charm AT lists.cs.illinois.edu

Subject: Charm++ parallel programming system

List archive

Re: [charm] Memory leak

From: Eric Bohm <ebohm AT illinois.edu>
To: James Bordner <jobordner AT gmail.com>
Cc: charm <charm AT lists.cs.illinois.edu>
Subject: Re: [charm] Memory leak
Date: Tue, 14 May 2019 15:46:36 -0500
Authentication-results: illinois.edu; spf=neutral smtp.mailfrom=ebohm AT illinois.edu; dkim=pass header.d=illinois-edu.20150623.gappssmtp.com header.s=20150623; dmarc=none header.from=illinois.edu

We don’t have a specific tool to track down charm message leaks. In general
practice we’ve found that Address Sanitizer and/or Valgrind are excellent
ways to track down those kinds of issues.

I’ll note that getting good results from valgrind requires a bit of command
line customization. I usually go with —log-file=vg%p.log —fair-sched=yes
—trace-children=yes in order to get the traces more cleanly separated and to
get good behavior from valgrind when the application is built on a multicore
charm target.

> On May 6, 2019, at 11:57 AM, James Bordner
> <jobordner AT gmail.com>
> wrote:
>
> Hi, I'm trying to track down a memory leak at machine-eth.c:392 ("msg =
> (char *)CmiAlloc(size);"). My assumption has been that my code is not
> deleting a CkMessage when received somewhere, but I don't see where in my
> code that could be. Does my assumption sound correct? Any suggestions for
> tracking this down?
>
> Thanks!
> James

[charm] Memory leak, James Bordner, 05/06/2019
- Re: [charm] Memory leak, Eric Bohm, 05/14/2019