AboutDownloadsDocumentsForumsWikiIssuesNews
Date:
2009-08-12 22:47
Priority:
3
State:
Open
Submitted by:
Siddharth Srinivasan (sshrinivasan)
Assigned to:
Peter Eastman (peastman)
Resolution:
None
Summary:
Segfault causes CUDA device to stop responding to openMM calls.

Detailed description
Running a large system (180k atoms) causes OpenMM to crash on a S1070 server, and an earlier bug report was filed (https://simtk.org/tracker/?func=detail&atid=435&aid=896&group_id=161). Even after updating the driver and applying the patch Peter mentioned, the problem is not solved, the simulation crashes. This is itself is a problem, but I realize that this is still in active development. Whats more of an issue is that all further OpenMM jobs, however small, refuse to run on this device after the first crash, though the device itself is OK (all CUDA examples can run on this device, but not the OpenMM MD code itself, or even the OpenMM CUDA tests). This requires me to reload the driver module each time to get the OpenMM code working again.

Add A Comment: Notepad

Comments:

Message  ↓
Date: 2010-07-09 17:35
Sender: Peter Eastman

Could you check whether this still happens with OpenMM 2.0 and CUDA 3.1?

Attached Files:

Changes

Field Old Value Date By
assigned_tonone2009-08-12 23:53sshrinivasan
Feedback