bofh@snoopy.virtual.net.au said:
> This is interesting. How fast is the I2O controller? Would it be
> possible to do some useful things on the i960 other than controlling
> hardware?
My company has a board that uses the i960RD on a PCI board we build.
We selected it because it has a CPU and PCI circuitry on a single chip.
The i960 in this case is aa 66MHZ (2X the PCI clock) and frankly is not
much good for other then shepherding connections around. It works very
well in our board because we have lots of autonomous devices on its
secondary side doing the image processing. I essentially hope to use it
to make the board appear more palattable to the host.
My experience with this I/O processor stuff is that it is a mess that has
rather limited appeal. The I/O processor and the requisite message passing
introduce latencies, and besides a 600MHz alpha is a *hell* of a lot faster
at processing TCP packets then an attached processor will ever be. Also,
cards nowadays tend to have just about the right amount of smarts, in order
to reduce the interrupt rate.
If anybody listening has written NT drivers, you might recognize in I2O
the cumbursome, buzzword-laden structure of a device driver interface.
NT has the HAL which is a failed attempt at device-independent drivers
(write a driver for a memory mapped device in x86 and alpha, and you will
know what I mean) that adds some nasty and wasteful overhead. I2O seems
headed in the same direction.
I say we deal with I2O when it arrives, but I would hardly consider it
any kind of a threat.
-- Steve Williams steve@icarus.com steve@picturel.com"The woods are lovely, dark and deep. But I have promises to keep, And lines to code before I sleep, And lines to code before I sleep."