Page 1 of 1

How to run yambo without using the virtual memory

Posted: Wed Mar 23, 2022 10:23 am
by jasonhan0710
Dear all,

I have run a BSE calculation on HPC, which has 192GB memory per node. However, when I tried to run BSE kernal process, it stopped and reported as follows

Code: Select all

  [02.01] Transition Groups build-up @q1
  ======================================
  [BSK] Transition groups (total):  1140
  [ERROR] STOP signal received while in[02.01] Transition Groups build-up @q1
  [ERROR]insufficient virtual memory
Then I logged onto the computation node and find only 30GB memory used by yambo with 160GB left. But it seems yambo have to use the virtual memory. Is there any input parameters to specified in order to avoid this problem? Thank you!

Best,
Jason

Re: How to run yambo without using the virtual memory

Posted: Wed Mar 23, 2022 10:52 am
by Daniele Varsano
Dear Jason,

it is possible that Yambo tries to allocate more memory than available, you can try to insect it by compiling the code using the flag --enable-memory-profile.
Next, it is also possible that the memory is not distributed properly and this can be controlled by setting the parallelization in the input file using the variables (BS_CPU,
BS_ROLEs= "k eh t" ) See here for an explanation:
https://www.attaccalite.com/yambo-input ... explained/

and Sec.8.5 of the Yambo paper:
https://iopscience.iop.org/article/10.1 ... cmab15d0s8

Best,

Daniele