The paper explores using large language models (LLMs) to replace traditional compiler backends, a process termed AI lowering. An LLM agent translates Triton kernels directly into NVIDIA PTX, achieving 0.83x–3.34x the performance of autotuned Triton on a variety of GPUs and ML kernels. The study also extends a PTX verifier to support modern GPU features, highlighting the potential for AI compilers to reduce engineering effort for new hardware.
By Fran\c{c}ois Costa, Charly Castes, Thomas Bourgeat, Azalia Mirhoseini
arXiv:2606. 02963v1 Announce Type: new Abstract: Production inference increasingly targets a heterogeneous mix of accelerators.
By Taras Sereda, Burak Bartan, Ankita Nayak, Tom St. John, Natalie Serrino, Zain Asgar
CUDA‑Harness is a framework that enables the generation and optimization of CUDA kernels directly from natural language. It introduces Intermediate‑Structured Generation to bridge high‑level semantics with low‑level kernel code, uses Synthesis‑Based Verification to mitigate reward hacking by providing isolated test data, and employs Feedback‑Adaptive Evolution to prioritize correctness while improving performance. Experiments show the approach generalizes across different large language models, hardware platforms, and even supports C‑to‑CUDA transpilation.
By Qi Fan, An Zou, Yehan Ma
arXiv:2607. 24762v1 Announce Type: new Abstract: Machine learning models are increasingly embedded in everyday software, and most of their runtime is spent in a small set of compute kernels such as matrix multiplication, convolution, and normalization.
By Joshua Brodsky, Dhravid Kumar, Savini Kashmira, Jayanaka Danatanarayana, Jason Mars, Krisztian Flautner, Lingjia Tang
arXiv:2604. 01489v2 Announce Type: replace Abstract: High-performance GPU kernels are critical to modern machine learning systems, yet developing them remains a manual, expert-driven process.
By Tara Saba, Zhiyang Chen, Jikai Jason Li, Anne Ouyang, Xujie Si, Fan Long
KernelOPT is a multi‑agent system that optimizes GPU kernels generated by compilers like PyTorch Inductor by treating compiled models as structured artifacts. It preserves vendor library calls and focuses on Triton sub‑kernels, using five profiling‑guided LLM agents and a four‑gate verification cascade to ensure correctness and performance before re‑stitching the model. On 250 KernelBench problems, KernelOPT achieves geometric mean speedups of 1.40×, 1.15×, and 1.07× over torch.compile at three optimization levels.
By Aheli Poddar, Sanskar Prasad, Arindam Samanta, Subha Chakraborty, Vishal Goyal, Rohit Singh Rathaur
arXiv:2607. 01590v1 Announce Type: new Abstract: Developing high-performance kernels for Neural Processing Units (NPUs) is a critical industry bottleneck, requiring developers to manually navigate implicit hardware constraints and strict memory hierarchies.
By Junyi Wen, Ruiyan Zhuang, Yongjia Xu, Pengtu Li, Rui Zou, Hongyi Chen, Chingman Wan, Puxu Yang, Wuhui Chen, Yanlin Wang
arXiv:2607. 16241v1 Announce Type: cross Abstract: Recent large language models (LLMs) can generate custom CUDA kernels that appear to outperform PyTorch on benchmarks such as KernelBench.
By Yunxiang Zhang (Xiangjun), Ping Yu (Xiangjun), Jianyu Wang (Xiangjun), Max (Xiangjun), Fan, Julian Reed, Azalia Mirhoseini, Will Su
arXiv:2512. 23236v4 Announce Type: replace-cross Abstract: Making deep learning recommendation model (DLRM) training and inference fast and efficient is important.
By Gang Liao, Hongsen Qin, Ying Wang, Alicia Golden, Michael Kuchnik, Yavuz Yetim, Jia Jiunn Ang, Chunli Fu, Yihan He, Samuel Hsia, Zewei Jiang, Dianshi Li, Uladzimir Pashkevich, Varna Puvvada, Feng Shi, Matt Steiner, Ruichao Xiao, Liyuan Li, Nathan Yan, Xiayu Yu, Zhou Fang, Roman Levenstein, Kunming Ho, Haishan Zhu, Alec Hammond, Richard Li, Ajit Mathews, Kaustubh Gondkar, Abdul Zainul-Abedin, Ketan Singh, Hongtao Yu, Wenyuan Chi, Barney Huang, Sean Zhang, Noah Weller, Zach Marine, Wyatt Cook, Carole-Jean Wu, Gaoxiang Liu
arXiv:2607. 20518v1 Announce Type: new Abstract: AI agents are now capable of writing, compiling, and iteratively optimizing low-level operator kernels on different hardware platforms.
By Xue-Jian Gao, Deng Pan, Yueming Su, Jiasheng Li, Bin Du, Fengming Zhu, Chengdi Ma, Junyi Fan, Qichen Liao, Chengqiu Hu, Xinxian Chen, Lingchao Zheng, Jun Li, Jiwei Yang, Yuwei Fan
KernelOPT is a multi-agent system that optimizes GPU kernels generated by compilers like PyTorch Inductor by treating compiled models as structured artifacts. It preserves vendor library calls and focuses on Triton sub-kernels, using five profiling-guided LLM agents and a four-gate verification cascade to filter and validate candidates. On 250 KernelBench problems, KernelOPT achieves geometric mean speedups of 1.40×, 1.15×, and 1.07× over torch.compile at different optimization levels.
arXiv:2606. 16231v1 Announce Type: cross Abstract: High-performance CUDA kernels are essential for scalable AI systems, while Large Language Models (LLMs) still struggle to generate correct kernels due to strict and implicit execution constraints.
By Wentao Chen, Jiace Zhu, Xing Zhe Chai, Zeng Qu, Qiaoling Xiao, Liucheng Duan, An Zou