> Hi,
>
[snip snip]
> > [me]
> >It would be nice to see how you could get MPI and DIPC working in the same
> >application.
>
> By "MPI", do you mean _any_ message passing mechanism? I thought you are
> refering to a specific software package that is also called MPI. In light of
> this new interpretation, I completely agree with your points in making the
> message passing primitives fatser in order to get better performance in
> systems like DIPC.
Well I was specificly refering to MPI, however, I would want to add the
nessassary features to make any message passing system faster. I would
think that MPI would be a smart test case.
For example, if I were to add functionality to the kernel that allowed a
faste-route from the NIC driver directly into a userspace message handler,
that would help just about any for of over the network message passing.
Doing so would require userspace support, I just suggest that MPI would be
a good place to start.
> -Kamran Karimi
>
Gregory Maxwell
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu