Stuck in the calculation of collisions

Questions and doubts about features of non linear optic in Yambo (yamb_nl)

Moderators: Davide Sangalli, claudio, myrta gruning

Post Reply
Liangting
Posts: 1
Joined: Wed Apr 17, 2024 8:57 am

Stuck in the calculation of collisions

Post by Liangting » Thu Jul 11, 2024 9:42 am

Dear developers,
I'm doing a repeated SHG calculation of ML-MoS2 by referring paper:PHYSICAL REVIEW B 89, 081102(R) (2014). I had troubles in calculating collisions database in BSE level.
At first, I used 2 (node) * 20 (ntasks-per-node) in our cluster to calculate collisions database, but all reported CPU files were stuck in the step [06.01] SEX+HARTREE (e-e correlation), the job stopped until the time limit (6 hours) of cluster.

Code: Select all

<01h-53m> P1-cn537: [06.01] SEX+HARTREE (e-e correlation)
<01h-53m> P1-cn537: [SEX+HARTREE] Plane waves (H,X,C) :  139  139  139
<01h-53m> P1-cn537: [MEMORY] Alloc WF%c( 1.472871 [Gb]) TOTAL:  1.582761 [Gb] (traced)  69.25200 [Mb] (memstat)
<01h-53m> P1-cn537: [PARALLEL distribution for Wave-Function states] Loaded/Total(Percentual):4410/6174(71%)
<01h-53m> P1-cn537: [FFT-SEX+HARTREE Collisions] Mesh size:   15   15   95
I guess the critical tags are #RT_CPU# and #RT_ROLEs# for Parallelization. so I try to add these two tags in input file,

Code: Select all

RT_CPU= "10.4.1"                  # [PARALLEL] CPUs for each role
RT_ROLEs= "k.b.q"                # [PARALLEL] CPUs roles (k,b,q,qp)
Then I found reported files CPU_1/2/3/4 stuck at 5% 5% 12% and 25%, and the other CPU files are in similar situations.

Code: Select all

<01h-54m> P1-cn519: Collisions |                                        | [000%] --(E) --(X)
<02h-50m> P1-cn519: Collisions |#                                       | [002%] 56m-17s(E) 01d-13h-31m(X)
<03h-52m> P1-cn519: Collisions |##                                      | [005%] 01h-58m(E) 01d-15h-28m(X)
My questions are
1. If I switch to a cluster with longer runtime limits, will the collisions calculation be able to complete?
2. how to distribute cores for parallelization in my calculation?
3. Could you please review the parameters in my input files to ensure they are reasonable?

Best,
Liangting
You do not have the required permissions to view the files attached to this post.

User avatar
Davide Sangalli
Posts: 629
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: Stuck in the calculation of collisions

Post by Davide Sangalli » Thu Jul 18, 2024 9:31 am

One quick suggestion is to try

Code: Select all

HXC_Potential= "SEX+HARTREE CVONLY"     # [SC] SC HXC Potential
It will neglect eh interaction of electrons between two different conduction bands and between two different valence bands (which are usually quite small), while including only eh-interactions between valence and conduction (as in bse at equilibrium).

This should reduce the computational effort.

Best,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

User avatar
claudio
Posts: 502
Joined: Tue Mar 31, 2009 11:33 pm
Location: Marseille
Contact:

Re: Stuck in the calculation of collisions

Post by claudio » Thu Jul 18, 2024 10:28 am

Dear Liangting

you can also try this approximation for the collisions:
https://www.yambo-code.eu/wiki/index.ph ... r_response

that is available in the branch tech-master of Yambo
https://github.com/yambo-code/yambo

best
Claudio
Claudio Attaccalite
[CNRS/ Aix-Marseille Université/ CINaM laborarory / TSN department
Campus de Luminy – Case 913
13288 MARSEILLE Cedex 09
web site: http://www.attaccalite.com

Post Reply