Initialize count for iseed_ca <= 0 in update_ca.F90

[NickSzapiro-NOAA]

In compiling UFS (ufs-community/ufs-weather-model#2703), there is a warning that:

/work/noaa/nems/nszapiro/tasks/gefs_ort/ufs-weather-model/stochastic_physics/update_ca.F90:397:47:
  397 |             count_trunc = iscale*(count/iscale)
      |                                               ^
Warning: 'count' may be used uninitialized [-Wmaybe-uninitialized]

It seems that initializing count (if iseed_ca <= 0) via system_clock was removed in #47 .

I'm not familiar with the internals/intent here. One option is to remove the separate conditions (as here). Another is to initialize count more appropriately.

fyi, any change here is expected to be bit-for-bit in UFS regression tests as iseed_ca>0.

[NickSzapiro-NOAA]

While we're here, the 7*mytile vs mytile is curious

[pjpegion]

@lisa-bengtsson can you take a look at these issues?

[lisa-bengtsson]

Hi Nick,

I would like to keep the distinction between iseed_ca <= 0, and the case when it has an integer larger than 0.

I don't recall why the count line is commented out - and I discussed it with Phil who did some updates to the initial seed generation to better match SPPT a few years ago, but we think you should just have to uncomment this line in update_ca.F90:

!call system_clock(count, count_rate, count_max)

In two places (for CA global and CA sgs).

The curious 7 is there simply to distinguish the seed between the sub-grid-scale CA, and the global CA.

Could you test if your warning goes away if you uncomment the call to system clock?

Thanks,
Lisa

[NickSzapiro-NOAA]

Sure. As you wish :) Thank you both for being so responsive

By calling system_clock first, the warning goes away. Noting as well that cpld_control_gefs and control_p8 RTs are bit-for-bit too (as expected from iseed_ca>0)