> Correction: SIGILL applies to Unix, Linux, and BSD (OS X). It is not caused
> by BOINC, SETI, or any other project. If memory serves, the kernel is
> responsible for signal management.

Thanks for the correction :-)

Zerjillo

---

> vendor_id : AuthenticAMD
> cpu family : 5
> model : 8
> model name : AMD-K6(tm) 3D processor
>
> Do anyone know if what version should I use?
>
> Thanks in advance
>
> Zerjillo

Zerj,

According to the AMD CPU identification instruction page, that CPU should be called a K6-2. (see below)

AMD-K6® Processor Model 6 “AMD-K6(tm) w/ multimedia extensions” No
AMD-K6 Processor Model 7 “AMD-K6(tm) w/ multimedia extensions” No
AMD-K6-2 Processor Model 8 “AMD-K6(tm) 3D Processor” No
AMD-K6-III Processor Model 9 “AMD-K6(tm) 3D+ Processor” No
AMD Athlon™ Processor Model 1 “AMD-K7(tm) Processor” No

---

Yes, from Ben's post it does look like it's a K6-2, but yet you say that client doesn't work either?

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

I tried each of k6, k6-2, i686 on my model 8 step 12 chip. i686 gave SIGILL as expected. Both k6 gave sigsegv. I am however using the k6-2 4.13 optimized version now with great improvement.

v4.13
w: 159
d: 395

v4.13opt
w: 426
d: 774

Also converted my PPro 200 dual to 4.19opt

v4.13
w: 101
d: 219

v4.19opt
w: 193
d: 373

Thanks so very much for your work. I'll be looking through the rest of your site now. :)

---

OK, since no one has managed to get any of the K6 clients working yet, I'm guessing that it's a bad compile. I'll recheck my notes and recompile them when I get a chance.

Thanks for the feedback, and I hope to have the problem fixed soon, although I don't have any K6 processors available for testing against.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

I'm following this pretty close and have several k6 chips, so I'll be testing for you. I'm also trying a build of my own to see what happens.

---

> OK, since no one has managed to get any of the K6 clients working yet, I'm
> guessing that it's a bad compile. I'll recheck my notes and recompile them
> when I get a chance.

Nope.
I've got your k6-2 client up and running on 2 hosts, as I stated earlier. No problems so far.

---

Hm, I'm curious. anyone have an idea whey I can't and rattelschneck can?

Here's what I get:

[root@morgain boinc]# ./boinc_4.19_k6-2-pc-linux-gnu
2005-01-30 17:58:00 [---] Starting BOINC client version 4.19 for i686-pc-linux-gnu
2005-01-30 17:58:00 [SETI@home] Project prefs: using separate prefs for home
2005-01-30 17:58:00 [---] Version Change Detected (4.13 -> 4.19); running CPU benchmarks
2005-01-30 17:58:00 [SETI@home] Host ID is 442735
2005-01-30 17:58:00 [---] General prefs: from SETI@home (last modified 2004-08-26 01:56:31)
2005-01-30 17:58:00 [---] General prefs: using separate prefs for home
SIGSEGV: segmentation violation
Exiting...
SIGSEGV: segmentation violation
Exiting...

on:

processor : 0
vendor_id : AuthenticAMD
cpu family : 5
model : 8
model name : AMD-K6(tm) 3D processor
stepping : 12
cpu MHz : 400.916
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 : 799.53

That's up to date [sic] RedHat 7.3.

---

Thanks for correcting me above rattelschneck :)

Here's a guess - could be related to the version of linux you guys are running. Maybe older versions have a conflict whilst newer versions work?

Could you guys please post what version of linux you're running and if the client works or not, and we can see if this holds any water - thanks.

If you wish to compile your own, it's not that difficult - detailed instructions are on my site (in my sig) and all you need to change is the -march= flag to whatever processor you have, e.g. -march=k6-2

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

Ned, I downloaded your new boinc client for Athlon's and have it running on both of my Athlon XP's. One is a 2200+, the other is a 2500+. both seem to be working just fine.

---

I stopped trying to compile when I got to seti and it couldn't find a lot of stuff. The needed software to build link is broken, or I would have resolved it, but I got bored and went on to other things for a while. When I get bored with those I'll come back and try to get the needed -devel- packages installed.

---

Hi Ned, and thanks for the new client :-)

OLD 4.19 i686 (standard download from here)
Dhrystone 1024.02 MIPS
Whetstone 522.45 MFLOPS

Your 4.19 pentium3
Dhrystone 2031.67 MIPS
Whetstone 991.97 MFLOPS

System:
CPU - P3 1Ghz (coppermine)
RAM - 512 MB
OS - Debian Sarge

Working like a champ so far, but I'll have to wait until the current WU is done to get an accurate crunch time for the new client. The old one took just over 8 hours per WU :-( Hopefully the new one will knock that down to about 6 hours or so.

