FFTW depends on memory alignment in order to take advantage of SIMD instructions. While this isn't a massive FFT, alignment is still important because when FFTW chooses the algorithm to use, it needs to consider the alignment. Ex: an algorithm with lots of SIMD instructions would be unusable on unaligned data. Since we have multiple audio channels to run FFTs on, FFTW can plan once and operate on
source not stored for this graph (policy: none)
nothing calls this directly
no outgoing calls
no test coverage detected