Results and Products
This project has the typical feature of a significant experimental computer science activity. The work we propose is novel and as such has a major research component which will be described in conventional journal and conference papers. However there are several major software system components which will produce artifacts of general value. These will consists of software realizations as well as innovative designs which will be documented by internal memos as well the external publications.
All the software we develop will be in the public domain and highlights of our work will include:
Technology Transfer
Electronic distribution and traditional scholarly activities will allow a broad based distribution. We pointed out above that extensive use of Java will help us maximize the ease of use and dissemination of our products. However we can also identify several special pro-active outreach (technology transfer) functions. Firstly we will work with DARPA and other funded projects from this BAA to establish connections with key users of our technology. These include commercial and academic designers of new hardware and software systems. We already have the connection with the PetaFlop studies where Fox has been an ongoing leading contributor. Another interesting (and perhaps unexpected) area is the HPCC education and training community as we believe our products will have significant pedagogical value stemming from their high level structure. Here we will target NSF, DoD modernization and DoE ASCI activities. We are well connected with these already. The particular selected applications will directly benefit from our work and this is illustrated by SAIC for ship design, John Hopkins for the virtual microscope and the NSF Black Hole collaboration for their simulation.
In a different way, we expect our project will produce and validate
a hierarchical machine description that we expect to have great
value in other HPCC arena including the design of extensions to
parallel environments like MPI and HPF.
Related Technology Transfer Activities
We have a good record for technology transfer with related HPCC activities. Syracuse's first DARPA award (1991-94) resulted in the Portland Group licensing the prototype HPF compiler produced and Portland Group is also using Maryland's advanced technology for compilation of irregular problems. We have also been active in the HPF forum for the initial (HPF1) and recent extension (HPF2) where again the Maryland work was critical in making clear that one can and should support irregular problems. Both Maryland and Syracuse are members of CRPC (Center for Research in Parallel Computation) which has an excellent record in technology transfer. We intend to incorporate the results of this performance project into our research compilers and as in the past we thereby will be able to transfer this technology to the commercial compiler industry.
Under the DARPA HPCD project, Rutgers has collaborated with the ship division of SAIC for the parallelization and improvement in robustness of the LAMP ship design codes. The parallel codes have been transferred to the Navy for their internal use. The LAMP system was recently used in the design of the revolutionary Arsenal Ship configuration developed by the General Dynamics/Raytheon/SAIC team and is a central application in this proposal. The proposed techniques can be used to assist this team and DARPA in selecting appropriate future machines based on the simulated performance study. The PYRROS system and related clustering techniques produced by Rutgers/UCSB under the HPCD project has been used by several other research groups including MIT, Berkeley, UMD, LIP, Tennessee, CMU, and NASA. The NSF is sponsoring interactions between LIP (France) and Rutgers/UCSB for studying automatic task graph generation systems to extend PYRROS techniques. UCSB is working with Navy NRAD for transferring parallel systems techniques developed in the NSF/DARPA/NASA sponsored digital library project on Navy parallel machines. UCSB is also sponsored by SUN MICRO for developing next generation SCI-based workstation clusters. The proposed performance technology will be used for them to evaluate/select new architectures.
Given this excellent record, we believe that we will be able to effectively technology transition the results of this project in both specific (particular applications) and general cases (broad based technology such as PETASIM and HLAM).