Date: Fri, 20 Jan 2023 15:55:43 +0000
In audio processing, lock-free queues are among the most frequently used data structures. Often, we use single-producer single-consumer queues as they have the desirable property to be wait-free, not just lock-free (a requirement for hard real time programming). But use cases for multi-producer multi-consumer queues exist also in this industry. Some companies roll their own; the two off-the-shelf solutions that I have seen being used in production are Boost and Moodycamel (https://github.com/cameron314/concurrentqueue). I found the Moodycamel one to have better performance than the Boost one.
Hope this is helpful.
On 18 Jan 2023, at 15:49, Honey Sukesan via SG14 <sg14_at_[hidden]> wrote:
Hi Scott,Thank you for your reply.Yes, it's a multithreaded program. I believe multiprocess constructs could be used along side threads as well. I've tried benchmarking using std::queue protected with a std:: unique_lock & condition_var for the same scenario. But performance results was not that great. Let me know what you think.Best,Honey_______________________________________________On Wed 18 Jan 2023, 8:01 a.m. Scott Michaud, <scott_at_[hidden]> wrote:You say that you have a producer-consumer scenario with multiple threads. Are they all in the same executable / process?On Tue, Jan 17, 2023 at 10:52 PM Honey Sukesan via SG14 <sg14_at_[hidden]> wrote:Hi all,_______________________________________________Hope all is well.This is my first email to the list. Not sure if it's the place to ask general technical questions. Please accept my apologies if I'm violating any rules of this study group.I would like to seek an opinion regarding boost:: interprocess:: message_queue vs boost::lockfree::queue.Has anyone used boost::lockfree::queue or any of Boost.LockFree data structures in any embedded platforms?I have a producer-consumer scenario involving multiple threads. Tried out a sample code using int values pushing to/consuming from the queue & analysed elapsed time between push/pop on a Linux x86. It's found boost::lockfree::queue gives better time performance results compared to boost:: interprocess:: message_queue.Haven't tried this on actual hardware target yet. Our hardware target is an ARM based processor running QNX RTOS.I'm listening to Herb's CppCon talk on Lock-free programming now.It would be great if any of you has any other recommendations/pros or cons I need to be aware of before I fix boost::lockfree::queue rather than boost:: interprocess:: message_queue in our production code?Thank you in advance,Best,HoneyHoney Sukesan
SG14 mailing list
SG14 mailing list
Received on 2023-01-20 15:55:48