Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
Word Length
83 views
Skip to first unread message
ken...@cix.compulink.co.uk
Mar 26, 2012, 1:20:14 PM3/26/12
to
When was word length standardised? I remember the computer books I have
dating from the early era expected people to be familiar with octal,
binary and hex notation. I have come across references to early
computers using something like a 40 bit word.
Ken Young
dating from the early era expected people to be familiar with octal,
binary and hex notation. I have come across references to early
computers using something like a 40 bit word.
Ken Young
Dan Espen
Mar 26, 2012, 1:37:03 PM3/26/12
to
8 bit bytes have become common, but a "word" can be
16, 32, or 64 bits on common architectures.
The most popular early era machine (the IBM 1401) didn't have
anything that resembled a word.
--
Dan Espen
Christian Brunschen
Mar 26, 2012, 1:50:04 PM3/26/12
to
In article <gKadnTD7G-fTOO3S...@giganews.com>,
It never was - and still isn't!
Even 'byte' length has varied over the years; and there has frequently
been a difference in lenght between 'bytes', 'words' and 'characters'.
These days, a 'byte' _usually_ refers to an 8-bit quantity, but character
sizes really depend on encoding.
>I remember the computer books I have
>dating from the early era expected people to be familiar with octal,
>binary and hex notation.
notation, and choice of base, is really unrelated to word size. Yes, octal
(3 bits per digit) fits nicely with architectures that have a word-size
that is a multiple of 3 bits, but has also been used in other contexts
where this a 'word' has not held an integer number of base-8 digits.
>I have come across references to early
>computers using something like a 40 bit word.
The 'IAS machine' architecture (see for instance
http://insar.stanford.edu/~lharcke/programming/ ) did indeed use a word
size of 40 bits - but instructions tended to be 20 bits, i.e., there were
2 instructions per word.
Other word sizes have been used, such as 36, 18 and 12 bits; and of course
16 bits. And in many such machines, there was no smaller individually
addressable unit, in contrast to today's prevalence of the 8-bit 'byte'
as the unit to address (with larger-sized accesses accessing multiple
bytes).
And indeed, not all machines have had specific 'word sizes' at all.
> Ken Young
// Christian Brunschen
It never was - and still isn't!
Even 'byte' length has varied over the years; and there has frequently
been a difference in lenght between 'bytes', 'words' and 'characters'.
These days, a 'byte' _usually_ refers to an 8-bit quantity, but character
sizes really depend on encoding.
>I remember the computer books I have
>dating from the early era expected people to be familiar with octal,
>binary and hex notation.
(3 bits per digit) fits nicely with architectures that have a word-size
that is a multiple of 3 bits, but has also been used in other contexts
where this a 'word' has not held an integer number of base-8 digits.
>I have come across references to early
>computers using something like a 40 bit word.
http://insar.stanford.edu/~lharcke/programming/ ) did indeed use a word
size of 40 bits - but instructions tended to be 20 bits, i.e., there were
2 instructions per word.
Other word sizes have been used, such as 36, 18 and 12 bits; and of course
16 bits. And in many such machines, there was no smaller individually
addressable unit, in contrast to today's prevalence of the 8-bit 'byte'
as the unit to address (with larger-sized accesses accessing multiple
bytes).
And indeed, not all machines have had specific 'word sizes' at all.
> Ken Young
// Christian Brunschen
Rod Speed
Mar 26, 2012, 4:43:14 PM3/26/12
to
ken...@cix.compulink.co.uk wrote
> When was word length standardised?
It never was.
> I remember the computer books I have dating from the early era
> expected people to be familiar with octal, binary and hex notation.
Hex showed up later.
> I have come across references to early
> computers using something like a 40 bit word.
Yeah, there there were lots of different ones, 60 bit, 18 bit, 12 bit, all sorts of things.
> When was word length standardised?
> I remember the computer books I have dating from the early era
> expected people to be familiar with octal, binary and hex notation.
> I have come across references to early
> computers using something like a 40 bit word.
David Dyer-Bennet
Mar 26, 2012, 5:05:20 PM3/26/12
to
general-purpose processors are byte-oriented. That became popular
roughly with the IBM 360 series in 1964. Modern systems don't have a
"word length" in anything like the classic sense.
I've worked on 12, 16, 32, and 36-bit word computers myself. I know of
60-bit systems as well. Plus computers that *aren't* word-orientd. My
first one had 20,000 decimal digits of core memory, for example (IBM
1620). For arithmetic, fields could be anything from two digits with no
upper limit (no limit in design; the available memory of course limited
what was possible in practice).
--
David Dyer-Bennet, dd...@dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info
David Dyer-Bennet
Mar 26, 2012, 5:08:13 PM3/26/12
to
lengths were variable, delimited by a dedicated memory bit. The bits in
each memory location were labeled C, B, A, 8, 4, 2, 1, M. C was a
check-bit, basically parity. 1,2,4,8 were numeric bits, and decimal
digits were represented by the right combination of those bits. B and A
were for alpha characters (combined with suitable 8421 settings) and
roughly corresponded to the top two rows on a punch card. M was the
word mark, which delimited fields for various instructions including
computation.
Message has been deleted
Dan Espen
Mar 26, 2012, 5:25:16 PM3/26/12
to
David Dyer-Bennet <dd...@dd-b.net> writes:
> Dan Espen <des...@verizon.net> writes:
>
>> ken...@cix.compulink.co.uk writes:
>>
>>> When was word length standardised? I remember the computer books I have
>>> dating from the early era expected people to be familiar with octal,
>>> binary and hex notation. I have come across references to early
>>> computers using something like a 40 bit word.
>>
>> I don't believe there is any standard for word length.
>>
>> 8 bit bytes have become common, but a "word" can be
>> 16, 32, or 64 bits on common architectures.
>>
>> The most popular early era machine (the IBM 1401) didn't have
>> anything that resembled a word.
>
> That was my second computer. Yeah, nothing like a word. Numeric field
> lengths were variable, delimited by a dedicated memory bit. The bits in
> each memory location were labeled C, B, A, 8, 4, 2, 1, M. C was a
> check-bit, basically parity. 1,2,4,8 were numeric bits, and decimal
> digits were represented by the right combination of those bits. B and A
> were for alpha characters (combined with suitable 8421 settings) and
> roughly corresponded to the top two rows on a punch card. M was the
> word mark, which delimited fields for various instructions including
> computation.
After working on 1401/1460 I worked at a place with an 8K 1440.
> Dan Espen <des...@verizon.net> writes:
>
>> ken...@cix.compulink.co.uk writes:
>>
>>> When was word length standardised? I remember the computer books I have
>>> dating from the early era expected people to be familiar with octal,
>>> binary and hex notation. I have come across references to early
>>> computers using something like a 40 bit word.
>>
>> I don't believe there is any standard for word length.
>>
>> 8 bit bytes have become common, but a "word" can be
>> 16, 32, or 64 bits on common architectures.
>>
>> The most popular early era machine (the IBM 1401) didn't have
>> anything that resembled a word.
>
> That was my second computer. Yeah, nothing like a word. Numeric field
> lengths were variable, delimited by a dedicated memory bit. The bits in
> each memory location were labeled C, B, A, 8, 4, 2, 1, M. C was a
> check-bit, basically parity. 1,2,4,8 were numeric bits, and decimal
> digits were represented by the right combination of those bits. B and A
> were for alpha characters (combined with suitable 8421 settings) and
> roughly corresponded to the top two rows on a punch card. M was the
> word mark, which delimited fields for various instructions including
> computation.
Sitting at the console one day, I keyed in a small routine to set a word
mark around position 10, then put a 1 at location 7999 and in low
storage I keyed in an add 7999 to 7999 and branch back to add.
So the program was doubling an 8000 digit number.
On the console you could clearly see the A and B registers cycle down
from 8000 to 10 before the add completed. That was one of the few times
I could watch an individual instruction execute.
Today, trying to add 2 8000 digit numbers isn't so easy.
--
Dan Espen
Peter Flass
Mar 26, 2012, 7:48:55 PM3/26/12
to
Is 32 or 64 the standard? Or maybe 128?
The reason for powers-of-two is to allow use of all the possible address
space the address bits give you.
--
Pete
Peter Flass
Mar 26, 2012, 7:51:40 PM3/26/12
to
On 3/26/2012 5:05 PM, David Dyer-Bennet wrote:
>
> Word length has never been standardized. However, basically all modern
> general-purpose processors are byte-oriented. That became popular
> roughly with the IBM 360 series in 1964. Modern systems don't have a
> "word length" in anything like the classic sense.
>
What is the definition. I'd take "word length" to be the size of a
>
> Word length has never been standardized. However, basically all modern
> general-purpose processors are byte-oriented. That became popular
> roughly with the IBM 360 series in 1964. Modern systems don't have a
> "word length" in anything like the classic sense.
>
register, for machines that have registers.
Peter Flass
Mar 26, 2012, 7:53:10 PM3/26/12
to
--
Pete
Bill Leary
Mar 26, 2012, 7:59:50 PM3/26/12
to
wrote in message news:gKadnTD7G-fTOO3S...@giganews.com...
What do you mean by "standardized"?
When I started into the business the word "word" (usually) meant the size of
a single memory access. Often also called the natural word size of the
machine. A hardware addressable location. For example 12 bits for the
PDP-8. Bytes also weren't always 8 bits. I worked briefly on a 36 bit
machine that had byte addressing instructions for word locations. SIx 6-bit
bytes per 36-bit word.
- Bill
When I started into the business the word "word" (usually) meant the size of
a single memory access. Often also called the natural word size of the
machine. A hardware addressable location. For example 12 bits for the
PDP-8. Bytes also weren't always 8 bits. I worked briefly on a 36 bit
machine that had byte addressing instructions for word locations. SIx 6-bit
bytes per 36-bit word.
- Bill
Walter Bushell
Mar 26, 2012, 8:01:04 PM3/26/12
to
thing was agreeing on the size of a byte. And now we have to think
about how many bytes in a character, with the various forms of Unicode.
It *never* ends.
--
This space unintentionally left blank.
Stan Barr
Mar 27, 2012, 3:06:30 AM3/27/12
to
On Mon, 26 Mar 2012 17:50:04 +0000 (UTC), Christian Brunschen
<c...@mer.df.lth.se> wrote:
>
> And indeed, not all machines have had specific 'word sizes' at all.
And the Russian SETUN computers muddied the waters by being ternary
<c...@mer.df.lth.se> wrote:
>
> And indeed, not all machines have had specific 'word sizes' at all.
rather than binary so didn't have "bits" at all :-)
--
Cheers,
Stan Barr plan.b .at. dsl .dot. pipex .dot. com
The future was never like this!
Ahem A Rivet's Shot
Mar 27, 2012, 3:10:48 AM3/27/12
to
On Mon, 26 Mar 2012 19:48:55 -0400
Peter Flass <Peter...@Yahoo.com> wrote:
> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
> > When was word length standardised? I remember the computer books I
> > have dating from the early era expected people to be familiar with
> > octal, binary and hex notation. I have come across references to early
> > computers using something like a 40 bit word.
> >
> > Ken Young
>
> 12, 24, 26, 40, 48, 60, pick any of the above.
Er 36, or your not playing with a full DEC.
Peter Flass <Peter...@Yahoo.com> wrote:
> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
> > When was word length standardised? I remember the computer books I
> > have dating from the early era expected people to be familiar with
> > octal, binary and hex notation. I have come across references to early
> > computers using something like a 40 bit word.
