Hello

I am getting this error on my two Linux64 Ubuntu boxes.

<core_client_version>6.4.5</core_client_version>
<![CDATA[
<message>
process exited with code 193 (0xc1, -63)
</message>
<stderr_txt>
boinc initialized
work files resolved, now working
-> lasievee_1.07_x86_64-pc-linux-gnu
-> -r
-> -f
-> 142230000
-> -c
-> 2000
-> -R
-> ../../projects/escatter11.fullerton.edu_nfs/S2p859.poly
-> -o
-> ../../projects/escatter11.fullerton.edu_nfs/S2p859_142230_0_0
SIGILL: illegal instruction
Stack trace (3 frames):
[0x43f78d]
[0x44e610]
[0x425152]

Exiting...

</stderr_txt>
]]>

One with BOINC 6.4.5 (Linux
2.6.31-11-generic)and the other with BOINC 6.6.36(Linux
2.6.28-15-generic).
I tried to solve this problem with a new install of Linux with no success.
Is there any solution for me or is this an app error?

Thank you for any help.
LB

SIGILL: illegal instruction

Do these computers have the very first generation Intel x86_64 processor? In their first generation processor, Intel left out an AMD64 prefetch instruction that the 64-bit optimized assembly uses. Intel of course included this instruction in all later versions of their x86_64 processors. Unfortunately, the removal of this instruction slows the code on all Opteron/Phenom systems while affecting very few Intel systems, so I really don't want to remove it. If Linux is set up to allow you to run 32-bit programs, then you can run the Linux 32-bit version of BOINC for the NFS@Home project.

Ahh,thanks.
Yes,these are Xeon Irwidale,one of the first 64bit series.
O.K. I will stick with XP64.
I wanted to run it for speed advantage.

Thank you.
LB

Greg, Do I read this that it's an unsupported op code in the CPU itself, so situation will exist with any 64bit OS because either 64b app uses the pre-fetch (Win or Linux)? Just hoping to get this clear in my head. Thanx.

---

- da shu @ HeliOS,
"A child's exposure to technology should never be predicated on an ability to afford it."

Ahh.. this explains why my two Pentium D dual-core machines finish all NFS work units immediately with a computation error. Luckily, it's only two machines out of a fairly large cruncher farm. Thanks for the info!

Yes, that's correct. It's an AMD64 op code that Intel did not implement in EM64T in their earliest 64-bit processors, so it is independent of OS if using 64-bit code. I believe it only affects the 64-bit capable 90nm Pentium 4 and Xeon Nocona and Irwindale. Later Intel processors implement the instruction.

However, there is no Windows x64 NFS@Home client, so Windows x64 BOINC will download the 32-bit Windows client, and it will run without problems.

Also, if you wanted to go to the trouble, you can install the 32-bit Linux version of BOINC, and use the Linux 32-bit client.

That makes sense, then... the two Pentium D machines are early 64-bit 90nm procs.. probably just two P4's in one can. They're not great performers, and I've got the project running on several other machines, so this is no big deal. I'm relieved to know that the problem isn't the procs actually going bad.
:)