[ale] Slackware user list?

[Jeff Walters]

On Wed, Aug 23, 2000 at 10:06:44AM -0400, Jim Kinney wrote:
> For most things, running is OK. However, the more closely the compiled
> binary matches with the cpu the more efficient it can run. Things compiled
> with i686 optomizations should run faster on a P6 than i386 stuff. The
> important word is "should". If you system is running OK, don't worry about
> it. If your system gets pounded on sometimes by user services,
> optomization may help relieve some of the system stress by letting a
> process get done quicker or with fewer resources. Still, though, the
> biggest performance hit is in the code itself. It the code is sloppy, no
> performance gains will be seen by compiling with P6 vs i386 flags.

I'm no gcc expert, so I'll defer to all the responses that say it doesn't
matter much for the average user.  This is actually good, since it simplifies
distribution selection.  I'm aware that the architecture of the i586 (actually
I'm running K6-2) is a lot different than the i386/i486.  It has some 
pipelining and branch prediction and other features that reduce the average 
number of clocks per instruction if the instructions are sequenced right; and I
assumed that optimized assembly code from the compiler could in most cases 
produce a noticeable performance difference.

But you are definitely right that a bad algorithm in the code can't be fixed by
any compiler unless it's somehow intelligent enough to rewrite it!  And that's
probably a much bigger perfomance issue.
--
To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message body.