> >
> > Ken Young
>
> 12, 24, 26, 40, 48, 60, pick any of the above.
--
Steve O'Hara-Smith | Directable Mirror Arrays
C:>WIN | A better way to focus the sun
The computer obeys and wins. | licences available see
You lose and Bill collects. | http://www.sohara.org/
Jean-Marc Bourguet
Mar 27, 2012, 3:36:55 AM3/27/12
to
Ahem A Rivet's Shot <ste...@eircom.net> writes:
> On Mon, 26 Mar 2012 19:48:55 -0400
> Peter Flass <Peter...@Yahoo.com> wrote:
>
>> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>> > When was word length standardised? I remember the computer books I
>> > have dating from the early era expected people to be familiar with
>> > octal, binary and hex notation. I have come across references to early
>> > computers using something like a 40 bit word.
>> >
>> > Ken Young
>>
>> 12, 24, 26, 40, 48, 60, pick any of the above.
>
> Er 36, or your not playing with a full DEC.
I wonder if 26 wasn't a typo for 36.
> On Mon, 26 Mar 2012 19:48:55 -0400
> Peter Flass <Peter...@Yahoo.com> wrote:
>
>> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>> > When was word length standardised? I remember the computer books I
>> > have dating from the early era expected people to be familiar with
>> > octal, binary and hex notation. I have come across references to early
>> > computers using something like a 40 bit word.
>> >
>> > Ken Young
>>
>> 12, 24, 26, 40, 48, 60, pick any of the above.
>
> Er 36, or your not playing with a full DEC.
Yours,
--
Jean-Marc
Stan Barr
Mar 27, 2012, 4:48:01 AM3/27/12
to
On Mon, 26 Mar 2012 19:48:55 -0400, Peter Flass <Peter...@Yahoo.com> wrote:
> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>> When was word length standardised? I remember the computer books I have
>> dating from the early era expected people to be familiar with octal,
>> binary and hex notation. I have come across references to early
>> computers using something like a 40 bit word.
>>
>> Ken Young
>
> 12, 24, 26, 40, 48, 60, pick any of the above. Is it standardized now?
> Is 32 or 64 the standard? Or maybe 128?
Historically, 18, 20, 22(?) and 36 too.
> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>> When was word length standardised? I remember the computer books I have
>> dating from the early era expected people to be familiar with octal,
>> binary and hex notation. I have come across references to early
>> computers using something like a 40 bit word.
>>
>> Ken Young
>
> 12, 24, 26, 40, 48, 60, pick any of the above. Is it standardized now?
> Is 32 or 64 the standard? Or maybe 128?
Rod Speed
Mar 27, 2012, 4:56:26 AM3/27/12
to
Jean-Marc Bourguet wrote
> Ahem A Rivet's Shot <ste...@eircom.net> wrote
>> Peter Flass <Peter...@Yahoo.com> wrote
> Ahem A Rivet's Shot <ste...@eircom.net> wrote
>> Peter Flass <Peter...@Yahoo.com> wrote
>>> ken...@cix.compulink.co.uk wrote
>>>> When was word length standardised? I remember the computer books
>>>> I have dating from the early era expected people to be familiar
>>>> with octal, binary and hex notation. I have come across references
>>>> to early computers using something like a 40 bit word.
>>>> When was word length standardised? I remember the computer books
>>>> I have dating from the early era expected people to be familiar
>>>> with octal, binary and hex notation. I have come across references
>>>> to early computers using something like a 40 bit word.
>>> 12, 24, 26, 40, 48, 60, pick any of the above.
>> Er 36, or your not playing with a full DEC.
> I wonder if 26 wasn't a typo for 36.
Even if it was, he's still not playing with a full DEC given that he left out 18.
>> Er 36, or your not playing with a full DEC.
> I wonder if 26 wasn't a typo for 36.
Peter Flass
Mar 27, 2012, 7:54:11 AM3/27/12
to
--
Pete
Peter Flass
Mar 27, 2012, 7:56:10 AM3/27/12
to
On 3/27/2012 3:10 AM, Ahem A Rivet's Shot wrote:
> On Mon, 26 Mar 2012 19:48:55 -0400
> Peter Flass<Peter...@Yahoo.com> wrote:
>
>> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>>> When was word length standardised? I remember the computer books I
>>> have dating from the early era expected people to be familiar with
>>> octal, binary and hex notation. I have come across references to early
>>> computers using something like a 40 bit word.
>>>
>>> Ken Young
>>
>> 12, 24, 26, 40, 48, 60, pick any of the above.
>
> Er 36, or your not playing with a full DEC.
>
Typo (26==36). I did leave out 18 (forgot) and 16 (too common to
> On Mon, 26 Mar 2012 19:48:55 -0400
> Peter Flass<Peter...@Yahoo.com> wrote:
>
>> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>>> When was word length standardised? I remember the computer books I
>>> have dating from the early era expected people to be familiar with
>>> octal, binary and hex notation. I have come across references to early
>>> computers using something like a 40 bit word.
>>>
>>> Ken Young
>>
>> 12, 24, 26, 40, 48, 60, pick any of the above.
>
> Er 36, or your not playing with a full DEC.
>
mention). Any others?
--
Pete
Walter Bushell
Mar 27, 2012, 8:09:45 AM3/27/12
to
In article <20120327081048....@eircom.net>,
or 12 or 18. Some even used 8 as a byte size, when they wanted to be
silly.
Ahem A Rivet's Shot <ste...@eircom.net> wrote:
> On Mon, 26 Mar 2012 19:48:55 -0400
> Peter Flass <Peter...@Yahoo.com> wrote:
>
> > On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
> > > When was word length standardised? I remember the computer books I
> > > have dating from the early era expected people to be familiar with
> > > octal, binary and hex notation. I have come across references to early
> > > computers using something like a 40 bit word.
> > >
> > > Ken Young
> >
> > 12, 24, 26, 40, 48, 60, pick any of the above.
>
> Er 36, or your not playing with a full DEC.
On some DEC computers you were allowed to set the byte size to 6 or 7
> On Mon, 26 Mar 2012 19:48:55 -0400
> Peter Flass <Peter...@Yahoo.com> wrote:
>
> > On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
> > > When was word length standardised? I remember the computer books I
> > > have dating from the early era expected people to be familiar with
> > > octal, binary and hex notation. I have come across references to early
> > > computers using something like a 40 bit word.
> > >
> > > Ken Young
> >
> > 12, 24, 26, 40, 48, 60, pick any of the above.
>
> Er 36, or your not playing with a full DEC.
or 12 or 18. Some even used 8 as a byte size, when they wanted to be
silly.
jmfbahciv
Mar 27, 2012, 9:07:44 AM3/27/12
to
if the byte size was 0. None of the PDP platforms insisted on
one flavor of byte nor did they insist on byte-only accessing.
/BAH
jmfbahciv
Mar 27, 2012, 9:07:55 AM3/27/12
to
weren't stored on a word basis.
/BAH
jmfbahciv
Mar 27, 2012, 9:07:56 AM3/27/12
to
/BAH
Dan Espen
Mar 27, 2012, 11:52:11 AM3/27/12
to
"-1" notation. Ie. 0 is 1. Therefore, there would be no way
to set byte size 0.
--
Dan Espen
Patrick Scheible
Mar 27, 2012, 11:56:02 AM3/27/12
to
-- Patrick
Stan Barr
Mar 27, 2012, 12:52:41 PM3/27/12
to
reasons I can't remember.
EDSAC was 18-bit of which only 17 could be used because of timing
constraints.
4004 microprocessor - 4 bits.
There are always a few oddities :-)
Walter Banks
Mar 27, 2012, 2:05:45 PM3/27/12
to
Stan Barr wrote:
> On Mon, 26 Mar 2012 19:48:55 -0400, Peter Flass <Peter...@Yahoo.com> wrote:
> > On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
> >> When was word length standardised? I remember the computer books I have
> >> dating from the early era expected people to be familiar with octal,
> >> binary and hex notation. I have come across references to early
> >> computers using something like a 40 bit word.
> >>
> >> Ken Young
> >
> > 12, 24, 26, 40, 48, 60, pick any of the above. Is it standardized now?
> > Is 32 or 64 the standard? Or maybe 128?
>
> Historically, 18, 20, 22(?) and 36 too.
13 Bits Westinghouse PRODAC 50
14 Bits (Instruction paths ) Microchip midrange
9 bits several small processors so 18 bits is double word size. I have only seen them in Special Purpose processors
Walter Banks
Ahem A Rivet's Shot
Mar 27, 2012, 1:22:31 PM3/27/12
to
On 27 Mar 2012 16:52:41 GMT
MC14500 - 1 bit!
Rod Speed
Mar 27, 2012, 2:44:37 PM3/27/12
to
Patrick Scheible wrote
> jmfbahciv <See....@aol.com> wrote
>> Walter Bushell wrote
With much more overhead.
> jmfbahciv <See....@aol.com> wrote
>> Walter Bushell wrote
>>> Ahem A Rivet's Shot <ste...@eircom.net> wrote
>>>> Peter Flass <Peter...@Yahoo.com> wrote
>>>>> ken...@cix.compulink.co.uk wrote
>>>>>> When was word length standardised? I remember the computer
>>>>>> books I have dating from the early era expected people to be
>>>>>> familiar with octal, binary and hex notation. I have come across
>>>>>> references to early computers using something like a 40 bit word.
>>>>>> When was word length standardised? I remember the computer
>>>>>> books I have dating from the early era expected people to be
>>>>>> familiar with octal, binary and hex notation. I have come across
>>>>>> references to early computers using something like a 40 bit word.
>>>>> 12, 24, 26, 40, 48, 60, pick any of the above.
>>>> Er 36, or your not playing with a full DEC.
>>> On some DEC computers you were allowed to set the byte size to 6 or 7 or
>>> 12 or 18. Some even used 8 as a byte size, when they wanted to be silly.
>> I could set the byte size from 1-36. I don't remember what happens
>> if the byte size was 0. None of the PDP platforms insisted on
>> one flavor of byte nor did they insist on byte-only accessing.
> I thought the PDP-11s dealt most easily with 8-bit bytes?
They didnt INSIST on it tho, you could uses others if you wanted to.
>>>> Er 36, or your not playing with a full DEC.
>>> On some DEC computers you were allowed to set the byte size to 6 or 7 or
>>> 12 or 18. Some even used 8 as a byte size, when they wanted to be silly.
>> I could set the byte size from 1-36. I don't remember what happens
>> if the byte size was 0. None of the PDP platforms insisted on
>> one flavor of byte nor did they insist on byte-only accessing.
> I thought the PDP-11s dealt most easily with 8-bit bytes?
With much more overhead.
Rod Speed
Mar 27, 2012, 2:48:59 PM3/27/12
to
Walter Banks wrote
> Stan Barr wrote
>> Peter Flass <Peter...@Yahoo.com> wrote
It is one of the less well known DEC lines tho.
> Stan Barr wrote
>> Peter Flass <Peter...@Yahoo.com> wrote
>>> ken...@cix.compulink.co.uk wrote
>>>> When was word length standardised? I remember the computer books
>>>> I have dating from the early era expected people to be familiar
>>>> with octal, binary and hex notation. I have come across references
>>>> to early computers using something like a 40 bit word.
>>>> When was word length standardised? I remember the computer books
>>>> I have dating from the early era expected people to be familiar
>>>> with octal, binary and hex notation. I have come across references
>>>> to early computers using something like a 40 bit word.
