segmentation faults

[jjkay]

Hello,

I made the mistake of updating RHEL 4...and now I am getting segmentation faults. I am quite new to this, so I would appreciate any advice on what anyone considers to be the most probable remedies for this situation. (I imagine that it occurs frequently.)

I am running RHEL 4 on a multi-processor EM64T machine; I was using ifort 9.1.036 before updating the OS. I was optimizing at O3 and parallelizing at the highest possible levels; now the best I can do is to get my program to execute half-way to the end using ifort 8.1.037.

Using the -traceback option I can trace the error to a particular subroutine; nothing out of the ordinary there. (The code, after all, used to run just fine before I updated the OS.) The only piece of information that appears helpful is that it lists one of the images as "libc.so.6", which I understand to be part of the "glibc" libraries. Therefore I suspect that updating the C libraries has caused something to freak out. Which versions of glibc are compatible with ifort 9.1 at present?

Any advice? I realize that this is sort of like asking a doctor to how perform brain surgery over the phone.

Thanks a lot,

Jeff

[Ron_Green]

Jeff,

Check and increase your stacksize limits. This is a shell setting. For bash:

ulimit -a

to list all the setting

ulimit -s unlimited

to unlimit the stack.

Unfortunately, you have an older 9.1 version. Any chance you can update that? The newer 9.1 compilers have a -heap-arrays option that should alleviate the seg faults w/o having to mess with the shell stack limits.

You should search through this forum for stack limits / stacksize and/or seg fault for more information.

ron