Unifying data and control transformations for distributed shared-memory machines

摘要: We present a unified approach to locality optimization that employs both data and control transformations. Data transformations include changing the array layout in memory. Control involve execution order of programs. have developed new techniques for compiler optimizations distributed shared-memory machines, although same can be used sequential machines with memory hierarchy.Our are based on an algebraic representation mappings model. pure transformation algorithm unifying While there has been much work transformations, opportunities largely neglected. In fact, advantage being applicable programs cannot optimized The algorithm, which performs simultaneously, offers improvement over obtained by applying separately.The experimental results using set applications parallel machine show improve performance significantly. These further analyzed metrics instrumentation simulation.