Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: [OMPI users] LS-DYNA profiling [was: OpenMPI Hangs, No Error]
From: Eugene Loh (eugene.loh_at_[hidden])
Date: 2010-07-14 12:16:21


I took the liberty of changing the subject line.

Yes, MPI_Barrier waits until all other processes in the communicator catch up.  So, long barrier time usually indicates there is some "load imbalance"... one or more processes reach the synchronization point well before the others.  Other communications might be instantaneous.  E.g., a "send" can take a short time, but the corresponding receive might be long (if the receiver was blocking on the message well before it was sent) and associated barriers might also show long waits.

But, there is presumably a lot more information available at your finger tips.  Try clicking on the MPI Timeline tab to see a timeline of what happened.  You can drag the mouse horizontally to zoom in on a time interval.  You should be able to see the load imbalances that are causing your long(ish) barrier times.

Or, you stay on the "MPI Chart" tab where you were already.  Drag vertically on the "MPI_Barrier" bar and click on the "Filter" glyph.  This will filter out all data except for MPI_Barrier calls.  Then, go to that timeline and the Barrier calls should stick out.  Or, stick with the MPI Chart and change the y axis to be "process" (or rank or something, I forget offhand what it's called).  Since you've filtered out everything but barrier, it should show you time spent per process but only for barrier calls.  This will show you if there is an overall imbalance in who is spending barrier time.  The processes with the most barrier time are those reaching the barriers first... the processes with least computational work (or in any case fastest for some other reason).  The processese with the least barrier time are the problem:  they're the processes holding everyone else up.

On the MPI Timeline tab, there are also message lines.  They'll help you understand the relationship between fast MPI_Sends and long MPI_Recvs... presumably, one process waits on another.  The message gets sent quickly, but it's sent "late".

I don't know LS-DYNA, so I don't know what opportunities you'll have to, for example, balance loads better for better performance.  But, synchronization delays in the execution due to load imbalances may be more of a problem than the performance of the data transfers.


, you can go to the right-hand panel and construct other charts.  E.g., "Functions" "2d Chart" X: "

Robert Walters wrote:
Also, I finally got some graphical output from Sun Studio Analyzer.

I see MPI_Recv and MPI_Wait taking a lot of time, but I would think that is ok, this program does heavy number crunching and I would expect it to need to Wait or wait to Receive very often since there is a decent amount of time between communications. Is this the correct assumption?

What does catch my eye is MPI_Barrier takes up a significant chunk of around 10%. I read that MPI_Barrier blocks the caller until all processes have called? Perhaps there is something fishy there that it is taking an awful long time for processes to call each other although MPI_Send is not taking very long so it makes me feel more comfortable about network communication.

Anyways, please have a look and let me know what you think could be the issue.

Regards,
Robert Walters

--- On Tue, 7/13/10, Robert Walters <raw19896@yahoo.com> wrote:

From: Robert Walters <raw19896@yahoo.com>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error
To: "Open MPI Users" <users@open-mpi.org>
Date: Tuesday, July 13, 2010, 10:42 PM

Naturally, a forgotten attachment.

An to edit that, it was compiled to be used with OpenMPI 1.4.1, but as I understand, 1.4.2 is just a bug fix of 1.4.1.

--- On Tue, 7/13/10, Robert Walters <raw19896@yahoo.com> wrote:

From: Robert Walters <raw19896@yahoo.com>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error
To: "Open MPI Users" <users@open-mpi.org>
Date: Tuesday, July 13, 2010, 10:38 PM

I think I forgot to mention earlier that the application I am using is pre-compiled. It is a finite element software called LS-DYNA. It is not open source and I likely cannot obtain the code it uses for MPP. This version I am using was specifically compiled, by the parent company, for OpenMPI 1.4.2 MPP operations.

I recently installed the Sun Studio 12.1 to attempt to analyze the situation. It seems to work partially. It will record various processes individually, which is cryptic. The function it fails on, though, is the MPI Tracing. It errors that "no MPI tracing data file in experiment, MPI Timeline and MPI Charts will not be available". Sometime during the analysis (about 10,000 iterations later, the VT_MAX_FLUSHES complains that there are too many i/o flushes and its not happy. I've increased this number in the environmental variable and killed the analysis before it had a chance to error but still no MPI Trace data is recorded. Not sure if you guys have heard of that happening or know any way to fix it...Did OpenMPI need to be configured/built for Sun Studio use?

I also noticed that from the data I do get back, there are two sets of functions for everything. There is mpi_recv and then my_recv_, both with the same % utilization time. The mpi one comes from your program's library and the my_recv_ one comes from my program. Is that typical or should the program I'm using be saying mpi_recv only? This data may be enough to help me see what's wrong so I will pass it along. Keep in mind this is percent time of total run time and not percent of MPI communication. I attached the information in a picture rather than me attempting to format a nice table in this nasty e-mail application.
I blacked out items that are related to LS-DYNA but afterward I just realized that I think every function with an _ at the end represents a command issuing from LS-DYNA.

These are my big spenders. The processes I did not include are in the bottom 4%. The processes that would be above these were the LS-DYNA applications at 100%. Like I mentioned earlier, there are two instances of every MPI command, and they carry the same percent usage. It's curious that this version, built for OpenMPI, uses different functions.

Just for a little more background info, OpenMPI is being launched from a local hard drive on each machine, but the LS-DYNA job files, and related data output files, are on a mounted drive on that machine, where the mounted drive is located on a different machine also in the cluster. We were thinking that might be an issue but it isn't writing enough data for me to think that would significantly decrease MPP performance.

I would like to make one last mention. That is that OpenMPI running 8 cores on a single node, with all the communication, works flawlessly. It works much faster than the Shared Memory Parallel (SMP) version of LS-DYNA that we currently have used scaled to 8 cores. LS-DYNA seems to be approximately 25% faster (don't quote me on that) when using the OpenMPI installation than when using the standard SMP, which is awesome. My point being that OpenMPI seems to be working fine, even with the screwy mounted drive. This leads me to continue to point at the network.

Anyhow, let me know if anything seems weird on the OpenMPI communication subroutines. I don't have any numbers to lean on from experience.

Sorry this e-mail was long. Thank you again for all of your help.

Regards,
Robert Walters

--- On Tue, 7/13/10, David Zhang <solarbikedz@gmail.com> wrote:

From: David Zhang <solarbikedz@gmail.com>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error
To: "Open MPI Users" <users@open-mpi.org>
Date: Tuesday, July 13, 2010, 9:42 PM

Like Ralph says, the slow down may not be coming from the kernel, but rather on waiting for messages.  What MPI send/recv commands are you using?

On Tue, Jul 13, 2010 at 11:53 AM, Ralph Castain <rhc@open-mpi.org> wrote:
I'm afraid that having 2 cores on a single machine will always outperform having 1 core on each machine if any communication is involved.

The most likely thing that is happening is that OMPI is polling waiting for messages to arrive. You might look closer at your code to try and optimize it better so that number-crunching can get more attention.

Others on this list are far more knowledgeable than I am about doing such things, so I'll let them take it from here. Glad it is now running!