Lines Matching refs:a

8 threads. Each logical thread is serviced for a *time slice* by a
9 physical thread. If a thread runs longer than a time slice, as most do,
25 reality of set-associative caches is a bit more complicated, but this is
26 not a cache primer.) When a logical thread gets its time slice, as it
27 references a piece of data for the first time, this data will be pulled
30 cache, and only take a few cycles. Such data is called "hot in cache".
31 Time slicing undoes this, because if a thread A finishes its time slice,
36 worse yet, the next time slice for thread A may be on a different
37 physical thread that has a different cache altogether.
40 Another cost is *lock preemption.* This happens if a thread acquires a
41 lock on a resource, and its time slice runs out before it releases the
42 lock. No matter how short a time the thread intended to hold the lock,
44 turn at a time slice to come up. Any other threads waiting on the lock