PDP-11 addressing question and a model round-up
>>>>> "Mike" == Mike Cesari <mcesari_at_comcast.net> writes:
Mike> On Feb 14, 2005, at 8:53 AM, Paul Koning wrote:
Mike> <snip>
>> Life gets more complicated in 22-bit unibus machines, because
>> there you have 22 bit physical (memory) addresses, yet 18 bit
>> addresses on the I/O bus (the Unibus). So those 18 bit addresses
>> are virtual, they are mapped to physical addresses via the Unibus
>> map. Q-bus systems avoid this hassle by using full 22 bit
>> physical addresses on the I/O bus. (And Unibus systems do that
>> partway -- Massbus adapters hook right into the memory so they
>> handle 22 bit addresses directly, not via the Unibus map.)
>>
Mike> WRT massbus -- 22 bit addresses only worked with RH70's on
Mike> 11/70's. RH11's on the unibus were strictly 18 bit.
True, because they were meant for 18 bit systems. I suppose you'd
have to use one if you wanted big disks on an 11/44... yuck.
Mike> <ramble> Something I remembered regarding DZ11's and interrupt
Mike> issues -- if you had the money, you could use COMM-IOP-DZ. This
Mike> consisted of up to 4 DZ11's, an M8206 (KMC-11B) and software to
Mike> handle the charcter-at-a-time interrupt issue. This combo wound
Mike> up being better regarding system resources than DH11's in our
Mike> testing at the time (early `80's). We also used these on
Mike> VAX78x's. It made DW780's not suck as much. (The DW780 was
Mike> *slower* than the 11/70 unibus. So we always had multiple DW's
Mike> on each system.) </ramble>
Interesting. I think that Comm-iop-dz was only supported with RSX.
Certainly not with RSTS. And RSTS did rather well with DH11s as I
recall. It even had a special mode (APT) for bulk I/O which was used
by DEC manufacturing -- they used RSTS systems to drive large
quantities of other systems being tested after final assembly.
What you say about the 780 is not surprising. There's a rumor that
the 11/74 was canceled because it made the 780 look foolish,
performance-wise.
paul
Received on Mon Feb 14 2005 - 11:05:37 GMT
This archive was generated by hypermail 2.3.0
: Fri Oct 10 2014 - 23:37:38 BST