| View previous topic :: View next topic |
| Author |
Message |
altrhombus
Tux's lil' helper
Joined: 25 Sep 2004
Posts: 92
|
 Posted: Tue Aug 22, 2006 5:37 am Post subject: K-6, K6-2, K6-3: What is what? |
 |
|
I think it's about time we figure out how to get those K6/K6-2/K6-3 processors working. I've been trying to build a K6-3 Gentoo box for a long time and have yet to figure out a solution that works.
Here's the problem: There's a nasty "bug" with AMD's old marketing (from bug 24379):
There are 6 different cores (that is based on the 'flags' field in /proc/cpuinfo):
[all also have 'fpu vme de pse tsc msr mce cx8']
mmx
mmx 3dnow syscall k6_mtrr
mmx 3dnow pge syscall mtrr
mmx 3dnow pge sep mtrr <-K6-2
mmx 3dnow pge syscall 3dnowext k6_mtrr <-K6-3
There are only 4 unique values for the 'model name' field:
AMD-K6(tm) 3D processor
AMD-K6(tm) 3D+ Processor
AMD-K6(tm)-III Processor
AMD-K6tm w/ multimedia extensions
How do we figure out which one is what? If you've got a K6 processor, post your cpuinfo here, as well as your make.conf if you got it to work. I'll start off with what I have (I'm not posting my make.conf as I can't get it to work--it keeps failing on compiles):
| Code: | cat /proc/cpuinfo
processor : 0
vendor_id : AuthenticAMD
cpu family : 5
model : 9
model name : AMD-K6(tm) 3D+ Processor
stepping : 1
cpu MHz : 451.031
cache size : 256 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr mce cx8 pge mmx syscall 3dnow k6_mtrr
bogomips : 903.74 |
The processor core is stamped with "AMD K6-3." It can't be a true K6-3 though because there's no 3dnowext. So it's a "3D+" processor, whatever that is.
Which -march or -mtune am I to use? So far it's failed with i586, k6-2, and k6-3...
Last edited by altrhombus on Wed Jan 06, 2021 3:41 am; edited 1 time in total |
|
| Back to top |
|
 |
Sachankara
l33t
Joined: 11 Jun 2004
Posts: 696
Location: Stockholm, Sweden
|
| Posted: Tue Aug 22, 2006 10:25 am Post subject: Re: K-6, K6-2, K6-3: What is what? |
|
|
Unless I remember it incorrectly, K6-3 is a K6-2 with speed throttling for laptops. You should use the i586 stage 3 tbz2 file. I'd recommend something like "-Os -mtune=i586 -pipe" for that CPU. MMX and 3DNow really doesn't add any performance to the applications on those old processors, unless they've been specifically written for a certain subset of CPU features (like mplayer).
_________________
Gentoo Hardened Linux 2.6.21 + svorak (Swedish dvorak) |
|
| Back to top |
|
|
altrhombus
Tux's lil' helper
Joined: 25 Sep 2004
Posts: 92
|
| Posted: Wed Aug 23, 2006 10:00 pm Post subject: |
|
|
@Sachankara
The -Os seemed to be my problem. I've been doing -O2 for a long time and it always failed. Last I checked it compiled further than gcc (where it usually failed) and failed at perl. I'm doing a emerge --resume to see if it was just another sporadic compile bug.
I'll post my make.conf when I'm at the PC again... |
|
| Back to top |
|
|
odessit
Apprentice
Joined: 01 Feb 2004
Posts: 180
Location: Current Residency - Server Room - Caution - Frostbite Imminent!
|
| Posted: Wed Aug 23, 2006 10:57 pm Post subject: Re: K-6, K6-2, K6-3: What is what? |
|
|
| Sachankara wrote: | | Unless I remember it incorrectly, K6-3 is a K6-2 with speed throttling for laptops. |
No, the chips that were typically used in laptops were the K6-2+ and K6-3+ same Super Socket 7 as desktop chips made them quite common in the desktops as well.
