The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <gonze@pcpm.ucl.ac.be>]

Dear all,

You might remember some discussion in the Forum mailing list,
about the Intel Fortran Compiler v7.0 for PCs ,
that unlike  previous versions, was causing some problems.
Michel Cote had succeeded to produce something that worked (thanks again !).

However, more systematic tests have shown that, beyond standard
Ground-state calculations, there are still serious problems with this
compiler ! (e.g some response function calculations
as well as stresses with spin-orbit, gives
sometimes numerically wrong results - a few percent ! - , without any notice).
The understanding of these problems has advanced (for example,
only the lowest optimized level of compilation treats correctly
the use of array segments as arguments of subroutines), but
at present, I strongly suggest people NOT to use IFCv7.0 .
(Use IFCv5 or v6, or the PGI compiler).

By the way, Bernard van Renterghem told me that IFCv7.1 was made
available a few days ago. It should be tested ...

(Hoping you did not spend months of CPU doing response-function
calculations using ABINIT with IFCv70 ;-)
Xavier