Two issues with CAM

[jlin5@atmos_uiuc_edu]

*1) The CAM link on http://www.ucar.edu/tools/models.jsp actually
leads to Biome-BGC.*


*2) When I test CAM on NCSA-SGI (Intel ia64 dual core cluster with
GNU/Linux OS and Intel Fortran Compiler 9.0), there are several
incompatibilities as shown below, followed by my changes.*


*In areaMod.F90, line 308: !dir$ inlinenever areaave*


*-*à changed to !dir$ attributes noinline :: areaave


*In histFileMod.F90, line 1231: !dir$ inlinenever hfields_normalize*


*-*à changed to !dir$ attributes noinline :: hfields_normalize


*In history.F90, line 6058: !dir$ INLINENEVER gen_hash_key*


*-*à changed to !dir$ attributes noinline :: gen_hash_key


*In SurfaceAlbedoMod.F90, line 490: !dir$ inlinenever SoilAlbedo*


*-*à changed to !dir$ attributes noinline :: SoilAlbedo


*In TridiagonalMod.F90, line 104: !dir$ unroll 0*


*-*à commented out (changed to ! !dir$ unroll 0)


*Without revision, the incompatibilities cause errors in the
compiling, but the executable is still generated. I am not sure if
these errors will have any consequences in the simulation. My revision
passed the test for Port Validation for CAM, though.*

 

[eaton]

The compiler directives "!dir$" are intended for use on the Cray-X1/X1E. I'm surprised that the intel compiler is recognizing them. These directives are unlikely to be of any benefit on the IA64 architecture. I'd recommend looking for a compiler option to turn off recognition of these directives. Given that the executable is producing a simulation that's passing the CAM port test I'm guessing that the directives aren't doing anything anyway.