QiMeng-MuPa: Mutual-Supervised Learning for Sequential-to-Parallel Code Translation
This addresses the challenge of automated parallel programming for HPC developers, offering a novel solution to functional equivalence, though it builds incrementally on back-translation methods.
The paper tackles the problem of ensuring functional equivalence in machine learning-based sequential-to-parallel code translation by proposing QiMeng-MuPa, a mutual-supervised learning framework with a Translator and Tester that iteratively improve each other. Experimental results show it improves Pass@1 by up to 28.91%, boosts Tester performance by 68.90%, and outperforms the previous state-of-the-art method CodeRosetta in BLEU and CodeBLEU scores.
The rise of GPU-based high-performance computing (HPC) has driven the widespread adoption of parallel programming models such as CUDA. Yet, the inherent complexity of parallel programming creates a demand for the automated sequential-to-parallel approaches. However, data scarcity poses a significant challenge for machine learning-based sequential-to-parallel code translation. Although recent back-translation methods show promise, they still fail to ensure functional equivalence in the translated code. In this paper, we propose \textbf{QiMeng-MuPa}, a novel \textbf{Mu}tual-Supervised Learning framework for Sequential-to-\textbf{Pa}rallel code translation, to address the functional equivalence issue. QiMeng-MuPa consists of two models, a Translator and a Tester. Through an iterative loop consisting of Co-verify and Co-evolve steps, the Translator and the Tester mutually generate data for each other and improve collectively. The Tester generates unit tests to verify and filter functionally equivalent translated code, thereby evolving the Translator, while the Translator generates translated code as augmented input to evolve the Tester. Experimental results demonstrate that QiMeng-MuPa significantly enhances the performance of the base models: when applied to Qwen2.5-Coder, it not only improves Pass@1 by up to 28.91% and boosts Tester performance by 68.90%, but also outperforms the previous state-of-the-art method CodeRosetta by 1.56 and 6.92 in BLEU and CodeBLEU scores, while achieving performance comparable to DeepSeek-R1 and GPT-4.1. Our code is available at https://github.com/kcxain/mupa.