Welcome, Guest
Username: Password: Remember me
Forum header

TOPIC: Running HIRLAM 7.4 fails using old RCR area

Running HIRLAM 7.4 fails using old RCR area 7 years 2 months ago #957

First of all, I checked that running Hirlam 7.4 with default configuration works.

Then I tried to run an experiment the old RCR-area (RCR_7.1 in Env_domain). In this case the problem size is much smaller

NLON=582 # areasize
NLAT=448

RCR_7.4)
NLON=1030 # areasize
NLAT=816

The run fails when going first time for the minimization in 4DVAR
IND=F, LLWINDPROF=F, LLSPGPS=F, LSPGPSTDEL=F
/
=========== gemini_tl kstep = 1 time = 1800. =======
ERROR: 0031-250 task 37: Trace/BPT trap
ERROR: 0031-250 task 55: Trace/BPT trap
ERROR: 0031-250 task 32: Trace/BPT trap
ERROR: 0031-250 task 33: Trace/BPT trap

I could not find a good explanation for this error message, but I suppose it is related to the memory handlin or OpemMP. Before that everything seems very normal. I tried also with LOWRESINCR=no with the same result.

I attach the whole log-file for this cycle. The experiment in ecgate is
/home/ms/fi/fne/hl_home/V74_COARS

Has anyone any clue how to continue?

Kalle

File Attachment:

File Name: HL_Cycle_2012012006-a504a5e71559129b135528a53fee69e9.gz
File Size: 163739

Re: Running HIRLAM 7.4 fails using old RCR area 7 years 2 months ago #958

  • Xiaohua Yang
  • Xiaohua Yang's Avatar
  • OFFLINE
  • Administrator
  • Posts: 195
  • Thank you received: 4
Kalle,

The problem is most likely in yocur onfiguration for 4DVAR at change of domain. Please look at the configuraiton of 4DVAR in 7.3 ) (~hlam/hirlam_release/7.3/scripts/ on ecgate) for RCR-7-1 domain. LOWRESINCR is often mandatory in 4DVAR, and the selection of minimisation resolution for 15km model can not be the same as those for 7 km model.

Xiaohua

Re:Running HIRLAM 7.4 fails using old RCR area 7 years 2 months ago #963

Thanks Xiaohua

The error message was so misleading that I started thinking memory or OpenMp problems instead checking the most logical possibility.

Anyhow, changing in Env_expdesc
ILRES=8,6 # resolution decrease for consecutive inner loops
# the ratio is relative to original model gridsize

to
ILRES=6,3 # resolution decrease for consecutive inner loops
# the ratio is relative to original model gridsize

helped to solve the problem.

Kalle
Time to create page: 0.078 seconds