I notice that the third argument when calling the function above is nptperstep. Within fft.f90, however,
https://github.com/seismology/mc_kernel/blob/master/src/fft.f90#L175-L206
suggests this number should be set to ntraces i.e.
"
!< This routines initializes a FFTw plan for 1D DFTs.
!! The time series is supposed to have length ntimes_in, and is stored along the first
!! dimension of a three-dimensional array. The other two dimensions are ndim and ntraces.
!! Internally, this array will always be converted to two-dimensional array of dimensions
!! [ntimes_in, ndim*ntraces]'."
This would make the parameter, ntraces_fft = [ndumps, sem_data%get_ndim()*nptperstep].
https://github.com/seismology/mc_kernel/blob/2dcbf021a25aa5ca4a040ea152ea98411f24b7d2/src/slave.f90#L143
I notice that the third argument when calling the function above is nptperstep. Within fft.f90, however,
https://github.com/seismology/mc_kernel/blob/master/src/fft.f90#L175-L206 suggests this number should be set to ntraces i.e. " !< This routines initializes a FFTw plan for 1D DFTs. !! The time series is supposed to have length ntimes_in, and is stored along the first !! dimension of a three-dimensional array. The other two dimensions are ndim and ntraces. !! Internally, this array will always be converted to two-dimensional array of dimensions !! [ntimes_in, ndim*ntraces]'."
This would make the parameter, ntraces_fft = [ndumps, sem_data%get_ndim()*nptperstep].
Is this okay?