charm AT lists.cs.illinois.edu

Subject: Charm++ parallel programming system

List archive

Re: [charm] [ppl] ampi: MPI_Info_dup fails

From: Jim Edwards <jedwards AT ucar.edu>
To: Gengbin Zheng <zhenggb AT gmail.com>
Cc: charm AT cs.uiuc.edu
Subject: Re: [charm] [ppl] ampi: MPI_Info_dup fails
Date: Tue, 12 Jul 2011 15:17:39 -0600
List-archive: <http://lists.cs.uiuc.edu/pipermail/charm>
List-id: CHARM parallel programming system <charm.cs.uiuc.edu>

On Tue, Jul 12, 2011 at 3:05 PM, Gengbin Zheng <zhenggb AT gmail.com> wrote:

Jim,

Is the lower level library (which calls MPI_INFO_* calls) also
compiled with AMPI?

No and it cannot be as it is an MPI-IO interface library - Is there any way I'm going to be able to get away with this?
My IO is called on a subset of the total tasks, my hope was to pin those tasks to specific processors while allowing
those not involved with IO to migrate for load balancing. What is the typical IO scheme for AMPI programs?

[charm] ampi: MPI_Info_dup fails, Jim Edwards, 07/12/2011
- Re: [charm] [ppl] ampi: MPI_Info_dup fails, Gengbin Zheng, 07/12/2011
  - Re: [charm] [ppl] ampi: MPI_Info_dup fails, Jim Edwards, 07/12/2011
    - Re: [charm] [ppl] ampi: MPI_Info_dup fails, Gengbin Zheng, 07/12/2011
      - Re: [charm] [ppl] ampi: MPI_Info_dup fails, Jim Edwards, 07/12/2011
        
        Re: [charm] [ppl] ampi: MPI_Info_dup fails, Jim Edwards, 07/13/2011
        
        Re: [charm] [ppl] ampi: MPI_Info_dup fails, Eduardo, 07/13/2011
        
        Re: [charm] [ppl] ampi: MPI_Info_dup fails, Jim Edwards, 07/13/2011
        Re: [charm] [ppl] ampi: MPI_Info_dup fails, Jim Edwards, 07/13/2011