>>> 12, 24, 26, 40, 48, 60, pick any of the above. Is it standardized
>>> now? Is 32 or 64 the standard? Or maybe 128?
>> Historically, 18, 20, 22(?) and 36 too.
> Add to that
> 13 Bits Westinghouse PRODAC 50
> 14 Bits (Instruction paths ) Microchip midrange
> 9 bits several small processors so 18 bits is double word size.
> I have only seen them in Special Purpose processors
Thats also what the PDP7/9/15 had, nothing like special purpose processors.
>>> now? Is 32 or 64 the standard? Or maybe 128?
>> Historically, 18, 20, 22(?) and 36 too.
> Add to that
> 13 Bits Westinghouse PRODAC 50
> 14 Bits (Instruction paths ) Microchip midrange
> 9 bits several small processors so 18 bits is double word size.
> I have only seen them in Special Purpose processors
It is one of the less well known DEC lines tho.
David Dyer-Bennet
Mar 27, 2012, 3:15:18 PM3/27/12
to
It used to mean the basic unit of memory access; but clearly that's the
byte in nearly every case now.
The IBM 1620 and IBM 1401 weren't word-oriented by any definition I've
heard. The PDP-8 was a 12-bit-word architecture. With the 11, we get
into byte-oriented memory, with larger arithmetic units, so it gets
complicated. Somebody who said the word size was 16 bits wouldn't be
clearly wrong.
--
David Dyer-Bennet, dd...@dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info
David Dyer-Bennet
Mar 27, 2012, 3:19:13 PM3/27/12
to
differently than we did back then). I remember manually encoding the
high and low address bits to get big tables (the middle bits were
sometimes used for something else; was there indirect addressing or
something?)
> So the program was doubling an 8000 digit number.
>
> On the console you could clearly see the A and B registers cycle down
> from 8000 to 10 before the add completed. That was one of the few times
> I could watch an individual instruction execute.
wouldn't show much any more I suppose.
> Today, trying to add 2 8000 digit numbers isn't so easy.
infinite-precision numbers, though.
David Dyer-Bennet
Mar 27, 2012, 3:25:06 PM3/27/12
to
Peter Flass <Peter...@Yahoo.com> writes:
> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>> When was word length standardised? I remember the computer books I have
>> dating from the early era expected people to be familiar with octal,
>> binary and hex notation. I have come across references to early
>> computers using something like a 40 bit word.
>>
>> Ken Young
>
> 12, 24, 26, 40, 48, 60, pick any of the above. Is it standardized
> now? Is 32 or 64 the standard? Or maybe 128?
36 (PDP-10), 18 (pdp-9)
> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>> When was word length standardised? I remember the computer books I have
>> dating from the early era expected people to be familiar with octal,
>> binary and hex notation. I have come across references to early
>> computers using something like a 40 bit word.
>>
>> Ken Young
>
> 12, 24, 26, 40, 48, 60, pick any of the above. Is it standardized
> now? Is 32 or 64 the standard? Or maybe 128?
(Is your "26" supposed to be 36? I can't think of a 26-bit
architecture, but my ignorance is large!)
The 7030 "Stretch" was 64 bit. As was the Illiac 4 and the Cray 1
The numbers get pretty boring after about 1964.
> The reason for powers-of-two is to allow use of all the possible
> address space the address bits give you.
one *instruction* in a word, though, with less of that being the
address.
Hey, Wikipedia says the Babbage Analytical engine had a word size of 50
decimal digits!
David Dyer-Bennet
Mar 27, 2012, 3:27:06 PM3/27/12
to
instructions did that. 1-bit bytes were sometimes useful for accessing
bitvectors. (One-word global byte pointers in the extended architecture
had a limited range of byte sizes, but the local byte-pointers and the
two-word globals supported all possible sizes.)
David Dyer-Bennet
Mar 27, 2012, 3:28:14 PM3/27/12
to
instruction set was widely desribed as "logically complete, but not
completely logical", that's probably right.
David Dyer-Bennet
Mar 27, 2012, 3:28:48 PM3/27/12
to
Patrick Scheible
Mar 27, 2012, 3:31:26 PM3/27/12
to
Peter Flass <Peter...@Yahoo.com> writes:
> On 3/26/2012 5:05 PM, David Dyer-Bennet wrote:
>>
>> Word length has never been standardized. However, basically all modern
>> general-purpose processors are byte-oriented. That became popular
>> roughly with the IBM 360 series in 1964. Modern systems don't have a
>> "word length" in anything like the classic sense.
>>
>
> What is the definition. I'd take "word length" to be the size of a
> register, for machines that have registers.
Yes, but what about those modern machines with two or more different
> On 3/26/2012 5:05 PM, David Dyer-Bennet wrote:
>>
>> Word length has never been standardized. However, basically all modern
>> general-purpose processors are byte-oriented. That became popular
>> roughly with the IBM 360 series in 1964. Modern systems don't have a
>> "word length" in anything like the classic sense.
>>
>
> What is the definition. I'd take "word length" to be the size of a
> register, for machines that have registers.
sizes of registers?
-- Patrick
David Dyer-Bennet
Mar 27, 2012, 3:36:39 PM3/27/12
to
first computer ever.)
(Instructions were 12 decimal digits long; two op-code, 5 address, 5
address or immediate data. There were probably some weird cases that
didn't quite conform to that structure, but not many. Gack, I still
remember THREE instructions for clearing core, and hence three numeric
opcodes; 260000800009 (using TF, Transmit Field), 160001000000 (using
TFM, Transmit Field Immediate), and.... Oops, I appear to have lost the
Transmit Record op-code. That one was "interesting" because it cleared
memory in the opposite order. Okay, Bitsavers to the rescue. TR is
31.)
Dan Espen
Mar 27, 2012, 3:57:55 PM3/27/12
to
Up until then, the computers adapted to common human usage, Ie.
they had native base 10 arithmetic and addressing.
On a 1401, the address was 3 characters which gets you to 999.
Then the BA bits were used in the first and last byte to get up
to 16,000. The middle bits identified one of 3 index registers
at fixed storage locations.
>> So the program was doubling an 8000 digit number.
>>
>> On the console you could clearly see the A and B registers cycle down
>> from 8000 to 10 before the add completed. That was one of the few times
>> I could watch an individual instruction execute.
>
> I really miss front-panel lights. Though at todays's speeds they
> wouldn't show much any more I suppose.
Why this isn't viewed as a marketing opportunity, I don't know.
>> Today, trying to add 2 8000 digit numbers isn't so easy.
>
> Not in machine language, anyway. Plenty of software packages support
> infinite-precision numbers, though.
--
Dan Espen
David Dyer-Bennet
Mar 27, 2012, 4:23:50 PM3/27/12
to
Dan Espen <des...@verizon.net> writes:
> David Dyer-Bennet <dd...@dd-b.net> writes:
>
>> Dan Espen <des...@verizon.net> writes:
>>
>>> After working on 1401/1460 I worked at a place with an 8K 1440.
>>> Sitting at the console one day, I keyed in a small routine to set a word
>>> mark around position 10, then put a 1 at location 7999 and in low
>>> storage I keyed in an add 7999 to 7999 and branch back to add.
>>
>> That's the size we had (though readers today will interpret "8K" a bit
>> differently than we did back then). I remember manually encoding the
>> high and low address bits to get big tables (the middle bits were
>> sometimes used for something else; was there indirect addressing or
>> something?)
>
> Yep. My first exposure to the term "K" came with System/360.
> Up until then, the computers adapted to common human usage, Ie.
> they had native base 10 arithmetic and addressing.
>
> On a 1401, the address was 3 characters which gets you to 999.
> Then the BA bits were used in the first and last byte to get up
> to 16,000. The middle bits identified one of 3 index registers
> at fixed storage locations.
Oh, of course, index registers. Much more basic, very important use.
> David Dyer-Bennet <dd...@dd-b.net> writes:
>
>> Dan Espen <des...@verizon.net> writes:
>>
>>> After working on 1401/1460 I worked at a place with an 8K 1440.
>>> Sitting at the console one day, I keyed in a small routine to set a word
>>> mark around position 10, then put a 1 at location 7999 and in low
>>> storage I keyed in an add 7999 to 7999 and branch back to add.
>>
>> That's the size we had (though readers today will interpret "8K" a bit
>> differently than we did back then). I remember manually encoding the
>> high and low address bits to get big tables (the middle bits were
>> sometimes used for something else; was there indirect addressing or
>> something?)
>
> Yep. My first exposure to the term "K" came with System/360.
> Up until then, the computers adapted to common human usage, Ie.
> they had native base 10 arithmetic and addressing.
>
> On a 1401, the address was 3 characters which gets you to 999.
> Then the BA bits were used in the first and last byte to get up
> to 16,000. The middle bits identified one of 3 index registers
> at fixed storage locations.
>>> So the program was doubling an 8000 digit number.
>>>
>>> On the console you could clearly see the A and B registers cycle down
>>> from 8000 to 10 before the add completed. That was one of the few times
>>> I could watch an individual instruction execute.
>>
>> I really miss front-panel lights. Though at todays's speeds they
>> wouldn't show much any more I suppose.
>
> Agree. All I have now is a disk access indicator light.
> Why this isn't viewed as a marketing opportunity, I don't know.
>>> Today, trying to add 2 8000 digit numbers isn't so easy.
>>
>> Not in machine language, anyway. Plenty of software packages support
>> infinite-precision numbers, though.
>
> I wonder how execution times would stack up.
I love living in the future!
Nick Spalding
Mar 27, 2012, 4:39:37 PM3/27/12
to
David Dyer-Bennet wrote, in <ylfkpqbx...@dd-b.net>
on Tue, 27 Mar 2012 15:23:50 -0500:
>Dan Espen <des...@verizon.net> writes:
>
>> On a 1401, the address was 3 characters which gets you to 999.
>> Then the BA bits were used in the first and last byte to get up
>> to 16,000. The middle bits identified one of 3 index registers
>> at fixed storage locations.
>
>Oh, of course, index registers. Much more basic, very important use.
85-87, 90-92 and 95-97 IIRC.
Its big brothers the 1410 and 7010 could address up to 100,000 though
the 1410 could only have up to 80,000. There were 15 index registers, 5
digits long occupying 25-99, addressed by the BA bits on digits 3 and 4
of the 5 digit addresses.
--
Nick Spalding
on Tue, 27 Mar 2012 15:23:50 -0500:
>Dan Espen <des...@verizon.net> writes:
>
>> On a 1401, the address was 3 characters which gets you to 999.
>> Then the BA bits were used in the first and last byte to get up
>> to 16,000. The middle bits identified one of 3 index registers
>> at fixed storage locations.
>
>Oh, of course, index registers. Much more basic, very important use.
Its big brothers the 1410 and 7010 could address up to 100,000 though
the 1410 could only have up to 80,000. There were 15 index registers, 5
digits long occupying 25-99, addressed by the BA bits on digits 3 and 4
of the 5 digit addresses.
--
Nick Spalding
Message has been deleted
Dan Espen
Mar 27, 2012, 5:39:02 PM3/27/12
to
David Dyer-Bennet <dd...@dd-b.net> writes:
> Dan Espen <des...@verizon.net> writes:
>
>> David Dyer-Bennet <dd...@dd-b.net> writes:
>>
>>> Dan Espen <des...@verizon.net> writes:
>>>
>>>> After working on 1401/1460 I worked at a place with an 8K 1440.
