3 May
2016
3 May
'16
10:39 p.m.
On Tue, 3 May 2016, Christian Kennedy wrote:
[snip]
No, they implemented new instructions "double word instructions" as
they called it in the 5605 programmers manual. These do not exist
on 1601, 1602 and 1602 but on 1602A, 1602B and 1650. As the name
suggests, they use two consecutive words in core and thus it is
easy to give a 16 bit address directly. There are also some
floating point instructions here and a bunch of "double precision
instructions" working on 32 bit operands. So essentially the
5605 has got a microcode based 32 bit extension.
I have not tried whether the DGC way works,but I know that software
for Novas with two banks of core does not work - e.g. one is re-
stricted to 32kW in using RDOS on the 1602B...
BTW: The 1602 and 1602B have a stack implemented but I do not
know (or have looked it up) whether this is compatible to some-
thing else...
Erik.
A 1602B can
hold 64kW of memory (only accessible via a banking
function proprietary to Rolm, so not usable e.g. for RDOS).
I thought the 16xx did it the same way Keronix and DCC did it -- by
limiting indirection to one level and then using the high order bit for
address rather than indicating indirection?