We demonstrate a system-level CMP architecture where optical cache memories are shared among multiple processing cores through optical buses. System-level simulations show 25–45% execution time improvement and significant capacity requirements reduction through simpler memory hierarchy.
展开▼