Analytic Models of Workload Behavior and Pipeline Performance

The evaluation of pipeline performance and the analysis of different design alternatives and cost/performance tradeoffs are a fundamental aspect of high-performance computer system design. This performance evaluation process requires accurate models of both the pipeline organization and the characteristics of the workload being executed. In this paper we derive general mathematical models and analyses of workload behavior and pipeline performance that can provide measures as accurate as detailed trace-driven simulations with the computational efficiency of analytic methods. 1 Introduction The use of pipelining, wherein the execution of multiple instructions is performed concurrently, has been and continues to be a fundamental and important technique in the design of high-performance computer systems (e.g., see [5]). This technique consists of breaking the processing of instructions into steps, or stages, each performing a different piece of the entire instruction, and simultaneously..