"Martijn Sipkema" <[EMAIL PROTECTED]> writes: > (the lock-free ringbuffer is a special case since the only > synchronizing that is done, i.e. read/write access to an int, is > done by the processor. I'm not convinced that this will work on all > architectures)
I can't think of any architecture on which concurrent reads and writes of a single int do not work. Do you know an example? Note that the ringbuffer does *not* require "strong ordering" of storage operations, just atomicity of reads and writes to a single location, so you either get the new value or the old one and not some mixed-up in-between garbage. There are many more important portability issues to worry about than that one. But comparison, assuming that mutexes always provide priority inheritance is highly questionable and non-portable. -- joq
