UNSTABLE_MACHINE and libcudart.so.2: wrong ELF class: ELFCLA

Moderators: slegrand, Site Moderators, PandeGroup

UNSTABLE_MACHINE and libcudart.so.2: wrong ELF class: ELFCLA

Postby anlayne » Tue Dec 22, 2009 9:55 am

I'm trying to get my 9600GT folding under Ubuntu 9.04. I installed the drivers and the wrapper as in the wiki, but I keep getting an exception thrown during GuardedRun which results in an UNSTABLE_MACHINE core shutdown.

Code: Select all
[09:42:28] Loaded queue successfully.
[09:42:28]
[09:42:28] + Processing work unit
[09:42:28] Core required: FahCore_11.exe
[09:42:28] Core found.
[09:42:28] Working on queue slot 09 [December 22 09:42:28 UTC]
[09:42:28] + Working ...
[09:42:28] - Calling '.\FahCore_11.exe -dir work/ -suffix 09 -checkpoint 3 -verbose -lifeline 8 -version 623'

[09:42:28] - Autosending finished units... [December 22 09:42:28 UTC]
[09:42:28] Trying to send all finished work units
[09:42:28] + No unsent completed units remaining.
[09:42:28] - Autosend completed
[09:42:29]
[09:42:29] *------------------------------*
[09:42:29] Folding@Home GPU Core
[09:42:29] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[09:42:29]
[09:42:29] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86
[09:42:29] Build host: amoeba
[09:42:29] Board Type: Nvidia
[09:42:29] Core      :
[09:42:29] Preparing to commence simulation
[09:42:29] - Looking at optimizations...
[09:42:29] DeleteFrameFiles: successfully deleted file=work/wudata_09.ckp
[09:42:29] - Created dyn
[09:42:29] - Files status OK
[09:42:29] Error: Missing work file=<>
[09:42:29]
[09:42:29] Folding@home Core Shutdown: MISSING_WORK_FILES
[09:42:33] CoreStatus = 74 (116)
[09:42:33] The core could not find the work files specified. Removing from queue
[09:42:33] Deleting current work unit & continuing...
[09:42:37] Trying to send all finished work units
[09:42:37] + No unsent completed units remaining.
[09:42:37] - Preparing to get new work unit...
[09:42:37] + Attempting to get work packet
[09:42:37] - Will indicate memory of 3895 MB
[09:42:37] - Detect CPU. Vendor: GenuineIntel, Family: 6, Model: 7, Stepping: 10
[09:42:37] - Connecting to assignment server
[09:42:37] Connecting to http://assign-GPU.stanford.edu:8080/
[09:42:37] Posted data.
[09:42:37] Initial: 43AB; - Successful: assigned to (171.67.108.11).
[09:42:37] + News From Folding@Home: Welcome to Folding@Home
[09:42:37] Loaded queue successfully.
[09:42:37] Connecting to http://171.67.108.11:8080/
[09:42:37] Posted data.
[09:42:37] Initial: 0000; - Receiving payload (expected size: 45903)
[09:42:38] - Downloaded at ~44 kB/s
[09:42:38] - Averaged speed for that direction ~69 kB/s
[09:42:38] + Received work.
[09:42:38] + Closed connections
[09:42:43]
[09:42:43] + Processing work unit
[09:42:43] Core required: FahCore_11.exe
[09:42:43] Core found.
[09:42:43] Working on queue slot 00 [December 22 09:42:43 UTC]
[09:42:43] + Working ...
[09:42:43] - Calling '.\FahCore_11.exe -dir work/ -suffix 00 -checkpoint 3 -verbose -lifeline 8 -version 623'

