Open MPI logo

Open MPI Development 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 devel] Question about barrier()
From: Jeff Squyres (jsquyres_at_[hidden])
Date: 2010-11-02 11:54:43


On Nov 2, 2010, at 10:11 AM, <ananda.mudar_at_[hidden]> <ananda.mudar_at_[hidden]> wrote:

> Yes, I am calling MPI_Init_thread() with MPI_THREAD_MULTIPLE. And I have enabled thread and mpi_threads while configuring OpenMPI. BTW, I am using OpenMPI 1.4.2.
>
> It will be helpful to know if this is a design constraint or some implementation defect.

When you use MPI_THREAD_MULTIPLE, we're activating locks that aren't necessary when you're MPI_THREAD_SINGLE. These locks therefore cost overhead time that you don't normally see (and potentially system time as opposed to user time). To date, we haven't spent a huge amount of time optimizing the THREAD_MULTIPLE code paths. There's a number of efforts going on that may help with this, but it will take some time.

-- 
Jeff Squyres
jsquyres_at_[hidden]
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/