REVIEW 1 cited by
Profile Guided Optimization without Profiles: A Machine Learning Approach
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Profile Guided Optimization without Profiles: A Machine Learning Approach
read the original 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.
Forward citations
Cited by 1 Pith paper
-
GRACE: Cluster-Specific Sequence Reuse for Compiler Auto-Tuning
GRACE pre-computes cluster-specific compiler-pass sequences via contrastive embeddings and evolution, cutting LLVM IR instruction count about 10% over opt -Oz in under 1 second per program.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.