16 Nov
2019
16 Nov
'19
1:23 p.m.
On 2019-Nov-16 09:42:47 +1100, Dave Horsfall <dave(a)horsfall.org> wrote:
On Fri, 15 Nov 2019, Arthur Krewat wrote:
More than just the instruction set - IBM published a formal description of
the S/360 (in APL in the IBM Systems Journal issue that announced the
S/360). The S/360 was (I believe) the first case where a company announced
a computer architecture (rather than an implementation) and implementations
were expected to precisely comply with the architecture (no more finding
undocumented instructions and side-effects and writing code that depended
on them). This meant that clone makers could build a clone that accurately
emulated a S/360.
How did Amdahl get away with making 360 clones? I
would have thought
that IBM would have crushed his bones into dust.
Clones in the way that they were able to run OS/360; that's about all that
I can remember. Hitachi also came out with a clone, as did no doubt many
other manufacturers; after all, the instruction set was public
knowledge... I dimly recall that some opcodes had undocumented
side-effects, so in
theory (and likely in practice) OS/360 could detect whether it was running
on a clone, and "fail to proceed" (in Rolls Royce terms).
AFAIR, the only "implementation defined" instruction was DIAGNOSE, OS/360
could presumably tell what it was running on by checking particular
DIAGNOSE function. (VM/370 was paravirtualised and used DIAGNOSE to
communicate with the hypervisor - CP).
In the early PC era, it was not uncommon for applications to verify they
were running on a genuine IBM PC by looking for the copyright notice in the
BIOS - which clone makers countered by placing a "not" before an equivalent
copyright notice.
--
Peter Jeremy