Crystallographic computing on an array processor

Abstract
To date, six computationally intense but frequently needed crystallographic applications programs have been developed for an array processor (Floating Point Systems model AP 190L) driven by a DEC 10 computer. In all of the applications attempted, it was possible to reduce the required processing time to at least an order of magnitude below that required by a large university scale computer (DEC 10) for the same problem. In fact, the rate-determining step in full-matrix least-squares refinement can be made to run 20–30 times faster in the array processor. For the refinement of proteins, one cycle with a space-group-general algorithm executes faster in the array processor than a typical small-molecule refinement cycle executes on the DEC 10. We conclude that many of the time-consuming operations frequently encountered in crystallographic computer programs can be handled very efficiently on an inexpensive array processor attached to a host computer.