Thanks for that reference, Raul. I'm glad to see the FORTH community
still plugging away. In the days when I implemented event-queues in
FORTH on a 6502 chip for central-heating control, I could have done
with something like that.

So the advantages I'm keenly aware of. The limitations I'd say have
mostly to do with contention for shared resources by near-simultaneous
events. You need something which can malloc quickly. Or not have to
malloc at all, preferably.

As I wrote what I did, I recalled IBM did market a thoroughgoing
event-driven computer once, for applications like managing door-locks
and security sensors. We're talking 70s or 80s now. It never broke out
of its market niche.

Maybe the biggest reason why event-driven computers never broke out
was the difficulty of recruiting programmers who understood them. We'd
all been von-Neumann-ized into feeling that the only way to resolve
multi-server-multi-queue shared-resource issues was to funnel
everything through one CPU.

Which suited IBM fine in those days, because the party line was that
everything needed a big-big central processor. Start distributing your
processing power outwards and the plug-compatibles would eat your
market from the edges inwards. As happened.


On Sat, Apr 30, 2011 at 2:10 PM, Raul Miller <[email protected]> wrote:
> On Sat, Apr 30, 2011 at 5:20 AM, Ian Clark <[email protected]> wrote:
>> But why-oh-why weren't hardware architectures made completely
>> event-driven in the first place?
>
> Some are.
>
> http://greenarraychips.com/ for example.
>
> Apparently this approach has some significant advantages, and some
> significant limitations.
>
> --
> Raul
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to