SBIR-STTR Award

Dataflow specification of applications for parallel computing was established over four decades ago and has been richly developed in academic research efforts. Defense Department R&D programs have resulted in computer programming languages and supporting software tools for both military and COTS hardware systems. Despite clearly demonstrated suitability and productivity gains, data flow programming has not been widely adopted by the parallel computing industry sector. One clear impediment to acceptance of this powerful programming technology is the lack of an industrial standard language. The existence of an industry wide standard data flow language is necessary precondition to industrial acceptance of data flow software technology for production code development. Research in the development of a data flow language, Dataflow Interchange Format (DIF), supporting non proprietary exchange of data flow graph specifications has been conducted at the University of Maryland. This research provides a technology base for development of a common data flow graph specification language capable of becoming the enabling standard for broad industrial use of dataflow software technology. This proposal is for full development of DIF as a potential standard and its supported open source distribution. DIF is expected to immediately provide high performance computing system production code developers the means to save and reuse high value data flow application specifications in a vendor neutral form and support from vendor and academic software tools. Early in phase II, DIF will support combined use of U.C. Berkeley''''s Ptolemy and MCCI''s Autocoding Toolset with its MATLAB import capability for parallel application design and prototyping. This capability is expected to be immediately useful in DARPA PCA and HPCS research programs and broadly useful in all DoD and commercial high performance computing based programs.

Keywords:
Dataflow, Dif, Ptolemy, Autocoding Toolset, Open Source, Vendor Neutral, Industry Standard, Matlab Import

The DIF project addresses a significant impediment to broad industry acceptance of dataflow technology in high performance computing, namely the lack of a standard language for dataflow application programs. Despite the rich body of Academic and Defense Department sponsored R&D in dataflow programming and computing technology, data flow programming has not been widely adopted by the parallel computing industry sector. Existing vendor proprietary and unsupported academic tools and environments are perceived as posing unacceptable technical and business risks. The establishment of a vendor independent language for exchange and reuse of legacy data flow graphs will reduce this risk, making it possible to preserve of high value data flow specifications independent of any particular vendor''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''s tools. Further, the use of tools from multiple vendors will be enabled through sharing data flow graph application programs. In phase I, the MCCI/UMD team demonstrated a Turing complete preliminary version of DIF that supported exchange of data flow graphs between the University of California Berkeley''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''s Ptolemy data flow simulation environment and MCCI''''''''''''''''s Autocoding Toolset®. This established the feasibility of a common dataflow language for all tools and environments supporting the mathematical model of data flow behavior. In phase I, requirements for a full performance version of DIF were established. Additionally the need for a common Domain Actor library, a library for functions underlying dataflow graph nodes was recognized. Our phase II effort will develop the full performance version of DIF. A primitive library supporting the VSIPL standard will implemented as the DIF Domain Actor library. The full performance version of DIF will be evaluated by Lockheed Martin, MS2 for suitability as a standard for high performance application specification. The Xilinx corporation will review DIF for potential inclusion in dataflow design tools for hardware design.

Keywords:
Dataflow Dif Ptolemy Autocoding Toolset® Open Source Vendor Neutral Industry Standard Java