Gamma ray/binary pulsar search "not available for your type of computer"

Cory Van Pelt
Cory Van Pelt
Joined: 25 Feb 05
Posts: 4
Credit: 1378687
RAC: 0
Topic 196207

I searched for these messages on Google and the forums and didn't see anything. I just upgraded my Linux machine to Fedora 15 and installed BOINC. I also installed a GEForce GTX 460 video card.

When I attached to Einstein@home, I got messages saying "Gamma-ray pulsar search #1 is not available for your type of computer" and the same thing for binary radio pulsar search. It's the exact same computer I've had (that was running these just fine for 2 years) but now has a video card and an upgraded operating system. It's now only running gravitational wave search. Any ideas what this means? Thanks.

Jord
Joined: 26 Jan 05
Posts: 2952
Credit: 5893653
RAC: 1

Gamma ray/binary pulsar search "not available for your type of c

When looking at your log messages, here, I see that it was a work request for your Nvidia GPU.

It tells you that there is "no app version available: APP#16 (einstein_S6Bucket) PLATFORM#7 (x86_64-pc-linux-gnu)", "no app version available: APP#17 (hsgamma_FGRP1) PLATFORM#7 (x86_64-pc-linux-gnu)" and "no app version available: APP#19 (einsteinbinary_BRP4) PLATFORM#7 (x86_64-pc-linux-gnu)".

Which is correct, as all the applications here at Einstein are still 32bit only. So you must make your BOINC either a 32bit version, or make your OS be capable of using 32bit libraries. Not all Linux distros can do this automatically. So, see http://boinc.berkeley.edu/wiki/Installing_BOINC#64_Bit_Considerations for more information on that.

Afterwards you may want to restart the computer, or at least restart BOINC.

Cory Van Pelt
Cory Van Pelt
Joined: 25 Feb 05
Posts: 4
Credit: 1378687
RAC: 0

Thanks for the tips! I

Thanks for the tips! I followed those instructions and it has stopped giving me that error message and downloaded some CUDA apps, but is now giving me instant computation errors on them. I don't know if it's related to the same problem so I'm going to search around.

Quote:

When looking at your log messages, here, I see that it was a work request for your Nvidia GPU.

It tells you that there is "no app version available: APP#16 (einstein_S6Bucket) PLATFORM#7 (x86_64-pc-linux-gnu)", "no app version available: APP#17 (hsgamma_FGRP1) PLATFORM#7 (x86_64-pc-linux-gnu)" and "no app version available: APP#19 (einsteinbinary_BRP4) PLATFORM#7 (x86_64-pc-linux-gnu)".

Which is correct, as all the applications here at Einstein are still 32bit only. So you must make your BOINC either a 32bit version, or make your OS be capable of using 32bit libraries. Not all Linux distros can do this automatically. So, see http://boinc.berkeley.edu/wiki/Installing_BOINC#64_Bit_Considerations for more information on that.

Afterwards you may want to restart the computer, or at least restart BOINC.

Jord
Joined: 26 Jan 05
Posts: 2952
Credit: 5893653
RAC: 1

Looks like you're missing a

Looks like you're missing a library file, or your BOINC can't find where it is installed:
../../projects/einstein.phys.uwm.edu/einsteinbinary_BRP4_1.23_i686-pc-linux-gnu__BRP4cuda32nv270: error while loading shared libraries: libcuda.so.1: cannot open shared object file: No such file or directory

Looks to me like it's a part pf your Nvidia videocard drivers.

I also see in the scheduler log "SCHEDULER_REQUEST::parse(): unrecognized: 0". Are you running multiple (BOINC) clients at the same time, and is this from a second version that may not know where you installed the videocard drivers to?

Cory Van Pelt
Cory Van Pelt
Joined: 25 Feb 05
Posts: 4
Credit: 1378687
RAC: 0

I'm only running one instance

I'm only running one instance of BOINC. It seems that Einstein@home is the only one having the problem, since every other GPU-capable project is crunching on the GPU just fine.

I did a YUM search for libcuda.so.1 and it returned some packages that I had already tried installing before (which caused a bunch of problems, including "no GPU" errors for all projects in BOINC).

Quote:

Looks like you're missing a library file, or your BOINC can't find where it is installed:
../../projects/einstein.phys.uwm.edu/einsteinbinary_BRP4_1.23_i686-pc-linux-gnu__BRP4cuda32nv270: error while loading shared libraries: libcuda.so.1: cannot open shared object file: No such file or directory

Looks to me like it's a part pf your Nvidia videocard drivers.

