[Wrf-users] WRFSI built with g95: success, but ...

[Leigh Orf]

Arnold Moene wrote:

|   Hi,
|   
|   I don't have the PG compiler available. But the first reason to try g95 
|   was that compilation with IFC resulted in a gridgen_model.exe that was 
|   only consuming more and more memory while it ran, until it died with a 
|   signal 9 or signal 11 (I have 768 Mb available, that should do, I think, 
|   for a small domain ;).

I will add my experience with ifort (formerly ifc) 8.1 here - I have had
nothing but trouble with the WRF quarter-circle hodograph case using
ifort / mpi on a Xeon Linux cluster. I can get it to compile but with
most hardware geometries (numbers of processors, i.e. mpirun -np X)
the model dies without much information. I have spent way to much time
messing with ifort/mpi and have given up on it. With PGF I do not have
these problems.


|   Andrés Calderón wrote:
|   > It is very interesting.
|   > 
|   > Do you have something about the performance? IFC vs PGI vs g95?
|   > 
|   > thanks,
|   > 
|   >  -- Andres Calderon


--
Leigh Orf
KG4ULP
Assistant Professor of Atmospheric Sciences
Department of Geography
Central Michigan University
Mount Pleasant, MI 48859
(989)774-1923