Re: Lets get this right (WAS RE:MOSIX and kernel mods)

Richard Gooch (rgooch@atnf.csiro.au)
Sun, 7 Mar 1999 19:19:33 +1100

Larry McVoy writes:
> : Just how smart have they made this optimisation, BTW? For the volume
> : rendering application, you have two conceptual transfers of data: the
> : volume data is "shipped" to worker nodes and they in turn "ship"
> : subimage data back to the scheduling node.
>
> If I recall correctly, they actually wacked the OS. They used /proc
> and made a way that another process could map a different process'
> address space. Sends can be implemented, on page sized quantities,
> as mapping a page COW into another process' address space.
>
> Zero copy for big data that wasn't touched, otherwise copy on write.

Sounds good for the different processes case. But I was thinking of
the shared VM case (threads), where you don't (ideally) even have to
do that.

Perhaps the best approach is to define extended send semantics to
"send and I don't mind if my buffer is modified". That would be much
simpler than stuffing around in the VM layer to do all kinds of
tricks. Tricks that probably won't work in a shared VM environment
anyway.

> For small data, it was just a copy and a wakeup.

How small is "small"?

Regards,

Richard....

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/