VASP/WANNIER90 interface out of date

Message

myliu · #1 Post by **myliu** » Sun Oct 27, 2013 4:48 pm

As of 14-Oct-2013, Wannier90 was updated to release 2.0.

VASP 5.3.3 can be compiled with -DVASP2WANNIER90 precompiler option without any errors. During runtime, VASP segfaults when the VASP/Wannier90 interface is invoked. The segfault occurs immediately after the first line,

Code: Select all

?num_wann?=?????....??!?set?to?NBANDS?by?VASP

has been written to wannier.win. No other wannier-related files are generated.

The issue is fixed by using the legacy release Wannier90 v1.2.

Please update the online manual/wiki to include this.

Cheers,
myliu

-----------------------
Some keywords for search engines:
DVASP2WANNIER90 LIB precompiler CPP WANNIER LWANNIER90 libwannier.a wannier wannier90 segfault core dumped

Paul_Brown · #2 Post by **Paul_Brown** » Mon Oct 28, 2013 4:03 pm

I have compiled the wannier version 2.0.0 along with vasp.5.3.3 using -DVASP2WANNIER90 flag. Upon running a test job I received an error suggesting that I haven't specified the num_wann flag; however, the flag is specified in the appropriate seedname.win file. Is it true that the current version of vasp cannot interface with this version of wannier90?

Cheers mate!

#3 Post by **admin** » Wed Oct 30, 2013 4:39 pm

The problem seems to be due to the new version of wannier90-2.0.0 (compiled with the current Intel compiler)

The variables list of wannier_setup (src/wannier_lib.F90) has two additions compared to wannier90-1.2:

proj_s_loc
proj_s_qaxis_loc

which are "optional, intent(out)".

VASP does not (yet) use these quantities, and as such, calls wannier_setup without them.
This should not be a problem, since these variables are declared as optionals, but apparently the ifort compiler messes up (we have not tried other compilers).

Solution:

1) delete these 2 parameters from the call list of wannier_setup in wannier_lib.F90
2) comment out all lines in wannier_setup referring to these variables,
i.e., their declaration and a single if-clause.

ckande · #4 Post by **ckande** » Sun Dec 08, 2013 8:23 pm

Thanks. Solved the problem.

leoant · #5 Post by **leoant** » Thu Jan 09, 2014 9:34 pm

Dear VASP & WANNIER users,

Here I have another problem when I try to use VASP-5.3.3 with Wannier90-2.0.

The number of projections are not correct if there is spin-orbit coupling. I have
tried to calculate bccFe, but using the 5.3.3+2.0 combo it does not work. While
using 5.2.12+1.2 it worked.

Have any of you experienced also this problem?

Thank you!