Kernel launch latency

[kouchy]

Hi @ProjectPhysX,

Thank you very much for this benchmark. It's very concise, exactly what I was looking for.

A small suggestion here: what do you think about adding "kernel launch latency" as a new metric?
In our research it is of interest because we want to run CPU code then GPU code then CPU code and so on.

If you are interested in it we can discuss more.

Thank you again for the good work!

Regards