CODEWITHSUNDEEP
network-programmingzeromqnonblocking
AberrantWolf
AberrantWolf

Reputation: 506

Using zmq_recv and ZMQ_NOBLOCK versus zmq_poll for one socket

There seem to be two main ways to asynchronously get data from your socket in ZeroMQ:

if(zmq_recv(&msg, ZMQ_NOBLOCK) == 0) { // could return EAGAIN or others
    // do stuff
}

...and...

if(zmq_poll(&pollItems[0], num, timeout)) {
    if(zmq_poll(&msg) == 0) {
        // do stuff
    }
}

If I only have one socket that I want to read from, is there a compelling reason I should opt for the version that uses zmq_poll rather than zmq_recv with ZMQ_NOBLOCK?

Upvotes: 1

Views: 2127

Answers (1)

nos
nos

Reputation: 229274

You would rarely do a zmq_recv(&msg, ZMQ_NOBLOCK) without using zmq_poll

Normally you would read in a loop;

while (!done) {
...
  zmq_recv(&msg, ZMQ_NOBLOCK);
... 
}

This is a busy wait, it uses CPU unnecessarily. zmq_poll() blocks, and don't consume CPU while it's waiting for something to happen.

If you only have one socket, and don't want to use zmq_poll, you would normally use a blocking zmq_recv, and set a ZMQ_RCVTIMEO so you don't block forever if something has gone bad at the sending side.

Upvotes: 2

Related Questions