Google, IBM & Meta Certificates – 40% Off
One plan covers every Professional Certificate on Coursera.
Unlock All Certificates
Watch an 11-minute conference talk exploring a solution for cross-architecture dynamic migration across heterogeneous AI acceleration systems. Learn how CMCC addresses the challenge of applications being locked into vendor-specific tool chains when using diverse AI accelerators like GPU, FPGA, NPU, and TPU. Discover their innovative approach that includes a unified abstraction layer for different accelerators, a cross-architecture compiler integrated with existing compilers, and an adaptive runtime supporting dynamic device identification. Understand how this solution enables seamless application migration between various accelerators without requiring code rewrites or changes to development practices. Presented by Sheng Wang and Qihui Zhao, the talk covers the system's architectural design, technical modules, open source initiatives, and includes a demonstration of their minimal system implementation.
![CNCF [Cloud Native Computing Foundation]](https://www.classcentral.com/cdn-cgi/image/height=40,format=auto,quality=85/images/logos/institutions/cncf-cloud-native-computing-foundation-hz.png){kind=small}
