ParEVO：通过自主进化合成不规则数据的代码：高性能并行性

The transition from sequential to parallel computing is crucial for modern high-performance applications, yet it is impeded by the steep learning curve associated with concurrent programming. This issue is particularly pronounced for irregular data structures, such as sparse graphs, unbalanced trees, and non-uniform meshes, where static scheduling fails and data dependencies are unpredictable. Current Large Language Models (LLMs) often perform poorly on these tasks, producing code that is susceptible to subtle race conditions, deadlocks, and sub-optimal scaling. ParEVO is introduced as a framework designed to synthesize high-performance parallel algorithms specifically for irregular data. Key contributions include the Parlay-Instruct Corpus, a curated dataset comprising 1,382 examples aimed at supporting this objective.