The Defense Advanced Research Projects Agency (DARPA) has released high productivity computing systems (HPCS) benchmarks as specifications of performance required to solve important problems. This work reports on the design, implementation, and benchmarking of the benchmark number two, dynamic programming (DP), as developed in a general purpose reconfigurable computer. DP methods are used to solve DNA sequencing problems. DP has been also used successfully for optimal compression of barcode and other data, for control problems such as depth vision systems, and for chess endgames, among others. Clearly, it is helpful to know the application background; however, it is more important to note the benchmark is the product of a set of applications. Nonetheless it is intended to be developed and tested in a manner that is independent of any particular application.

We implement and report performance numbers of an entire molecular dynamics application in floating point on reconfigurable hardware achieving sustainable speedup and scaling with a novel technique based on adaptive processing elements.

Molecular electronics can be developed if we are able to program a random arrangement of molecules or a field-programmable molecular random array. The ansatz that small molecules can be programmed needs to be demonstrated; this means characterizing the smallest molecular system with programmable features. We demonstrate that even two molecules in a series conformation can have multivalued responses and, thus, is able to be programmed; we also indicate how to extend this programmability to other molecular circuit conformations. Current programs for the calculation of current-voltage characteristics of electronic circuits, needed for such demonstrations, are only capable of predicting single-valued characteristics. We present results from our ab initio procedures that couple the molecular approaches with a practical analysis of molecular circuits having strong nonlinearities.