I'm trying to output tendencies following the instructions in How to control OpenIFS output#Physicaltendenciesandfluxes(budget)output(PEXTRA). However setting lbud23 to true causes the model to crash near the start of the run (after setting all the other required parameters). This is crashing for the t21test setup in both CY38 and CY40. Is there some incompatibility for the tendencies with the t21test or some other variables I need to set in the namelist?

The error messages also show that they are failing in different routines. For CY38 I get

  14:26:20 STEP    0 H=   0:00 +CPU=  1.932

JSETSIG: sl->active = 0

signal_harakiri(SIGALRM=14): New handler installed at 0xac2fe1; old preserved at 0x0

***Received signal = 8 and ActivatED SIGALRM=14 and calling alarm(10), time =    1.62

[myproc#1,tid#1,pid#31486,signal#8(SIGFPE)]: Received signal :: 268MB (heap), 152MB (rss), 0MB (stack), 0 (paging), nsigs 1, time     1.62

tid#1 starting drhook traceback, time =    1.62

[myproc#1,tid#1,pid#31486]:  MASTER

[myproc#1,tid#1,pid#31486]:   CNT0<1>

[myproc#1,tid#1,pid#31486]:    CNT1

[myproc#1,tid#1,pid#31486]:     CNT2

[myproc#1,tid#1,pid#31486]:      CNT3

[myproc#1,tid#1,pid#31486]:       CNT4

[myproc#1,tid#1,pid#31486]:        STEPO

[myproc#1,tid#1,pid#31486]:         SCAN2H

[myproc#1,tid#1,pid#31486]:          SCAN2M

[myproc#1,tid#1,pid#31486]:           GP_MODEL

[myproc#1,tid#1,pid#31486]:            CPG

[myproc#1,tid#1,pid#31486]:             >OMP-CPG         1        (1025)

[myproc#1,tid#1,pid#31486]:              CPG_DIA

[myproc#1,tid#1,pid#31486]:               GPRH


and for CY40 I get

  12:37:17 STEP    0 H=   0:00 +CPU=  1.668

  MAX U WIND=   306.13097669574682     

  MAX V WIND=   324.93554918973513     

  MAX U WIND=   306.32652430590468     

  MAX V WIND=   348.93744277983961     

  MAX V WIND=   353.52689949818011     

  MAX V WIND=   353.54319779446274     

  MAX V WIND=   324.01019139888371     

  MAX U WIND=   317.31737041482785     

  MAX U WIND=   315.59021166357309     

  MAX V WIND=   407.45745811027041     

  MAX V WIND=   386.04271784454772     

  MAX V WIND=   259.56161483696303     

  MAX V WIND=   276.03331499316238     

  MAX V WIND=   395.65145298527545     

JSETSIG: sl->active = 0

[/network/aopp/cirrus/pred/saffin/oifs/oifs40r1v1.1/src/ifsaux/support/drhook.c-l1029] signal_harakiri(SIGALRM=14): New handler installed at 0xadfd15; old preserved at (nil)

***Received signal = 8 and ActivatED SIGALRM=14 and calling alarm(10), time =    1.52

[myproc#1,tid#2,pid#27214,signal#8(SIGFPE)]: Received signal :: 333MB (heap), 143MB (rss), 0MB (stack), 0 (paging), nsigs 1, time     1.52

tid#2 starting drhook traceback, time =    1.52

[myproc#1,tid#1,pid#27214]:  MASTER

[myproc#1,tid#1,pid#27214]:   CNT0<1>

[myproc#1,tid#1,pid#27214]:    CNT1

[myproc#1,tid#1,pid#27214]:     CNT2

[myproc#1,tid#1,pid#27214]:      CNT3

[myproc#1,tid#1,pid#27214]:       CNT4

[myproc#1,tid#1,pid#27214]:        STEPO

[myproc#1,tid#1,pid#27214]:         SCAN2M

[myproc#1,tid#1,pid#27214]:          GP_MODEL

[myproc#1,tid#1,pid#27214]:           EC_PHYS_DRV

[myproc#1,tid#2,pid#27214]:            >OMP-PHYSICS              (1001)

[myproc#1,tid#2,pid#27214]:             EC_PHYS

[myproc#1,tid#2,pid#27214]:              CALLPAR

[myproc#1,tid#2,pid#27214]:               TURBULENCE_LAYER

[myproc#1,tid#2,pid#27214]:                VDFOUTER

[myproc#1,tid#2,pid#27214]:                 VDFMAIN




5 Comments

  1. Unknown User (nagc)

    Hi Leo,

    Those errors are indicative that the LBUD23 diagnostics are not enabled correctly in the namelist, and there is some memory overwriting going on.

    It is very important the model arrays are sized correctly if you enable the LBUD23 diagnostics. Could you please check you have changed all the relevant namelists?

    &NAEPHY
  LBUD23=true,
&NAMDPHY
   NVEXTR=24,                              ! set number of tendency output fields (see table)
   NCEXTR=<number of full model levels>,   ! edit to correctly set number of full model levels e.g. 60, 91, 137
&NAMPHYDS
    NVEXTRAGB(1:24)=91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114

    The NAMDPHY setting are crucial. If they are not set, or NVEXTR is set to the wrong number, the model will crash as you have seen.

    If that's not it, please attach a copy of your namelist to this post and I'll take a look and try it myself.

    Thanks, Glenn

    p.s. to attach a file, look under the '+' menu at the top of the comment box and select 'Filles and images'.

  2. Unknown User (gbs2)

    Hi Glenn,


    I think I have set these variables correctly following the tutorial. NVEXTR=24 and NCEXTR=19.

    I've attached the namelists for CY38

    fort.4


    and CY40

    fort.4


    I think they just have the default options for the t21test with some changes to model output.


    Thanks,

    Leo


  3. Unknown User (nagc)

    Ok, I'll try it. It's possible the problem is related to the T21 test. That's just intended as a small test, it doesn't run the model the same way as for higher resolution as it uses the explicit dynamics, not the semi-Lagrangian scheme, but the physics will be the same.

    Glenn

  4. Unknown User (gbs2)

    "It's possible the problem is related to the T21 test." This seems to be the case. I just successfully output model tendencies using a higher resolution set up I've been sent.

    It would still be good to get this to run for the t21test if possible. Would that be able to be run using the semi-Lagrangian scheme. I have LSLAG set to TRUE, is the other switch here LTWOTL=TRUE? (Still gives the same error though).

    Leo

    1. Unknown User (nagc)

      Hi Leo,

      I suspected that might be the case. I don't want to spend time getting lbud23 to work with the t21 test case. Other than a useful technical test after openifs is compiled, t21 is not of any practical scientific use. There might be some namelist switches for the physics which need enabling.

      Cheers,  Glenn