> I give it up:
> I've attached 2 more ideas, but one of them is a bit slow, and the other
> one requires cmpxchg (ie only 486 and above).
Interesting. Here's what I'm doing (it's subtle, but cute and efficient
-- 1 instruction and an untaken branch for both readers and writers):
count is initially set to a BIAS (I'm using 0x01000000). Readers
decrement count by 1 and see a positive value if successful. Writers
decrement by the BIAS and see a result of zero when acquired.
This makes the recovery case a bit more challenging, but easily solved
with a second value used to decide who is the next writer (and thus
responsible for maintaining the bias while waiting).
-ben
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/