1 |
On Mon, 2003-04-28 at 08:30, Mark Gordon wrote: |
2 |
[cut] |
3 |
> > >>I've been testing http://www.pixelbeat.org/scripts/gcccpuopt |
4 |
> > >>over the last while with good results. Essentially it's |
5 |
> > >>a Linux specific script that only helps on x86 at present |
6 |
> > >>but can be updated to other architectures easily. |
7 |
[cut] |
8 |
> > >>It would be beneficial IMHO for gentoo to run it at |
9 |
> > >>installation time and update /etc/make.conf appropriately? |
10 |
[cut] |
11 |
> It needs extending even for the x86 compatible world. |
12 |
> |
13 |
> markg@brenda markg $ ./gcccpuopt |
14 |
> -march=pentium2 -mmmx |
15 |
[cut] |
16 |
|
17 |
I think this would probably be more useful if it dumped out the CHOST |
18 |
line in addition to the optimizations. This would help with people that |
19 |
have the more uncommon processors to set the correct CHOST. I seen more |
20 |
than a few times K6* users have had CHOST="i686-pc-linux-gnu", when it |
21 |
should be i586-pc-linux-gnu. |
22 |
|
23 |
so it'd print this for a K6-2: |
24 |
|
25 |
$ ./gcccpuopt |
26 |
CHOST="i586-pc-linux-gnu" |
27 |
CFLAGS="-march=k6-2" |
28 |
|
29 |
|
30 |
So it could easily be pasted into make.conf. (-mmmx is redundant) |
31 |
|
32 |
-- |
33 |
Chris PeBenito |
34 |
<pebenito@g.o> |
35 |
Developer, Gentoo Linux |
36 |
Hardened Gentoo Project |
37 |
|
38 |
"Engineering does not require science. Science helps |
39 |
a lot, but people built perfectly good brick walls |
40 |
long before they knew why cement works."-Alan Cox |
41 |
|
42 |
Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE6AF9243 |
43 |
Key fingerprint = B0E6 877A 883F A57A 8E6A CB00 BC8E E42D E6AF 9243 |