| 1 |
On 8/30/06, Vladimir G. Ivanovic <vgivanovic@×××××××.net> wrote: |
| 2 |
> On Mon, 2006-08-21 at 20:06 -0700, Richard Fish wrote: |
| 3 |
> > -Os seems to take much less time and memory to compile than -O2 or -O3. And being a |
| 4 |
> > ~arch user, time-to-compile is a nice thing to reduce. |
| 5 |
> |
| 6 |
> It is possible that one of my usual CFLAGS options (-Wall -march=k8 |
| 7 |
> -msse3 -pipe -O2 -ffast-math) overwhelms any -Os/-O2 execution time |
| 8 |
|
| 9 |
Well it looks like I'll have to retract that statement in relation to |
| 10 |
-Os/-O2. My current tests with -march=pentium4 -fweb -pipe |
| 11 |
-fomit-frame-pointer for my laptop or -march=athlon64 -fweb -pipe for |
| 12 |
my AMD box show a <5% difference in compile time between -Os and -O2. |
| 13 |
|
| 14 |
I think there was a bigger difference in previous versions of gcc, but |
| 15 |
I can't seem to find any notes or anything to support that belief... |
| 16 |
|
| 17 |
Cheers, |
| 18 |
-Richard |
| 19 |
-- |
| 20 |
gentoo-user@g.o mailing list |
Archives