Speculative execution of code is becoming a key technique fornenhancing the performance of pipeline processors. We study schemes thatnpredict the execution path of a program based on the history of branchnexecutions. Building on previous work, we present a model for analyzingnthe effective speedup from pipelining using various schemes fornspeculative execution. We follow this with stochastic analyses ofnvarious speculative execution schemes. Finally, we conclude withnsimulations covering several of the settings we study
展开▼
