Warp Processors: Dynamic Hardware/Software Partitioning

Traditional microprocessor software bits represent sequential instructions that are executed by a programmable microprocessor. A computation may execute faster on an FPGA than as sequential instructions on a microprocessor because a circuit allows concurrency from the bit to the process level.

Prof. Frank Vahid at UCR has invented a WARP processor, a microprocessor that allows the dynamic and transparent partitioning of an executing software’s binary kernels into customized FPGA circuits resulting in 2-100 times speed up over executing on microprocessors. 

Prof. Vahid’s dynamic approach allows techniques associated with dynamic software optimization to be applied to hardware/software partitioning. The profiler, compiler and synthesis tools are entirely on-chip, so that warp processor partitioning does not require extra designer effort or disruption to standard tool flow.

Prof. Vahid’s invention has immense commercial applications as almost any kind of microprocessor-based technology can utilize the benefits of warp processing including video and audio processing, bioinformatics, mainframe computers and even TV’s.