Page 1 of 1

possible bug in scala.F

Posted: Tue Feb 22, 2022 1:38 am
by sergey_lisenkov1
Gfortran complains to the following line (#1420) in scala.F:

#ifdef ELPA
! call ELPA 2-stage solver
IF (ELPA) THEN
....

1420 | IF (ELPA) THEN
| 1
Error: IF clause at (1) requires a scalar LOGICAL expression

I believe it should be "LELPA" instead of "ELPA". Am I correct?

Intel compiler does not complain, btw.

Related question - is it recommended to use ELPA in vasp 6.3 or not? I don't see that ELPA mentions anywhere except one makefile in arch/

Sergey

Re: possible bug in scala.F

Posted: Tue Mar 01, 2022 6:08 am
by ferenc_karsai
Thank you very much for pointing that out. That indeed is a bug and we are going to incorporate the fix.
Why it didn't pop up is because there is a precompiler macro ELPA that is most likely interpreted as an integer in a boolean context.
The intel compiler is usually not very strict so it accepts it.

It is recommended to use ELPA in VASP 6.3. The NEC machines for example ELPA seems to perform very well.