About PCIL
Real-time signal processing applications frequently
must
meet stringent latency, throughput, reliability, power
consumption, and cost constraints. In the last ten years,
commercially available, off-the-shelf processors have
made
parallel and distributed computing an attractive
option for such
applications as adaptive beamforming,
target recognition, and
navigation. However, the scarcity
of cost-effective, efficient
procedures for designing
parallel algorithms remains a serious
barrier to its viability.
Our research focuses on the
development of parallel
algorithm synthesis methodologies for
designing new,
highly efficient, and portable algorithms. The
goal is to
develop tunable approaches that permits algorithm
designers to reap the benefits of both high performance
and
portability on a variety of computer architectures.
|
Current projects include a Playstation 3 cluster, a GPU vector processing system, and general code optimization. |
