Ah, the bulldozer problem was also simple. Basically when you deleted the mpn/x86_64/sqr_basecase.asm file it reverted to using mpn/sqr_basecase.c. This file also only works up to SQR_KARATSUBA_THRESHOLD, which happens to be 31.
So this again is expected behaviour. We'll come to the question of which files are faster for bulldozer later on. For now there is only one choice, since I already determined that the GMP sqr_basecase.asm is fastest for k10 and friends (MPIR treats bulldozer as one of them for now). Bill. On 17 February 2014 23:02, Bill Hart <goodwillh...@googlemail.com> wrote: > Ah, that makes sense! Phew! > > So the sqr_basecase code on k7 is only designed to run up to > SQR_KARATSUBA_THRESHOLD, which happens to be 50. So this is not unexpected > behaviour after all. > > Bill. > > > On 17 February 2014 22:50, Jean-Pierre Flori <jpfl...@gmail.com> wrote: > >> >> >> On Monday, February 17, 2014 10:45:01 PM UTC+1, Bill Hart wrote: >>> >>> Hmm, that one is a mystery. I compiled for atom on my machine and it >>> works fine. >>> >>> In fact, for some strange reason the version of ubuntu install is 32 bit. >> I just checked and the proc is an ATOM N450 and it should support Intel >> 64... >> So i guess I'll be reinstalling it, but that will take some time. >> >> >> > Can I just check, it segfaulted *before* you removed mpn/x86_64/atom/ >> sqr_basecase.as >> >> The assembly file used was mpn/x86/k7/sqr_basecase.asm and it segfaults >> with this file (testing right now without). >> > > -- 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.
