Compiling methods for the Fresh Breeze processor

Abstract

The Fresh Breeze project is designing a new parallel architecture that is designed to serve as a model for high-performance computing while maintaining good software programming principles. One of the unique aspects of this architecture is a write-once memory model with hardware-based garbage collection and a global address space. Due to this memory immutability, new designs are needed before code generation can be developed for the Fresh Breeze compiler. This paper outlines designs for arrays and the method calling protocol. These designs are analyzed with respect to running times, possible alternatives, and algorithm performance.