Dear all Let me give some quick remarks on this
2012/11/1 Guangping Zhang <[email protected]> > Dear Nick and other expert users, > > I have read the manual of truck-424 about the TranSIESTA part and the > update document for TranSIESTA/TBtrans, but I have yet some questions. > > (1) the update document say NPoles + NCircle + NLine and/or NumPoints if a > voltage applied have better to be divisible by the number of processor.Does > this mean the TranSIESTA parallels on the energy points above? Can it > parallel on basis orbitals and k-points as is siesta? > No, not yet. Only parallel across energy points. If the number of processors is not divisable by the number of energy points, some processors will do nothing in the last parallel segment (this has not changed since the last version). > > (2) The TBTrans now can parallel over the energy points spectified by > TS.TBT.NPoints and the k-points. As the energy points parallelisation is > full and the k-points parallelisation is partial, is it better to have > TS.TBT.NPoints divisible by the number of processor prior to the > k-points? Because a divisible k-points by the number of k-poins is adviced > in previous version. > One should only consider TS.TBT.NPoints divisible by the number of processors. It is "partially" parallelised across k-points, but rather it is an internal fast method of calculating the Hamiltonian in advance and is not "really" parallel. > > (3) Whether the new added subroutines of AtomPDOS and COOP are also > parallelized? > Yes, they are. > > (4) Whether the output data for AtomPDOS and COOP are averaged over the > k-points? > No, as you will see in the data files for this, they are output like <systemlabel>.TRANS, i.e. with k-point header and weight in the header segments. > > Please give some hints and instructions. > > Best regards. > > Guang-Ping > > 2012-11-01 > ------------------------------ > Guangping Zhang >
