Compiler Assisted Stream Register File Optimization via Loop Transformations

Stream processors are gaining popularity in many scientific and embedded applications. Stream Register File (SRF) is a critical resource in a stream processor, as all data should be stored in SRF for executing. Research into advanced compiler technology for stream languages and architectures is still at its infancy. This project plans to study loop transformation techniques during compilation to optimize both the utilization and performance of the SRF.