[DFTB-Plus-User] Questions about compiling DFTB+MPI-NEGF and running 2D Carbon Material Tutorial

[Ben Hourahine]

Dear Yuefei,

On 06/11/17 00:20, yh46 wrote:
> Hello Dear friends,
> I was compiling DFTB+ on a server, the version is r4729 or r4732. The
> source codes are in a file named 'dftb+mpi-negf.r4732', but when I
> compile the code and run the binary dftb+, it tells me it is '
> Unofficial release (r4729)'. So I am not sure about the version.
>

The number comes from the version software control software used for
that branch of the code. The revisions r4729 and r4732 are essentially
the same (there are minor cosmetic changes in some printed message).

> First question is about compiling, I saw 3 README files in the
> directory: README, README.PARALLEL and README.NEGF, respectively. When
> I followed the README.NEGF to compile the code, there were some
> SEGMENTATION FAULT while running, happening both when I use single
> core or multiple cores. So I compiled the code again without the
> instructions in README.NEGF or README.PARALLEL, used
> /sysmakes/make.x86_64-linux-ifort as template and changed some links
> to library according to my system. Then I successfully compiled and it
> can run some transport calculation from the autotest directories. It
> can also run on multiple cores. In the output, I saw ' INIT MPI-NEGF
> ON           4 NODES', so I think it is doing the job parallel.
>
> So does that mean the compiled dftb+ binary has both MPI and NEGF
> extension? It is quite confusing because I didn't follow README.NEGF.
> So is it decided by the source code, and no matter what setting you
> use in the makefile, you will get both MPI and NEGF?
>

The NEGF code is constructed on top of a version of the MPI code. Both
codes are MPI parallel, but the main difference for compilation is that
the NEGF code can also use OpenMP parallelism for some operations (this
is the reason for the extra -openmp flags).

If you have compiled dftb+mpi-negf.r4732 successfully using an MPI aware
Fortran compiler, this should contain the NGEF routines (mpif90 is a
wrapper for the usual compiler but adds in options specific for MPI).

> Second question is that when I use this to run the 'Tutorial:
> Electronic Structure and Electron Transport in 2D Carbon Materials'
[cut]
> So why would the output files be missing or not correctly written?
> Thank you very much!

It seem that you probably have an earlier version of the source code
than some of the features used in the tutorial. I'm less familiar with
this part of the code, so off hand don't know at which version number
these features were added here.

Regards

Ben

-- 
      Dr. B. Hourahine, SUPA, Department of Physics,
    University of Strathclyde, John Anderson Building,
            107 Rottenrow, Glasgow G4 0NG, UK.
    +44 141 548 2325, benjamin.hourahine at strath.ac.uk

2013/14 THE Awards Entrepreneurial University of the Year
      2012/13 THE Awards UK University of the Year

   The University of Strathclyde is a charitable body,
        registered in Scotland, number SC015263

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20171106/a6b4039b/attachment.html>