7 Dec
2014
7 Dec
'14
12:10 p.m.
On 12/06/2014 10:20 PM, Chuck Guzis wrote:
On 12/06/2014 06:55 PM, Noel Chiappa wrote:
Let's see; IBM 1620--basic memory size=20,000 digits,
increments of 20,000 digits up to 60,000. IBM 705, 7080,...
What is curious is the marketing numbers used for the
memory size. 65K, 131K, etc. I believe IBM was guilty of
this in their S/360 marketing literature.
Say, this brings up a question I've always wondered about.
On the emulation
microcode for the 360 series that ran decimal machine code,
such as
the 1401, did they translate every address from decimal to
binary before
accessing the memory, or did they just use the bare BCD
address, leaving
little and big holes all over the binary address space?
Obviously, the
translation to binary would take up a lot of time, and on
machines with
enough memory it didn't make so much difference. But, the
holes all over
would play hob with channels accessing the memory. But, at
least the 1401
series apparently had I/O buffers at fixed memory addresses,
so maybe
that wasn't such a big deal.
Anybody know?
Jon
From:
Jerome H. Fine
I would point out that one reason that memory may
have stayed with 2 **
30 is that memory is usually produced and sold in
multiples of 2 ** 30
these days
Main memory has pretty much _always_ been sold in blocks
that were exact
powers of two, for obvious reasons (at least, powers of
two of the word size
of the machine in question)...