currently the throughput of one message end to end is about 50 micro-seconds, two queues (input_q and outgoing_q) take more than 10 micro-seconds out of the 50...
we are using this queue for trading platform's backend server.
the queue itself will cost about 6 micro-seconds to enqueue and dequeue a message, if we want to introduce a new layer into the design...
we cannot have reader thread keep running without waiting, and we want to have better throughput by read any messages enqueued as soon as possible, currently we use pthread_condwait() and...
one question for the idea of one local queue per writer, how can we avoid the reader thread to avoid blocking by a writer which is a bit free?
and we...
we have designed a FIFO queue with pthread mutex and its conditionWait(), inside the queue, we use std::deque, which is quite efficient. and we only enqueue & dequeue pointers to objects,...
One question: I cannot tell why a float equal to 333.33 cannot be print correctly to screen? And after we run ff = f1 + f2; the result is not 333.33? does it mean that result of f1 + f2 will be...
the link provided by exterminator explain one more factor, a++ and ++a can be evaluated at the same time, which can mess up the memory. :ehh: so maybe undefined behavior will be more correct than...
I think it is supposed to let programmer only modify the define of SOFTVERSION so that all other part of the codes know that now version changes to a new one.