FFTW¶

FFTW is a C library for computing discrete Fourier transforms (DFT) in one or more dimensions, supporting real, complex, and even/odd (DCT/DST) data of arbitrary size.

Loading the module¶

module load FFTW/3.3.10-GCC-12.3.0

FFTW is also available through the MPI-aware variant FFTW.MPI and as part of Intel MKL (imkl-FFTW). To see all installed versions across clusters, check the software catalog or run module spider FFTW.

Compiling against FFTW¶

After loading the module, the headers and libraries are available through standard paths. Link with -lfftw3 (double precision), -lfftw3f (single precision), or -lfftw3_omp / -lfftw3_mpi for threaded/MPI variants:

# Serial
gcc -o fft_example fft_example.c -lfftw3 -lm

# OpenMP threaded
gcc -fopenmp -o fft_example fft_example.c -lfftw3_omp -lfftw3 -lm

# MPI (use the FFTW.MPI module)
mpicc -o fft_example fft_example.c -lfftw3_mpi -lfftw3 -lm

#include <fftw3.h>
#include <stdio.h>

int main() {
    int N = 8;
    double *in = fftw_malloc(sizeof(double) * N);
    fftw_complex *out = fftw_malloc(sizeof(fftw_complex) * (N/2 + 1));

    for (int i = 0; i < N; i++)
        in[i] = i + 1;

    fftw_plan plan = fftw_plan_dft_r2c_1d(N, in, out, FFTW_ESTIMATE);
    fftw_execute(plan);

    printf("Output:\n");
    for (int i = 0; i < N/2 + 1; i++)
        printf("  (%f, %f)\n", out[i][0], out[i][1]);

    fftw_destroy_plan(plan);
    fftw_free(in);
    fftw_free(out);
    return 0;
}

Batch job example¶

#!/bin/bash
#SBATCH --qos=regular
#SBATCH --job-name=fftw
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=1
#SBATCH --mem=4G
#SBATCH --time=00:30:00
#SBATCH --output=%x.%j.out
#SBATCH --error=%x.%j.err

module load FFTW/3.3.10-GCC-12.3.0

./fft_example

For MPI programs, adjust --ntasks and launch with srun.

See also¶