why is FC5 still compiled at -O2

[Mike McCarty]

sean wrote:
> On Tue, 28 Mar 2006 11:56:16 -0500 (EST)
> howarth at bromo.msbb.uc.edu (Jack Howarth) wrote:
> 
> 
>>    Is there any plans to eventually move the builds of Fedora from
>>-O2 to -O3? It would seem that we are leaving some performance
>>optimization laying on the table by not doing so.

Eh? When I worked on gcc many years ago, it *needed* some optimization,
or the produced code was truly horrible. But few applications benefit
very much from the kinds of optimizations performed by -O3 and not by
-O2.

> Actually there is some serious talk of having FC6 userland compiled
> with -Os.   Which seems counterintuitive but actually ends up giving
> better performance overall on modern CPU's than any of the 
> alternatives.

That's interesting, but not (to me, at least) counterintuitive.
-Os would have a tendency to put a little bit more into cache,
and with today's processors, running from cache is the name of
the game in regards to speed. Many instructions actually execute
in 0 cycles, regardless of what the manual says, whereas cache
read misses take (5,1,1,1) or the like at *bus* (not CPU) speeds.

Even today's faster RAM with (1,1,1,1) can't compete, even when
the policy lets the processor execute as soon as the needed bytes are
there instead of waiting for the entire cache line to fill.

Mike
-- 
p="p=%c%s%c;main(){printf(p,34,p,34);}";main(){printf(p,34,p,34);}
This message made from 100% recycled bits.
You have found the bank of Larn.
I can explain it for you, but I can't understand it for you.
I speak only for myself, and I am unanimous in that!