Compilation for Reconfigurable Computing

Pedro Diniz and Mary Hall


Summary


Adaptive computing systems consisting of configurable computing components offer significant performance advantages when tailored to the specific computational demands of particular applications, e.g., automatic target recognition (template matching), morphology, sonar beamforming, Infosec (cryptography). However, developing programs for execution on configurable computing units (CCUs) is extremely cumbersome, demanding that system software developers also assume the role of hardware designers. At present, most systems require low-level, assembler-like VHDL coding. While a few high-level tools have been developed, these have been narrowly focused on a single application (e.g., ATR machine at UCLA) or a specific configurable architecture. The absence of general-purpose, high-level programming tools for adaptive computing has prevented this technology from being adopted or even explored by more than a highly specialized and narrow group of users.

Keywords:

Reconfigurable Architectures.


Related Publications