I did do an actual lock-free MPMC ring buffer implementation as an exercise. I used that to make blocking bounded queues using various synchronization mechanism

I did do an actual lock-free MPMC ring buffer implementation as an exercise. I used that to make blocking bounded queues using various synchronization mechanisms, mutex/condvars and eventcounts among others. The eventcount version runs about 8x faster than the mutex version.