1 paper
Zhenyu Bai, Dan Wu, Pranav Dangi +3
Current approaches to scheduling workloads on heterogeneous systems with specialized accelerators often rely on manual partitioning, offloading tasks with specific compute patterns…