[Wien] MPI parallelization and block cyclic data layout in lapw1

[Robert Laskowski]

hi,
as you said this is block cyclic data distribution used by scalapack. 
Each process set up its own part of the H and S matrixes, thus there is no communication 
between processes during execution of the hamiltonian set up routines (hamilt, hns, horb). 
For the layout description read scalapack documentation.

regards

Robert


On Friday 20 February 2009 03:35:49 MingWenmei wrote:
> 
> Dear all,
> 
>  
> 
> I am studying the source code of Wien2k. I am eager to understand how the MPI parallelization was implemented in lapw1. Especially, can anyone give me some detailed reference on how the hamilton(H) and overlap(S) matrix elements are mapped onto the two dimension processor grid to form the block cyclic data layout in the subroutine Hamilt? 
> 
>  
> 
> Thanks very much. 
> 
>  
> 
> Wenmei Ming 
> 
> 
> _________________________________________________________________
> 新版手机MSN，满足你多彩需求！参加抢鲜体验活动，领取特色奖品！
> http://mobile.msn.com.cn/



-- 
Dr Robert Laskowski

Vienna University of Technology, 
Institute of Materials Chemistry, 
Getreidemarkt 9/165-TC, A-1060 Vienna

tel. +43 1 58801 15686
Fax  +43 1 58801 15698