I have now written tuning code for the FFT. The code is in the tune directory of the fft module in the program tune_fft.c. It prints a valid fft_tuning.h to stdout.
It relies on the *nix timing functions, but presumably Windows has a similar function. The actual timer doesn't matter too much. So long as the time increases for longer times and there is about millisecond resolution or so then it should work. It does give slightly different values on each run, and if we were tuning for arches for mpir we might increase the iters by a factor of 10 in each half of the code and just wait 10 mins for the better tuning values. But for most purposes the values produced are good enough. Only squaring code to be done now. I am not going to bother with separate tuning values for this as they are unlikely to differ much. I will try to fix the squaring facility over the weekend. Unfortunately this will require modifying some of the existing functions. I'll try to make a diff of all the changes so the same changes can be made in mpir. Bill. -- You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
