Page 1 of 1

Memory buildup from HSE06 calculations

Posted: Mon Dec 09, 2013 9:51 pm
by tanggara
Dear VASP admin,

We are testing HSE06 calculations on VASP 5.2.11, 5.2.12, and 5.3. As the optimization continues, the memory gradually increased upward to 50-80 GB of RAM. I am working closely with our computational resource computing and they say that the most intense memory usage came from fock_acc subroutine (makes sense since HSE06 alter the Fock operator!). My question will be: is this normal for the memory to build upward to 50-80 GB of RAM for each optimization?

Memory buildup from HSE06 calculations

Posted: Tue Dec 10, 2013 10:16 am
by alex
Hi tanggara,

I'm observing a similar behaviour for long lasting PBE0 calculations. After 10 optimisation steps (about a month of wall time) the machine reboots because memory is out. I suspect a memory leak in the code but I did not bother to wirte a note up to now. Since I got several steps memory should be deallocated again.

Maybe there is a memory leak in the code?

Cheers,

alex

Memory buildup from HSE06 calculations

Posted: Wed Dec 18, 2013 8:31 pm
by tanggara
Thank you Alex for letting us know that we are not the only one who are having this problem! Here are two snapshots of the memory buildup: https://app.box.com/s/opf5cfof15x2fz0co7ye and https://app.box.com/s/8rq805zbhjq7g149rw6d.

Memory buildup from HSE06 calculations

Posted: Fri Feb 14, 2014 1:17 pm
by admin
sorry, but you don't provide sufficient information to answer your question.

Memory buildup from HSE06 calculations

Posted: Thu Mar 27, 2014 8:43 pm
by tanggara
After going back and forth between me and vasp via e-mail, we figured out the problem. The memory increase issue will be implemented in the next vasp-release according to them. I am running HSE right now using the patch that vasp gives and the memory increase has gone away.

Memory buildup from HSE06 calculations

Posted: Fri Mar 28, 2014 7:39 am
by alex
Hi tanggara,

thanks for the update!

Cheers,

alex

Re: Memory buildup from HSE06 calculations

Posted: Wed Dec 10, 2014 9:16 am
by zzhlax
tanggara wrote:After going back and forth between me and vasp via e-mail, we figured out the problem. The memory increase issue will be implemented in the next vasp-release according to them. I am running HSE right now using the patch that vasp gives and the memory increase has gone away.
Hi tanggara,

I'm using vasp version 5.2.2 and find the memory buildup problem within my HSE06 calculations. I am trying to fix this problem. Could you please send me your patch file? My email is zzhlax@mail.xjtu.edu.cn. Thanks very much.

zzhlax