31 Oct
2010
31 Oct
'10
9:42 a.m.
Date: Sat, 30 Oct 2010 13:10:50 +0200
From: bqt at softjar.se
To: cctalk at classiccmp.org
Subject: Re: Memory mapped I/O
On 2010-10-30 01:12, ard at p850ug1.demon.co.uk (Tony Duell) wrote:
Good point.
Was the PDP-11 the first computer that did not have dedicated I/O
instructions then? Thus relying on memory mapped I/O, instead of just
having it as a potential (as any computer do). Did anyone before the
PDP-11 actually utilize memory mapped I/O before? The fact that lots of
machines do it today (even the x86 I guess) could just be a legacy of
the PDP-11 making it popular?
(Oh, and I'm not counting the IOT instruction on the PDP-11 as an I/O
instruction... :-) )
Hi
One thing to consider is that I/O on many of the larger machines included
a lot more that just a simple read and write. Various states, status and controls
were built into the I/O itself. Even interrupt and dma were more integrated
into the I/O hardware. Special conditional branches had there own special
instructions.
The I/O hardware would often handle many hardware specific operations
that today we tend to let the processor handle.
Dwight
> >
> Just because a CPU architecture has IO instructions doesn't mean you
> > > can't do memory-mapped I/O.
I beleive Steve Ciarcia said in one
of the Circuit Cellar articles in
Byte many year ago that any processor that could access memory could have
memory-mapped I/O. I would add that in the case of a system (rather
than a bare processor),
it helps if there's some spare space in the memroy map:-).
Indeed. :-)