The surprising thing is that almost all the speedup in our benchmark came from updating to the latest versions of mpz/gcd.c and mpz/gcdext.c from GMP-5.1.3.
Updating to their gcd/gcdext mpn code seemed to make little difference, even if I temporarily used their tuning values. This is surprising since clearly a lot of effort has gone into improving the mpn level code in GMP. I don't see a whole lot going on in the mpz code, so maybe the old mpz code had to work hard to make use of the old mpn code. Bill. On 27 February 2014 18:57, Bill Hart <goodwillh...@googlemail.com> wrote: > So gcd and xgcd has now caught up with GMP and even overtaken it (for > large sizes). > > I think it will be the jacobi code that will be affected by the crappy > mpn_mulmod_bmn1 I added. But I'll figure out how to use the fft for the > latter before the release. > > Bill. > > > On 27 February 2014 17:50, Brian Gladman <b...@gladman.plus.com> wrote: > >> On 27/02/2014 16:05, Bill Hart wrote: >> > I have updated to the latest version of the gcd and xgcd code from GMP. >> > >> > This required rather a lot of changes. >> > >> > Basically the files added/affected are: >> > >> > mpn/generic/gcd* >> > mpn/generic/matrix* >> > mpn/generic/hgcd* >> > mpn/generic/jacobi* >> > gmp-impl.h >> > >> > I also had to add a rudimentary mpn_mulmod_bnm1 function which is in >> > mpn/generic/mulmod_2expm1.c for now. >> > >> > Brian, I guess the Windows builds will now be out of date. >> >> They are now up to date in my repo for the Windows Visual Studio 2012 >> and 2013 builds. I am no longer maintaining the 2010 build but the >> Windows build generator (mpir_config.py) is still available for anyone >> who wants to use this Visual Studio version. >> >> The Windows command line build was Jason's 'baby' but Case kindly did >> some maintenance work on it some time ago. It might need quite a bit >> more work with all the recent changes though. >> >> Brian >> >> -- >> You received this message because you are subscribed to the Google Groups >> "mpir-devel" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to mpir-devel+unsubscr...@googlegroups.com. >> To post to this group, send email to mpir-devel@googlegroups.com. >> Visit this group at http://groups.google.com/group/mpir-devel. >> For more options, visit https://groups.google.com/groups/opt_out. >> > > -- You received this message because you are subscribed to the Google Groups "mpir-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to mpir-devel+unsubscr...@googlegroups.com. To post to this group, send email to mpir-devel@googlegroups.com. Visit this group at http://groups.google.com/group/mpir-devel. For more options, visit https://groups.google.com/groups/opt_out.
