number of atoms

[martinspenke]

Dear Daniele,

Up to how many atoms can yambo_4.0.1_rev_89 handle ?
Is there any record regarding this ?

Bests
Martin

[Daniele Varsano]

Dear Martin,
cannot tell you a number, it really depends on the available machines. The are published papers with systems of hundreds of atoms.
yambo_4.0.x has been coded in order to have a good scaling on hpc machines and has been demonstrated to scale up to tenth of thousands cpu.
When setting up very big calculations on many cpu a careful setting of the parallelization variable is needed in order to properly balance the workload among cpus.

Best,
Daniele

[martinspenke]

Dear Daniele,

many thanks, one further question :

How can i properly control the output printed by yambo_4.0.1 ?
Is there any way to command the program to handle data internally to speed up calculation and save place and memory ?
For instance in a GW calculation in folder Real_Axis, is it possible to circumvent the huge databases generated by Yambo as of ndb.dipoles_fragment_35 and ndb.em1d_fragment_29 ?
I would like to have directly the QP corrections without all these data, if possible. Any keyword to enforce this ?

Bests
Martin

[Daniele Varsano]

Dear Martin,
yes something can be done.
When you prepare your input add the verbosity "-V io".
You will find in your input file the keyword:

Code: Select all

DBsIOoff= "none"             # [IO] Space-separated list of DB with NO I/O. DB=(DIP,X,HF,COLLs,J,GF,CARRIERs,WF,W,SC,BS)

Here you can tell to the code to not print some of the databse "e.g. DIP".
Of course they will rest in memory, this is the price to pay.
About fragmentation we are working in these days in order to add option to have fragmented or not fragmented databases as both strategies have pro and cons.
This option should be ready very soon.

Daniele