As I understand it, these techniques are oriented towards producer-consumer queues where there is only 1 producer and only 1 consumer. My software does IO in this manner. What is confusing is that these schemes don't scale beyond multiple cores for obvious reasons (1 consumer, 1 producer).
I understand that I can reduce the few ms latency when acquiring a std::mutex, and perhaps this is valuable for high frequency trading, but I am not sure if this will improve performance in a palpable manner.
Does anybody have any good uses cases for lock-free programming where the goal is to scale beyond a single producer-consumer scheme?
I understand that I can reduce the few ms latency when acquiring a std::mutex, and perhaps this is valuable for high frequency trading, but I am not sure if this will improve performance in a palpable manner.
Does anybody have any good uses cases for lock-free programming where the goal is to scale beyond a single producer-consumer scheme?