Re: Multicore-CUDA NAMD output hangs

From: Mitchell Gleed (
Date: Tue Mar 03 2015 - 10:34:20 CST

I thought I would update this thread with a couple more things we've tried
to no avail:
- changing -O2 to -O1 in the Linux-x86_64-icc.arch file and recompiling
- using a communications thread with the ++commthread option and using 23
cpus instead of 24

Also, non-CUDA multicore builds run without hanging on these nodes.

Could this be some sort of race condition? I haven't been able to find
these options, but is there a way to build namd with debugging options or
to increase verbosity of output? Any ideas about resolving the issue, ideas
for other troubleshooting steps, or other relevant feedback would be
greatly appreciated.

This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:20:56 CST