I have pushed another round of improvements to the parallel VecScatter code. There is now a universal parallel VecScatterBegin/End templated over block size. Also further optimizations in generation and use of the scatters.
Be on the look out for subtle bugs in scatters. Barry
