Intel® 64 and IA-32 Architectures Optimization Reference Manual

APPLICATION PERFORMANCE TOOLSA.1.6.1Interprocedural Optimization (IPO)You can use the /Qip (-ip in Linux and Mac OS) option to analyze your code and applyoptimizations between procedures within each source file. Use multifile IPO with/Qipo (-ipo in Linux and Mac OS) to enable the optimizations between procedures inseparate source files.A.1.6.2Profile-Guided Optimization (PGO)Creates an instrumented program from your source code and special code from thecompiler. Each time this instrumented code is executed, the compiler generates adynamic information file. When you compile a second time, the dynamic informationfiles are merged into a summary file. Using the profile information in this file, thecompiler attempts to optimize the execution of the most heavily travelled paths inthe program.Profile-guided optimization is particularly beneficial for the Pentium 4 and Intel Xeonprocessor family. It greatly enhances the optimization decisions the compiler makesregarding instruction cache utilization and memory paging. Also, because PGO usesexecution-time information to guide the optimizations, branch-prediction can besignificantly enhanced by reordering branches and basic blocks to keep the mostcommonly used paths in the microarchitecture pipeline, as well as generating theappropriate branch-hints for the processor.When you use PGO, consider the following guidelines:• Minimize the changes to your program after instrumented execution and beforefeedback compilation. During feedback compilation, the compiler ignoresdynamic information for functions modified after that information was generated.NOTEThe compiler issues a warning that the dynamic informationcorresponds to a modified function.• Repeat the instrumentation compilation if you make many changes to yoursource files after execution and before feedback compilation.For more on code optimization options, see the Intel C++ Compiler documentation.A.1.7Auto-Generation of Vectorized CodeThis section covers several high-level language examples that programmers canuse Intel Compiler to generate vectorized code automatically.A-6