Should we apply old-school multi-core scheduling to GPUs? | Not Hacker News!