[zeromq-dev] (remote/local)_lat.cpp versus the other languages

Martin Sustrik sustrik at fastmq.com
Wed Apr 8 09:32:30 CEST 2009


Hi Michael,

> Maybe I am missing something, but is there a reason why remote_lat and
> local_lat seem to be switched in the cpp versions compared to the
> other languages?  This is particularly confusing because this is the
> version any documentation discusses, but when one attempts to
> experiment with the other languages the performance test calls are
> reversed.

True. C++ version of the tests tries to be smart and uses 
synchronisation messages to get "run local first, remote afterwards" 
behaviour. Perf tests for other languages are simpler, using no 
synchronisation and thus require following treatment:

1. latency test - run remote first, local afterwards
2. throughput test - run local first, remote afterwards

You are right, it's confusing. We should drop the synchronisation code 
from C++ tests in the future.

> I am glad to see that the CLR binding works well under Mono, though.

:)

Martin



More information about the zeromq-dev mailing list