Archer: Towards Agentic Review for Compiler Optimizations

2026-07-02

AI Summary

Modern compilers undergo frequent updates, yet manual code review struggles to comprehensively scrutinize all optimization patches, often leading to undetected semantic bugs or delayed integration. To address this challenge, this work proposes Archer—the first automated agent-based tool specifically designed for reviewing compiler optimizations. Archer innovatively integrates obligation-guided analysis with a deterministic verification mechanism, accepting only those bug reports substantiated by executable evidence to ensure reliability and reproducibility. Implemented on LLVM, Archer identified semantic errors in 21% of 70 recent open pull requests and 11% of 328 already merged ones, substantially revealing critical gaps in current review practices and demonstrating its effectiveness and practical utility.

Abstract

Modern compilers are frequently updated, but expert review capacity is highly limited, leading to delayed integration and, in some cases, subtle semantic bugs entering the compiler codebase. Automating the code review process with modern general code review agents may be feasible, but it faces critical challenges due to compiler complexity. In this paper, we use LLVM as our target compiler and present Archer, the first automated agentic code review tool for compiler optimizations. Archer constrains the agentic review process from both ends by using obligations to guide analysis and a deterministic validation guard to admit only findings backed by executable evidence. We evaluated Archer on 70 open PRs and 328 closed PRs in LLVM from the last two months. The review results are shocking and concerning: Archer discovers that 21% of open PRs and 11% of closed PRs are buggy, i.e, introducing semantic bugs such as miscompilations in LLVM. Our findings expose a critical gap in the capacity for critical review in large compiler projects and demonstrate the practical value of Archer as an additional reviewer.

Problem

• compiler optimizations
• code review
• semantic bugs
• LLVM
• miscompilations

Innovation

• agentic code review
• compiler optimization
• LLVM
• deterministic validation
• obligation-guided analysis