Nathaniel Wen:
Hi, everyone:
I'm reading the source code of "AbstractQueuedSynchronizer", getting a little confused of the reason why "fast path" is faster than full enq in the comment "Try the fast path of enq; backup to full enq on failure".
David Holmes:
What version of the code are you looking at? I can’t see anything like that in current OpenJDK sources.
Nathaniel Wen:
My JDK version is 1.8
private Node addWaiter(Node mode) {
Node node = new Node(Thread.currentThread(), mode);
// Try the fast path of enq; backup to full enq on failure
Node pred = tail;
if (pred != null) {
node.prev = pred;
if (compareAndSetTail(pred, node)) {
pred.next = node;
return node;
}
}
enq(node);
return node;
}
David Holmes:
The fast path just does a direct enqueue to the existing tail. The enq(), in the worst case, has to initialize the queue, and in general deals with contention due to concurrent enqueue attempts.
The code is different now.
Nathaniel Wen:
The full enq will loop only once when there is no contention and the tail isn't null. The reason why "fast-path" is faster is because in most cases the tail isn't null. So "if (tail != null)" is more efficient than "if (tail == null) ... else ..." in this very case, am I right?
David Holmes:
The old structure was intended to be more amenable to inlining the fast-path code IIRC. The new code doesn’t bother AFAICS.
Martin Buchholz:
Alternatively: initializeSyncQueue is now a separate cold-code method.
(All of this is unlikely to matter much)
网友评论