Decoupled Architectures for Complexity-Effective General Purpose Processors

Decoupled architectures have previously been investigated in the context of high performance scientific computing. For general purpose computing, however, superscalar processors have proven to be flexible in providing high performance across a wide range of applications. To achieve this goal, these architectures have incorporated enormous amounts of complexity to obtain modest performance improvements. Looking towards the future, complexity will be of primary importance, and these inefficient designs will not be tolerated. This work investigates decoupled architectures as a complexity-effective means of obtaining high performance for general purpose workloads. A general survey of decoupled architectures is presented, as well as some proposals for incorporating decoupled architectures into a general purpose computing environment.