Messages in this thread | | | From | Daniel Mehrmann <> | Subject | Re: [PATCH 2.4/2.5] Athlon CFLAGS | Date | Mon, 9 Sep 2002 02:13:10 +0200 |
| |
On Monday 09 September 2002 01:18, Dave Jones wrote: > On Sun, Sep 08, 2002 at 09:28:11PM +0200, Daniel Mehrmann wrote: > > Hi Alan, > > > > i add for the AMD Athlon family some optimize compilerflags. > > Gcc 3.1 and 3.2 support more specific Athlon instructions as > > 3.0 or 2.95x. This patch for 2.4.19, 2.4.20-pre5 and 2.5.33 > > set a new "-march" flag: > > > > Athlon TB/Duron += -march=athlon-tbird > > Athlon XP/Athlon4/Duron += -march=athlon-xp > > Athlon MP += -march=athlon-mp > > I thought these were all just gcc aliases for the same options ? > It's been a while since I looked at the gcc option parser, so > I've forgotten exactly what happens, but at least you missed the > bogus athlon-4 option. > > Are the gains between all these options really worth the added > complexity ? > > Dave
Hi Dave,
yes, you`re right with the athlon-4 option. Well, first thing, the mobile athlon have the same core as XP (Palomino) expect some "speed scheudle". I never see that we support mobile chips. So i think it`s enough that we put Athlon4 into the "XP group". I think too the new core "Thoroughbread" should use the "XP group".
I readed the gcc documentation, gcc-3.2 only, very deep. This was the idea for this patch. Then i looked fast into gcc-3.1/3.0/2.95x. I believe that the compiler create own code for *every* chip-release. chip-release as: athlon-tbird, athlon-xp, ...
Also take a look into the binary code and size. It`s different.
chears, Daniel
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |