Joaquim

@SRHudson wrote: > I believe that the l2stell input file is too tiny to really see the impact of the speed improvements that I > intend to pursue, and whether...

More over, along these runs I did profiling, and found that the ma00aa routine increases percentage of total cpu time as Fourier resolution increases. This means that code speedup relies...

@SRHudson not sure I understand your Comment One, > the only speed improvements that I made to ma00aa (beyond the re-looping and eliminating divisions > as suggested by Sam) was...

@SRHudson OK I understand, although somehow I do not see any red terms when I open > http://w3.pppl.gov/~shudson/Spec/ma00aa.pdf

No the times do not agree. This is how I do it, you can try: 1) compile the code with the profiling option, e.g. > make CC=intel_prof 2) run the...

OK I tried the Ltiming option and also got that mp00ac takes most of the time (by far: 10 times more than any other subroutine). However, it is weird because...

If I understand correctly, there is an ongoing "speedup effort" which translates into modifications of the ma00aa branch (which should change to a more generic name since it seems that...

The ma00aa branch has been merged into master and deleted. I would wait until the NAG_REPLACE branch reaches a point in which the code runs well (from the discussions, it...

I added Erol Balkovic to the reviewing, since he is currently testing free-boundary test cases. I pass the review torch to him and Chris, and then I am guessing it...