High Productivity Computing Systems for Command and Control
Abstract
The most significant issue underlying all future command and control (C2) architectures is the ability to develop software that can harness the next generation of processors. Multicore processors, scaling into thousands of processors per chip will soon be prevalent in all C2 systems. The success of C2 systems will depend on our ability to adapt to the new processor technology. Existing C2 systems that implement scientific codes for image processing and many other applications have been a dominant user of high performance computers (HPCs) for several decades. However, increasingly diverse C2 applications are now also being adapted to HPCs, due to dropping prices and increased availability. The goal of the DARPA High Productivity Computing Systems (HPCS) program is to develop high performance computers that are substantially easier to program, thereby reducing software development cost and time to solution. We employ a publication/subscription information management (PSIM) system in a case study to compare new HPCS approaches to parallel code implementation with existing techniques. The PSIM system requires intensive CPU cycles and communications bandwidth, for brokering XML information objects between publishers and corresponding subscribers. The study compares two new HPCS languages, Chapel (Cray) and X10 (IBM), with the Message Passing Interface (MPI) standard.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2008
- Accession Number
- ADA486979
Entities
People
- Christopher Flynn
- Scott Spetka
Organizations
- SUNY Polytechnic Institute