[Wien] Parallel execution of hybrid functional calculation

[西村 真一]

Thank you for your kind reply.

I have just started to examine effectiveness of the HF calculation in my system.
> I don't really know. Anyway, the execution of the hf module will take
> between 1 and 3 orders of magnitude more time than lapw1 and lapw2.

I know that expensive calculations are not always effective.
However, at initial stage of calculation, 
many trials are necessary before higher accuracy calculation.
So disabling the parallel execution of initial stage calculation consumes too much time
and reserved cores in job-queing systems.

Thanks,
Shinichi

On 2013/09/01, at 18:10, tran at theochem.tuwien.ac.at wrote:

> I don't really know. Anyway, the execution of the hf module will take
> between 1 and 3 orders of magnitude more time than lapw1 and lapw2.
> Therefore, to run lapw1 and lapw2 for FBZ in parallel will barely
> change the total execution time.
> 
> F. Tran
> 
> On Sun, 1 Sep 2013, tran at theochem.tuwien.ac.at wrote:
> 
>> Hi,
>> 
>> Thank you for your quick reply.
>> 
>>> No, with hybrid functionals the k-point parallelization is for the
>>> k-points in case.klist_ibz only.
>> How about the fine-grained parallelization?
>> I can put the $para to the lines for case.klist_fbz calculations.
>> Does  this cause some problems?
>> 
>> Shinichi
>> 
>> On 2013/09/01, at 17:31, tran at theochem.tuwien.ac.at wrote:
>> 
>>> Hi,
>>> No, with hybrid functionals the k-point parallelization is for the
>>> k-points in case.klist_ibz only.
>>> Another thing:
>>> If you start a hybrid calculation without case.vectorhf,
>>> case.energyhf and case.weighhf in your directory, then in order to
>>> generate these files, during the 1st iteration lapw1 and lapw2 will
>>> be executed one additional time compared to the next iterations.
>>> F. Tran
>>> On Sun, 1 Sep 2013,   wrote:
>>>> Dear WIEN2k users,
>>>> I am trying to perform the hybrid functionals calculation with
>>>> -hf of option like:
>>>> $ runsp -hf -p.
>>>> The version of WIEN2k is 13.1.
>>>> With the -hf execution lapw1 and lapw2 with the FBZ-klist do not run in 
>> parallel mode,
>>>> while the normal runsp runs in parallel mode (MPI/k-point),
>>>> I checked the runsp script, and found that
>>>> the lines for the calculations with FBZ-klist do not have the $para 
>> variable.
>>>> We can not run the FBZ calculations with any parallel options?
>>>> Thanks in advance,
>>>> Shinichi Nishimura
>>>> _______________________________________________
>>>> Wien mailing list
>>>> Wien at zeus.theochem.tuwien.ac.at
>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>> SEARCH the MAILING-LIST at: 
>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>> _______________________________________________
>>> Wien mailing list
>>> Wien at zeus.theochem.tuwien.ac.at
>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>> SEARCH the MAILING-LIST at: 
>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html