16 Sep
2020
16 Sep
'20
3:48 p.m.
Re:
On Wed, Sep 16, 2020 at 11:29 AM Lee Courtney <leec2124 at gmail.com> wrote:
I believe it was a performance issue. The APL was so
slow without the
microcode assist that the system was unusable.
Close. Without the microcode, they had no apparent way of even halfway
efficiently implementing a large virtual memory mechanism
(other than using multiple 64KB data segments, and the number of those was
severely limited, system-wide). In no way would any HP 3000 (at that time)
run APL\3000 at all without the microcode (#1).
With the microcode, they managed to get a *very slow* virtual memory
system. Marginally adequate for a few light-weight users, it brought the
HP 3000 Series III (a megahertz machine, IIRC) to its knees with any heavy
use. Which is why HP was sued over APL\3000. When HP won (apparently
because the judge decided the person suing should have been able to
determine the performance problem before specifying/buying an HP 3000), HP
immediately dropped APL\3000. (It had been explained to me that they
didn't want to drop it during the case.)
Stan
1. Gavin Scott's method lets it run, but by simulating the missing
instructions when they are called and cause an interrupt.