Re: CPU optimization
:: Mircea Luca writes:
>> I guess that recompiling every package is not worth the
>> trouble, but recompiling multimedia and CPU intensive ones may
>> be a major gain (say, MPEG players and such).
> Yes,I was thinking more like X ,KDE ,Gnome and multimedia so I guess
> the few things left won't take that much time.:-).
Is it safe to recompile X with different flags? IIRC, X 3.3.6
shouldn't even be compiled with gcc 2.95.2 (the Makefile would look
for gcc272), because 2.95 would miscompile it. (I could be wrong).
>> BTW, you can use -O6 for your programs. If I remember
>> correctly, it will fallback to whatever is the highest mode of
>> optimization (somebody please correct me if I'm wrong -- I
>> read this a long time ago and didn't find it listed in the
>> current documentation now; the funny thing is that it's not
>> mentioned in the current potato manpage).
>>
>> I also don't know if -O6 does the same thing as -O9.
>>
>> []s, Roger...
> I've read something on a archive of a gcc mailing list but apparently it
> doesn't relate to the gcc we use-don't remember the link ,I
> researched this for a few hours.I got stuck at the debian/rules when
> I decided to ask for help.
Yes... I saw a few Makefiles using -O9 and -O10.
I've made some tests here modifying pentium-builder. I think the
prolem with changing it is that, if I call
gcc -O2 -O3
That'll be the same as "gcc -O3"
So it'll work if you set an env. variable to "O3" and override the
-O2, but will not work the other way araound (you can't force it to
use a lower optimization level than the one in the command line).
That's because -O2 will just enable a set of optimizations, and -O3
will enable a superset of -O2.
Unless the code was changed in order to strip the previous "-O"
setting from the arguments (Joey, if you're reading this: is it
nonsense?)...
Anyway, the difference between -O2 and -O3 seems to be loop-unrolling;
I'm not sure if it's worth recompiling everything with -O3...
J.
--
Jeronimo Pellegrini
Institute of Computing - Unicamp - Brazil
http://www.ic.unicamp.br/~jeronimo
mailto:jeronimo@ic.unicamp.br mailto:pellegrini@iname.com
Reply to: