23 Jan
2011
23 Jan
'11
2:30 p.m.
On 2011 Jan 23, at 1:16 PM, Evan Koblentz wrote:
The 1948 modifications were to use the function tables (hundreds of
manually-set switches) as an instruction ROM. This didn't really make
it a 'stored-program' machine in terms of what the concept meant even
back then (data and modifiable instructions in one R/W memory), it was
a half-way step. In modern terms it would probably be described as a
ROM-based Harvard-architecture machine at that point.
Some core was later added, but that of course would have to be after
core was developed (>1952). Exactly how the architecture would be
described at that point I don't know.
When was this stored-program (in the normal
sense) system added?
Before or after other stored-program machines such as EDSAC
Manchester Mk1, etc?
Good question. Conventional wisdom says it happened in 1948, which is
the same year as the Manchester computer. Wikipedia (FWIW) says the
Manchester computer first operated on June 21. I don't know when in
1948 ENIAC got its own stored-program ability. But .... here's something interesting to ruffle
the Manchester fans'
feathers: http://www.columbia.edu/acis/history/acm-meeting.jpg. It's
a memo stating that, at the first meeting of the ACM in 1947, the
unfinished EDVAC computer was working and demonstrated. EDVAC was a
stored-program machine too. So if the memo's claims are true, then
it's possible this function was show pre-Manchester.
(Personally I think that argument is pointless. It's not like the
Manchester team didn't do any testing; I have to assume that the June
21 day of running its first program was not actually the first day it
was turned on and tested. Still, the memo linked above is talking
about March 1947, which is a long time before June 1948. So it just
might hold water.)
