Re: Message passing in user-land

[Niels Möller]

rreale@iol.it writes:

> But how much performance imprevement would we achieve if we created such
> a mechanism, only for pure data?

The fastest way to do it that I can imagine, for the simplest case of
mutually trusting processes, is to have a (possibly) short message
queue, protected by a mutex and some condition variables, all in th
eshared memory page. It's not obvious to me if that will be slower or
faster than L4 ipc that can pass the data in registers.

Furthermore, I have never implemented mutexes and condition variables,
but I wouldn't be surprised if some syscall is needed for conditions
to propagate between threads or processes.

Regards,
/Niels