Description
In the design of SPUR, a high-performance multiprocessor workstation, the need for large "snooping" caches suggests a new approach to virtual address translation. By performing this translation in each processor's virtual cache, the need for separate translation lookaside buffers is eliminated. Trace-driven simulations show that normal cache behavior is only minimally effected, and that unless an extremely large and complex TLB were built, using a separate device would actually reduce system performance.