Repository navigation
converting the CLUBB portable layer - #1718
bstephens82 wants to merge 70 commits into
Conversation
… stored in src/atmos_phys/schemes/clubb/clubb
and namelist entries and adding clubb3_run subroutine to clubb_tend_cam
and updating atmos_phys path
| path = src/atmos_phys | ||
| url = https://github.com/bstephens82/atmospheric_physics | ||
| fxtag = 1dc98193cbdd90d3e8bcfbe6ca416f6560e17d4c |
There was a problem hiding this comment.
Reminder for me - this needs to be updated before tag assignment
jtruesdal
left a comment
jtruesdal
left a comment
There was a problem hiding this comment.
This looks good cam keeps namelist, pbuf and outfld and clubb and clubb mf goes to atmos_phys. Just a few simple change requests. I hope in phase 2 we can get rid of the many many many pbuf arrays passed through the interface just for outfld. Was Adam able to cut the number of these down . If he hasn't looked at that yet it might help to do it at this stage to help with generating and maintaining the meta data for the atmos_phys side. Just fyi, I checked out your code and noticed that the src/atmos_phys pointer hasn't been updated.
modified: src/atmos_phys (new commits)
gitfleximod will work around that but anyone using submodule updates (me sometimes) will get the old link. You should just run git add src/atmos_phys to allow it to also checkout with the git method.
Ok thanks, I think I've handled all of these now including the src/atmos_phys pointer. |
… for tag prep Resolve the atmos_phys conflict by pointing src/atmos_phys at a local merge of atmos_phys0_31_000 (856127273a) into clubb_mf_ccpp (e8aef8366): c257e1c0a3c9f71953192886cc1a165ed7cf3732. That commit exists only locally; an ESCOMP atmos_phys tag is needed before the CAM tag. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
c257e1c0a3c9 is the local merge of atmos_phys0_31_000 into clubb_mf_ccpp, now pushed to jtruesdal/atmospheric_physics branch pr1718-tagprep. An ESCOMP atmos_phys tag is still needed before the CAM tag. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The BAS calday/tropp_days declarations and the do n = 1,12 loop were outside #ifdef CLUBB_SGS, but n is declared inside it, so non-CLUBB builds (e.g. PC4) failed with "This name does not have a type [N]". calday and tropp_days are only used in the CLUBB call, so move both blocks inside the CLUBB_SGS ifdefs. No change to the compiled code for CLUBB builds. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
clubb_ini_cam sets the wp2/up2/vp2/rtp2/thlp2 pbuf fields from these tolerances but no longer imported them (cam6_4_215 did), so nvfortran stopped with S-0038 "has not been explicitly declared". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
cam6_4_207 (e976e11) dropped the latvap/cpair scaling on these diagnostics and changed their units to kg/kg m/s and K m/s. The refactor brought the scaling back, so the history fields no longer matched their units or the baselines. Diagnostic output only; model state is unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…/th_zt) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The read of clubb_stats_nl opens the hard-coded "atm_in". In cam6_4_215 that happened only inside stats_init_clubb, under stats_metadata%l_stats; the refactor made it unconditional, so multi-instance runs (atm_in_NNNN) aborted with "file not found ... atm_in". The values read are only used when l_stats is true, so restore the guard. No answer change. Reviewed by an independent reviewer subagent against cam6_4_215: APPROVE. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Pass errmsg/errflg to clubb2_run and abort on error with the same endrun block used after clubb1_run and clubb3_run. Update atmos_phys to 182bd17a, which adds those outputs to clubb2_run. No answer change. Reviewed by an independent reviewer subagent: APPROVE. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Drop the never-used calday/tropp_days calculation and its declarations, stop passing calday, tropp_days, lat, phis and pint to clubb3_run, and remove the now-unused get_curr_calday/get_calday imports. Update atmos_phys to 2b845b8e, which removes those clubb3_run arguments (atmos_phys PR ESCOMP#452 review). No answer change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
clubb_tend_cam fetches the CLUBB-MF pbuf pointers only when do_clubb_mf is true, but passes all of them to clubb1_run. With do_clubb_mf false they were undefined pointer actual arguments, which NAG's runtime checking rejects. When do_clubb_mf is false, associate the 34 MF-only pointers (MF memory, macmic accumulators, and the deep-convection hookup fields including the integer jt/maxg/ideep) with zero-size local targets of matching type and rank, and fetch tpert unconditionally since it is always registered. clubb1_run only touches these arrays under do_clubb_mf, so this is bit-for-bit. Reviewed by an independent reviewer subagent: APPROVE. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…gprep b3b90a6 and f700525 make the same clubb2_run/clubb3_run call changes as 5349684 and cb6c30c; clubb_intr.F90 merged with no change to ours. Resolve .gitmodules to keep url = jtruesdal/atmospheric_physics with fxtag = 1cb28d99, and point the atmos_phys gitlink at 1cb28d99, the merge of d8df985 into 2b845b8 (tree identical to 2b845b8). No source changes relative to the tested a7db32b. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
@bstephens82 I merged in the changes from my (Claude's really) regression test updates. There were some unexpected problems that needed to be updated and were all fixed. The issue for cleaning up clubb_intr when mf is not active will be updated with some findings from these changes. I don't think you will need to commit anything more before this is tagged but if you do please pull first to merge my (Claude's) fixes into your update. I also will note on the ChangeLog that the Silhs test is the only baseline difference due to the removal of the ice supersat flag. I assume that was taken into account when you got rid of it but just checking now (@nusbaume maybe you know). I'm not sure of the science effect on the SILHS users. |
…term) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This PR converts the portable layer for CLUBB, by gathering sections of code from clubb_intr.F90 and incorporating them into new subroutines in the atmospheric_physics repo. It also removes clubb_mf.F90 from CAM/src/physics/cam and into the atmospheric_physics repo, and changes the way that CLUBB is checked out so that it gets checked out into atmos_phys/schemes/clubb.
We also decided to remove the flags clubb_do_icesuper and clubb_do_liqsupersat, so those have been removed from namelist files along with their associated blocks of code that were previously in clubb_intr.F90.