On Wed, 7 Feb 2001, Roman Shterenzon wrote:
> For CFLAGS (long battle, huh?) I attached a patch to configure.
> You have -funroll-all-loops in the configure script, but in gcc(1) manpage it
> Perform the optimization of loop unrolling. This
> is done for all loops. This usually makes programs
> run more slowly.
> Why would you want xine to run more slowly?
I had done some benchmarks a while ago and found that xine runs faster
with this optimization set - I guess it's due to xine's massive
cache-locale behaviour and it really spends a lot of time in inner loops.
But if you benchmark xine without that flag to run faster, please drop me
> Why do you have -static there?
> As usual, I removed the -O3. In FreeBSD, user supplied CFLAGS should have
> optimization flags.
hum ... all these -fsomething flags fall into that category as well, so
wouldn't we have to remove them as well?
You say _should_ have - do all FreeBSD users know about that?
> Is it possible to check (I think I've seen it in samba) if CFLAGS have
> -Osomething and if not, only then define -O3, it would be great!
agreed - could you please mail me that patch?
BTW: I removed that -O3 now for FreeBSD, but remember that xine will crash
if compiled this way without "-O?" coming in (the mmx code won't work for
example, win32codec support AFAIK as well).
Perhaps we should add an FAQ entry for this one at least describing how to
supply that flag from "the outside" ?
Get latest updates about Open Source Projects, Conferences and News.