NOTE: FFTW MUST be installed on the build machine and fftw.h must be visible for this to work.
NOTE to Maintainer (me): Update the version number in DESCRIPTION and configure.ac prior to uploading to CRAN. This must be done in two places.
I appreciate all contributions, pull requests, and help. I would like to keep this version ready to upload to CRAN. That means if a pull request does not pass R CMD check --as-cran with the latest version of R, I will only spend a short bit of time trying to debug it before removing it.
Please note that if the --as-cran check generates a WARNING or a NOTE then it is not a clean pass, and it is not ready to upload to CRAN. If I cannot quickly debug the Warning or Note, then I will remove the change.
In general, I do not keep a seperate development version and changes are uploaded to CRAN once they are integrated.
R package wrapping fftw. It has 1d, 2d, 3d, multivariate fftw, and other tools.
Alternative manual build instructions for general Linux machine:
- download and unzip to a folder called fftwtools-master
- from the parent folder:
- R CMD build fftwtools-master/
- R CMD check fftwtools_*.tar.gz (optional)
- R CMD INSTALL fftwtools_*.tar.gz
Note: The version number will change. Compiling this from source will require a C compiler [I used gcc], math library, and fftw development libraries installed.
See the folder speedTrials for an example of testing for which data length fftw is generally faster. On my laptop I found fftw faster at approximately >= 2^17 data points, and mvfftw faster at approximately >= 2^16 data points using five columns of data.
You can use fftw directly if your OS does not provide it as a package. http://www.fftw.org/
Note: Windows users please see: https://cran.r-project.org/bin/windows/Rtools/
Mac users please see: https://mac.r-project.org/tools/
