- Category
- Developer Tools
- Rank
- No. 3642Tools index
Previous survey · No. 3420 ·
- Type
- TOOL
- Use case
- Coding
- GitHub
- 6 stars
- Date
About
Jevopt uses the Jev model to decide, at each discretionary LLVM IR call site, whether a function call should be inlined, aiming to reduce compiled binary size for contexts like embedded systems. Evaluated on the 19 Embench 1.0 programs, it produced 7.87% larger aggregate .text size than clang -Oz but beat clang -Oz on 7 of 19 programs, including a 58% reduction on Statemate.
At a glance
Jevopt uses Jev to help decide which C or C++ function calls Clang should inline. It is worth evaluating if binary size matters and you can compare generated code under controlled compiler settings.
Good fit
Developers experimenting with compiler decisions on small C or C++ workloads.
Consider something else when
Keep a conventional compiler setup if builds must be offline or you cannot verify the modified output.
Start here
Choose a synthetic program with a known test suite. Compare binary size and runtime with the same pinned LLVM baseline, and verify that both builds pass the tests.
Cost and requirements
The project targets Linux and LLVM 21, with Python 3.12 or newer, build tools and a TypeSafe API key.
Limits to know
A smaller binary can run slower or differ in behavior. Check which program information reaches the service before using proprietary code.
Sources checked 2026-11-05. Source-based assessment.
Why it made the leaderboard
A concrete, evaluated example of replacing a hand-tuned compiler heuristic with a model-driven decision at a specific optimization point, useful for anyone exploring LLM-guided compiler optimization.
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.