Peppo
2006-03-13, 22:40:48
[...]
> > I worked on the Safeguard software from 1969 to 1975 [...]
> What was it written it? What did it run on?
Almost all of it was written in assembly language -- at the time the
largest programming project ever written -- maybe 4,000 man-years.
(It came in on time, on budget, and on spec, and there aren't very
many programming projects of ANY size that did that! And as far as we
knew it would have worked perfectly against the threat it was designed
for, ten years before! <grin>)
The reason for ass'y was simply for speed. As I've told a lot of my
programming projects, the design goal was, "If we can shave another
micro-second out of an inner loop, then we can keep 652 balls in the
air at one time instead of 651. Hey! Now we can save Pittsburgh!"
Many of the inner "loops" were NOT, in fact loops, but were written
out linearly for speed. (ANY tradeoff of memory for speed was
legitimate - I can remember design-change meetings where the proposal
was to save a millisecond.)
It ran on a multi-CPU mainframe "farm" custom-built by Univac. Six or
eight CPUs, as I recall, with task dispatching as required. Note that
SMP systems weren't exactly common in the late 60's, let alone
fault-tolerant ones. We used to show off when the generals came
through by asking them to disconnect any data cable or pull any power
plug out of the wall, and the software would reconfigure on the fly
for the loss of a CPU, memory unit, console, or whatever!!! For
transient errors, it could recover from an error interrupt ANYWHERE
without a problem - and that included the first-level interrupt
routines themselves! IIRC, we could recover from a trap in over 99%
of the SECOND-LEVEL interrupt-handler code!
[...]
http://www.paineless.id.au/missiles/Computers.html
Wahnsinn... Ich kann das irgendwie gar nicht vorstellen...
Ich meine, wenn man Software für ein ABM System schreibt, da bleit sicher
kein Platz für Bugs... Denn im Ernstfal, würden diese Bugs millionen Tote forden... :eek:
Mit welchen Aufwand ist sowas möglich?
Und vor allem, wie? :uponder:
Edit:
Falls das hier die falsche Unterforum wäre, bitte dementsprechend verschieben... :)
> > I worked on the Safeguard software from 1969 to 1975 [...]
> What was it written it? What did it run on?
Almost all of it was written in assembly language -- at the time the
largest programming project ever written -- maybe 4,000 man-years.
(It came in on time, on budget, and on spec, and there aren't very
many programming projects of ANY size that did that! And as far as we
knew it would have worked perfectly against the threat it was designed
for, ten years before! <grin>)
The reason for ass'y was simply for speed. As I've told a lot of my
programming projects, the design goal was, "If we can shave another
micro-second out of an inner loop, then we can keep 652 balls in the
air at one time instead of 651. Hey! Now we can save Pittsburgh!"
Many of the inner "loops" were NOT, in fact loops, but were written
out linearly for speed. (ANY tradeoff of memory for speed was
legitimate - I can remember design-change meetings where the proposal
was to save a millisecond.)
It ran on a multi-CPU mainframe "farm" custom-built by Univac. Six or
eight CPUs, as I recall, with task dispatching as required. Note that
SMP systems weren't exactly common in the late 60's, let alone
fault-tolerant ones. We used to show off when the generals came
through by asking them to disconnect any data cable or pull any power
plug out of the wall, and the software would reconfigure on the fly
for the loss of a CPU, memory unit, console, or whatever!!! For
transient errors, it could recover from an error interrupt ANYWHERE
without a problem - and that included the first-level interrupt
routines themselves! IIRC, we could recover from a trap in over 99%
of the SECOND-LEVEL interrupt-handler code!
[...]
http://www.paineless.id.au/missiles/Computers.html
Wahnsinn... Ich kann das irgendwie gar nicht vorstellen...
Ich meine, wenn man Software für ein ABM System schreibt, da bleit sicher
kein Platz für Bugs... Denn im Ernstfal, würden diese Bugs millionen Tote forden... :eek:
Mit welchen Aufwand ist sowas möglich?
Und vor allem, wie? :uponder:
Edit:
Falls das hier die falsche Unterforum wäre, bitte dementsprechend verschieben... :)