FastROCS TK 0.3.0
=================

* Added ShapeDatabaseAddMols.py script to add molecules to a running
  ShapeDatabaseServer.

* Added ShapeDatabaseIsLoaded.py script to report whether the initial
  database load has completed.

* Added argument to ShapeDatabaseClient.py for the number of hits to
  return, defaults to 100.

* Redesigned how the BestShapeOverlay.py script runs to be more useful
  for doing evaluations.

* Restricting OEShapeDatabase to only run on one type of device and
  prefer to run on a GPU, and fall back to a CPU device only if there
  is no GPU present. Used to be that on OS X the search would run on
  the GPU and CPU simultaneously, however, the results of the overlay
  could differ depending on whether they were sent to the GPU or the
  CPU. Due to dynamic load balancing between the CPU and GPU this
  could change the rank of the results returned in an apparently
  non-deterministic manner. It is not pleasant for an end-user to run
  the same query multiple times and get different results so we
  restrict searches to either CPU or GPU, not both now. Note, this
  only would occur on OS X. The NVidia implementation doesn't support
  a CPU device, so this never occurred there.
