Quote:
Originally Posted by henryzz
Probably not too hard to substitute in an alternate ecm method. It is also worth bearing in mind that the majority of time is spent on larger numbers than this thread covers.
I suspect that switching to the CADO siever is probably a better use of time though.
|
Agreed on the CADO note. However I think on most jobs, microecm is completely sufficient. Anything with mfba/r of 64 or less can use microecm as a complete replacement for mpqs. For really big jobs or jobs with three large primes on one side, then tinyecm would need to be used as well as Till said.
So I went ahead and experimented with this, and indeed it does help! microecm is almost 10x faster than the assembly mpqs in ggnfs. But, that portion of the job is pretty small compared to the whole. So overall the speed increase is not that big. Still, it is something.
edit:
more details
here