[Simh] [SimH] Networking support

Anders Magnusson ragge at ludd.ltu.se
Sun Mar 13 05:35:59 EDT 2016


Den 2016-03-12 kl. 20:50, skrev Paul Koning:
>> On Mar 12, 2016, at 12:25 PM, Anders Magnusson <ragge at ludd.ltu.se> wrote:
>>
>> Den 2016-03-12 kl. 17:45, skrev Clem Cole:
>>> FYI:  CDC and Cray's often used HyperChannel adapters; but I suspect have long lost the info on it (very funky SW interface).  Plus I doubt I still have the code we developed for it (the HyperChannel was the other side of the Tektronix TCP/IP for VMS implementation we did in the late 1970's).  My memory is that we dedicated a PP to talking to NOS, so there was very little OS code.    We spliced it into the RJE system and never did much beyond FTP services for it; when I worked on it.   I still talk with the guy that did the the PP work for NOS (Stan Smith whom I will ask).   The Cray port was done by the NCAR guys working with our CDC and VMS code; but that was after I was involved in the work.
>>>
>>> BTW: @ LCC we did some work with DG on their UNIX port; I somehow seem to remember that the Eclipse family used the original AMD ethernet chip set on their network adapter.
>> DG-UX or MV/UX?
>> The DG ethernet card has a 82586 on board.  I have two of those cards, but unfortunately no programming specs. Anyone that has?
> I have a copy of the datasheet -- I can email it if you like (2.4 MB).  It's the chip used on the DEC Pro series DECNA Ethernet card.  It's a truly horrible design, with errors that were well known as things to avoid 15 years earlier.
>
> In the DECNA, you program it directly -- the card is basically just a bus bridge plus a small local memory for the chip to talk to.  If the DG boards are like that, the 82586 datasheet will serve.  But if there's machinery in between that exposes a different API, you'd need those specs instead, of course.
>
> Also: I've seen a NetBSD driver for it, which may also help understand how that chip works.
I am unfortunately too familiar with the 82586 :-/  Never been a fan of 
the Intel network chips.
But I assume DG has abstracted the programming interface on this card 
beyond recognition, at least based on the number of other chips on the card.

It seemed to be quite common to add another API to standard chips, the 
DELUA for example has a LANCE chip but you do not program the LANCE 
directly.

-- Ragge



More information about the Simh mailing list