[09:42:43]
[09:42:43] *------------------------------*
[09:42:43] Folding@Home GPU Core
[09:42:43] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[09:42:43]
[09:42:43] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86
[09:42:43] Build host: amoeba
[09:42:43] Board Type: Nvidia
[09:42:43] Core      :
[09:42:43] Preparing to commence simulation
[09:42:43] - Looking at optimizations...
[09:42:43] DeleteFrameFiles: successfully deleted file=work/wudata_00.ckp
[09:42:43] - Created dyn
[09:42:43] - Files status OK
[09:42:43] - Expanded 45391 -> 251112 (decompressed 553.2 percent)
[09:42:43] Called DecompressByteArray: compressed_data_size=45391 data_size=251112, decompressed_data_size=251112 diff=0
[09:42:43] - Digital signature verified
[09:42:43]
[09:42:43] Project: 5771 (Run 10, Clone 82, Gen 1085)
[09:42:43]
[09:42:43] Assembly optimizations on if available.
[09:42:43] Entering M.D.
[09:42:49] Tpr hash work/wudata_00.tpr:  295002744 4263430377 2446665794 549264233 2904704546
[09:42:49]
[09:42:49] Calling fah_main args: 14 usage=100
[09:42:49]
Reading file work/wudata_00.tpr, VERSION 3.1.4 (single precision)
Reading file work/wudata_00.tpr, VERSION 3.1.4 (single precision)
Reading sasa-enabled ir 0 0
Initializing Nvidia gpu library
Run: exception thrown during GuardedRun
[09:42:49] Run: exception thrown during GuardedRun
[09:42:49] Run: exception thrown in GuardedRun -- Gromacs cannot continue further.
[09:42:49] Going to send back what have done -- stepsTotalG=0
[09:42:49] Work fraction=0.0000 steps=0.
[09:42:53] logfile size=4945 infoLength=4945 edr=0 trr=23
[09:42:53] + Opened results file
[09:42:53] - Writing 5481 bytes of core data to disk...
[09:42:53] Done: 4969 -> 1863 (compressed to 37.4 percent)
[09:42:53]   ... Done.
[09:42:53] DeleteFrameFiles: successfully deleted file=work/wudata_00.ckp
[09:42:53]
[09:42:53] Folding@home Core Shutdown: UNSTABLE_MACHINE


Looking around this forum it seems that UNSTABLE_MACHINE might indicate bad hardware, so I downloaded Memtest80. Unfortunately, when I run it I get

Code: Select all
./memtestG80: error while loading shared libraries: libcudart.so.2: wrong ELF class: ELFCLASS32


I thought this is the place to ask because it mentions libcudart.so.2. I'm using the 190.18 nvidia driver and 2.3 cuda toolkit. Thanks for the help.
anlayne
 
Posts: 2
Joined: Tue Dec 22, 2009 9:42 am

Re: UNSTABLE_MACHINE and libcudart.so.2: wrong ELF class: ELFCLA

Postby biodoc » Tue Dec 22, 2009 10:36 pm

Known issue. You need to install the 195.17 drivers (cuda 3.0). http://forums.nvidia.com/index.php?showtopic=149959

Keep the cuda 2.3 toolkit though! Check out the cuda 3.0 thread in this forum for more info. :)
User avatar
biodoc
 
Posts: 48
Joined: Sun Jan 06, 2008 10:15 am

Re: UNSTABLE_MACHINE and libcudart.so.2: wrong ELF class: ELFCLA

Postby ihaque » Wed Dec 23, 2009 1:24 am

It sounds like you downloaded the 64-bit Linux version of MemtestG80, but have the 32-bit CUDA drivers installed. Can you try with the 32-bit build?
User avatar
ihaque
Pande Group Member
 
Posts: 239
Joined: Mon Dec 03, 2007 4:20 am
Location: Stanford

Re: UNSTABLE_MACHINE and libcudart.so.2: wrong ELF class: ELFCLA

Postby anlayne » Wed Dec 23, 2009 2:07 am

Actually, biodoc's suggestion worked, so thanks to biodoc. Some time later, I will probably try the 32-bit MemtestG80. I downloaded the 64-bit because that is what my os is, but I didn't think that the bits of the cuda toolkit mattered. I guess I should have thought of that. Thanks.
anlayne
 
Posts: 2
Joined: Tue Dec 22, 2009 9:42 am


Return to unOfficial Linux GPU (WINE wrapper) (3rd party support)

Who is online

Users browsing this forum: No registered users and 1 guest