Changelog 2011 07 21

This is the changelog for Alpine3D 2011-07-21

• code cleanup

  • lots of code cleanup (the Fortran legacy code is now gone)
  • code cleanup in snowdrift, using Grid3DObjects instead of 1D grids (leading to much clearer code)
  • a generic grid "customOutput" is available to output any gridded parameter
  • more features are delegated to MeteoIO (grids splitting/re-assembling, which is therefore consistent between Snowpack and Ebalance)
  • cleanup in the SMSCG scripts

• bug fixes

  • fixed some date handling bugs
  • bug fixes in snowdrift
  • small bugs in the Makefile fixed (colorgcc in latest Ubuntu, wrong behavior of cp)
  • fixed some possible segfaults as well as cases where a file could have been left open
  • fixes for albedos of old snow (so that old snow would not converge to a wrong value, but to glacier ice instead)
  • extra checks in SnowInterface (checking the size of the various data structures, so that we always talk about the same domain, adding more "consts" to guarantee that we don't silently change some variables)
  • some wrong pointer handling have been fixed (in ebalance and snowpack)

• improvements

  • keeping in sync with MeteoIO
  • better error messages and better information messages
  • unnecessary data exchange have been removed.
  • improvements in the timing information given by the modules
  • support for DEM with holes in the radiation model as well as Snowpack
  • ILWR spatial interpolations are now performed by MeteoIO (and therefore user-configurable)
  • hourly meteo data outputted for each special point
  • the hydrological output of glaciers is now consistent with what other people do (even if our previous approach made more sense)
  • water pixels are ignored for snowpack
  • lots of work in snowdrift

• parallel version

  • new radiation model, that is 10% faster when running sequential. It is also available in a parallel version, then leading to a x8~9 overall speedup (!!) compared to previous parallel simulations. This means that a simulation that took 8 days when running parallel now takes 1 days on the same system!
  • the (heavier) snowpack workers are created first, and then the ebalance workers. This should lead to better load balancing.
  • trying to achieve well balanced workers (by making sure that each worker has at most 1 column more than the smallest worker)
  • improved SMSCG scripts so a simulation crashing would not be fully lost

• tools

  • script for generating sno files from land use codes
  • glacier layers were previously not properly characterized in the usual sno files. This has been fixed in the sno files generator