Re: backward architecture, The Design of Design

According to Scott Lurndal <slp53@pacbell.net>:

I think it was an obvious idea at the time, even before IBM started

>
The Burroughs B100/200/300 systems were just that, multiple models
with different performance using a common cpu architecture.  Early
1960's.

I'm looking at the B100/200/300 manuals on bitsavers and I dunno.

For one thing it appears that the models were the same machine adding
more hardware components as you went up the scale. More importantly
the larger machines had more instructions than the smaller ones. You
could run B100 code on a B300 but not vice versa.

IBM made the decimal and floating point instruction sets optional on
the smaller models but if you got a 360/30 with both (the universal
instruction set) it had precisely the same instruction set as a 360/65
or /75. You could write code on a /75 and so long as there was enough
memory and it didn't have timing dependencies you could count on it to
run on a /30.

The big innovation in the 360 architecture was that the models were
not just upward compatible, which was familiar from newer machines
running code for older machines, but also downward compatible. They
could do all their software development on a /65 and not have to worry
about whether it would work on a /50 or /40 or /30 or the later /25
and /22.  If it worked on one it would work on all of them.

There were a few minor variations like the 360/20 which was a 16 bit
subsset with simpler I/O, the 360/44 which was a scientific subset
with realtime extensions and the 360/91 which had slightly different
floating point rounding and some strangeness with imprecise
interrupts, but even there there was a great deal of compatibility.

I wrote Fortran programs on the 360/91 using the same Fortran G and H
compilers that ran on every other model and they worked the same.



--
Regards,
John Levine, johnl@taugh.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. https://jl.ly