Robert craig
|
Re: Additional interface with QNET
|
Robert craig
12/18/2009 11:20 AM
post44140
|
Re: Additional interface with QNET
I seem to remember that there was a rapidio implementation developed for some Freescale parts (never officially release)
that included a pseudo-ethernet over RapidIO driver. It might be worthwhile talking to your sales rep. to see if you
can get the source.
R.
|
|
|
Andrew Boyd(deleted)
|
RE: Additional interface with QNET
|
Andrew Boyd(deleted)
12/18/2009 1:07 PM
post44145
|
RE: Additional interface with QNET
> run Qnet over IP
Right.
Most people use Qnet in a local, direct-connected
fashion. This works well for most applications
because of the thousands of small messages that
they send. For decent performance, you need a small
RTT.
Anyways, the key to making your topology work is
if your driver emulates ethernet, complete with
the hoary old 14 byte MAC header of two addresses
with a protocol type.
See, if your driver looks just like an ethernet
device to the higher-level protocols - ie the
protocol transmits and receives stock ethernet
packets - then all sorts of stuff starts to work,
like Qnet over IP. See the bind option.
Qnet over IP encapsulates the qnet packet in an
IP packet and thus allows it to be routed. Of
course, there is a performance hit to this - see
RTT mumblings above - but you don't get something
for nothing.
--
aboyd
|
|
|