Large language models show promise for code generation, yet their effectiveness in producing high-performance GPU kernels remains insufficiently understood. Recent agentic workflows improve over direct prompting through specialized generation roles and execution feedback, but our characterization of a representative workflow reveals failures in optimization-strategy selection, hardware-specific instruction realization, and shape-dependent parameter selection. We present AlphaCu, a transformation-driven synthesis framework for LLM-based GPU kernel generation. AlphaCu treats kernel generation as structured synthesis over explicit optimization decisions rather than end-to-end code emission. Specifically, AlphaCu represents these decisions as scoped transformation passes maintained across agent-coordinated synthesis stages, enabling compatibility checking and failure attribution. AlphaCu binds selected hardware-sensitive operations to pre-validated microkernels and uses target-device measurements to revise transformation choices and tune runtime configurations. We evaluate AlphaCu on 104 test cases covering single and fused operators across four GPU architectures: NVIDIA H100, A100, V100, and AMD MI210. Under comparable generation budgets, AlphaCu achieves 100% functional correctness on our benchmark suite and delivers an average speedup of 5.64× over prior LLM workflows, while remaining competitive with vendor libraries and tensor compilers.

@inproceedings{paperbot12564,
  title = {AlphaCu: A Transformation-Driven Synthesis Framework for LLM-Based GPU Kernel Generation},
  author = {Yangjie Zhou and Xing Ma and Zihan Liu and Qing Wang and Weihao Cui and Wu Sun and Hanjing Wang and Binhang Qi and Jingwen Leng and Yun Lin and Minyi Guo and Jin Song Dong},
  booktitle = {ASE},
  year = {2026},
  doi = {10.1145/3832783.3837433}
}