[developers] -limit, -memlimit, -nsolutions, packing

Yi Zhang yzhang at coli.uni-sb.de
Mon Oct 20 19:55:51 CEST 2008


hi Bart.,

> I heard that, in order for selective unpacking to happen at all,
> nsolutions has to be set. Is this true (wiki!)?
yes, see http://wiki.delph-in.net/moin/PetSelectiveUnpacking
here is the actual code that decide which unpacking algorithm to use.


     if ((opt_packing & PACKING_SELUNPACK)
          && nsolutions > 0
          && Grammar->sm()) {
        nres = unpack_selectively(trees, upedgelimit, nsolutions,
                                  UnpackTime, readings);
      } else { // unpack exhaustively
        nres = unpack_exhaustively(trees, upedgelimit, UnpackTime, readings);
      }


for your case, i think there is an extra factor. do you actually have
a statistical disambiguation model (Grammar->sm()) ? if not, it still
falls back to exhaustive unpacking. and the timeout checking is
currently not done properly for exhaustive unpacking (extra checking
must be made in function unpack() in item.cpp). I will try to fix it.

as to the -memlimit, maybe someone else can answer it.

best,
yi



More information about the developers mailing list