Performance

The easiest way to make KDIS faster is to compile it with optimization. Using CMake, you need to specify the release build type.

% cmake -DCMAKE_BUILD_TYPE=Release ../KDIS

(there is something similar to be done for windows)

I always suggest that you profile the code prior to making any other changes, as performance optimisations are non-intuitive and rarely work like you think.

If you are dead-reckoning a lot of PDU's (say to provide to a visual system at 60hz), and you are using an i386 or x86_64 processor, enabling SMD instruction generation will make the code much, much faster. Google for how to do that with your compiler. You should not need to touch any of the source code to get this speedup, just change the compilation options.

Missing PDU's is a common problem. Most cases, increasing the OS socket buffer size is not the solution. Best solution is to create a thread to read from the socket. Make this thread blocking, run it at a real time priority and scheduling discipline (man sched_setscheduler). You will need to use locks for synchronization between threads, but you should never 'miss' a PDU coming in.