>>>> Sitting at the console one day, I keyed in a small routine to set a word
>>>> mark around position 10, then put a 1 at location 7999 and in low
>>>> storage I keyed in an add 7999 to 7999 and branch back to add.
>>>
>>> That's the size we had (though readers today will interpret "8K" a bit
>>> differently than we did back then). I remember manually encoding the
>>> high and low address bits to get big tables (the middle bits were
>>> sometimes used for something else; was there indirect addressing or
>>> something?)
>>
>> Yep. My first exposure to the term "K" came with System/360.
>> Up until then, the computers adapted to common human usage, Ie.
>> they had native base 10 arithmetic and addressing.
>>
>> On a 1401, the address was 3 characters which gets you to 999.
>> Then the BA bits were used in the first and last byte to get up
>> to 16,000. The middle bits identified one of 3 index registers
>> at fixed storage locations.
>
> Oh, of course, index registers. Much more basic, very important use.
Toward the end of my 1401 programming era, I came to regard the index
> Dan Espen <des...@verizon.net> writes:
>
>> David Dyer-Bennet <dd...@dd-b.net> writes:
>>
>>> Dan Espen <des...@verizon.net> writes:
>>>
>>>> After working on 1401/1460 I worked at a place with an 8K 1440.
>>>> Sitting at the console one day, I keyed in a small routine to set a word
>>>> mark around position 10, then put a 1 at location 7999 and in low
>>>> storage I keyed in an add 7999 to 7999 and branch back to add.
>>>
>>> That's the size we had (though readers today will interpret "8K" a bit
>>> differently than we did back then). I remember manually encoding the
>>> high and low address bits to get big tables (the middle bits were
>>> sometimes used for something else; was there indirect addressing or
>>> something?)
>>
>> Yep. My first exposure to the term "K" came with System/360.
>> Up until then, the computers adapted to common human usage, Ie.
>> they had native base 10 arithmetic and addressing.
>>
>> On a 1401, the address was 3 characters which gets you to 999.
>> Then the BA bits were used in the first and last byte to get up
>> to 16,000. The middle bits identified one of 3 index registers
>> at fixed storage locations.
>
> Oh, of course, index registers. Much more basic, very important use.
registers as less necessary.
Using Store A register (SAR) you could search a table a bit more
efficiently than with index registers.
Just SAR into the instruction that moved the current entry out
of the table. No index to fool with and when you found the
matching entry it was already where you wanted it.
--
Dan Espen
Message has been deleted
Message has been deleted
Rich Alderson
Mar 27, 2012, 7:50:19 PM3/27/12
to
Dan Espen <des...@verizon.net> writes:
> jmfbahciv <See....@aol.com> writes:
>> Walter Bushell wrote:
>>> In article <20120327081048....@eircom.net>,
>>> Ahem A Rivet's Shot <ste...@eircom.net> wrote:
>>>> On Mon, 26 Mar 2012 19:48:55 -0400
>>>> Peter Flass <Peter...@Yahoo.com> wrote:
>>>>> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>>>>>> When was word length standardised? I remember the computer books I
>>>>>> have dating from the early era expected people to be familiar with
>>>>>> octal, binary and hex notation. I have come across references to early
>>>>>> computers using something like a 40 bit word.
> jmfbahciv <See....@aol.com> writes:
>> Walter Bushell wrote:
>>> In article <20120327081048....@eircom.net>,
>>> Ahem A Rivet's Shot <ste...@eircom.net> wrote:
>>>> On Mon, 26 Mar 2012 19:48:55 -0400
>>>> Peter Flass <Peter...@Yahoo.com> wrote:
>>>>> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>>>>>> When was word length standardised? I remember the computer books I
>>>>>> have dating from the early era expected people to be familiar with
>>>>>> octal, binary and hex notation. I have come across references to early
>>>>>> computers using something like a 40 bit word.
>>>>> 12, 24, 26, 40, 48, 60, pick any of the above.
>>>> Er 36, or your not playing with a full DEC.
>>> On some DEC computers you were allowed to set the byte size to 6 or 7
>>> or 12 or 18. Some even used 8 as a byte size, when they wanted to be
>>> silly.
Only the PDP-6/PDP-10 family provides this facility. 7-bit bytes are used for
>>>> Er 36, or your not playing with a full DEC.
>>> On some DEC computers you were allowed to set the byte size to 6 or 7
>>> or 12 or 18. Some even used 8 as a byte size, when they wanted to be
>>> silly.
ASCII text (packed 5 characters per word), 6 bits for a 64-character subset.
Other byte sizes are often used in data structure manipulations. One might use
an 18-bit byte to pick up a value which does not align with the left or right
half of the word.
>> I could set the byte size from 1-36. I don't remember what happens
>> if the byte size was 0. None of the PDP platforms insisted on
>> one flavor of byte nor did they insist on byte-only accessing.
2 8-bit bytes when she makes that claim about "none of the PDP platforms",
of course.
> If you could set it to 1-36, then I'd guess byte size was in
> "-1" notation. Ie. 0 is 1. Therefore, there would be no way
> to set byte size 0.
The format of a PDP-10 section-local[1] byte pointer is
| P | S | |I X Y |
+______+______+_+_+____+________________________+
|pppppp|ssssss|0|i|xxxx|yyyyyyyyyyyyyyyyyyyyyyyy|
+------+------+-+-+----+------------------------+
0 5 6 1 1 1 1 1 1 3
1 2 3 4 7 8 5
The architecture requires that the size field S be non-zero. The position
field P is the location of the rightmost bit of the byte within the word,
represented as the number of bits from the right end of the word; a value of 44
(36) is special, being 1 bit to the left of the high-order bit 0, allowing the
incrementing byte instructions to begin at the leftmost point in the first word
of a string without any schratzing around. If an increment would move beyond
of the current word, the next byte is taken from the leftmost byte position in
the following word.[2]
Bit 12 is required to be 0.
The effective address of the word containing the byte is E (calculated as if in
an instruction, from the Y address in the right half word and if present, the
contents of the index register X and/or the indirection bit I).
[1] The original PDP-10 architecture used 18-bit addresses (allowing a neat
implementation of Lisp CONS cells) for a 256KW space. The second
generation processor presented 256KW spaces out of a 22-bit (4MW) physical
address space to each process in the system (including the monitor). The
third generation defined "extended addressing" of 30 bits and implemented
22 physical, 23 virtual, so "global byte pointers" were defined to allow 30
bits of address. These wre either 2 words, the second with the address, or
1-word pointers in which P+S were encoded in 6-bit values in the range
45-76 representing both position and size for 5 common byte sizes (7, 6, 8,
9 and 18).
[2] For example, if TEXT is located at address 1337, the pseudo-op POINT 7,TEXT
will be assembled as 440700001337. Incrementing byte instructions will
move through the string picking up the bytes at 35, 26, 17, 10, and 1 in
word 1337, then bytes at 35... in word 1340, etc. etc. usw. ktl.
--
Rich Alderson ne...@alderson.users.panix.com
the russet leaves of an autumn oak/inspire once again the failed poet/
to take up his pen/and essay to place his meagre words upon the page...
Rich Alderson
Mar 27, 2012, 8:03:08 PM3/27/12
to
what you meant, but it sounds as though you were claiming 9-bit words for them.
> It is one of the less well known DEC lines tho.
continued to manufacture them for the next decade and a half, and they were
popular enough with their user community to cause very hard feelings when the
line was canceled, I think you're overstating the case.
And were it not for the PDP-7, we might not have a certain universal operating
system in anything like its current form. Thompson and Ritchie *wanted* a
PDP-10, after all.
Message has been deleted
Rich Alderson
Mar 27, 2012, 8:16:26 PM3/27/12
to
David Dyer-Bennet <dd...@dd-b.net> writes:
> With the 11, we get
> into byte-oriented memory, with larger arithmetic units, so it gets
> complicated. Somebody who said the word size was 16 bits wouldn't be
> clearly wrong.
If you mean the PDP-11, they not only "wouldn't be clearly wrong", they'd be
> With the 11, we get
> into byte-oriented memory, with larger arithmetic units, so it gets
> complicated. Somebody who said the word size was 16 bits wouldn't be
> clearly wrong.
correct.
Warren Adams
Mar 27, 2012, 9:07:48 PM3/27/12
to
On 3/27/2012 6:56 AM, Peter Flass wrote:
register, 13 bit address. And memory was only addressed by 20 bit words.
> On 3/27/2012 3:10 AM, Ahem A Rivet's Shot wrote:
>> On Mon, 26 Mar 2012 19:48:55 -0400
>> Peter Flass<Peter...@Yahoo.com> wrote:
>>
>> On Mon, 26 Mar 2012 19:48:55 -0400
>> Peter Flass<Peter...@Yahoo.com> wrote:
>>
>>> On 3/26/2012 1:20 PM, ken...@cix.compulink.co.uk wrote:
>>>> When was word length standardised? I remember the computer books I
>>>> have dating from the early era expected people to be familiar with
>>>> octal, binary and hex notation. I have come across references to early
>>>> computers using something like a 40 bit word.
>>>>
>>>> Ken Young
>>>> When was word length standardised? I remember the computer books I
>>>> have dating from the early era expected people to be familiar with
>>>> octal, binary and hex notation. I have come across references to early
>>>> computers using something like a 40 bit word.
>>>>
>>>
>>> 12, 24, 26, 40, 48, 60, pick any of the above.
>>
>>> 12, 24, 26, 40, 48, 60, pick any of the above.
>>
>> Er 36, or your not playing with a full DEC.
>>
>
>>
>
> Typo (26==36). I did leave out 18 (forgot) and 16 (too common to
> mention). Any others?
>
GE 200 series, 235,etc., were 20 bit words. 5 bit op code, 2 bit index
> mention). Any others?
>
register, 13 bit address. And memory was only addressed by 20 bit words.
Rod Speed
Mar 27, 2012, 9:38:18 PM3/27/12
to
Rich Alderson wrote
> Rod Speed <rod.sp...@gmail.com> wrote
> That may be what you meant,
Its also what I said.
> but it sounds as though you were claiming 9-bit words for them.
Nope.
>> It is one of the less well known DEC lines tho.
> Hmm. Given that their first computer product was an 18-bit system,
> and they continued to manufacture them for the next decade and a
> half, and they were popular enough with their user community to cause
> very hard feelings when the line was canceled, I think you're overstating the case.
Nope. While most who didnt use them have heard of the 8, 11, 10 and the
others with the same word size, they mostly havent heard of the 7/9/15
because it never did sell anything like as well as those other lines did.
> And were it not for the PDP-7, we might not have a certain
> universal operating system in anything like its current form.
> Thompson and Ritchie *wanted* a PDP-10, after all.
Sure, but thats a different issue to what those who never used one have heard of.
> Rod Speed <rod.sp...@gmail.com> wrote
>> Walter Banks wrote
>>> Stan Barr wrote
>>>> Peter Flass <Peter...@Yahoo.com> wrote
>>>>> ken...@cix.compulink.co.uk wrote
>>>>>> When was word length standardised? I remember the computer
>>>>>> books I have dating from the early era expected people to be
>>>>>> familiar with octal, binary and hex notation. I have come across
>>>>>> references to early computers using something like a 40 bit word.
>>>>> 12, 24, 26, 40, 48, 60, pick any of the above. Is it standardized
>>>>> now? Is 32 or 64 the standard? Or maybe 128?
>>>> Historically, 18, 20, 22(?) and 36 too.
>>> Add to that
>>> 13 Bits Westinghouse PRODAC 50
>>> 14 Bits (Instruction paths ) Microchip midrange
>>> 9 bits several small processors so 18 bits is double word size.