both of them have good speed improvements over the non-plus versions (better cache, better float/integer performance)
-Os -mtune=i586 -pipe - sounds reasonable |
|
| Back to top |
|
|
altrhombus
Tux's lil' helper
Joined: 25 Sep 2004
Posts: 92
|
| Posted: Thu Aug 24, 2006 5:14 am Post subject: |
|
|
| Code: | # These settings were set by the catalyst build script that automatically built this stage
# Please consult /etc/make.conf.example for a more detailed example
# Last edited: 8/21/06 by jaker
CHOST="i586-pc-linux-gnu"
#
CFLAGS="-march=k6-2"
#
CFLAGS="${CFLAGS} -Os"
CFLAGS="${CFLAGS} -fomit-frame-pointer"
CFLAGS="${CFLAGS} -pipe"
#
CXXFLAGS="${CFLAGS}"
#
#LDFLAGS="-Wl,-O1 -Wl,--sort-common -s"
#
ACCEPT_KEYWORDS="x86"
PORTAGE_TMPDIR=/var/tmp
PORTDIR=/usr/portage
DISTDIR=${PORTDIR}/distfiles
PKGDIR=${PORTDIR}/packages
PORT_LOGDIR=/var/log/portage
PORTDIR_OVERLAY=/usr/local/portage
MAKEOPTS="-j2"
GENTOO_MIRRORS="http://gentoo.chem.wisc.edu/gentoo http://gentoo.osuosl.org http://www.ibiblio.org/pub/Linux/distributions/gentoo"
SYNC="rsync://rsync.gentoo.org/gentoo-portage"
RSYNC_RETRIES="3"
RSYNC_TIMEOUT=180
PORTAGE_NICENESS=3
AUTOCLEAN="yes"
FEATURES="distlocks sandbox userpriv usersandbox" |
This seems to be working for me with random fails on long compiles (but a emerge --resume fixes it the second time) |
|
| Back to top |
|
|
depontius
Advocate
Joined: 05 May 2004
Posts: 3525
|
| Posted: Sat Aug 26, 2006 5:35 pm Post subject: Re: K-6, K6-2, K6-3: What is what? |
|
|
| jakertberry wrote: | I think it's about time we figure out how to get those K6/K6-2/K6-3 processors working. I've been trying to build a K6-3 Gentoo box for a long time and have yet to figure out a solution that works.
Here's the problem: There's a nasty "bug" with AMD's old marketing (from bug 24379):
There are 6 different cores (that is based on the 'flags' field in /proc/cpuinfo):
[all also have 'fpu vme de pse tsc msr mce cx8']
mmx
mmx 3dnow syscall k6_mtrr
mmx 3dnow pge syscall mtrr
mmx 3dnow pge sep mtrr <-K6-2
mmx 3dnow pge syscall 3dnowext k6_mtrr <-K6-3
There are only 4 unique values for the 'model name' field:
AMD-K6(tm) 3D processor
AMD-K6(tm) 3D+ Processor
AMD-K6(tm)-III Processor
AMD-K6tm w/ multimedia extensions
How do we figure out which one is what? If you've got a K6 processor, post your cpuinfo here, as well as your make.conf if you got it to work. I'll start off with what I have (I'm not posting my make.conf as I can't get it to work--it keeps failing on compiles):
| Code: | cat /proc/cpuinfo
processor : 0
vendor_id : AuthenticAMD
cpu family : 5
model : 9
model name : AMD-K6(tm) 3D+ Processor
stepping : 1
cpu MHz : 451.031
cache size : 256 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr mce cx8 pge mmx syscall 3dnow k6_mtrr
bogomips : 903.74 |
The processor core is stamped with "AMD K6-3." It can't be a true K6-3 though because there's no 3dnowext. So it's a "3D+" processor, whatever that is.
Which -march or -mtune am I to use? So far it's failed with i586, k6-2, and k6-3...
~Jaker |
I'm running the same CPU core as you, and always thought of it as the "old" k6-3, the 2.4V part, as opposed to the newer 2.2V parts and the k6-2+ or k6-3+. I've run Gentoo happily on it for years, except that it overheats easily. Right now it's my server, running Gentoo Hardened. Here are some relevant lines from /etc/make.conf:
| Code: | USE="-gnome -esd -eds -kde -arts threads nptl nptlonly pic hardened"
CFLAGS="-march=k6-3 -O2 -pipe"
#CFLAGS="-O2 -pipe"
CHOST="i586-pc-linux-gnu"
CXXFLAGS="${CFLAGS}" |
The commented CFLAGS line is because I couldn't get the recent busybox to emerge, and thought it might have been a problem with k6-3, so figured I'd try it without '-march='. No go, but one of these days I'll dig harder, or there will be a newer ebuild come out.
