[Wien] Tetra NaN problem solved

[Parker, David S.]

Dear All: The instructions below were helpful in solving a problem
With tetra producing some DOS and partial DOS values as 'NaN' despite the lack
Of an 'NaN' in the corresponding case.qtl file.  To fix just add the flag ("-fp-model precise") below at the end
Of this line in the tetra Makefile, i.e.

FOPT     = -FR -mp1 -w -prec_div -pc80 -pad -ip -DINTEL_VML -traceback -fp-model precise

Note that the other optimization flags may vary depending upon your compiler, arthitecture, etc.

And then compile using "make" as usual.

Good luck - David Parker






Dear colleagues,
when using tetra (WIEN2k 9.1), compiled with the Intel Fortran compiler
11.0, occasionally the calculation of DOS and partial DOSs results in
'NaN' for single energy values. This is a matter of numerical precision.
When compiling tetra with the option:
-fp-model precise
, which disables optimization that can change the result of
floating-point calculations, the problem wasn't observed up to now.
The effect of this compiler option on the performance of the program
seems to be marginal.

Best regards

Ulrich Wedig

--
-----------------------------------------------------------------
Dr. Ulrich Wedig                              Tel. 0711/6891535
Max-Planck-Institut fuer Festkoerperforschung FAX  0711/6891502
Heisenbergstr. 1
70569 Stuttgart                               U.Wedig at fkf.mpg.de
-----------------------------------------------------------------