Can't use "-march=athlon". Which is better, "-march-k6" or "march-i686"

[Giscardo]

I had to downgrade from gcc3.0 back to gcc2.95.3 under Slack8.0 because KDE 2.1.2 apparently wont compile under gcc-3.0. I guess gcc-2.95.3 doesn't recognize the 'athlon' value for the '-march' switch. I was just wondering if it's better to use '-march=k6' or '-march=i686' if I have an Athlon 1.36Ghz CPU. Which one would give me better performance. Please don't reply if you don't actually know the answer, I would rather not use one over the other based one someone else's conjecture.

[Malakin]

I've never actually tested it myself but everyones always said to use "-686" for an athlon and that would make sense to me. Try a small test program if you're really concerned about it. -ffast-math can also make a big difference.

KDE 2.2 beta 1 still doesn't support gcc 3 according to their announcement notes (http://www.kde.org/announcements/announce-2.2-beta1.html), maybe the next beta will.

[Giscardo]

Do you have any suggestions as to what programs i can use to benchmark the differences? I am interested in make a project of this issue, and would like to test all sorts of apps, video, sound, 3d, and whatever else i can. What programs could I use? My ideas so far are:

LMbench (http://www.bitmover.com/lmbench/), (for an example of it's use visit Linux Benchmark (http://euclid.nmu.edu/~benchmark/)).

Well that's the only thing I can think of right now, but I'll do some searching. I'd really appreciate any input, and suggestions for benchmarks to use.

[Malakin]

If you wait a couple months I'm sure everything important like KDE will compile fine on gcc 3.0, but if it's your intention to see how much faster programs run with different compiler options on an Athlon then that's certainly a worthy project :)

I like to see real programs used for benchmarks. Here's some ideas.
Game - quakeforge
(http://sourceforge.net/projects/quake/)
Video Player - mplayer
(http://mplayer.dev.hu/homepage/)
Graphics Software - GIMP
(http://www.gimp.org)