Slow Haydock

Various technical topics such as parallelism and efficiency, netCDF problems, the Yambo code structure itself, are posted here.

Moderators: Davide Sangalli, andrea.ferretti, myrta gruning, andrea marini, Daniele Varsano, Conor Hogan, Nicola Spallanzani

Post Reply
batman
Posts: 12
Joined: Sun Jun 20, 2021 2:04 pm

Slow Haydock

Post by batman » Tue Sep 05, 2023 1:44 pm

Dear Developers,

The Haydock solver from github master branch is very slow (nearly x15 times slower) than 5.1.2 version. I am attaching log and report files (input is same for both) when computed using two different Yambo versions. Could you please point out what is going wrong?

Best regards,
Murali
You do not have the required permissions to view the files attached to this post.
Muralidhar Nalabothula
Doctoral student at Department of Physics and Materials Science,
Université du Luxembourg

User avatar
Davide Sangalli
Posts: 617
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: Slow Haydock

Post by Davide Sangalli » Fri Sep 08, 2023 12:05 pm

Dear Murali,
thanks for reporting.

If you check the log. The solver slows down every 10 iterations. It is due to the writing of the Haydock results (output files) on the fly. This is handled via a series of subroutines.

Code: Select all

K_OUTPUT --> ELEMENTAL_dump --> X_OUTPUT_driver --> OUTPUT_driver --> OUTPUT_driver --> msg
(both complex and super slow. Don't ask me why, not my doing ...)

I'd suggest first to change this line in K_Haydock

Code: Select all

if ( mod(it,max(1,Haydock_iterIO/10))==0) call K_OUTPUT(iq,W,"override dump Haydock close",OBS)
to

Code: Select all

if ( mod(it,max(1,Haydock_iterIO))==0) call K_OUTPUT(iq,W,"override dump Haydock close",OBS)
(I'll probably also fix this in the main repo.)

Next you can play with the input variable

Code: Select all

BSHayItrIO
Default is 100 (this is why the code slows done every 10 iterations).

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/

Post Reply