Ignore:
Timestamp:
Feb 9, 2013, 5:48:51 PM (12 years ago)
Author:
aslmd
Message:

LMDZ.GENERIC (and LMDZ.UNIVERSAL)

  • Optimized calculations for continuum (done for H2 and He, to be done for others)
    • new common bilinear interpolation routine (bilinearbig)
    • optimization: only one calculation is actually needed

to find indexes of wavelength for bilinear interpolation
... because this will not change with level and integration step!

  • optimization: use while loop in bilinearbig
  • completely similar results obtained (test case for a gas giant, many simulated days)

NB: those changes really improve gcm speed (factor 2.2 for whole model!)

continuum was very expensive, now very cheap
--> e.g. 1 day, 25 dyn ts, 5 phys ts
--> before: 243 seconds (including 120 seconds for continuum bilinear interpolation)
--> after: 108 seconds

  • Corrected a bug: Continuum in inifis instead of continuum

... until now, most users (unbeknownst to them) were running with the continuum by default!

  • Cosmetic changes in optcv (mostly spaces and line breaks)

... so that comparisons with optci are easy e.g. through vimdiff

File:
1 edited

Legend:

Unmodified
Added
Removed
  • trunk/LMDZ.GENERIC/README

    r869 r873  
    878878- Users can still use e.g. H2_ but H2 also works
    879879- Simplified condense_cloud so that igas_CO2 is used directly
     880
     881== 09/02/2013 == AS
     882- Optimized calculations for continuum (done for H2 and He, to be done for others)
     883      - new common bilinear interpolation routine (bilinearbig)
     884      - optimization: only one calculation is actually needed
     885                      to find indexes of wavelength for bilinear interpolation
     886                      ... because this will not change with level and integration step!
     887      - optimization: use while loop in bilinearbig
     888      - completely similar results obtained (test case for a gas giant, many simulated days)
     889  NB: those changes really improve gcm speed (factor 2.2 for whole model!)
     890      continuum was very expensive, now very cheap
     891      --> e.g. 1 day, 25 dyn ts, 5 phys ts
     892      --> before: 243 seconds (including 120 seconds for continuum bilinear interpolation)
     893      --> after: 108 seconds
     894- Corrected a bug: Continuum in inifis instead of continuum
     895    ... until now, most users (unbeknownst to them) were running with the continuum by default!
     896- Cosmetic changes in optcv (mostly spaces and line breaks)
     897    ... so that comparisons with optci are easy e.g. through vimdiff
Note: See TracChangeset for help on using the changeset viewer.