>>> I have only seen them in Special Purpose processors
>> Thats also what the PDP7/9/15 had, nothing like special purpose processors.
> These, along with the PDP-1 and PDP-4, were all 18-bit computers.
What I said.
>>> Stan Barr wrote
>>>> Peter Flass <Peter...@Yahoo.com> wrote
>>>>> ken...@cix.compulink.co.uk wrote
>>>>>> When was word length standardised? I remember the computer
>>>>>> books I have dating from the early era expected people to be
>>>>>> familiar with octal, binary and hex notation. I have come across
>>>>>> references to early computers using something like a 40 bit word.
>>>>> 12, 24, 26, 40, 48, 60, pick any of the above. Is it standardized
>>>>> now? Is 32 or 64 the standard? Or maybe 128?
>>>> Historically, 18, 20, 22(?) and 36 too.
>>> Add to that
>>> 13 Bits Westinghouse PRODAC 50
>>> 14 Bits (Instruction paths ) Microchip midrange
>>> 9 bits several small processors so 18 bits is double word size.
>>> I have only seen them in Special Purpose processors
>> Thats also what the PDP7/9/15 had, nothing like special purpose processors.
> These, along with the PDP-1 and PDP-4, were all 18-bit computers.
> That may be what you meant,
> but it sounds as though you were claiming 9-bit words for them.
>> It is one of the less well known DEC lines tho.
> Hmm. Given that their first computer product was an 18-bit system,
> and they continued to manufacture them for the next decade and a
> half, and they were popular enough with their user community to cause
> very hard feelings when the line was canceled, I think you're overstating the case.
others with the same word size, they mostly havent heard of the 7/9/15
because it never did sell anything like as well as those other lines did.
> And were it not for the PDP-7, we might not have a certain
> universal operating system in anything like its current form.
> Thompson and Ritchie *wanted* a PDP-10, after all.
Dan Espen
Mar 27, 2012, 9:49:46 PM3/27/12
to
>> If you could set it to 1-36, then I'd guess byte size was in
>> "-1" notation. Ie. 0 is 1. Therefore, there would be no way
>> to set byte size 0.
>
> Your guess would be wrong.
>
> The format of a PDP-10 section-local[1] byte pointer is
>
> | P | S | |I X Y |
> +______+______+_+_+____+________________________+
> |pppppp|ssssss|0|i|xxxx|yyyyyyyyyyyyyyyyyyyyyyyy|
> +------+------+-+-+----+------------------------+
> 0 5 6 1 1 1 1 1 1 3
> 1 2 3 4 7 8 5
>
> The architecture requires that the size field S be non-zero. The position
> field P is the location of the rightmost bit of the byte within the word,
> represented as the number of bits from the right end of the word; a value of 44
> (36) is special, being 1 bit to the left of the high-order bit 0, allowing the
> incrementing byte instructions to begin at the leftmost point in the first word
> of a string without any schratzing around. If an increment would move beyond
> of the current word, the next byte is taken from the leftmost byte position in
> the following word.[2]
Well, I was smart enough to say I was guessing.
>> "-1" notation. Ie. 0 is 1. Therefore, there would be no way
>> to set byte size 0.
>
> Your guess would be wrong.
>
> The format of a PDP-10 section-local[1] byte pointer is
>
> | P | S | |I X Y |
> +______+______+_+_+____+________________________+
> |pppppp|ssssss|0|i|xxxx|yyyyyyyyyyyyyyyyyyyyyyyy|
> +------+------+-+-+----+------------------------+
> 0 5 6 1 1 1 1 1 1 3
> 1 2 3 4 7 8 5
>
> The architecture requires that the size field S be non-zero. The position
> field P is the location of the rightmost bit of the byte within the word,
> represented as the number of bits from the right end of the word; a value of 44
> (36) is special, being 1 bit to the left of the high-order bit 0, allowing the
> incrementing byte instructions to begin at the leftmost point in the first word
> of a string without any schratzing around. If an increment would move beyond
> of the current word, the next byte is taken from the leftmost byte position in
> the following word.[2]
Not that I would have expected something as complicated as what you
describe. I don't pretend to understand everything you've explained
but 6 bits for an offset (P), 6 bits for a size (S) with zero excluded
and a special value?
The 1401 presented a simple architecture. Most people could completely
understand everything about the machine. S/360 made complete
understanding something difficult. Mr. Wheeler conquered it, and I
probably could have but I'd guess less than 1% of programmers really
understand most of S/360.
You've presented only a glimpse of PDP I'm sure but that looks like a
big mess with not much to recommend it.
--
Dan Espen
Rod Speed
Mar 27, 2012, 11:08:06 PM3/27/12
to
Dan Espen wrote
> Rich Alderson <ne...@alderson.users.panix.com> wrote
The 11 and Vax was nothing like that messy/bizarre
> Rich Alderson <ne...@alderson.users.panix.com> wrote
Bill Leary
Mar 28, 2012, 2:37:28 AM3/28/12
to
((...portions omitted..))
"David Dyer-Bennet" wrote in message news:ylfkzkb1...@dd-b.net...
mean it's word size is a byte.
> With the 11, we get into byte-oriented memory, with larger arithmetic
> units, so it gets complicated. Somebody who said the word size was
> 16 bits wouldn't be clearly wrong.
The word size, memory width, of the PDP-11 is 16 bits. It's instruction set
provides for byte addressing. Some machines of the era also provided for
hardware support of bytes. If you wrote a byte it didn't have to do a
read/modify/write cycle to the word containing the byte but could actually
write just the upper or lower byte. I seem to recall that some of the early
Intel 16 bit CPUs provided for this as well.
But yes, it can get complicated. Was the word size of the Motorola MC68000
16 or 32 bits? 16-bit memory, 32 bit registers. And after deciding that,
how about the 68008 with its 8 bit memory?
- Bill
"David Dyer-Bennet" wrote in message news:ylfkzkb1...@dd-b.net...
> Peter Flass <Peter...@Yahoo.com> writes:
>> What is the definition. I'd take "word length" to be the size of a
>> register, for machines that have registers.
>
> That's one possible definition for one class of machine, certainly.
>
> It used to mean the basic unit of memory access; but clearly that's
> the byte in nearly every case now.
I'd have to disagree with that. The memory being byte addressable doesn't
>> What is the definition. I'd take "word length" to be the size of a
>> register, for machines that have registers.
>
> That's one possible definition for one class of machine, certainly.
>
> It used to mean the basic unit of memory access; but clearly that's
> the byte in nearly every case now.
mean it's word size is a byte.
> With the 11, we get into byte-oriented memory, with larger arithmetic
> units, so it gets complicated. Somebody who said the word size was
> 16 bits wouldn't be clearly wrong.
provides for byte addressing. Some machines of the era also provided for
hardware support of bytes. If you wrote a byte it didn't have to do a
read/modify/write cycle to the word containing the byte but could actually
write just the upper or lower byte. I seem to recall that some of the early
Intel 16 bit CPUs provided for this as well.
But yes, it can get complicated. Was the word size of the Motorola MC68000
16 or 32 bits? 16-bit memory, 32 bit registers. And after deciding that,
how about the 68008 with its 8 bit memory?
- Bill
Message has been deleted
Jean-Marc Bourguet
Mar 28, 2012, 5:02:46 AM3/28/12
to
"Bill Leary" <Bill_...@msn.com> writes:
> But yes, it can get complicated. Was the word size of the Motorola MC68000
> 16 or 32 bits? 16-bit memory, 32 bit registers. And after deciding that,
> how about the 68008 with its 8 bit memory?
If you go and ask on comp.arch, the most common definition is the size
> But yes, it can get complicated. Was the word size of the Motorola MC68000
> 16 or 32 bits? 16-bit memory, 32 bit registers. And after deciding that,
> how about the 68008 with its 8 bit memory?
of integer/fixed point registers, not any other size (neither
architecturally -- address, index, floating point, vector registers, or
visible address spaces -- nor implementation -- internal or external
busses of various kind). But confusion regularly occur, when people are
comming from a culture where several have the same value and found out
an architecture where the unification doesn't occur or when marketting
departments find out usefull to use it as a merit value.
I think the 68000 familly is the major example of a deviation from the
common definition. It is a 32 bits architecture but was often called a
16 bits -- and 8 bits for the 68008 -- one, probably more because it was
in those market segments they wanted to compete than for any technical
reasons.
Yours,
--
Jean-Marc
Message has been deleted
Message has been deleted
Christian Brunschen
Mar 28, 2012, 6:11:03 AM3/28/12
to
In article <F6udnaqIzoUxLO_S...@giganews.com>,
40-bit computers (like the ILLIAC) based on the IAS machine design? 40-bit
word size ... but 20-bit instruction size. So each word held two
instructions!
For one of the Swedish computers based on this design, SMIL, this meant
that there were two variants on jump instructions - one to jump to the
left instruction in a word, one to ump to the right instruction. However,
loading & storing data to/from registers is generally done on a whole-word
basis. There _were_ also data transfer variants that would load
information shifted left or right by 20 bits, so you could load the left
half of a word in memory into the right half of a register, but the number
format (fized-point initially, later extended to also allow a
floating-point format) is 40 bits in size, 1 whole word.
Anyway, this just goes to show that even from the start, the concept of a
processor's "word size" was ... fluent at best.
Regarding the 680xx processors, I seem to recall that the 68000 was
referred to as a '16/32' bit processor - 32 bit internally, with a 16-bit
external bus. The 68008 was, analogously, referred to as an 8/32 bit
processor or even an 8/16/32-bit processor. Of course, the 68k family have
always had memory addressable in 8-bit quantities. The 68000 however does
not expose the lowest address bit (A0), whereas the 68008 does (see the
M68000 User Manual,
http://www.freescale.com/files/32bit/doc/ref_manual/MC68000UM.pdf): the
68000 expected 16-bit-wide memory and peripherals; the 68008 was
introduced to allow less-expensive 8-bit-wide memory and peripherals to be
used, as well as allowing less-expensive boards to be built (since they
required less board space, fewer connections to be routed, etc).
The 68020 exposes all address bits and allows the external 'port' (memory
or similar) to specify its width (8, 16 or 32 bits), and adapts itself -
using the low address bits if/when necessary (see the MC68020 Manual,
http://www.freescale.com/files/32bit/doc/ref_manual/MC68020UM.pdf).
Similarly, of course, the 8088 was an 8/16 bit processor - 16 bits
internally, 8 bits externally - and a variant of the 16-bit-all-the-way
8086 processor (much like the 68008 is a variant of the 68000).
> - Bill
// Christian
Bill Leary <Bill_...@msn.com> wrote:
>But yes, it can get complicated. Was the word size of the Motorola MC68000
>16 or 32 bits? 16-bit memory, 32 bit registers. And after deciding that,
>how about the 68008 with its 8 bit memory?
Well, things have been complicated right form the start. Those early
>But yes, it can get complicated. Was the word size of the Motorola MC68000
>16 or 32 bits? 16-bit memory, 32 bit registers. And after deciding that,
>how about the 68008 with its 8 bit memory?
40-bit computers (like the ILLIAC) based on the IAS machine design? 40-bit
word size ... but 20-bit instruction size. So each word held two
instructions!
