Will a -march=k6 -Os Kernel & system run on duron? -Yes!

JohnY · Guru Joined: 26 Jul 2002 Posts: 305

I currently have 2 machines running Gentoo:
1. a K6-III+ @ 616mHz 112mHz fsb running an ~x86 1.2 gentoo with gcc 2.95.3 - started in the very early gentoo days

2. an A7V running at 2020 mHz (tbred-b) ~x86 1.4 gentoo

I aquired another A7V for $31 and I have an older spitfire duron 700mHz and a morgan duron 1.3gHz. I would like to move the hard drive from the K6-III+ fic503+ to the new 2nd A7V.

The question is will a gcc 2.95.3 system compiled with -march=k6 -Os run on either or both durons? I know that I could install fresh or even clone the other A7V but I would like to try a gentoo upgrade just for the hell of it

I actually use the K6-III+ system more (because it has a nice 19" Samsung monitor and a w2k install that I need occassionally). I could even clone the drive and put it on the duron just to be safe instead of daring if need be.....

I have both 2.4.20-r6 gentoo-sources and 2.6.0-test6-mm1 kernels running on the k6-III+ .... So do you think they will boot on either or both flavor durons?

JohnY

JohnY · Guru Joined: 26 Jul 2002 Posts: 305

No one has any thoughts on the above ? The second A7V is booted and running with an old 700mHz Duron spitefire. I'll move the drive tonight or tomorrow and give it a try... (found out my voodoo3500 agp is dead ;( )

JohnY

PowerFactor · Posted: Thu Oct 02, 2003 5:19 am Post subject:

I see no reason why it wouldn't. I've compiled systems optomized for a k6 (and also k6-2 and i486) from stage1 in a chroot on my XP with no problems. You of course may need to recompile your kernel to get dma support etc. with the new motherboard chipset if it's different, but that shouldn't stop it from booting.

From my experience compiling those k6 systems though I'd say it really will be worthwhile performancewise to recompile for the duron. Of couse the effect might not be as dramatic on a spitfire duron as it is on an athlonXP. For the morgan duron it would probably be close though.

JohnY · Guru Joined: 26 Jul 2002 Posts: 305

Thanks for that info .... I can't get the morgan 1.3gHz duron to boot on this A7V 1.02 board yet (booted on other A7V 1.02 thou).

I plan on not only recompiling but doing an in place upgrade from gentoo 1.2 ~x86 to gentoo 1.4 ~x86. This k6-III+ gentoo box started life in the very early days of gentoo

(I think before 1.0 but i can't remember for sure - and before that RHat) and I already have both 2.95.3 and 3.2.3 gcc on it ready to go

, but I have never tried booting a k6 optimized system on a newer AMD chip. Good to hear that it worked for you.

Thanks,

JohnY

JohnY · Guru Joined: 26 Jul 2002 Posts: 305

Well I just moved the drive from a k6-III+ 616 mHz fic503+ 320mb mvp3 via chipset, to an A7V kt133 700mHz duron w/ 512mb ram running at 133mHz. my G400 16mb, aha3940 dual scsi, and tulip nic stayed the same.... changed from cs4237 isa sound to ensoniq 1370 pci. The gentoo is a 1.2 gcc2.95.3 ~x86.

I'm happy to report that the new machine just booted fine with a 2.6.0-test5-mm1 kernel and aside from the 3940 going into pio instead of memory mapped mode, and then having to recompile my kernel for the new alsa ens1370 modules. It's running fine. Bogomips went from 1089 to 1398 and glxgears -info went from 500 to 800 fps with the original -march=k6 -Os CHOST i586 kernel and to 813 fps with the new CHOST i686 -march=k6 -Os kernel.

I gather that with gcc 2.95.3 you till need to use -march k6 with a spitfire (original) duron cpu (at least according to freehackers.org).

I guess it's time to start the in place upgrade from gentoo 1.2 to gentoo 1.4 and at least gcc 3.2.3 .....

I'm so surprized that the box just came fine .... I'm betting that the win2k on the drive won't fair as well

JohnY

Lovechild · Posted: Fri Oct 03, 2003 10:37 am Post subject:

the -Os being faster was true on GCC 2.95.3 where -Os was actually better at optimizing than -O3/-O2 in many cases, this isn't true for GCC3.x though.

On GCC 3.x I would recommend using -O2 since -O3 is prone to in some cases causing minor breakage and it is slower in many cases since it produces larger code and is more prone to cache misses, I personally can't feel the difference between -O2 and -O3 in runtime but startup times are radically slower with -O3 as are compile times.

word of the wise seems to be -O2 for GCC3.x and -Os for GCC 2.95.3.

JohnY · Guru Joined: 26 Jul 2002 Posts: 305

LC,

Lovechild · Posted: Fri Oct 03, 2003 1:36 pm Post subject: