Re: non-parallel NAMD compilation

From: Gengbin Zheng (gzheng_at_ks.uiuc.edu)
Date: Mon Mar 19 2007 - 13:59:49 CDT

Hi,

 There are many ways to compile NAMD. so you don't have to compile MPI
version of Charm++/NAMD, and bypassing MPI_Init.
 For example, you can compile UDP (net version) of Charm++, which uses
standard UDP sockets. Having that compiled, you can run NAMD in a
standalone fashion just like sequential:
  ./namd2 ./config

Gengbin

Xiuping Tao wrote:

> Hello Dr. Zheng,
>
> I have a question to consult with you:
>
> While NAMD is good as a parallel code, I need non-parallel NAMD. I
> want to wrap
> NAMD's main( ) function (renamed as NAMD_main( ) ) and call it from
> my main( )
> which uses MPI_Init( ) and does the call from its slave processors.
> Therefore I want
> MPI_Init( ) not used inside NAMD again.
>
> Another reason for doing this is I need to call NAMD_main( ) many
> times from any
> slave processor. Between the calls I need to do data analysis. If
> MPI_Init( ) is used
> inside NAMD, I can only call NAMD_main( ) once.
>
> My question is: how to bypass MPI_Init( ) NAMD and compile it on a
> Linux amd64
> cluster?
>
> Thank you in advance for your assistance,
>
> Xiuping Tao
>
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:46:06 CST