|
Boost : |
From: Peter Dimov (pdimov_at_[hidden])
Date: 2004-09-07 07:19:23
Joel Salomon wrote:
> John Torjo wrote:
>> Use mutexes.
> and
>> In my experience, I hardly need just a container to be thead-safe. I
>> usually need more data in addition to the container (which needs to
>> be thread-safe). For instance, an extra latest_access_time, out_file,
>> used_for_the_first_time, etc.
>>
>> Thus, to me, the benefit of having a thread-safe container is very
>> small.
> One thread-safe container that might be useful is a channel - a queue
> that one thread writes to and another reads to, the reader blocking
> when the channel is empty.
Right. This is the _only_ thread-safe(*) container (that I'm aware of) that
is generally useful. Even here, there is a choice between non-blocking and
blocking semantics (which preclude a lock-free implementation.) Having a max
size implies blocking on push, of course.
(*) For the implied definition of "thread safe" that we are using in this
context ("strong" thread safety).
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk