This may have been mentioned in lecture, but wouldn't using a reader/writer lock solve this problem? Also, for data structures in general, won't hand-over-hand locking with reader/writer locks give one very similar memory semantics as transactional memory (on a high level)?
HingOn
i can see how reader/writer lock provides isolation (since it does block all readers before the writer is done) and atomicity (since each writer gain exclusive access to the node).
ChandlerBing
@scedarbaum I think we are doing something similar in this slide. We wish to modify contents of 3 and 4, but the contents of 1 and 2 are only read, not written. Hence, the two transactions can occur concurrently since there are no read-write or write-write conflicts.
This may have been mentioned in lecture, but wouldn't using a reader/writer lock solve this problem? Also, for data structures in general, won't hand-over-hand locking with reader/writer locks give one very similar memory semantics as transactional memory (on a high level)?
i can see how reader/writer lock provides isolation (since it does block all readers before the writer is done) and atomicity (since each writer gain exclusive access to the node).
@scedarbaum I think we are doing something similar in this slide. We wish to modify contents of 3 and 4, but the contents of 1 and 2 are only read, not written. Hence, the two transactions can occur concurrently since there are no read-write or write-write conflicts.