QiMeng-Kernel: Macro-Thinking Micro-Coding Paradigm for LLM-Based High-Performance GPU Kernel Generation
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914170290569216 |
|---|---|
| author | Zhu, Xinguo Peng, Shaohui Guo, Jiaming Chen, Yunji Guo, Qi Wen, Yuanbo Qin, Hang Chen, Ruizhi Zhou, Qirui Gao, Ke Wu, Yanjun Zhao, Chen Li, Ling |
| author_facet | Zhu, Xinguo Peng, Shaohui Guo, Jiaming Chen, Yunji Guo, Qi Wen, Yuanbo Qin, Hang Chen, Ruizhi Zhou, Qirui Gao, Ke Wu, Yanjun Zhao, Chen Li, Ling |
| contents | Developing high-performance GPU kernels is critical for AI and scientific computing, but remains challenging due to its reliance on expert crafting and poor portability. While LLMs offer promise for automation, both general-purpose and finetuned LLMs suffer from two fundamental and conflicting limitations: correctness and efficiency. The key reason is that existing LLM-based approaches directly generate the entire optimized low-level programs, requiring exploration of an extremely vast space encompassing both optimization policies and implementation codes. To address the challenge of exploring an intractable space, we propose Macro Thinking Micro Coding (MTMC), a hierarchical framework inspired by the staged optimization strategy of human experts. It decouples optimization strategy from implementation details, ensuring efficiency through high-level strategy and correctness through low-level implementation. Specifically, Macro Thinking employs reinforcement learning to guide lightweight LLMs in efficiently exploring and learning semantic optimization strategies that maximize hardware utilization. Micro Coding leverages general-purpose LLMs to incrementally implement the stepwise optimization proposals from Macro Thinking, avoiding full-kernel generation errors. Together, they effectively navigate the vast optimization space and intricate implementation details, enabling LLMs for high-performance GPU kernel generation. Comprehensive results on widely adopted benchmarks demonstrate the superior performance of MTMC on GPU kernel generation in both accuracy and running time. On KernelBench, MTMC achieves near 100% and 70% accuracy at Levels 1-2 and 3, over 50% than SOTA general-purpose and domain-finetuned LLMs, with up to 7.3x speedup over LLMs, and 2.2x over expert-optimized PyTorch Eager kernels. On the more challenging TritonBench, MTMC attains up to 59.64% accuracy and 34x speedup. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_20100 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | QiMeng-Kernel: Macro-Thinking Micro-Coding Paradigm for LLM-Based High-Performance GPU Kernel Generation Zhu, Xinguo Peng, Shaohui Guo, Jiaming Chen, Yunji Guo, Qi Wen, Yuanbo Qin, Hang Chen, Ruizhi Zhou, Qirui Gao, Ke Wu, Yanjun Zhao, Chen Li, Ling Distributed, Parallel, and Cluster Computing Computation and Language Developing high-performance GPU kernels is critical for AI and scientific computing, but remains challenging due to its reliance on expert crafting and poor portability. While LLMs offer promise for automation, both general-purpose and finetuned LLMs suffer from two fundamental and conflicting limitations: correctness and efficiency. The key reason is that existing LLM-based approaches directly generate the entire optimized low-level programs, requiring exploration of an extremely vast space encompassing both optimization policies and implementation codes. To address the challenge of exploring an intractable space, we propose Macro Thinking Micro Coding (MTMC), a hierarchical framework inspired by the staged optimization strategy of human experts. It decouples optimization strategy from implementation details, ensuring efficiency through high-level strategy and correctness through low-level implementation. Specifically, Macro Thinking employs reinforcement learning to guide lightweight LLMs in efficiently exploring and learning semantic optimization strategies that maximize hardware utilization. Micro Coding leverages general-purpose LLMs to incrementally implement the stepwise optimization proposals from Macro Thinking, avoiding full-kernel generation errors. Together, they effectively navigate the vast optimization space and intricate implementation details, enabling LLMs for high-performance GPU kernel generation. Comprehensive results on widely adopted benchmarks demonstrate the superior performance of MTMC on GPU kernel generation in both accuracy and running time. On KernelBench, MTMC achieves near 100% and 70% accuracy at Levels 1-2 and 3, over 50% than SOTA general-purpose and domain-finetuned LLMs, with up to 7.3x speedup over LLMs, and 2.2x over expert-optimized PyTorch Eager kernels. On the more challenging TritonBench, MTMC attains up to 59.64% accuracy and 34x speedup. |
| title | QiMeng-Kernel: Macro-Thinking Micro-Coding Paradigm for LLM-Based High-Performance GPU Kernel Generation |
| topic | Distributed, Parallel, and Cluster Computing Computation and Language |
| url | https://arxiv.org/abs/2511.20100 |