From: Norman Geist (norman.geist_at_uni-greifswald.de)
Date: Thu Oct 04 2012 - 00:51:13 CDT
The good thing is, Iâ€™m also using a machine with Ubuntu and two Tesla C2075, and it works like a charm. So it must be something regarding your setup.
1st of all you should make sure that your namd is using the libcudart shipped with it. Therefore try ldd namd2 to check which one it is dynamically linking.
2nd, if this doesnâ€™t point something out, check if CUDA4.2 isnâ€™t too much for the current namd version, means Iâ€™m not quite sure. And if your driver is installed properly. Also check the output of nvidia-smi.
And finally, what you should try in all cases: Is some other CUDA binary (maybe a test prog from the SDK) running on your GPUs ??
Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im Auftrag von ???
Gesendet: Mittwoch, 3. Oktober 2012 05:34
Betreff: namd-l: Cuda errors in Ubuntu12amd64
Namd is running in Ubuntu12amd64 with cuda4.2,tesla c2075.
When I run the command
./../../charmrun ++local +p 8 ../../../namd2 +idlepoll +devices 0 x.conf | tee md.log
the errors are below:
OPENING EXTENDED SYSTEM TRAJECTORY FILE
MINIMIZER SLOWLY MOVING 39025 ATOMS WITH BAD CONTACTS DOWNHILL
FATAL ERROR: CUDA error in cuda_check_remote_progress on Pe 4 (ly-Precision-WorkStation-T5500 device 0): unspecified launch failure
I know someone mailed this problem, but how to solve it?
This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:22:39 CST