A scend K ernel G en: LLM -Driven Kernel Generation for NPU s
Xinzi Cao · Jianyang Zhai · Pengfei Li · Zhiheng Hu · Cen Yan · Mubingxu · Guanghuan Fang · Bin She · Jiayu Li · Yihan Su · Dongyang Tao · Feidiao Yang · Chang-Dong Wang · Yutong Lu · Weicheng Xue · Bin Zhou · Yonghong Tian
摘要
Neural Processing Units (NPUs) are critical for AI infrastructure, yet developing kernels remains a bottleneck due to the complexity of vendor-specific Domain-Specific Languages (DSLs). While LLMs excel in general coding, they fail to meet the stringent constraints of NPU development, showing a n...