[MITgcm-devel] Beaufort experiment with gfortran (4.7.0) and -O3

[Jean-Michel Campin]

Hi,

I have now a version (4.7.0) of gfortran/gcc similar to the one
Torge was using when reporting -O3 problems in this thread:
[MITgcm-devel] Beaufort experiment on mac os x
http://mitgcm.org/pipermail/mitgcm-devel/2012-March/005354.html
but this is not on a mac-os but on Linux FC.17.

I dowdloaded the beaufort-sea set-up and compiled using linux_amd64_gfortran
(no '-ieee', -O3, empty NOOPTFILES list) with the old version of ini_maskis_etc.F 
(the one which caused Pb, before Martin changed it), but it run fine. 
I also added "-ftree-vectorize" to FOPTIM, since this flag is used 
in darwin_amd64_gfortran, but it also run fine.

In conclusion, the compiler problem with optimisation (-O3) for ini_maskis_etc.F
is not just related to the version of gfortran (here 4.7.0) but also 
to the system (darwin/linux FC17).

Cheers,
Jean-Michel