TOWARDS AUTOMATED KERNEL GENERATION IN THE ERA OF LLMS: A SURVEY

The performance of modern AI systems is fundamentally constrained by the quality of their underlying GPU kernels, which dominate the cost of large-scale training and inference. Achieving near-optimal kernels requires expert-level understanding of GPU architectures, programming models, and intricate hardware-software trade-offs, making kernel engineering a critical but notoriously time-consuming and non-scalable process. Recent advances in large language models (LLMs) and agentic LLM systems have opened new possibilities for automating kernel generation and optimization. LLMs are well-suited to compress expert-level kernel engineering knowledge that is difficult to formalize and reuse, while agentic LLM systems further enable scalable optimization by casting kernel development as an iterative, feedback-driven loop involving code generation and performance-guided refinement. Together, they offer a practical pathway toward scalable and automated kernel engineering. Rapid progress has been made in this area. However, the field remains fragmented, lacking a systematic perspective for LLM-driven kernel generation. This survey addresses this gap by providing a structured overview of existing approaches, spanning LLM-based approaches and agentic optimization workflows, and systematically compiling the datasets and benchmarks that underpin learning and evaluation in this domain. Moreover, key open challenges and future research directions are further outlined, aiming to establish a comprehensive reference for the next generation of automated kernel optimization.