Date: Thu, 9 Jan 2003 17:55:41 -0500 (EST) From: Mikhail Teterin <mi@corbulon.video-collage.com> To: Andrew Gallatin <gallatin@cs.duke.edu> Cc: Mikhail Teterin <mi+mx@aldan.algebra.com>, Alexander Leidinger <Alexander@Leidinger.net>, cvs-all@FreeBSD.ORG, cvs-committers@FreeBSD.ORG, marius@alchemy.franken.de Subject: Re: cvs commit: src/sys/i386/i386 mp_machdep.c Message-ID: <200301092255.h09MtfHY025902@corbulon.video-collage.com> In-Reply-To: <15901.64772.844070.407901@grasshopper.cs.duke.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
> Mikhail Teterin writes: > > You may wish to take a look at > > > > http://www.FreeBSD.org/cgi/query-pr.cgi?pr=bin/43299 > > > > Especially, the follow up to it, where using SSE2 appears to slow things > > down -- at least for double values. > > > > -mi > Strange. The intel compiler is slower too. But at least it gets the > right answer, which is more than gcc can do (unless O0 is used) As I note in my follow up, gcc now gives the right answer too on my system. I suspect -- thanks to the commit I quote there. How recent is your system? > icc -O3 -tpp7 -xW: (P4) > > 2^2.1 is 4.28709 > 11^-2.1 is 0.00650243 > 5.77 real 5.68 user 0.02 sys > > > icc -O3 -tpp6 -xK: (PIII) > > 2^2.1 is 4.28709 > 11^-2.1 is 0.00650243 > 5.38 real 5.13 user 0.00 sys > > gcc -O3 -march=pentium4 > 2^2.1 is 0.5 > 11^-2.1 is 0.0909091 > 0.63 real 0.62 user 0.00 sys Yep, this lighting speed and incorrectness is what I was seeing, when I submitted the PR. > gcc -O3 -march=pentium3 > > 2^2.1 is 4.28709 > 11^-2.1 is 0.00650243 > 6.68 real 6.50 user 0.01 sys > I still build my system with CPUTYPE=p3, so I think my libs are OK. My example in there explicitly avoids using -lm anyway :-) -mi To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200301092255.h09MtfHY025902>