Test performance of CPU-bound Luv programs based on Luv.Integration

[aantron]

See ocsigen/lwt#622 for a description of an analogous problem in Lwt.

Basically, for promise libraries driven by Luv that choose to defer most callbacks to the next tick, there is a risk that select/etc. is called between every tick. If the program is CPU-bound (no I/O completes most of the time), that is a very wasteful call to a syscall on every tick. It will considerably slow down a Luv-based loop, relative to an ordinary synchronous loop.

[aantron]

Luv.Integration is not in Luv anymore and has never been released. It turned out to have been more of a demo.