Hugepages error

General discussion of the Cambridge quantum Monte Carlo code CASINO; how to install and setup; how to use it; what it does; applications.
Post Reply
Kayahan
Posts: 22
Joined: Fri Jun 07, 2013 5:56 pm

Hugepages error

Post by Kayahan »

Dear CASINO Users,

I have been trying to accumulate expval_data in molecular systems, but I get hugepages error in one of the computers I use to run my jobs. Expval_data file I have is around 2gb size. When I use smaller expval_data file, I don't get this error. Some lines from the error is following:

Code: Select all

[...]

-Output from node #1025--
 
ERROR : Allocation problem in SETUP_MOL_DENSITY (expval_mol_den DMC arrays) on node #1025.
 
CASINO internal traceback:
Problem detected at SETUP

[...]
libhugetlbfs [nid04383:31397]: WARNING: New heap segment map at 0x1002dc00000 failed: Cannot allocate memory
[...]
It seems like craype-hugepages2M module is also loaded in the module list. Do you think it could be related to the way CASINO is compiled or should I check more system related issues?

Thanks,
Kayahan
Mike Towler
Posts: 239
Joined: Thu May 30, 2013 11:03 pm
Location: Florence
Contact:

Re: Hugepages error

Post by Mike Towler »

Hi Kayahan,

More information required please. How big is the wave function file? Are you using shared memory? How much memory and how many cores per node?
How are you running the CASINO job? Can I see the arch file you're using to compile and run CASINO. What happens if you modify the arch file to not use hugepages at all? etc. etc.

As you may have seen in recent threads here, the density for molecular systems was implemented recently by Blazej using a method which can lead to the use of large amounts of memory.. It may just be that you genuinely are running out of memory, or there could be some bug in the Cray hugepages module (Googling reveals some hints of the latter..).

Let me know.

Mike
Post Reply