SIMULATION OF VECTOR COMPUTING FOR MULTI-USER DETECTION IN SPREAD-SPECTRUM COMMUNICATIONS

摘要

Multi-user detection in wireless spread-spectrum communications relies on matrix algorithms that are well suited for vector computing. Although commodity microprocessors now support vector processing, further study is needed to develop vector architectures that are optimized for multiuser detection. This paper describes enhancements to the SimpleScalar simulation framework (Burger and Austin, 1997) to support research on vector architectures. Annotated instructions are used to represent vector operations, and the main simulation loop is modified to recognize and interpret these annotated vector instructions. Experimental results indicate that vector simulation is a factor of 4 to 16 times faster than scalar simulation by virtue of a dramatically lower instruction count.