For one of the Swedish computers based on this design, SMIL, this meant
that there were two variants on jump instructions - one to jump to the
left instruction in a word, one to ump to the right instruction. However,
loading & storing data to/from registers is generally done on a whole-word
basis. There _were_ also data transfer variants that would load
information shifted left or right by 20 bits, so you could load the left
half of a word in memory into the right half of a register, but the number
format (fized-point initially, later extended to also allow a
floating-point format) is 40 bits in size, 1 whole word.
Anyway, this just goes to show that even from the start, the concept of a
processor's "word size" was ... fluent at best.
Regarding the 680xx processors, I seem to recall that the 68000 was
referred to as a '16/32' bit processor - 32 bit internally, with a 16-bit
external bus. The 68008 was, analogously, referred to as an 8/32 bit
processor or even an 8/16/32-bit processor. Of course, the 68k family have
always had memory addressable in 8-bit quantities. The 68000 however does
not expose the lowest address bit (A0), whereas the 68008 does (see the
M68000 User Manual,
http://www.freescale.com/files/32bit/doc/ref_manual/MC68000UM.pdf): the
68000 expected 16-bit-wide memory and peripherals; the 68008 was
introduced to allow less-expensive 8-bit-wide memory and peripherals to be
used, as well as allowing less-expensive boards to be built (since they
required less board space, fewer connections to be routed, etc).
The 68020 exposes all address bits and allows the external 'port' (memory
or similar) to specify its width (8, 16 or 32 bits), and adapts itself -
using the low address bits if/when necessary (see the MC68020 Manual,
http://www.freescale.com/files/32bit/doc/ref_manual/MC68020UM.pdf).
Similarly, of course, the 8088 was an 8/16 bit processor - 16 bits
internally, 8 bits externally - and a variant of the 16-bit-all-the-way
8086 processor (much like the 68008 is a variant of the 68000).
> - Bill
// Christian
Peter Flass
Mar 28, 2012, 7:58:58 AM3/28/12
to
On 3/27/2012 3:27 PM, David Dyer-Bennet wrote:
1-bit bytes were sometimes useful for accessing
> bitvectors.
That's one thing I miss a lot with today's architectures. If you want
to move some odd number of bits from some offset in one location to a
different offset somewhere else you're going to do a lot of shifting and
masking instead of a simple load end store. Now try repeating this
operation over a series of words or bytes.
--
Pete
1-bit bytes were sometimes useful for accessing
> bitvectors.
That's one thing I miss a lot with today's architectures. If you want
to move some odd number of bits from some offset in one location to a
different offset somewhere else you're going to do a lot of shifting and
masking instead of a simple load end store. Now try repeating this
operation over a series of words or bytes.
--
Pete
Ahem A Rivet's Shot
Mar 28, 2012, 7:54:05 AM3/28/12
to
On Wed, 28 Mar 2012 11:24:46 +0200
Jean-Marc Bourguet <j...@bourguet.org> wrote:
> The problem is that what is best depend on the intended usage. One code
> point per glyph is better for some (composition system for instance). A
> more structured approach is better for other. Various trade-off have
> been made in the past. And as absolute compatibility is often a
> prerequisite for an unifying system to succeed, Unicode has all the
> mistake done in any coding system -- and add some to make good measure.
It is also the only coding system that can be used across such a
wide range of scripts. Warts and all it's the best we've got.
--
Steve O'Hara-Smith | Directable Mirror Arrays
C:>WIN | A better way to focus the sun
The computer obeys and wins. | licences available see
You lose and Bill collects. | http://www.sohara.org/
Jean-Marc Bourguet <j...@bourguet.org> wrote:
> The problem is that what is best depend on the intended usage. One code
> point per glyph is better for some (composition system for instance). A
> more structured approach is better for other. Various trade-off have
> been made in the past. And as absolute compatibility is often a
> prerequisite for an unifying system to succeed, Unicode has all the
> mistake done in any coding system -- and add some to make good measure.
It is also the only coding system that can be used across such a
wide range of scripts. Warts and all it's the best we've got.
--
Steve O'Hara-Smith | Directable Mirror Arrays
C:>WIN | A better way to focus the sun
The computer obeys and wins. | licences available see
You lose and Bill collects. | http://www.sohara.org/
Peter Flass
Mar 28, 2012, 8:10:10 AM3/28/12
to
On 3/28/2012 2:37 AM, Bill Leary wrote:
> ((...portions omitted..))
> "David Dyer-Bennet" wrote in message news:ylfkzkb1...@dd-b.net...
>> Peter Flass <Peter...@Yahoo.com> writes:
>>> What is the definition. I'd take "word length" to be the size of a
>>> register, for machines that have registers.
>>
>> That's one possible definition for one class of machine, certainly.
>>
>> It used to mean the basic unit of memory access; but clearly that's
>> the byte in nearly every case now.
>
> I'd have to disagree with that. The memory being byte addressable
> doesn't mean it's word size is a byte.
Truth. This slipped by me, but the 360 was byte-addressed, but word
> ((...portions omitted..))
> "David Dyer-Bennet" wrote in message news:ylfkzkb1...@dd-b.net...
>> Peter Flass <Peter...@Yahoo.com> writes:
>>> What is the definition. I'd take "word length" to be the size of a
>>> register, for machines that have registers.
>>
>> That's one possible definition for one class of machine, certainly.
>>
>> It used to mean the basic unit of memory access; but clearly that's
>> the byte in nearly every case now.
>
> I'd have to disagree with that. The memory being byte addressable
> doesn't mean it's word size is a byte.
operands had to be on word (32-bit) boundaries. It wasn't until the 370
that this restriction was relaxed.
I haven't commented on the PDP-11 because I'm not an authority, but
didn't it have byte addressing but similar boundary restrictions?
--
Pete
Peter Flass
Mar 28, 2012, 8:13:06 AM3/28/12
to
On 3/28/2012 5:24 AM, Jean-Marc Bourguet wrote:
>
> Accent are a secondary sort key in French
And not considered at all for sorting German, AFAIK.
--
Pete
>
> Accent are a secondary sort key in French
And not considered at all for sorting German, AFAIK.
--
Pete
Peter Flass
Mar 28, 2012, 8:16:50 AM3/28/12
to
On 3/28/2012 6:11 AM, Christian Brunschen wrote:
> In article<F6udnaqIzoUxLO_S...@giganews.com>,
> Bill Leary<Bill_...@msn.com> wrote:
>> But yes, it can get complicated. Was the word size of the Motorola MC68000
>> 16 or 32 bits? 16-bit memory, 32 bit registers. And after deciding that,
>> how about the 68008 with its 8 bit memory?
>
> Well, things have been complicated right form the start. Those early
> 40-bit computers (like the ILLIAC) based on the IAS machine design? 40-bit
> word size ... but 20-bit instruction size. So each word held two
> instructions!
This was common too. Burroughs 5000, etc. used (AIR) a 48-bit word, with
> In article<F6udnaqIzoUxLO_S...@giganews.com>,
> Bill Leary<Bill_...@msn.com> wrote:
>> But yes, it can get complicated. Was the word size of the Motorola MC68000
>> 16 or 32 bits? 16-bit memory, 32 bit registers. And after deciding that,
>> how about the 68008 with its 8 bit memory?
>
> Well, things have been complicated right form the start. Those early
> 40-bit computers (like the ILLIAC) based on the IAS machine design? 40-bit
> word size ... but 20-bit instruction size. So each word held two
> instructions!
multiple instruction "syllables" per word. The instructions were very
short because they were mostly zero-address.
I believe the CDC 6400, 660, etc. series, with a 60-bit word, also
stored multiple instructions in a word.
--
Pete
Jean-Marc Bourguet
Mar 28, 2012, 8:13:30 AM3/28/12
to
Ahem A Rivet's Shot <ste...@eircom.net> writes:
> On Wed, 28 Mar 2012 11:24:46 +0200
> Jean-Marc Bourguet <j...@bourguet.org> wrote:
>
>> The problem is that what is best depend on the intended usage. One code
>> point per glyph is better for some (composition system for instance). A
>> more structured approach is better for other. Various trade-off have
>> been made in the past. And as absolute compatibility is often a
>> prerequisite for an unifying system to succeed, Unicode has all the
>> mistake done in any coding system -- and add some to make good measure.
>
> It is also the only coding system that can be used across such a
> wide range of scripts. Warts and all it's the best we've got.
Agree. Well, there is ISO 2022, which is a framework to integrate
> On Wed, 28 Mar 2012 11:24:46 +0200
> Jean-Marc Bourguet <j...@bourguet.org> wrote:
>
>> The problem is that what is best depend on the intended usage. One code
>> point per glyph is better for some (composition system for instance). A
>> more structured approach is better for other. Various trade-off have
>> been made in the past. And as absolute compatibility is often a
>> prerequisite for an unifying system to succeed, Unicode has all the
>> mistake done in any coding system -- and add some to make good measure.
>
> It is also the only coding system that can be used across such a
> wide range of scripts. Warts and all it's the best we've got.
several charsets respecting some constraints -- excepted Unicode, most
of them do --, but it is worse than Unicode.
Yours,
--
Jean-Marc
Erik Trulsson
Mar 28, 2012, 9:38:13 AM3/28/12
to
then (early/mid 80's) the most common definition of "bit-ness" seemed
to be the width of the data bus rather than the width of internal
registers.
--
<Insert your favourite quote here.>
Erik Trulsson
ertr...@student.uu.se
David Dyer-Bennet
Mar 28, 2012, 10:46:10 AM3/28/12
to
subroutine linkage, but I don't remember using SAR. Could have just
forgotten, though, since I did a LOT of stuff using tables.
--
David Dyer-Bennet, dd...@dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info
David Dyer-Bennet
Mar 28, 2012, 11:00:06 AM3/28/12
to
Dan Espen <des...@verizon.net> writes:
> Rich Alderson <ne...@alderson.users.panix.com> writes:
>
>> Dan Espen <des...@verizon.net> writes:
>>
>>> If you could set it to 1-36, then I'd guess byte size was in
>>> "-1" notation. Ie. 0 is 1. Therefore, there would be no way
>>> to set byte size 0.
>>
>> Your guess would be wrong.
>>
>> The format of a PDP-10 section-local[1] byte pointer is
>>
>> | P | S | |I X Y |
>> +______+______+_+_+____+________________________+
>> |pppppp|ssssss|0|i|xxxx|yyyyyyyyyyyyyyyyyyyyyyyy|
>> +------+------+-+-+----+------------------------+
>> 0 5 6 1 1 1 1 1 1 3
>> 1 2 3 4 7 8 5
>>
>> The architecture requires that the size field S be non-zero. The position
>> field P is the location of the rightmost bit of the byte within the word,
>> represented as the number of bits from the right end of the word; a value of 44
>> (36) is special, being 1 bit to the left of the high-order bit 0, allowing the
>> incrementing byte instructions to begin at the leftmost point in the first word
>> of a string without any schratzing around. If an increment would move beyond
>> of the current word, the next byte is taken from the leftmost byte position in
>> the following word.[2]
>
> Well, I was smart enough to say I was guessing.
Yeah, and I bought it.
> Rich Alderson <ne...@alderson.users.panix.com> writes:
>
>> Dan Espen <des...@verizon.net> writes:
>>
>>> If you could set it to 1-36, then I'd guess byte size was in
>>> "-1" notation. Ie. 0 is 1. Therefore, there would be no way
>>> to set byte size 0.
>>
>> Your guess would be wrong.