I also see in the scheduler log "SCHEDULER_REQUEST::parse(): unrecognized: 0". Are you running multiple (BOINC) clients at the same time, and is this from a second version that may not know where you installed the videocard drivers to?


Jord
Joined: 26 Jan 05
Posts: 2952
Credit: 5893653
RAC: 1

My help is exhausted here.

My help is exhausted here. Sorry. I don't know any further.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 779908412
RAC: 1191986

Hi! on the Linux console,

Hi!

on the Linux console, you can do this :

 whereis libcuda

Which on my system returns

libcuda: /usr/lib/libcuda.so /usr/lib64/libcuda.so 

whcih means that both the 32 bit version (in /usr/lib) and the 64 bit version (in /usr/lib64) are installed, which is the way you'd want it.

In those directories, there should be "aliases" or links which are named "libcuda.so.1" :

 whereis libcuda.so.1       libcuda.so: /usr/lib/libcuda.so.1 /usr/lib/libcuda.so /usr/lib64/libcuda.so.1 /usr/lib64/libcuda.so

also a listing of all the libcuda*so* files might help:

 ls -la /usr/lib/libcuda*so*
lrwxrwxrwx 1 root root      12  3. Aug 2011  /usr/lib/libcuda.so -> libcuda.so.1
lrwxrwxrwx 1 root root      17  3. Aug 2011  /usr/lib/libcuda.so.1 -> libcuda.so.280.13
-rwxr-xr-x 1 root root 8833884  3. Aug 2011  /usr/lib/libcuda.so.280.13



The actual library is the one with the driver version number (here 280.13, quite old, your's might be newer).

Let us know what the situation is on your system. Libcuda is supplied with the NVIDIA driver, not with BOINC or the einstein apps. If you are missing the 32 bit version of libcuda, you need to reinstall the NVIDIA driver package and make sure the 32 bit compatibility libs get installed as well.

CU
HB

Cory Van Pelt
Cory Van Pelt
Joined: 25 Feb 05
Posts: 4
Credit: 1378687
RAC: 0

Thanks for the answer! That

Thanks for the answer! That looks like my problem... a "whereis libcuda" returns nothing at all (I get "libcuda:") I've installed the latest drivers on my Fedora 15 system using the akmod-nvidia instructions I found online. That lets me run Milkyway@home and GPUgrid using the GPU; Einstein is the only one not working.

This may not be the appropriate place for it, but can you tell me how you got your libcuda libraries installed? I don't understand why they wouldn't come with the package that I installed to get the NVIDIA drivers (by the way, 280.13 is the version number installed on mine today). Thanks for your help!

Quote:

Hi!

on the Linux console, you can do this :

 whereis libcuda

Which on my system returns

libcuda: /usr/lib/libcuda.so /usr/lib64/libcuda.so 

whcih means that both the 32 bit version (in /usr/lib) and the 64 bit version (in /usr/lib64) are installed, which is the way you'd want it.

In those directories, there should be "aliases" or links which are named "libcuda.so.1" :

 whereis libcuda.so.1       libcuda.so: /usr/lib/libcuda.so.1 /usr/lib/libcuda.so /usr/lib64/libcuda.so.1 /usr/lib64/libcuda.so

also a listing of all the libcuda*so* files might help:

 ls -la /usr/lib/libcuda*so*
lrwxrwxrwx 1 root root      12  3. Aug 2011  /usr/lib/libcuda.so -> libcuda.so.1
lrwxrwxrwx 1 root root      17  3. Aug 2011  /usr/lib/libcuda.so.1 -> libcuda.so.280.13
-rwxr-xr-x 1 root root 8833884  3. Aug 2011  /usr/lib/libcuda.so.280.13


The actual library is the one with the driver version number (here 280.13, quite old, your's might be newer).

Let us know what the situation is on your system. Libcuda is supplied with the NVIDIA driver, not with BOINC or the einstein apps. If you are missing the 32 bit version of libcuda, you need to reinstall the NVIDIA driver package and make sure the 32 bit compatibility libs get installed as well.

CU
HB

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 779908412
RAC: 1191986

Hi! RE: but can you

Hi!

Quote:

but can you tell me how you got your libcuda libraries installed?

I downloaded my cuda drivers from NVIDIA site, not from the Linux distribution repositories. That requires compilation of a kernel module on driver installation tho (fully automatic by the installer, but requires other packages you might need to load first...

If you try this, stay away from version 295.x as this one has a bug that can crash apps when the gfx card powers down the monitor for energy saving.

HBE

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.