Profile Guided Optimization without Profiles: A Machine Learning Approach

TL;DR

This paper introduces a machine learning-based method to predict branch probabilities for compiler optimization, eliminating the need for profile data and improving program performance without profiling overhead.

Contribution

A novel statistical approach that infers branch probabilities from large binary datasets, integrated into LLVM to enhance optimization without profile collection.

Findings

Achieves performance gains over non-profiled compilation.

Requires no profiling runs during deployment.

Has negligible impact on compilation time.

Abstract

Profile guided optimization is an effective technique for improving the optimization ability of compilers based on dynamic behavior, but collecting profile data is expensive, cumbersome, and requires regular updating to remain fresh. We present a novel statistical approach to inferring branch probabilities that improves the performance of programs that are compiled without profile guided optimizations. We perform offline training using information that is collected from a large corpus of binaries that have branch probabilities information. The learned model is used by the compiler to predict the branch probabilities of regular uninstrumented programs, which the compiler can then use to inform optimization decisions. We integrate our technique directly in LLVM, supplementing the existing human-engineered compiler heuristics. We evaluate our technique on a suite of benchmarks, demonstrating some gains over compiling without profile information. In deployment, our technique requires no profiling runs and has negligible effect on compilation time.