Bill Hart wrote: > I stand corrected. Indeed, x86/nehalem's gmp-mparam.h is
/* Generated by tuneup.c, 2009-10-07, gcc 4.3 */ while x86/pentium4/sse2's for example is /* Generated by tuneup.c, 2014-03-24, gcc 4.7 */ (I thought you'd removed tuning values for architectures you didn't get fresh tuning values for, or replaced them by those of similar ones.) > Anyway, I'm going to set up some 32 bit VM's some time so we can at > least make sure things pass on such machines. Ideally, "unexpected" tuning values should raise an error (failed assertion) of course. > Supporting 32 bits is not such a high priority when we don't do anything > to support the latest Intel and AMD chips, but people seem to be testing > on those architectures more than they used to. At least Ubuntu (only recently?) stopped recommending to install 32-bit OSs on 64-bit hardware... ;-) -leif P.S.: With MPIR 2.7.0.alpha13 on nehalem-pc-linux-gnu ABI=32 (on Linux x86_64, Haswell) all tests pass for me with GCC 4.4.3 at least. (Haven't tried more recent versions.) > On 12 September 2015 at 18:20, leif <not.rea...@online.de > <mailto:not.rea...@online.de>> wrote: > > 'Bill Hart' via mpir-devel wrote: > > The bug appears to have been caused by no tuning values being set on 32 > > bit machines. Apparently we didn't have any users with 32 bit Linux when > > we asked for tuning values a couple of years back, so none have been set > > for x86/nehalem for example. > > I'm pretty sure you did get some, although not from a Nehalem (in 32-bit > mode). > > I personally did test all 2.7.0 alphas on 32-bit Linux, but on a Pentium > 4. (A machine I'm incidentally going to shut down again tomorrow, so > let me know soon in case there's anything I should test before doing > so...) > > > -leif > > > The bug goes away if they are set. > > > > Bill. > > > > On 12 September 2015 at 16:10, Bill Hart <goodwillh...@googlemail.com > <mailto:goodwillh...@googlemail.com> > > <mailto:goodwillh...@googlemail.com > <mailto:goodwillh...@googlemail.com>>> wrote: > > > > Hi Brian (and others), > > > > is it possible to reproduce this bug on Windows 32? > > > > I don't have access to a 32 bit machine at the moment and I'm > > working night and day on a big software release, so just can look > > into this bug for about 2 weeks. > > > > Can anyone else see what's going wrong? It should be in the > basecase > > range. > > > > http://trac.sagemath.org/ticket/18546#comment:21 > > > > I'd have thought that code was pretty bulletproof... > > > > Bill. -- 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/d/optout.