I tried running the ppro (i686) client on my machine and got "SIGSEGV: segmentation violation Exiting..." Any ideas?
Info:
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 1
model name : Pentium Pro
stepping : 9
cpu MHz : 199.436
cache size : 256 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 2
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov
bogomips : 398.13

AFACT, it's Unix's/Linux's/BSD's way of saying "Something went wrong, and I don't know who to blame".

If everyone is having SIG problems, then maybe v4.19 isn't as ready as we thought...

---

> Could you guys please post what version of linux you're running and if the
> client works or not, and we can see if this holds any water - thanks.

OK,
i'm runnig KNOPPIX V3.7 kernel 2.4.27. Since the system I refered to is a stack system, all of the stack nodes runs the same version. They are netbooted from the stackmaster via PXE, dhcp, tftp and nfs. What differs is the version of the boinc client. I'm using your 4.19_k6-2, 4.19_pentium3 and 4.19_i686 actually (the stack nodes differs in hardware).

For all who don't know KNOPPIX, it is a bootable Linux cdrom which is pretty complete and useful. Please go here to see what it is all about.
KNOPPIX iso images can be downloaded from several mirrors and via bittorrent.
Maybe it is a option for some of you to find out wheter Ned's client runs on this. Get you a copy of the iso image, burn it to cdrom, boot into it, copy the client over and have a try.

rattelschneck

---

>
> Working like a champ so far, but I'll have to wait until the current WU is
> done to get an accurate crunch time for the new client. The old one took just
> over 8 hours per WU :-( Hopefully the new one will knock that down to about 6
> hours or so.
>
>

It's only the BOINC client, so the only thing that will improve is your benchmark score and thus your requested credit (making it more in keeping with credit requested by Windows machines). I have not been able to achieve any real level of improvement by recompiling the SETI application with optimizations so your processing times should not change.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

> > Could you guys please post what version of linux you're running and if
> the
> > client works or not, and we can see if this holds any water - thanks.
>
> OK,
> i'm runnig KNOPPIX V3.7 kernel 2.4.27. Since the system I refered to is a
> stack system, all of the stack nodes runs the same version. They are netbooted
> from the stackmaster via PXE, dhcp, tftp and nfs. What differs is the version
> of the boinc client. I'm using your 4.19_k6-2, 4.19_pentium3 and 4.19_i686
> actually (the stack nodes differs in hardware).
>
>
> rattelschneck
>


Thanks rattelschneck. I think some of the other folks above may be running older versions (like RH 7.x etc) on their older processors and that's where the problems may be coming from. Newer versions of linux don't seem to be causing any trouble.

If people who are having problems with my client could post what version of linux they are running please, then we could tell if this is the case.

Thanks,

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

I have functioning with 4.19opt Dual PPro with FC1
I have crashing SIGSEGV 4.19opt on K6-2 with RH7.3
I have functioning with 4.13opt on same K6-2 RH7.3
Still working on building my own version for the RH7.3

---

> AFACT, it's Unix's/Linux's/BSD's way of saying "Something went wrong, and I
> don't know who to blame".

SIGSEV is a Segmentation Violation... essentially the software is trying to do things it shouldn't do, like access memory that hasn't been allocated to it. It's equivilent to the "This program has performed an illegal operation and will be terminated" in Win32 land.

> If everyone is having SIG problems, then maybe v4.19 isn't as ready as we
> thought...

Looking here, it looks to me like it's the K6-using folk. The AMD K6 series of CPUs is really less than ideal for doing math-intensive work like SETI@Home or any BOINC project. The CPU is lacking in several areas... and lacks really good compiler support, as well. Something optimized for the chip isn't really that well optimized for it in reality.

It also doesn't support the full i686 instruction set. I do believe most BOINC projects are assuming their participants are on i686 class CPUs... so many worker applications may produce SIGILL for you.

Another thing to keep in mind... as you optimize the boinc client itself, you're essentially speeding up the benchmarks. The optimization can and will produce a less stable binary (more apt to die with seg faults)... and all it will do for you is have you *claim* (not be granted) more credit for your workunits. It's a tradeoff, stability for a few extra claimed points.

4.19, as copmpiled by Berkeley, is what you should judge to determine 4.19's stability and readiness for prime-time. I like that Ned is providing these optimized clients... it's good to have the options here... but don't judge 4.19's fitness from it. You wouldn't say Ford Mustangs aren't ready for prime-time after a super-beefed-up Mustang burns through its brakes every 10,000 miles, would you? :-)

---

Rob

Looking here, it looks to me like it's the K6-using folk....
... but don't judge 4.19's fitness from it.
Y'know what? You're right. Berkeley should make the "fit-for-running" call and clone CPUs hadn't ocurred to me .o0(Weren't K6s opcode-for-opcode with i686?)

---