Is it possible to run replica exchange simulation on BlueGeneL?

From: Dong Luo (us917_at_yahoo.com)
Date: Fri Feb 12 2010 - 11:47:21 CST

We want to use T-REMD to sample the conformation of systems with large size (90k-180k atoms). I checked the replica exchange tcl files come with NAMD. It looks to me BlueGeneL is not supported straitforward. Basically, the tcl script need figure out the node list and feed it to NAMD. But on BlueGeneL, I keep get this kind of error:
no such variable
    (read trace on "env(HOST_FILE)")

or

no such variable
    (read trace on "env(LSB_HOSTS)")

So either both HOST_FILE and LSB_HOSTS are not run time einviroments variables on this platform or the format is not correct. I can't figure out the correct variable and don't know if such variable exist or not. Normally the BlueGeneL is using mpirun to invoke namd2 and they seem to handle the node list automatically without extra scripts. Now for replica exchange method, how can I resolve this?

I also need suggestions on the simulation itself. Does it make sense to apply T-REMD on our large systems? Can the conformations be sampled efficiently this way?

Thanks.

Dong

      

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:46 CST