Regression test failure for COSP version 2

8 views
Skip to first unread message

Jason Cole

unread,
Jun 24, 2019, 4:27:18 PM6/24/19
to COSP users group

Hi all,


  We are attempting to inline COSPv2 into our AGCM (CanAM5).  Before doing so we are testing the offline version of the code (commit 8fc670e) on our system.  We have been unable to pass the supplied regression test with most of the large differences occuring in the CALIPSO simulator.  The output from the regression test is below.  


We get similar regression results if we use gfortran (Ubuntu 4.8.4-2ubuntu1~14.04.3) or ifort (ifort version 16.0.1) on our Cray CS-400 (Ubuntu 14.04) and using ifort (ifort version 16.0.1) on our XC40 (the target machine for CanAM5).


Before investing effort into debugging these issues, would it be possible to get confirmation that this commit will pass the regression test?  If it does then I would be interested if anyone has had similar issues and if so how they dealt with them.


Thanks,


Jason


Result of regression test

==========================================================


python test_cosp2Imp.py data/outputs/UKMO/cosp2_output_um.ref.nc /space/hall2/sitestore/eccc/crd/ccrn/users/jcl001/COSPV2_TEST/cospV2_gfortran/driver/data/outputs/UKMO/cosp2_output_um.nc


############################################################################################
Treating relative differences less than 0.0010000000% as insignificant
  atb532_perp:         17.37 % of values differ, relative range:  -9.45e-05 to  1.87e-04
  atb532:              17.46 % of values differ, relative range:  -7.03e-05 to  5.91e-05
  cfadLidarsr532:       0.00 % of values differ, relative range:  -1.00e+00 to -1.00e+00
  lidarBetaMol532:     19.47 % of values differ, relative range:  -7.03e-05 to  5.92e-05
  clopaquecalipso:     22.22 % of values differ, relative range:   5.26e-02 to  2.00e+00
  clthincalipso:       36.60 % of values differ, relative range:  -1.00e+00 to -5.56e-02
  clzopaquecalipso:     32.68 % of values differ, relative range:  -1.00e+00 to  7.00e+00
  clcalipsoopaque:      1.18 % of values differ, relative range:   5.26e-02 to  1.00e+00
  clcalipsothin:        3.22 % of values differ, relative range:  -1.00e+00 to -5.56e-02
  clcalipsozopaque:      1.49 % of values differ, relative range:  -1.00e+00 to  6.00e-01
  clcalipsoopacity:      9.80 % of values differ, relative range:  -1.00e+00 to  1.70e+00
  clopaquetemp:        33.33 % of values differ, relative range:  -1.00e+00 to  1.12e-01
  clthintemp:          35.95 % of values differ, relative range:  -4.56e+27 to  7.74e-02
  clzopaquetemp:       32.68 % of values differ, relative range:  -1.00e+00 to  1.47e-01
  clopaquemeanz:       33.33 % of values differ, relative range:  -1.00e+00 to  3.25e+00
  clthinmeanz:         35.95 % of values differ, relative range:  -1.39e+27 to  1.29e+00
  clthinemis:           2.61 % of values differ, relative range:  -1.46e+30 to -1.13e-05
  clopaquemeanzse:     33.33 % of values differ, relative range:  -1.00e+00 to  3.25e+00
  clthinmeanzse:       35.95 % of values differ, relative range:  -1.39e+27 to  1.29e+00
  clzopaquecalipsose:     32.68 % of values differ, relative range:  -1.00e+00 to  7.00e+00
  lidarBetaMol532gr:     18.49 % of values differ, relative range:  -6.98e-05 to  5.90e-05
  atb532gr:            14.73 % of values differ, relative range:  -6.98e-05 to  5.90e-05
  lidarBetaMol355:      4.04 % of values differ, relative range:  -2.21e-05 to  1.54e-05
  atb355:               4.42 % of values differ, relative range:  -3.81e-05 to  4.95e-05
  dbze94:               0.35 % of values differ, relative range:  -5.25e-04 to  3.57e-04
  boxtauisccp:          2.22 % of values differ, relative range:  -3.02e-05 to  1.18e-05
  boxptopisccp:         2.32 % of values differ, relative range:  -3.82e-05 to  9.82e-05
All other fields match.

Dustin Swales - NOAA Affiliate

unread,
Jun 24, 2019, 5:18:23 PM6/24/19
to COSP users group
Hi Jason,

Thanks for bringing this to our attention. I was able to replicate the differences you see in the regression test. I tracked this down to the reference file, which I didn't update with the latest bug-fix (small fix in the Calipso thin/opaque cloud diagnostics). I will update the reference file in the repository.

Cheers,
Dustin 
Reply all
Reply to author
Forward
0 new messages