1 |
On Mon, 09 Jun 2008 06:43:41 +0100 |
2 |
Graham Murray <graham@×××××××××××.uk> wrote: |
3 |
|
4 |
> Andrew Gaydenko <a@××××××××.com> writes: |
5 |
> |
6 |
> > gcc-4.3.1 is umasked now. I'm on ~amd64 with Core 2 Duo CPU. Is it |
7 |
> > sufficient just to set |
8 |
> > |
9 |
> > CFLAGS="-O2 -march=core2 -mtune=core2 -pipe" |
10 |
> > |
11 |
> > instead of |
12 |
> > |
13 |
> > CFLAGS="-O2 -march=native -mtune=native -pipe" |
14 |
> > |
15 |
> > which I have in make.conf now? Are there any other things to do? |
16 |
> |
17 |
> Should the two not be equivalent? On a core2 processor, should |
18 |
> -march=native not generate code for the core2? |
19 |
|
20 |
If we are going to get picky about this, it should be noted the -march |
21 |
implies -mtune, at least according to the gcc man pages, and I will |
22 |
trust them to know. |
23 |
|
24 |
I can't remember which will actually be implemented if both are |
25 |
specified, but march is a essentially a stronger version of mtune which |
26 |
breaks compatibility with other CPUs within the same ARCH type. (i.e. |
27 |
march=pentium4 may not run on a pentium3, but should run faster than |
28 |
mtune on a pentium4) |
29 |
|
30 |
RobbieAB |
31 |
-- |
32 |
gentoo-user@l.g.o mailing list |