Hi Dmitry,
Thank you for your reply, I just added the mpmc_queue_pack<> which is a template containing a group of mpmc_queue<>s into a single queue, which gets the mpmc queue to be closer to wait-free.
On the mailing list I hope to get feedback, if there are bugs in the code, and comparison benchmarks of the queue compare to other queues.
In short the mpmc_queue<> is a fast, lockless bounded queue, using double word CAS, can be use between processes.
Thanks,
Erez