Alan Antonuk

I've had a chance to do some performance testing using the amqp_producer.c example program: Running on a 64-bit Linux 3.13.0-49-generic #83-Ubuntu SMP kernel, 2xIntel(R) Core(TM)2 Duo CPU T9400 @ 2.53GHz...

That's an interesting solution. I'll give it a try on my end see what kind of performance difference it makes.

After looking at it a bit more, `TCP_CORK` (or equivalent) is a viable way to improve performance by reducing the number of packets transmitted per publish. Some implementation notes: -...

While there is a small amount of overhead in extra syscalls, the performance overhead of the syscall itself is small in comparison to network write (we're talking ~100ns on most...

From my comments in the PR #272: Evaluating this running on Ubuntu 14.04.02 running in a VM on Mac OS X 10.11, talking to RabbitMQ 3.5.2 running different machine Mac...

Understood, thanks for the explanation. I believe that once I can get #272 cleaned up and pushed to master this will improve things for your use case (feel free to...

No. Not without changing the way AMQP works. (This doesn't stop you from putting more data in one message though).

#272 is now in the master branch. This will improve things for those using non-SSL sockets on Linux and BSDs that support TCP_NOPUSH. I have #273 open to do a...

Unfortunately the C-style varags is rather primitive: it only has a method to get the 'next' element, if you run off the end of the provided arguments for example, you...

Do you get the same behavior when you compile against `HEAD`? There have been a number of improvements since v0.11.0 to the build-system that hopefully improve behaviors when added as...