_________________
.sigs waste space and bandwidth |
|
| Back to top |
|
|
cac
n00b
Joined: 05 Jun 2006
Posts: 29
Location: Columbus, Ohio
|
| Posted: Fri Sep 01, 2006 1:57 pm Post subject: |
|
|
It is a Compaq Persario, exact model escapes me at the moment, but has a K6-2 inside:
| Code: | cat /proc/cpuinfo
processor : 0
vendor_id : AuthenticAMD
cpu family : 5
model : 8
model name : AMD-K6(tm) 3D processor
stepping : 12
cpu MHz : 500.318
cache size : 64 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr mce cx8 pge mmx syscall 3dnow k6_mtrr
bogomips : 1001.94 |
make.conf has:
| Code: | CHOST="i586-pc-linux-gnu"
CFLAGS="-Os -mtune=k6-2 -pipe"
CXXFLAGS="${CFLAGS}" |
I had been running solid for nearly two + (maybe 3 +) years; originially installed it during the gentoo 1.4 days. Now during the move to 2006.1 and gcc 4.1.1 I am having problems getting gcc to compile. Trying out CFLAGS="-O2 -mtune=k6-2 -pipe". I use this as my server, and had dabbled with the hardened stuff before, but never truely configured that properly. Thought this would make a great time to set it up along with the move to 2006.1, but now I fear that I am making too many changes at once. Ehh, I'll figure it out eventually; thats the fun of gentoo  . |
|
| Back to top |
|
|
pteppic
l33t
Joined: 28 Nov 2005
Posts: 781
|
| Posted: Fri Sep 01, 2006 3:19 pm Post subject: |
|
|
My router box is a | Code: | processor : 0
vendor_id : AuthenticAMD
cpu family : 5
model : 8
model name : AMD-K6(tm) 3D processor
stepping : 12
cpu MHz : 400.912
cache size : 64 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr mce cx8 pge mmx syscall 3dnow k6_mtrr
bogomips : 802.81 |
Being the router it never compiles X or the like, but has never had any trouble (except when using distcc).
Relavent make.conf lines | Code: | CFLAGS="-march=k6-2 -O2 -pipe -fomit-frame-pointer"
CHOST="i586-pc-linux-gnu"
CXXFLAGS="${CFLAGS}"
USE=" ipv6 nptl nptlonly -sdl -mppe-mppc -alsa -arts -avi -gnome -gstreamer -gtk -gtk2 -kde -qt -X -png -jpeg" |
|
|
| Back to top |
|
|
depontius
Advocate
Joined: 05 May 2004
Posts: 3525
|
| Posted: Fri Sep 01, 2006 11:37 pm Post subject: |
|
|
My impression is that mtune=k6-? optimizes for the k6-?, but emits instructions that will still work on i586, whereas march=k6-? may emit k6-? specific instructions that may not work on i586.
I also thought I heard that mtune was generally deprecated in favor of march, but given my definition, that wouldn't seem so.
_________________
.sigs waste space and bandwidth |
|
| Back to top |
|
|
cac
n00b
Joined: 05 Jun 2006
Posts: 29
Location: Columbus, Ohio
|
| Posted: Wed Sep 13, 2006 1:13 pm Post subject: |
|
|
| depontius wrote: | My impression is that mtune=k6-? optimizes for the k6-?, but emits instructions that will still work on i586, whereas march=k6-? may emit k6-? specific instructions that may not work on i586.
I also thought I heard that mtune was generally deprecated in favor of march, but given my definition, that wouldn't seem so. |
Actually it is slightly backwards from what you described; march=k6-? will break compatibility with other x86 archs, while mtune=k6-? will keep compatibility with x86 archs. mcpu was deprecated in favor of mtune, but AFAIK mtune and march are mutually exclusive. Obviously, if you want to take full advantage of a CPU, then you want to use march, but for odd-ball CPUs, like the K6-* it is generally thought to keep out gremlins. Maybe this thinking has changed? 
Either case, I have used mcpu/mtune for several years, and never any breakage because of that. I was finally able to get up to date with 2006.1/gcc4.1.1 after weeks of compiling and re-compiling. I think my problems lied with the use of hardened USE flag. I think I read somewhere that that is causing problems with the latest gcc. |
|
| Back to top |
|
|
depontius
Advocate
Joined: 05 May 2004
Posts: 3525
|
| Posted: Thu Sep 14, 2006 1:42 am Post subject: |
|
|
Not to be nyah-nyah, but I've been using march for a while, too. OTOH, I haven't been able to build busybox, but taking march=k6-3 out doesn't fix that.
_________________
.sigs waste space and bandwidth |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Kernel & Hardware
