Re: .i368 vs .i586

[Havoc Pennington <hp redhat com>]

"Poletti, Don" <don.poletti@comverse-in.com> writes:
> Doe anybody know how much of a performance inprovement there is
> to compiling gnome for 586 (pentium).
> 

 - most GUI operations are basically instant even on a fairly sluggish 
   machine; the ones that aren't are often dependent on the 
   video card or memory bandwidth
 - the most noticeable speed problem I see is program startup
   time (which many people have blamed on lots of shared libraries,
   but I don't know really)
 - the other noticeably slow thing typically is flipping desktops,
   I think this is usually just a bunch of paging apps in and out
   and kernel context switches as the new desktop gets a bunch of
   expose events

Anyway I guess what I'm saying is, the GNOME code itself probably
isn't causing many visible speed problems, it's mostly systemwide
issues like swapping to disk and X server efficiency. Heck, if
optimizations make binaries noticeably larger, they could create more
page faults and slow things down. In any case I wouldn't expect 
user-perceptible speed differences, though you could possibly get 
a nice placebo effect. ;-)

If you're running calculations or computations of some kind, then it's
a whole different story of course.

Havoc