Published October 2, 2026
| Version v1
Dataset
Open
Reproduction of Optimization Bugs across GCC and LLVM
Authors/Creators
Description
This dataset is used to understand the optimization bugs across GCC and LLVM. Particularly, it contains scripts and prompts to generate PoC using GPT-4o for those bugs without original repository PoC.
The bug information is derived from publicly available GCC Bugzilla and LLVM issue records identified through prior compiler testing papers. The top level source paper directories preserve the relationship between the collected bugs and the papers from which their identifiers were obtained.
Online GCC and LLVM records may change after this artifact is released; the files in this repository preserve the information collected for the study.