launch_bounds__(FFT::max_threads_per_block) __global__ void convolution_kernel(typename FFT::value_type* data) {
    using complex_type = typename FFT::value_type;
    using scalar_type  = typename complex_type::value_type;

    // Local array for thread
    complex_type thread_data[FFT::storage_size];

    // ID of FFT in CUDA block, in range [0; FFT::ffts_per_block)
    const unsigned int local_fft_id = threadIdx.y;
    // Load data from global memory to registers
    example::io<FFT>::load(data, thread_data, local_fft_id);

    // Execute FFT
    extern __shared__ complex_type shared_mem[];
    FFT().execute(thread_data, shared_mem);

    // Scale values
    scalar_type scale = 1.0 / cufftdx::size_of<FFT>::value;
    for (unsigned int i = 0; i < FFT::elements_per_thread; i++) {
        thread_data[i].x *= scale;
        thread_data[i].y *= scale;
    }

    // Execute inverse FFT
    IFFT().execute(thread_data, shared_mem);

    // Save results
    example::io<FFT>::store(thread_data, data, local_fft_id);
}

// This example demonstrates how to use cuFFTDx t operform a convolution using one-dimensional FFTs.
//
// One block is run, it calculates two 128-point convolutions by first doing forward FFT, then
// applying pointwise operation, and ending with inverse FFT.
// Data is generated on host, copied to device buffer, and then results are copied back to host.
template<unsigned int Arch>
void convolution() {
    using namespace cufftdx;

    static constexpr unsigned int ffts_per_block = 2;
    static constexpr unsigned int fft_size       = 128;
    // FFT_base defined common options for FFT and IFFT. FFT_base is not a complete FFT description.
    // In order to complete FFT description directions are specified: forward for FFT, inverse for IFFT.
    using FFT_base     = decltype(Block() + Size<fft_size>() + Type<fft_type::c2c>() + Precision<float>() +
                              ElementsPerThread<8>() + FFTsPerBlock<ffts_per_block>() + SM<Arch>());
    using FFT          = decltype(FFT_base() + Direction<fft_direction::forward>());
    using IFFT         = decltype(FFT_base() + Direction<fft_direction::inverse>());
    using complex_type = typename FFT::value_type;

    // Allocate managed memory for input/output
    complex_type* data;
    auto          size       = ffts_per_block * fft_size;
    auto          size_bytes = size * sizeof(complex_type);
    CUDA_CHECK_AND_EXIT(cudaMallocManaged(&data, size_bytes));
    for (size_t i = 0; i < size; i++) {
        data[i] = complex_type {float(i), -float(i)};
    }

    std::cout << "input [1st FFT]:\n";
    for (size_t i = 0; i < fft_size; i++) {
        std::cout << data[i].x << " " << data[i].y << std::endl;
    }

    // Increase max shared memory if needed
    CUDA_CHECK_AND_EXIT(cudaFuncSetAttribute(
        convolution_kernel<FFT, IFFT>,
        cudaFuncAttributeMaxDynamicSharedMemorySize,
        FFT::shared_memory_size));

    // Invokes convolution kernel with FFT::block_dim threads in CUDA block
    convolution_kernel<FFT, IFFT><<<1, FFT::block_dim, FFT::shared_memory_size>>>(data);
    CUDA_CHECK_AND_EXIT(cudaPeekAtLastError());
    CUDA_CHECK_AND_EXIT(cudaDeviceSynchronize());

    std::cout << "output [1st FFT]:\n";
    for (size_t i = 0; i < fft_size; i++) {
        std::cout << data[i].x << " " << data[i].y << std::endl;
    }

    CUDA_CHECK_AND_EXIT(cudaFree(data));
    std::cout << "Success" << std::endl;
}

template<unsigned int Arch>
struct convolution_functor {
    void operator()() { return convolution<Arch>(); }
};

int main(int, char**) {
    return example::sm_runner<convolution_functor>();
}