At CAPSL, we are interested in alternative approaches of processor architectures for parallel computing as a departure from the mainly centralized control paradigm that dominated the traditional microprocessor architecture evolution. In other words, we can employ a fully decentralized control paradigm - that is a processor chip is controlled by a large number of fine-grain threads, each enabled by individual events due to program control and data flow. As a result, the hardware architecture consists of a large grid of small processing cells and their local memories. Several research projects have been initiated at CAPSL along this direction which are listed below.

EARTH (Efficient Architecture for Running THreads)

CARE (Compiler Assisted Reorder Engine)

DIMES (Delaware Iterative Multiprocessor Emulation System)

PERCS (Productive, Easy-to-use, Reliable Computing System)