Re: Replica exchange problem

From: Josh Vermaas (joshua.vermaas_at_gmail.com)
Date: Wed Sep 02 2020 - 17:10:12 CDT

Hi Julian,

Don't use a multicore build of NAMD for replica exchange, as that isn't
supported. The netlrts build will be fine for this purpose, and you'd call
it with charmrun. On a single node, it'd be something like this:

charmrun ++local +p 12 +ppn 1 namd2 configfile.namd +replicas 12 +stdout
output_site/%d/job0.%d.log

That will buy you a 12 replicas, each with a single processor assigned to
their computation. This WILL run slowly, just FYI.

-Josh

On Wed, Sep 2, 2020 at 12:16 PM Julian David Baquero Contreras <
jdbaqueroc_at_unal.edu.co> wrote:

> Hi,
>
> I have been trying to perform a protein-ligand calculation using replica
> exchange, since I did the system on the CHARM-GUI web. But I get the
> following error:
> Charm ++: standalone mode (not using charmrun)
> Charm ++> Running in Multicore mode: 12 threads (PEs)
> ------- Partition 0 Processor 0 Exiting: Called CmiAbort ------
> Reason: + partitions other than 1 is not allowed for multicore build
>
> [0] Stack Traceback:
> [0: 0] namd2 0x150e6f7
> [0: 1] namd2 0x516dd3
> [0: 2] namd2 0x50c7b2
> [0: 3] libc.so.6 0x7fbfa0df909b __libc_start_main
> [0: 4] namd2 0x411935
> Segment violation
> I am doing this from a single node since I have the NAMD version
> 2.14_Linux-x86_64-multicore. The command I use is the following:
> namd2 + p12 + replicas 12 fep_site.conf --source FEP_remd_softcore.namd +
> stdout output_site /% d / job0.% d.log
> If I add the charmrun command the exact same thing happens how could I
> perform this calculation using this single node?
>
> Thank you very much, I look forward to a prompt reply.
>
> --
> Julián David Baquero Contreras.
> Químico.
> Universidad Nacional de Colombia.
> Bogotá, Colombia.
>

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:09 CST