From: David Hardy (dhardy_at_ks.uiuc.edu)
Date: Tue May 30 2017 - 11:11:07 CDT
Unfortunately we didn't find this bug until after the release of 2.12. The startup code for the new GPU kernels prevented MSM from being properly initialized. We have fixed the issue, but for now it is available only as source or in the nightly builds.
You should be able to use your 2.12 version of NAMD with MSM together with the old GPU kernels: in the config file set useCUDA2=no .
-- David J. Hardy, Ph.D. Theoretical and Computational Biophysics Beckman Institute, University of Illinois dhardy_at_ks.uiuc.edu http://www.ks.uiuc.edu/~dhardy/ > On May 30, 2017, at 10:12 AM, Shadrina Maria <shadrinams_at_gmail.com> wrote: > > Hello, > > I use NAMD_2.12_Win64-multicore-CUDA to run some simulations on my local computer. > Usually, it works perfectly well. But when I set up Multilevel Summation Method (MSM), NAMD crashes without an obvious reason. The last step in the output file is "TCL: Minimizing for 18000 steps" with no warnings or errors. > > When I set MSM off or replace it by PME, NAMD runs without crashing. The same job with MSM runs well on NAMD 2.12b1 for Linux-x86_64-MPI. > I noticed that the output file from the Windows machine misses all information about MSM (between Startup phase 6 and Startup phase 7) that the output from the Limux machine have. > > I also tried to run the corresponding tutorial (http://www.ks.uiuc.edu/Training/Tutorials/namd/namd-tutorial-unix-html/node8.html <http://www.ks.uiuc.edu/Training/Tutorials/namd/namd-tutorial-unix-html/node8.html>) and got the same problem on my Windows computer. > > Does anybody have an idea what can be wrong or what should I check to solve the problem? > > Thank you in advance. > > Best regards, > Maria. > >
This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:21:19 CST