Very high memory usage in Epsilon calculations

[박기명(일반대학원 물리학과)]

Dear BerkeleyGW users,

I am currently struggling with very high memory usage in my epsilon calculations.

While reading the BerkeleyGW papers (https://doi.org/10.1016/j.cpc.2011.12.006), I found references to a comm_disk option in the epsilon code, which seems to reduce memory usage by reading wavefunctions from disk. However, in my BerkeleyGW 4.0 installation, I do not see this option in the input file(source code) or in the documentation.

Has comm_disk been removed or renamed in recent versions (e.g., 4.0)?
If so, is there a recommended replacement or alternative option?

Best,
GiMyung Park

[Zhenglu Li]

Dear GiMyung,

You should still be able to use the flags.  The online manual misses flags.

Best,

Zhenglu

---

From: 박기명(일반대학원 물리학과) <dus...@yonsei.ac.kr>
Sent: Thursday, December 4, 2025 3:09 AM
To: BerkeleyGW Help <he...@berkeleygw.org>
Subject: Very high memory usage in Epsilon calculations

 

Dear BerkeleyGW users, I am currently struggling with very high memory usage in my epsilon calculations. While reading the BerkeleyGW papers (https: //doi. org/10. 1016/j. cpc. 2011. 12. 006), I found references to a comm_disk option in the epsilon

--
You received this message because you are subscribed to the Google Groups "BerkeleyGW Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to help+uns...@berkeleygw.org.
To view this discussion visit https://groups.google.com/a/berkeleygw.org/d/msgid/help/50ffd89c-a671-4090-9fed-d4e2565b6ff7n%40berkeleygw.org.

[Mauro Del Ben]

Hi GiMyung,

set this flag in input: 

max_mem_nv_block_algo XXX

With XXX being the max memory in Gb you have available per MPI task. 

Best

-M

To view this discussion visit https://groups.google.com/a/berkeleygw.org/d/msgid/help/BYAPR07MB5093B4BBEDB4D45C7398BAC5CEA6A%40BYAPR07MB5093.namprd07.prod.outlook.com.

[mike marchywka]

I'm not sure about your situation but I had a case where running with or without the mpirun made a big difference in memory usage.

Also, with a solid state drive on Ubuntu, i could just create massive amounts of virtual memory and didn't even notice it. In another

case it still seems to run ok with a lot of virtual memory but it looked like it was thrashing based on a lot of apparent swapping.