In theory, can_read() is more efficient in cases where you have a large number of clients that remains mostly constant between calls to can_read(), since add() and remove() should be performing only bit operations, wheras IO::Select::select() re-creates the bit mask every time.
In practice, I wouldn't be able to guess which one is faster. Graham Barr (the author of IO::Select) does implement add() and remove() using bit operations, however, he has a lot of rather inefficient code between add(), remove() and the code that actually does the bit operations.
Personally, I make the bit fields myself. The code isn't that complicated, and I handle things that can_read() does not such EINTR. add() is implemented in terms of vec(...) = 1, etc.
In reply to Re: IO::Select: select vs. can_read
by MarkM
in thread IO::Select: select vs. can_read
by Coruscate
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |