https://bugzilla.novell.com/show_bug.cgi?id=186074 guru@unixtech.be changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |guru@unixtech.be ------- Comment #34 from guru@unixtech.be 2006-10-11 15:36 MST ------- lame is just a ridiculous example, as it does 99.9% mathematical computing and almost no I/O. Typical applications do a lot more I/O than raw CPU processing and, hence, you probably won't see any measurable performance increase by tuning it for i686. Guys, provide some *factual data* about those optimizations (i.e. real benchmarks) for typical applications (KDE, Xorg, k3b, amarok, firefox, thunderbird, gimp, ...). Yes, -Bdirect and similar approaches give a much better startup time. Yes, using the latest stable GCC usually produces binaries with better performance. But building for i686 instead of i586 ? And athlon instead of i586/i686 ? Proof ? Where ? Piotrek, please buy a clue before saying that building the whole distro tree for i686 and i586 is a neglible effort. It isn't. Measure the size of the i586 repository, have a chat with some of the mirror maintainers (especially with Eberhard Moenkeberg who hosts the master mirror at ftp.gwdg.de), they'll tell you about cache misses, about hardware requirements to effectively host SUSE Linux CD ISOs, DVD ISOs, FTP trees, online updates, etc... IMO if there *is* enough proof that optimizing for i686 actually does provide a significant performance benefit (haven't seen such a thing yet though), then we should completely switch from i586 to i686 (i.e. no more i586) as adding another architecture tree is probably unbareable. Whether that's a good strategic option, I don't know. OTOH, 10.1 requires at least 256MB of memory, so very old boxes are not usable with SUSE Linux anyway. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.