On Wed, 16 Mar 2005 13:39:29 -0500, Joe Landman <landman at scalableinformatics.com> wrote: > It should. Basically mpiblast partitions the input and the databases. NCBI > blast could use partitioned databases. NCBI threads each search, rather than > across searches that MPIblast does. On the risk of going completely offtopic, does anybody know more on this _very_ intersting article ? http://www4.ncsu.edu/~xma/pubs/pioblast.pdf In it Heshan Lin, Xiaosong Ma, Praveen Chandramohan, Al Geist, and Nagiza Samatova of NCSU & ORNL describe their way of making mpiBLAST a lot faster. No code available yet I think, but it would be nice to know whether people think this research has any change to get integrated into mpiBLAST. -- Dag, Jan