Charm++ parallel programming system

["Ortega, Bob" <bobo AT mail.smu.edu>]

Ronak,

 

Thank you for the quick reply.

 

Well, I’m using srun to run NAMD.  Here’s the command,

 

date;time srun -n 36 -N 2 -p fp-gpgpu-3 --mem=36GB ./namd2.prj stmv/stmv.namd >namd2.prj.fp-gpgpu-3.6.log;date

 

How can I submit a similar charmrun command targeting 36 processors, 2 nodes, the fp-gpgpu-3 queue partition, 36GB of memory and +logsize of 10000000?

 

Oh, I’m not getting the exception anymore and unfortunately, during that run, I didn’t log the results to a file.

 

If it occurs again, I’ll forward the log file.

 

Thanks,

Bob

 

From: Ronak Buch <rabuch2 AT illinois.edu>
Date: Thursday, December 10, 2020 at 2:09 PM
To: "Ortega, Bob" <bobo AT mail.smu.edu>
Cc: "charm AT lists.cs.illinois.edu" <charm AT lists.cs.illinois.edu>, Nitin Bhat <nitin AT hpccharm.com>
Subject: Re: [charm] FW: Projections

 

Hi Bob,

 

Regarding the +logsize parameter, it is a runtime parameter, not a compile time parameter, so you shouldn't add it to the Makefile, you should add to your run command (e.g. ./charmrun +p2 ./namd <namd input file name> +logsize 10000000).

 

Regarding the exception you're seeing, I'm not sure why that's happening, it's likely due to some issue in initialization. Would it be possible for you to share the generated logs for debugging?

 

Thanks,

Ronak

 

On Thu, Dec 10, 2020 at 12:36 PM Ortega, Bob <bobo AT mail.smu.edu> wrote:

 

 

From: "Ortega, Bob" <bobo AT mail.smu.edu>
Date: Thursday, December 10, 2020 at 11:24 AM
To: "charm AT lists.cs.illinois.edu" <charm AT lists.cs.illinois.edu>
Cc: Nitin Bhat <nitin AT hpccharm.com>
Subject: FW: Projections

 

Nitin Bhat was kind enough to review my questions about some errors and messages I am receiving while using/running NAMD/Charm with projections enabled.

I am including some email messages I sent to Nitin about these issues. Let me know how I might resolve these issues and any references that may help to clarify

proper use of projections to be able to take further advantage of it’s capabilities.

 

Thanks,

Bob

 

 

 

From: Nitin Bhat <nitin AT hpccharm.com>
Date: Thursday, December 10, 2020 at 10:55 AM
To: "Ortega, Bob" <bobo AT mail.smu.edu>
Subject: Re: Projections

 

Hi Bob, 

 

I am just reading your latest emails about the issues that you’re seeing on projections. 

 

Can you reach out to the Charm mailing list (charm AT lists.cs.illinois.edu) with both the issues that you’re seeing?  (This one and the previous java exception that you saw when you launched projections). The folks who work with (and develop) projections will be able to better address those issues. 

 

Thanks,

Nitin

On Dec 10, 2020, at 8:52 AM, Ortega, Bob <bobo AT mail.smu.edu> wrote:

 

Nitin,

 

Thanks again for your support.  I’m now trying to find out how to use the following runtime option,

 

+logsize NUM

 

Because when I run NAMD.PRJ binary, at the end of the output I get this message:

 

*************************************************************

Warning: Projections log flushed to disk 101 times on 36 cores: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35.

Warning: The performance data is likely invalid, unless the flushes have been explicitly synchronized by your program.

Warning: This may be fixed by specifying a larger +logsize (current value 1000000).

 

I thought that perhaps this was entered into the Makefile under the projections section, so I put it there with this line,

 

+logsize 10000000

 

But I still am getting the Warning message.

 

Thanks,

Bob

 

Nitin,

 

As noted in an earlier email, I was successful running projections for traces generated by a run with 18 processors and 1 node.  But when I tried with 180 processors and 10 nodes, I get the following error when trying to run projections:

 

 

Do you know what could be the problem here?

 

Thanks,

Bob