>>
>> The format of a PDP-10 section-local[1] byte pointer is
>>
>> | P | S | |I X Y |
>> +______+______+_+_+____+________________________+
>> |pppppp|ssssss|0|i|xxxx|yyyyyyyyyyyyyyyyyyyyyyyy|
>> +------+------+-+-+----+------------------------+
>> 0 5 6 1 1 1 1 1 1 3
>> 1 2 3 4 7 8 5
>>
>> The architecture requires that the size field S be non-zero. The position
>> field P is the location of the rightmost bit of the byte within the word,
>> represented as the number of bits from the right end of the word; a value of 44
>> (36) is special, being 1 bit to the left of the high-order bit 0, allowing the
>> incrementing byte instructions to begin at the leftmost point in the first word
>> of a string without any schratzing around. If an increment would move beyond
>> of the current word, the next byte is taken from the leftmost byte position in
>> the following word.[2]
>
> Well, I was smart enough to say I was guessing.
I should have thought more (or remembered better); the range of values
needed for size and position aren't a power of two, so there are going
to be extra values available in the field, so my normal assumptions
about efficient use of values don't precisely apply.
> Not that I would have expected something as complicated as what you
> describe. I don't pretend to understand everything you've explained
> but 6 bits for an offset (P), 6 bits for a size (S) with zero excluded
> and a special value?
to represent it? If your answer is fewer than 6, how would you do it?
:-). Similarly, size has to represent the range 1-36 at least.
(Since a basic memory address is 18 bits, it's pretty much a given that
an original-generation byte pointer had to be one 36-bit word, so they
had space to play with. As they proved with the one-word global byte
pointer, you CAN get many of the useful features into smaller space, but
there was no point originally. AND the original KA was hardwired, not
micro-coded. AND the overall instruction architecture tended towards
being logically complete, so why would they arbitrarily limit the
positions or sizes of bytes?)
> The 1401 presented a simple architecture. Most people could completely
> understand everything about the machine. S/360 made complete
> understanding something difficult. Mr. Wheeler conquered it, and I
> probably could have but I'd guess less than 1% of programmers really
> understand most of S/360.
>
> You've presented only a glimpse of PDP I'm sure but that looks like a
> big mess with not much to recommend it.
only, though, I didn't do much with privileged instructions), and it was
quite easy to understand. Much simpler than the PDP-11 for example.
Getting a detailed explanation of byte pointers as your introduction is
not the easiest way to approach it :-). Many basic concepts used there
are used everywhere else, too, so if you start from simple stuff, you'll
understand a lot of what's going on when you get to byte pointers and
only have to add the new bits.
David Dyer-Bennet
Mar 28, 2012, 11:02:02 AM3/28/12
to
Rich Alderson <ne...@alderson.users.panix.com> writes:
> David Dyer-Bennet <dd...@dd-b.net> writes:
>
>> With the 11, we get
>> into byte-oriented memory, with larger arithmetic units, so it gets
>> complicated. Somebody who said the word size was 16 bits wouldn't be
>> clearly wrong.
>
> If you mean the PDP-11, they not only "wouldn't be clearly wrong", they'd be
> correct.
Yeah, okay. I'd been thinking through enough different architectures,
> David Dyer-Bennet <dd...@dd-b.net> writes:
>
>> With the 11, we get
>> into byte-oriented memory, with larger arithmetic units, so it gets
>> complicated. Somebody who said the word size was 16 bits wouldn't be
>> clearly wrong.
>
> If you mean the PDP-11, they not only "wouldn't be clearly wrong", they'd be
> correct.
and remembering the arguments about what word size even meant (as
computers stopped being basically word-oriented), that I was being
cautious. But yeah, I think the PDP-11 does qualify as 16-bit by pretty
much all sane definitions.
David Dyer-Bennet
Mar 28, 2012, 11:03:44 AM3/28/12
to
"Bill Leary" <Bill_...@msn.com> writes:
> ((...portions omitted..))
> "David Dyer-Bennet" wrote in message news:ylfkzkb1...@dd-b.net...
>> Peter Flass <Peter...@Yahoo.com> writes:
>>> What is the definition. I'd take "word length" to be the size of a
>>> register, for machines that have registers.
>>
>> That's one possible definition for one class of machine, certainly.
>>
>> It used to mean the basic unit of memory access; but clearly that's
>> the byte in nearly every case now.
>
> I'd have to disagree with that. The memory being byte addressable
> doesn't mean it's word size is a byte.
Yes, that's what I was saying.
> ((...portions omitted..))
> "David Dyer-Bennet" wrote in message news:ylfkzkb1...@dd-b.net...
>> Peter Flass <Peter...@Yahoo.com> writes:
>>> What is the definition. I'd take "word length" to be the size of a
>>> register, for machines that have registers.
>>
>> That's one possible definition for one class of machine, certainly.
>>
>> It used to mean the basic unit of memory access; but clearly that's
>> the byte in nearly every case now.
>
> I'd have to disagree with that. The memory being byte addressable
> doesn't mean it's word size is a byte.
But that *used to* be the standard definition of "word size". As
architectures changed, the definition changed some, and eventually we
mostly gave up on the idea.
>> With the 11, we get into byte-oriented memory, with larger arithmetic
>> units, so it gets complicated. Somebody who said the word size was
>> 16 bits wouldn't be clearly wrong.
>
> The word size, memory width, of the PDP-11 is 16 bits. It's
> instruction set provides for byte addressing. Some machines of the
> era also provided for hardware support of bytes. If you wrote a byte
> it didn't have to do a read/modify/write cycle to the word containing
> the byte but could actually write just the upper or lower byte. I
> seem to recall that some of the early Intel 16 bit CPUs provided for
> this as well.
>
> But yes, it can get complicated. Was the word size of the Motorola
> MC68000 16 or 32 bits? 16-bit memory, 32 bit registers. And after
> deciding that, how about the 68008 with its 8 bit memory?
basic unit we evaluated computers by, as I remember it.
Message has been deleted
Charles Richmond
Mar 28, 2012, 11:15:43 AM3/28/12
to
"Rich Alderson" <ne...@alderson.users.panix.com> wrote in message
news:mddzkb1...@panix5.panix.com...
>
> [snip...] [snip...]
> [snip...]
--
+<><><><><><><><><><><><><><><><><><><>+
| Charles Richmond nume...@aquaporin4.com |
+<><><><><><><><><><><><><><><><><><><>+
news:mddzkb1...@panix5.panix.com...
>
> [snip...] [snip...]
> [snip...]
>
> And were it not for the PDP-7, we might not have a certain universal
> operating
> system in anything like its current form. Thompson and Ritchie *wanted* a
> PDP-10, after all.
>
Back then, we *all* wanted our own PDP-10!!! :-)
> And were it not for the PDP-7, we might not have a certain universal
> operating
> system in anything like its current form. Thompson and Ritchie *wanted* a
> PDP-10, after all.
>
--
+<><><><><><><><><><><><><><><><><><><>+
| Charles Richmond nume...@aquaporin4.com |
+<><><><><><><><><><><><><><><><><><><>+
Stan Barr
Mar 28, 2012, 11:26:35 AM3/28/12
to
On Tue, 27 Mar 2012 12:31:26 -0700, Patrick Scheible <k...@zipcon.net> wrote:
> Peter Flass <Peter...@Yahoo.com> writes:
>
>> On 3/26/2012 5:05 PM, David Dyer-Bennet wrote:
>>>
>>> Word length has never been standardized. However, basically all modern
>>> general-purpose processors are byte-oriented. That became popular
>>> roughly with the IBM 360 series in 1964. Modern systems don't have a
>>> "word length" in anything like the classic sense.
> sizes of registers?
The Z80 had both 8 and 16 bit registers.
Even the venerable SC/MP had an 8-bit acc and 16-bit pointer registers.
--
Cheers,
Stan Barr plan.b .at. dsl .dot. pipex .dot. com
The future was never like this!
> Peter Flass <Peter...@Yahoo.com> writes:
>
>> On 3/26/2012 5:05 PM, David Dyer-Bennet wrote:
>>>
>>> Word length has never been standardized. However, basically all modern
>>> general-purpose processors are byte-oriented. That became popular
>>> roughly with the IBM 360 series in 1964. Modern systems don't have a
>>> "word length" in anything like the classic sense.
>>>
>>
>> What is the definition. I'd take "word length" to be the size of a
>> register, for machines that have registers.
>
> Yes, but what about those modern machines with two or more different
>>
>> What is the definition. I'd take "word length" to be the size of a
>> register, for machines that have registers.
>
> sizes of registers?
The Z80 had both 8 and 16 bit registers.
Even the venerable SC/MP had an 8-bit acc and 16-bit pointer registers.
--
Cheers,
Stan Barr plan.b .at. dsl .dot. pipex .dot. com
The future was never like this!
Dan Espen
Mar 28, 2012, 11:46:30 AM3/28/12
to
I remember at first wondering what it was for.
I'm sure I can't produce working code but it went something like:
Set MLC source address to end of table.
MOVE_LOOP:
move table_entry to target_area
SAR *-4? (change table_entry address in prior move instruction)
Compare target_area to desired_key
BNE MOVE_LOOP
* DESIRED TABLE ENTRY FOUND
A table with a key and value simply needs a chained move.
The A register is always left pointing to the next lower
entry.
There's another trick using move with word marks instead of
just move that can make tables really small but I'd have to
look at the bottom of the pile of my old listings to describe it.
--
Dan Espen
ken...@cix.compulink.co.uk
Mar 28, 2012, 11:51:22 AM3/28/12
to
In article <F6udnaqIzoUxLO_S...@giganews.com>,
address space. The only computer I recall using it was the Sinclair QL.
The 68000 was limited to 24 bit addressing IIRC as only 24 address lines
were implemented to reduce package size. X86 implemented 20 bit
addressing with either 8 or 16 bit data bus. The Z8000 IIRC came with 16
and 8 bit versions that were not binary compatible and in the 8 bit case
could only address 64K. Again IIRC the 8 bit version of the National
Semiconductors 16 bit processor had an 8080 mode. There may have been
another early 16 bit processor I have forgotten, I am not sure what TI
did after the 9400. The problem was not the chip but DILP chips limiting
the number of contacts possible.
By the way when was bit slicing depreciated?
Ken Young
Bill_...@msn.com (Bill Leary) wrote:
> 16-bit memory, 32 bit registers. And after deciding that,
> how about the 68008 with its 8 bit memory?
The 68008 had an 8 bit data bus with 32 bit registers and IIRC a 1MB
> 16-bit memory, 32 bit registers. And after deciding that,
> how about the 68008 with its 8 bit memory?
address space. The only computer I recall using it was the Sinclair QL.
The 68000 was limited to 24 bit addressing IIRC as only 24 address lines
were implemented to reduce package size. X86 implemented 20 bit
addressing with either 8 or 16 bit data bus. The Z8000 IIRC came with 16
and 8 bit versions that were not binary compatible and in the 8 bit case
could only address 64K. Again IIRC the 8 bit version of the National
Semiconductors 16 bit processor had an 8080 mode. There may have been
another early 16 bit processor I have forgotten, I am not sure what TI
did after the 9400. The problem was not the chip but DILP chips limiting
the number of contacts possible.
By the way when was bit slicing depreciated?
Ken Young
Dan Espen
Mar 28, 2012, 11:55:55 AM3/28/12
to
David Dyer-Bennet <dd...@dd-b.net> writes:
> The offset has to support the range 0-35. How many bits would *you* use
> to represent it?
Agreed, missed that completely.
> The offset has to support the range 0-35. How many bits would *you* use
> to represent it?
