>>: pavel@atrey.karlin.mff.cuni.cz (Pavel Machek) said:
>> Hmm - BTW what is the maximum throughput (IP) to be expected from PLIP?
My usual throughput is 42 kB/s NFS copying of files.
>: Dave@imladris.demon.co.uk (David Woodhouse) writes:
> Do we currently use DMA for this kind of thing, when it's available?
No, not yet -- feel free to add appropriate support!
> Also, I'd _love_ to run a getty on a bidirectional parallel port.
> Is that going to be feasible in the near future?
No one's undertook such task yet -- feel free to do!
I'm ready to help/test.
The problem is that the driver architecture would be completely different:
* the current PLIP cable is a 5-bit full-duplex cable, connecting
one computer's output-only (usually) data lines
to the other one's read and read-write control lines.
* In current driver, IO is done by IRQ (packet start) then
polling (packet reception), with 4 data bits (nibbles) and 1 control bit
to acknowledge every nibble.
The architecture is quite suboptimal, but very robust,
and works on just any PC parallel port (I used in 0.99.15 times
to connect a 486dx33 laptop to a 386dx33 desktop),
without timing dependencies.
* a "standard" bidirectional cable for use with bidir/ecp/epp
(does such thing exist?) would be half-duplex, hence require
a bit of a negociation protocol with the control lines.
Or it would be asymetric, with one computer using control lines,
the other one using data lines (in a reversible manner, perhaps).
Anyway, looks like non-trivial work ahead.
Plus to optimize the line, the driver could make timings
of send/receive throughput and delay, so as to send data
in tightly-timed packets (possibly with DMA),
without having to ACK every individual nibble.
* I might be wrong -- check docs and ask parport/plip gurus
for exact details...
## Faré | VN: Уng-Vû Bân | Join the TUNES project! http://www.tunes.org/ ##
## FR: François-René Rideau | TUNES is a Useful, Not Expedient System ##
## Reflection&Cybernethics | Project for a Free Reflective Computing System ##
An apple every eight hours will keep three doctors away.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu