View previous topic :: View next topic |
Author |
Message |
deviljelly Apprentice
Joined: 19 Feb 2003 Posts: 193 Location: Netherlands
|
Posted: Fri Feb 21, 2003 6:47 pm Post subject: Oppps merginf all ~x86 ebuilds, what can expect |
|
|
Opps,
I managed to get ACCEPT_KEYWORDS="~x86" in make.conf before I ran emerge -u world.
So noe I'm merging all of the "unstable" ebuilds including gcc3.2.2 and xfree86 4.2.99!!
What can I expect??? Is it going to be stable
P.S. My CFLAGS are:
-O3 -mcpu=athlon-xp -march=athlon-xp -pipe -fomit-frame-pointer -ffast-math -funroll-loops -mmmx -msse -m3dnow -fforce-addr -mfpmath=sse -frerun-cse-after-loop -frerun-loop-opt -falign-functions=4 -maccumulate-outgoing-args -fprefetch-loop-arrays |
|
Back to top |
|
|
-JeaN- Apprentice
Joined: 25 Nov 2002 Posts: 211 Location: PaRiS :D
|
Posted: Fri Feb 21, 2003 6:52 pm Post subject: |
|
|
Hello,
It should work good, if it doesn't anyways you'll find answers here i think
Problems could arise if you don't etc-update..
Also in your CFLAGS, you just need -march, -mcpu then is not needed, and the -funroll-loops is not such a advised optimisation..
Good luck and let us know how it goes |
|
Back to top |
|
|
pjp Administrator
Joined: 16 Apr 2002 Posts: 20067
|
Posted: Fri Feb 21, 2003 6:55 pm Post subject: |
|
|
Moved from Installing Gentoo. _________________ Quis separabit? Quo animo? |
|
Back to top |
|
|
modal Apprentice
Joined: 02 Oct 2002 Posts: 277
|
Posted: Sat Feb 22, 2003 2:15 am Post subject: |
|
|
why is the -funroll-loops not such an advised optimisation? |
|
Back to top |
|
|
clockwise Apprentice
Joined: 24 Aug 2002 Posts: 152 Location: uk
|
Posted: Sat Feb 22, 2003 2:14 pm Post subject: |
|
|
take a look at cflags central. off the top of my head i think the reasoning is; your hard-drive is the slowest part of your machine, therefore load times for programs are your worst bottleneck for performance. so, while unrolling loops makes for marginally faster code it also makes for larger code and therefore a longer load time.
i use (in a possibly redundant way):
Code: | CFLAGS="-march=pentium3 -Os -pipe -fomit-frame-pointer -msse -mmmx -mfpmath=sse -w |
_________________ "if an injury has to be done to a man it should be so severe that his vengeance need not be feared." - niccolò machiavelli |
|
Back to top |
|
|
Gnufsh Guru
Joined: 28 Dec 2002 Posts: 400 Location: Portland, OR
|
Posted: Tue Feb 25, 2003 6:36 am Post subject: |
|
|
The main penalty of loop rolling size problems can be a cpu issue also, as larger code means more cache misses, reducing performance. |
|
Back to top |
|
|
|