missing timer in RtmMod.F90

[worleyph]

There is a missing "call t_stopf('rtmi_grid')" that needs to added to the end of the subroutine Rtmini in RtmMod.F90 . This only causes problems when enabling the highest level of performance profiling (which we never do), but I discovered this when testing this option for some of the CIME work. The same error is in the CESM version, btw.

I'd like someone to drop this in next time that RTM is modified, or as part of some other pull request. It is not a productive use of anyone's time to do a pull request just for this bug fix. I can help with putting in the fix and describing how to test it.

[worleyph]

Still needs to be fixed.

[bishtgautam]

How does one enable "the highest level of performance profiling"? In other words, how can I make sure the fix that I add is correct?

[worleyph]

You can add

profile_detail_limit = 10

to user_nl_cpl. Let me check this again, and I'll tell you exactly where to put it.

[worleyph]

Okay, the fix is to add

     call t_stopf('rtmi_grid')

at line 430 in ./components/rtm/src/riverroute/RtmMod.F90, as in

 ...
     if (masterproc) then
        write(iulog,*)'RTM netcdf river direction file successfully read '
        call shr_sys_flush(iulog)
     endif

     call t_stopf('rtmi_grid')

     !-------------------------------------------------------
     ! Set dwnstrm_index from rdirc values
     !-------------------------------------------------------
 ...

If you then run with

 profile_detail_limit = 10

in user_nl_cpl the timing output file (./timing/XXX.stats) should indicate something like

 rtmi_grid                                    164      164 1.640000e+02   3.014999e+01 ...

in contrast to the current output of

 rtmi_grid                                      0        0 0.000000e+00   0.000000e+00     ...

There are other indicators as well, including at the end of the run, the original will indicate that the rtmi_grid timer was never stopped.

I can send you the udpated file as well, if you wish.