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] Calling MPI_Test() too many times results in a time spike
From: Jeff Squyres (jsquyres_at_[hidden])
Date: 2010-12-15 21:13:39


On Dec 15, 2010, at 9:10 PM, Ioannis Papadopoulos wrote:

> I agree that MPI_Test() has to do some progress, but as you can see I am only sending one message and I busy wait on it - since there is nothing else to do and no other incoming traffic, I would expect no difference among MPI_Test() calls, apart from the last one (the one that will notify me that my message arrived).

There's no other incoming traffic *for MPI*. But we're trapping down into the kernel (albeit in a non-blocking manner) and therefore the OS may stall us for longer than expected due to OS issues.

Have you eliminated all other sources of jitter? Are you binding to specific cores? Have you turned off all unused daemons? Do you see the same behavior with shared memory transport (which should make jitter issues more pronounced)? ...etc.

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