Implementation,Optimization and Evaluation of Scientific Applications on the FT64 Stream Processor
Yang Xue-jun · Computer Engineering and Science · 2008
The stream architecture is one of the emerging architectures that adapt well to the development of VLSI technologies, its efficiency to scientific applications is widely concerned. In this paper, we select a data-intensive program MG from the NASA Parallel Benchmarks and explore its implementation and optimization on the FT64 stream processor, which is the first implementation of a 64-bit stream processor for scientific computing. The evaluation shows that FT64 achieves a comparable performance to the Itanium 2 processor after optimizations towards on-chip memory hierarchies.