RE: RE: [Scalex86 - linux] FW: segfaults in mm_malloc

From: Tim Sirianni (tim_at_scalemp.com)
Date: Fri Jul 01 2005 - 17:46:03 CDT

J. Valencia,

Thanks for your reply. I'd already run into, and solved, the stack-limit
problem, but it might help someone else.

I've rebuilt NAMD inserting "#include <stdlib.h>" lines in all the .c files,
as is apparently required to get a prototype for malloc(). I'm currently
testing this version to see if the SIGSEGV's are resolved.

Tim

-----Original Message-----
From: "J. Valencia" [mailto:jonathan_at_ibt.unam.mx]
Sent: Friday, July 01, 2005 9:40 AM
To: namd-l_at_ks.uiuc.edu
Cc: Tim Sirianni
Subject: Re: namd-l: RE: [Scalex86 - linux] FW: segfaults in mm_malloc

Hello. I just checked your post on NAMD Wiki. I don't run NAMD on my box,
but
some recurrently aborted runs and the message "Segmentation fault" is what I
used to get using CHARMM. I'm running Fedora 3 on x86_64 architecture and,
according to CHARMM users/developers, this is an issue with the way Fedora
handles memory.

The solution they came up with is to increase the maximum stack size. If you
are
also using Fedora and run the command "ulimit -s" you will probably get a
number
like 10240, so all you would have to do is set it to "unlimited" (or to a
really
big number) with "ulimit -s unlimited" (or "ulimit -s some-big-number")
before
running NAMD.

Hope this helps.

J. Valencia

Quoting Tim Sirianni <tim_at_scalemp.com>:

> Brian and Dave,
>
> Thanks for your replies.
>
> I understand what you're saying, but I fear it applies to
> 64-address-bit
> systems with 32-bit integers (e.g., Itanium). Since I'm running on an
> x86
> system where sizeof(void *) == sizeof(int), I don't think that missing
> prototypes could be the cause of this problem. Plus, I see the
> SIGSEGVs
> only occasionally - maybe one run in 3 fails.
>
> I've used both icc and gcc and see the problem with either compiler.
>
> Maybe I've misunderstood what you were trying to tell me, please correct
> me
> if so.
>
> Thanks,
> Tim

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:40:55 CST