How to specify the postion of a restart calcultion?

[Guangfu Luo]

Dear friends,

In a usual case, we just need Yambo to restart a calculation from where it stops. But sometimes if we want to do some convergence tests, we could need Yambo to restart a calculation from a certain point we interest (since the parameters before the restart point have already converged).

Could you tell me how to do change the relevant files in the RESTART folder? Or is there any other way to do this?


Thanks very much

[Daniele Varsano]

Dear Guangfu Luo,

actually I can't understand exactly what do you want to restart.
What kind of calculation are you doing?
Please, if you post a specific example of what you need, we can try to answer.

Cheers,

Daniele

[Guangfu Luo]

Dear Daniele,

I am sorry for not explaining myself clear. Let me put it in a specific example.

Say I want to calculate the BS spectra of graphene, and surely I need to do the convergence test of many parameters at first. I have done several tests and have known the converged parameters for the [Xs] and [BSK] step. Now I have only parameters in the [BSS] step need to be tested. So in this situation, how can I reuse the output files of the former steps (then I just need to do the [BSS] step calculation)?

In Abinit, we can specify the calculation from a certain DATASET (and reuse the results of former DATASETs) by using “ndtset” and “jdtset”. By intuition, I believe Yambo can do the same thing, but the question is how.

Thanks for your explanation.


p.s. I just realize that this could be fulfilled by making a new yambo.in, which has only parameters of the the [BSS] step. Is this correct?

[Daniele Varsano]

OK I get the point.
Yambo does it automatically. In the sense that read the database
you have produced in the previous run. If it is compatible with your input
parameters (i.e. the converged one you calculated the BSK) it does not repeat
the calculation and will perform the diagonalisation. Depending of the parameters
you are changing you will see in the report a WARNING or an ERR. In the case it
produces an error, it will recalculate the BSK runlevel. Otherwise it will not.
In order to be sure, you can remove the file containing the diagonalized matrix
in your script:
rm ./SAVE/*ndb.BS_diago_ or ./SAVE/db_haydock.

Hope it helps,

Daniele

[Guangfu Luo]

Dear Daniele,

I got it and have successfully tested your idea. Thanks very much.