SCALABLE CROSS-MODULE OPTIMIZATION

Large applications are typically partitioned into separately compiled modules. Large performance gains in these applications are available b y optimizing across module boundaries. One barrier to applying cross-m odule optimization (CMO) to large applications is the potentially enor mous amount of time and space consumed by the optimization process. We describe a framework for scalable CMO that provides large gains in pe rformance on applications that contain millions of lines of code. Two major techniques are described. First, careful management of in-memory data structures results in sub-linear memory occupancy when compared to the number of lines of code being optimized. Second, profile data i s used to focus optimization effort on the performance-critical portio ns of applications. We also present practical issues that arise in dep loying this framework in a production environment. These issues includ e debuggability and compatibility with existing development tools, suc h as make. Our framework is deployed in Hewlett-Packard's (HP) UNIX co mpiler products and speeds up shipped independent software vendors' ap plications by as much as 71%.