|
From: <sl...@bl...> - 2004-09-05 20:10:30
|
Nuno Lucas <lu...@nl...> writes: > Joe Wells, dando pulos de alegria, escreveu : > > sl...@bl... (Joe Wells (reverse mailbox letters to reply)) writes: > >>"gboutwel" <gbo...@pr...> writes: > >>>The only significant change > >>>between 0.6.1 and 0719 in regards to FP operations was that flop20 > >>>didn't pass and now does. > >> > >>Where do I get flops20 so I can try it on my machine? Is it available > >>as a Gentoo ebuild? "emerge search flops" and "emerge search float" > >>do not reveal it. > > Does anyone know where I get this floating point test suite? It has > > been referred to as "fops20" and "flop20" in e-mail on this mailing > > list. > > Dan Aloni gave me a link to it but I don't remember anymore. You > should find it easily with google, but you can also get it at > http://nlucas.homeip.net/colinux/test/flops20.zip (it is my home pc, > so it is possible the net be down or something). Thanks very much for the pointer to flops20! > After some googling I found this last year forums about the pentium 4 > CFLAGS problems with floating point: > > http://www.gentoo.org/news/en/gwn/20030401-newsletter.xml#doc_chap4 > > It seems one way of testing if you have a gentoo configuration that > activates this gcc bugs is executing: > > # python -c 'int(10.1); int(10000.3); int(1.2)' This runs fine for me. > If it doesn't show nothing (no error) it is because your configuration > is ok (or you have a mixed compiled environment, off course). After looking at this earlier discussion, it is clear to me that it is unrelated to the problem I am reporting. The key differences are: 1. The earlier discussion is from 2003-04 and earlier (more than 16 months ago) and refers to GCC version 3.2. In many places, it states that the bugs are going to be fixed in GCC 3.3, and I am now using GCC 3.3.4. 2. The earlier discussion does not observe any connection with linking with libpthreads. The problem I am reporting only happens to programs that are linked with libpthreads. 3. The earlier discussion does not refer to any difference between Linux kernel versions. It also (obviously) does not mention coLinux. The problem I am reporting only happens with pre-0.6.2 coLinux snapshots using Linux kernel 2.6.7 and does not happen with coLinux 0.6.1 using Linux kernel 2.4.26. Thanks for pointing out the earlier discussion though. -- Joe |