timers calls as in CESM about cice

[eclare108213]

Info from @dabail10 in email:

The timers we use are from the General Purpose Timing Library (GPTL) that Jim Rozinski wrote. I >tried to wrap calls from the existing timers to this library. Kind of like this in ice_timers.F90:

#ifdef CESMCOUPLED
call t_startf('ICE:'//all_timers(timer_id)%name)
call t_stampf(wall, usr, sys)
all_timers(timer_id)%block_cycles1(block_id) = wall
#else
all_timers(timer_id)%block_cycles1(block_id) = MPI_WTIME()
#endif

The GPTL timers work better with OpenMP threading.

Is GPTL generally available on most/all machines, or is it something that users would be responsible for installing themselves? If the latter, then I'd be inclined to stick with what we've got.

[dabail10]

Finally looking at this with CICE6. I don't know the answer about the GPTL package, but will look into it.

[dabail10]

Ok. We still install GPTL as a part of the CESM build.

[eclare108213]

I'd like to close this task without any changes, since GPTL is a package that would need to be installed. Does anyone disagree with that?

[apcraig]

I agree that we can close this and stick with what we have. I think MPI_WTIME works well enough.

Having said that, we could probably add the gptl code to CICE and treat it as CICE code in the standalone mode. I'm not sure this is a good idea, but it's an option. We would probably still have to support ifdef-ing it out when building/running in CESM and other models that might have their own copy of gptl.

I'm not in favor of having to build separate libraries as part of CICE standalone at this time. It might be something we eventually have to do, but I don't think gptl is worth making that effort.

[dabail10]

The advantage of GPTL is that it works well with OMP threading. However, I am happy to leave it out of CICE6.

[eclare108213]

I'll close the issue. We can change it later if the need arises.