Scheduling – First Principles
Thread(T1) cpu – sync – I/O – kernel scheduler
FCFS, Highest static priority, highest dynamic priority, thread whose memory contents are in the cpu cache
Memory hierarchy refresher
cpu – l1-chache (1-2 cycles)- l2-chache(～10cycles) -> memory(～100 cycles)
cache affinity scheduling
T1 descheduled, t1 rescheduled
interventing threads on P1
FCFS: Ignores affinity for fairness
Fixed proccessor: Ti always on P fixed
Last processor: Ti on P last
Minimum Intervening: Ti -> Pj ^ Imi
implementation issues
queue based
-global que
-affinity-based local queues
Ti’s priority = BPi + agei + affinityi
-Determines poisition in the queue
Performance
Figures of merit
-throughput -> system centric
-response time -> user centric
-variance -> user centric
Multicore multithreaded processors
