The Most Addictive Python and SQL Courses
Python, Prompt Engineering, Data Science — Build the Skills Employers Want Now
Overview
Google, IBM & Meta Certificates — All 10,000+ Courses at 40% Off
One annual plan covers every course and certificate on Coursera. 40% off for a limited time.
Get Full Access
This technical talk explores how UADK (Unified Acceleration Development Kit) has evolved from supporting only hardware accelerators to now including CPU-based accelerators such as Cryptographic Extensions (CE) and Scalable Vector Extensions (SVE) for optimized performance. Learn about the current limitations where only one accelerator type can be used at a time, and discover the roadmap for implementing heterogeneous scheduling that would enable simultaneous use of multiple accelerator types. Understand the technical approach to intelligently distribute workloads across hardware accelerators, CPU instruction-based accelerators, and future acceleration technologies based on their specific capabilities and the nature of processing tasks, ultimately aiming to maximize overall system performance through optimal resource utilization.
Syllabus
LIS25 214 UADK Heterogeneous Scheduling for Hardware and CPU Based Accelerators
Taught by
LinaroOrg