optimizations

[Auke Kok]

just as a reference, I have done some intensive scripting in the past of
performance-testing and benchmarking with povray, so if you're interested
in increasing computational speed, you might want to read:

  http://lunar-linux.org/?q=node/view/289

PS surprisingly, on my P4 the winning optimization is

  "-O3 -mtune=pentium4 -march=pentium4"

makes you think again about all those dangerous flags, doesn't it ? :^)

sofar



On Mon, October 18, 2004 7:03 am, Woody Gilk said:
> been looking around the net for some good gcc CFLAGS and i think i've
> found the best set for AthlonXP users.
>
> -O2 -mcpu=athlon-xp -march=athlon-xp -fomit-frame-pointer -mmx -msse
> -dnow -mfpmath=sse
>
> there's actually a noticable difference between using -O3 and -O2 with
> -fomit-frame-pointer... strangely -O3 is slower. from what i gathered,
> -O3 creates bigger binaries, so the speed increase is offset greatly
> by the load time, so -O2 is actually faster because the binary is
> smaller... or something to that effect.
>
> a word of caution: -fomit-frame-point frees up a register in most
> compiles, creating faster binaries, but it also (generally) breaks
> debugging, so if you do any debugging at all, be aware that it could
> break your debugging process. :)
>
> just thought some people might be interesting. i recommend everyone
> who uses -O3 try -O2 and see if it makes a difference. i don't have
> hard stats, but my desktop definately feels more responsive after
> recompiling my commonly uses GTK+ apps.
>
> -woody (shadowhand)
> ----------------------------------------------------------
> wgilk design and blog [ http://www.wgilk.com ]
> follower of the fat penguin [ http://www.linux.org ]
> _______________________________________________
> Lunar mailing list
> Lunar at lunar-linux.org
> http://lunar-linux.org/mailman/listinfo/lunar
>
>