Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |  

This web mail archive is frozen.

This page is part of a frozen web archive of this mailing list.

You can still navigate around this archive, but know that no new mails have been added to it since July of 2016.

Click here to be taken to the new web archives of this list; it includes all the mails that are in this frozen archive plus all new mails that have been sent to the list since it was migrated to the new archives.

Subject: Re: [OMPI users] Roadrunner blasts past the petaflop mark with Open MPI
From: Jeff Squyres (jsquyres_at_[hidden])
Date: 2008-06-17 10:26:52


On Jun 17, 2008, at 10:18 AM, Brad Benton wrote:

> The main run-time parameters used are as follows:
>
> - For scalable process launch
> routed = binomial

Note that this is a new parameter for OMPI v1.3 -- it has to do with
how non-MPI communications are sent around (e.g., OMPI startup
coordination). More documentation will be coming along on the new
ORTE-level controls for v1.3.

> - To reduce the overhead for IB connections:
> btl_openib_receive_queues = P,128,64,32,32,32:S,2048,1024,128,32:S,
> 12288,1024,128,32:S,65536,1024,128,32

This is also a new parameter for v1.3 -- see the "sneak peak" v1.3
screencast in the videos section (in the OpenFabrics subsection) on
the OMPI web site for a description of what it is/does.

> - Bind processes to computational cores:
> opal_paffinity_alone = 1
>
> - Leave buffers pinned and registered
> mpi_leave_pinned = 1

These params exist in v1.2.

-- 
Jeff Squyres
Cisco Systems