← Back to Model Beat
4Hardware·5h ago

Hawk: Harnessing Hardware-Aware Knowledge for High-Performance NPU Kernel Generation

Researchers have introduced Hawk, a framework designed to automate the generation of high-performance kernels for Neural Processing Units. By integrating hardware-specific constraints into the development process, the system aims to reduce the manual labor currently required to optimize code for complex memory hierarchies.

Covered by 1 source

  • AarXiv CS.AIJunyi Wen, Ruiyan Zhuang, Yongjia Xu, Pengtu Li, Rui Zou, Hongyi Chen, Chingman Wan, Puxu Yang, Wuhui Chen, Yanlin Wang5h ago

Related stories

HardwareAnthropic in Talks With Samsung for Custom AI Chip: InformationJul 2 · 3 sourcesHardwareHow NVIDIA’s Inference Software Stack Powers the Lowest Token CostJun 30 · 4 sourcesHardwareMeta Is Planning a Cloud Business to Sell AI Computing PowerJul 1 · 7 sourcesHardwareMeituan's LongCat-2.0 shows China can train massive AI models without NvidiaJun 30 · 2 sources