[MITgcm-devel] problems in some testing

[Jean-Michel Campin]

iHi Constantinos,

Regarding all the different compiler/platform (daily) testing you are
running, I would like to propose some improvement to increase the number
of "pass" : 

batsi: getting 100% fail since few days, and looks like your code is
not clean (fail to link because of ini_procs_new.F, which does not 
exist in the CVS repository).

blueice: 
I would like to get something tested with mpi (sp5+mpi is getting
100% fail since Oct 2007).
>From the error msg, I have the impression that the command  
"mpirun.lsf mitgcmuv" by itself does not specify to run on 2 processors
(as if using the traditional mpirun without the option "-np 2").
Could you confirm this ?

And regarding the sp5-32 we are getting 100% fails at the compilation 
stage since March, but was working before. Looks like weird things happen 
(with the preprocessor) when tying to compile seawater.F (checked in at the 
end of Feb), and I would be curious to see what is in this "seawater.f" file
after the CPP command ; also, do we need "-qsmp=omp" for non multi-threaded test ?
>
> xlf95_r -q32 -qfixed=132  -O3 -qsmp=omp -qstrict -Q -qarch=auto -qtune=auto -qcache=auto -qmaxmem=-1 -c seawater.f
> "seawater.f", line 1.0: 1515-005 (S) Continuation line is not permitted at beginning of program or after INCLUDE, EJECT, @PROCESS statements or comment directive.  Line is ignored.
> "seawater.f", line 2.0: 1515-005 (S) Continuation line is not permitted at beginning of program or after INCLUDE, EJECT, @PROCESS statements or comment directive.  Line is ignored.
> "seawater.f", line 304.1: 1515-048 (I) One or more lines in this statement specifies an OMP conditional compilation sentinel and at least one line does not.  This may cause this statement to behave differently when it is not compiled with -qsmp=omp or -qcclines.  Please make sure this was intended.
> "seawater.f", line 305.1: 1515-048 (I) One or more lines in this statement specifies an OMP conditional compilation sentinel and at least one line does not.  This may cause this statement to behave differently when it is not compiled with -qsmp=omp or -qcclines.  Please make sure this was intended.
> "seawater.f", line 306.1: 1515-048 (I) One or more lines in this statement specifies an OMP conditional compilation sentinel and at least one line does not.  This may cause this statement to behave differently when it is not compiled with -qsmp=omp or -qcclines.  Please make sure this was intended.
> ** find_rho_scalar   === End of Compilation 1 ===
> ** sw_ptmp   === End of Compilation 2 ===
> ** sw_temp   === End of Compilation 3 ===
> ** sw_adtg   === End of Compilation 4 ===
> 1501-511  Compilation failed for file seawater.f.
> gmake: *** [seawater.o] Error 1

aces open64 : may be we can stop this one since I am running a 
similar one (with trapuv) which is getting a better "score" (62/70 
instead of 46/70) ; some differences are know (e.g. ideal_2D_oce: 
bug in the compiler optimization level ; or MLAdjust.1.leith: missing 
netcdf) and at the end, I don't think it adds much to the testing 
in general. If you prefer to keep it, then I suggest to change the optimisation
level and to get the netcdf to work.

Thanks,

Jean-Michel