From: Sriram Adhyapak (sriram_at_multicorewareinc.com)
Date: Mon Jul 27 2015 - 20:36:14 CDT
The tests "alanin" and "tclforces" fail when using the pre-built binary
"NAMD_2.10_Linux-x86_64-multicore-CUDA".
Any build without CUDA (for e.g. "NAMD_2.10_Linux-x86_64-multicore" does
not fail, for the same INPUT files).
I have included the INPUT config files for the 2 test cases (as attachment).
For example, if my directory structure is
-+NAMD_TEST
---+ NAMD_2.10_Linux-x86_64-multicore-CUDA
---+ test
------+ alanin
---------+ alanin.pdb
---------+ alanin.params
---------+ alanin.psf
---------+ alanin.namd
(similar for tclforces)
The command I use to run the tests are ...
cd NAMD_TEST/NAMD_2.10_Linux-x86_64-multicore-CUDA
./charmrun +p1 ./namd2 ../test/alanin/alanin.namd
The error I get is ...
..
..
ERROR: Atom 52 velocity is 71020.6 27433.4 53425.4 (limit is 12000, atom 51
of 66 on patch 0 pe 0)
ERROR: Atom 53 velocity is -998786 -330130 -833407 (limit is 12000, atom 52
of 66 on patch 0 pe 0)
ERROR: Atom 58 velocity is 2730.34 -13357.2 -11486.8 (limit is 12000, atom
57 of 66 on patch 0 pe 0)
ERROR: Atom 60 velocity is -3275.25 14454 12426.7 (limit is 12000, atom 59
of 66 on patch 0 pe 0)
ERROR: Atoms moving too fast; simulation has become unstable (25 atoms on
patch 0 pe 0).
=======================================
Any help to solve this issue is greatly appreciated!!
---Sriram Adhyapak
This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:22:00 CST