I reached the (my) conclusion that RCU doesn't really help with my implementation of a concurrent hash.
My design is based upon the ideas presented in A lock-free hash table. (Note. the premise is that the implementation does not require internal locks --though there will be some internal spin-waits.)
That doesn't mean that the user will not have the ability to lock both individual key/value pairs for those situations where the new value of a key is derived from the old value. I'm also reaching the conclusion that I will have to provide the ability to lock the entire hash.
In reply to Re^2: Design thoughts: iterator invalidation
by BrowserUk
in thread Design thoughts: iterator invalidation
by BrowserUk
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |