[QE-developers] Parallelizing linear response calculations over optimal number of pools

[Lorenzo Paulatto]

> In fact, the bottleneck stems from when the wavefunction is read from 
> the SCF calculation through "PW/src/read_file_new.f90”. There, 
> "divide_et_impera” is called which fails if -npools is larger than 
> nkstot from the SCF.

It does not fail anymore in version 6.6 of the code which should be out 
soon, although this specific case has not been tested very much. While 
waiting, you can use the tricks of running pw with more k-points than 
necessary, than specify nk1, nk2 ad nk3 in the input of phonon.

cheers



-- 
Lorenzo Paulatto - Paris