--
Dan Espen
Charlie Gibbs
Mar 28, 2012, 12:40:40 PM3/28/12
to
In article <ick425k...@home.home>, des...@verizon.net (Dan Espen)
Where else but in the magical world of marketing can you
add 32 to 32 and get 65?
>> I really miss front-panel lights. Though at todays's speeds they
>> wouldn't show much any more I suppose.
>
> Agree. All I have now is a disk access indicator light.
> Why this isn't viewed as a marketing opportunity, I don't know.
It seems to be seen as an opportunity on everything _except_
computers. Just about every gadget has far too many blinkenlights -
most of them that irritating high-intensity blue - the purpose of
which is to distract you from what's actually going on.
(Why are people so infatuated with using bright blue lights?
Why does a dog lick his balls? Because he can.)
On the other hand, many user interfaces have the equivalent
of all those irritating blinkenlights. So they're still there,
they've just moved from the front panel to the monitor screen -
and morphed into things like in-your-face popup messages, etc.
--
/~\ cgi...@kltpzyxm.invalid (Charlie Gibbs)
\ / I'm really at ac.dekanfrus if you read it the right way.
X Top-posted messages will probably be ignored. See RFC1855.
/ \ HTML will DEFINITELY be ignored. Join the ASCII ribbon campaign!
writes:
> David Dyer-Bennet <dd...@dd-b.net> writes:
>
>> That's the size we had (though readers today will interpret "8K" a
>> bit differently than we did back then). I remember manually encoding
>> the high and low address bits to get big tables (the middle bits were
>> sometimes used for something else; was there indirect addressing or
>> something?)
>
> Yep. My first exposure to the term "K" came with System/360.
> Up until then, the computers adapted to common human usage, Ie.
> they had native base 10 arithmetic and addressing.
I defined the two as "salesman's K" vs. "programmer's K".
> David Dyer-Bennet <dd...@dd-b.net> writes:
>
>> That's the size we had (though readers today will interpret "8K" a
>> bit differently than we did back then). I remember manually encoding
>> the high and low address bits to get big tables (the middle bits were
>> sometimes used for something else; was there indirect addressing or
>> something?)
>
> Yep. My first exposure to the term "K" came with System/360.
> Up until then, the computers adapted to common human usage, Ie.
> they had native base 10 arithmetic and addressing.
Where else but in the magical world of marketing can you
add 32 to 32 and get 65?
>> I really miss front-panel lights. Though at todays's speeds they
>> wouldn't show much any more I suppose.
>
> Agree. All I have now is a disk access indicator light.
> Why this isn't viewed as a marketing opportunity, I don't know.
It seems to be seen as an opportunity on everything _except_
computers. Just about every gadget has far too many blinkenlights -
most of them that irritating high-intensity blue - the purpose of
which is to distract you from what's actually going on.
(Why are people so infatuated with using bright blue lights?
Why does a dog lick his balls? Because he can.)
On the other hand, many user interfaces have the equivalent
of all those irritating blinkenlights. So they're still there,
they've just moved from the front panel to the monitor screen -
and morphed into things like in-your-face popup messages, etc.
--
/~\ cgi...@kltpzyxm.invalid (Charlie Gibbs)
\ / I'm really at ac.dekanfrus if you read it the right way.
X Top-posted messages will probably be ignored. See RFC1855.
/ \ HTML will DEFINITELY be ignored. Join the ASCII ribbon campaign!
Charlie Gibbs
Mar 28, 2012, 12:44:44 PM3/28/12
to
In article <jkuu5m$ikl$1...@dont-email.me>, Peter...@Yahoo.com
I once had to read a tape that must have been written by such a system.
Each data field took up exactly as many bits as it needed, and fields
were strung together without regard for byte boundaries. For instance,
a field with 16 possible values took 4 bits; if the next field had, say,
3 possible values it took the next 2 bits, etc. I almost wore out the
shift instructions on the machine with which I processed it.
Each data field took up exactly as many bits as it needed, and fields
were strung together without regard for byte boundaries. For instance,
a field with 16 possible values took 4 bits; if the next field had, say,
3 possible values it took the next 2 bits, etc. I almost wore out the
shift instructions on the machine with which I processed it.
Dan Espen
Mar 28, 2012, 12:31:54 PM3/28/12
to
"Charlie Gibbs" <cgi...@kltpzyxm.invalid> writes:
> (Why are people so infatuated with using bright blue lights?
Some studies have shown that blue is the most common favorite color.
> (Why are people so infatuated with using bright blue lights?
http://en.wikipedia.org/wiki/Color_psychology
The hues in the blue range have been shown to have the highest
preference amongst people.
There are some hazards unique to blue light:
http://www.mdsupport.org/library/hazard.html#blue
I know I find the blue indicators most visually appealing
but I would expect a good blinken-lights display to be
multi-colored.
--
Dan Espen
Walter Bushell
Mar 28, 2012, 12:32:18 PM3/28/12
to
In article <ylfkhax8...@dd-b.net>,
Anyway the byte pointers did everything you could want (except for
allowing bytes to overlap words, which would have been nice, except
probably not worth paying for).
--
This space unintentionally left blank.
allowing bytes to overlap words, which would have been nice, except
probably not worth paying for).
--
This space unintentionally left blank.
Ahem A Rivet's Shot
Mar 28, 2012, 12:50:35 PM3/28/12
to
On Wed, 28 Mar 2012 10:51:22 -0500
ken...@cix.compulink.co.uk wrote:
> By the way when was bit slicing depreciated?
The TI-990 was bit slice mini (1973), the TMS9900 (1976) was a
ken...@cix.compulink.co.uk wrote:
> By the way when was bit slicing depreciated?
single chip CPU largely compatible with it, so I'd put the end of bitslice
somewhere around the mid to late 70s.
Patrick Scheible
Mar 28, 2012, 1:14:16 PM3/28/12
to
Dan Espen <des...@verizon.net> writes:
> Rich Alderson <ne...@alderson.users.panix.com> writes:
>
>> Dan Espen <des...@verizon.net> writes:
>>
>>> If you could set it to 1-36, then I'd guess byte size was in
>>> "-1" notation. Ie. 0 is 1. Therefore, there would be no way
>>> to set byte size 0.
>>
>> Your guess would be wrong.
>>
>> The format of a PDP-10 section-local[1] byte pointer is
>>
>> | P | S | |I X Y |
>> +______+______+_+_+____+________________________+
>> |pppppp|ssssss|0|i|xxxx|yyyyyyyyyyyyyyyyyyyyyyyy|
>> +------+------+-+-+----+------------------------+
>> 0 5 6 1 1 1 1 1 1 3
>> 1 2 3 4 7 8 5
>>
>> The architecture requires that the size field S be non-zero. The position
>> field P is the location of the rightmost bit of the byte within the word,
>> represented as the number of bits from the right end of the word; a value of 44
>> (36) is special, being 1 bit to the left of the high-order bit 0, allowing the
>> incrementing byte instructions to begin at the leftmost point in the first word
>> of a string without any schratzing around. If an increment would move beyond
>> of the current word, the next byte is taken from the leftmost byte position in
>> the following word.[2]
>
> Well, I was smart enough to say I was guessing.
>
> Rich Alderson <ne...@alderson.users.panix.com> writes:
>
>> Dan Espen <des...@verizon.net> writes:
>>
>>> If you could set it to 1-36, then I'd guess byte size was in
>>> "-1" notation. Ie. 0 is 1. Therefore, there would be no way
>>> to set byte size 0.
>>
>> Your guess would be wrong.
>>
>> The format of a PDP-10 section-local[1] byte pointer is
>>
>> | P | S | |I X Y |
>> +______+______+_+_+____+________________________+
>> |pppppp|ssssss|0|i|xxxx|yyyyyyyyyyyyyyyyyyyyyyyy|
>> +------+------+-+-+----+------------------------+
>> 0 5 6 1 1 1 1 1 1 3
>> 1 2 3 4 7 8 5
>>
>> The architecture requires that the size field S be non-zero. The position
>> field P is the location of the rightmost bit of the byte within the word,
>> represented as the number of bits from the right end of the word; a value of 44
>> (36) is special, being 1 bit to the left of the high-order bit 0, allowing the
>> incrementing byte instructions to begin at the leftmost point in the first word
>> of a string without any schratzing around. If an increment would move beyond
>> of the current word, the next byte is taken from the leftmost byte position in
>> the following word.[2]
>
> Well, I was smart enough to say I was guessing.
>
> Not that I would have expected something as complicated as what you
> describe. I don't pretend to understand everything you've explained
> but 6 bits for an offset (P), 6 bits for a size (S) with zero excluded
> and a special value?
Yes. The indirect bit (I), index register (X), and word (Y) were used
> describe. I don't pretend to understand everything you've explained
> but 6 bits for an offset (P), 6 bits for a size (S) with zero excluded
> and a special value?
the same way in every instruction as well as the byte pointers, so
although they look complicated at first the assembly language programmer
got proficient with them quickly. They are combined to calculate one
effective address:
if I=0 and X=0, effective address is Y
if I=0 and X!=0, effective address = Y + (X)
if I=1, find the word in memory addressed by Y + (X) and perform
effective address calculation on that word's I, X, and Y fields.
Address calculation continued until a word is found with I=0.
Note that an 18-bit word address like Y would refer to a register,
if it was in the range 0-15.
This is way simpler than the eleventy-one addressing modes of the Vax or
the Vax's contemporaries, but it's easy to do displacement, indexed,
jump tables, or whatever other data structures you need.
The byte pointers were simpler to use than to describe. In normal
cases, you set up the byte pointer once with an assembler macro and the
load and deposit byte instructions did the right thing with it.
> The 1401 presented a simple architecture. Most people could completely
> understand everything about the machine.
It was famous for the instruction set diagram that fit on one
letter-sized piece of paper. See page A-6 and A-7 in the processor
reference manual at
http://bitsavers.org/pdf/dec/pdp10/TOPS10_softwareNotebooks/vol05/AA-H391A-TK_DECsystem-10_DECSYSTEM-20_Processor_Reference_Jun1982.pdf
Extended addressing got more complicated, but hardly anyone needed
extended addressing unless they were programming the monitor.
> S/360 made complete understanding something difficult. Mr. Wheeler
> conquered it, and I probably could have but I'd guess less than 1% of
> programmers really understand most of S/360.
> You've presented only a glimpse of PDP I'm sure but that looks like a
> big mess with not much to recommend it.
(except for extended addressing). There's nothing superfluous to take
out of the byte pointers: they find a word the same way every
instruction finds a word, they use a position within that word and a
size. Operating on different size bytes is a huge win; ASCII characters
were normally represented in 7 bits; identifiers in assembly language,
file names, command names in the exec were in 6 bits; communicating with
PDP-11 land often used 8 bits. And the same instructions could be used
for packing and unpacking non-character data as well. The Vax would
probably come up with 27 separate instructions to do the same things,
only not as generally.
-- Patrick
ken...@cix.compulink.co.uk
Mar 28, 2012, 1:54:18 PM3/28/12
to
In article <9tgotb...@mid.individual.net>, pla...@dsl.pipex.com (Stan
accumulator (A) was the only register that did not apply to. Still every
8 bit micro had sat least one 16 bit register (program counter) to allow
a 64K address space.
Ken Young
Barr) wrote:
> The Z80 had both 8 and 16 bit registers.
The 8080 and the Z80 had 8 bit registers that could be paired. The
> The Z80 had both 8 and 16 bit registers.
accumulator (A) was the only register that did not apply to. Still every
8 bit micro had sat least one 16 bit register (program counter) to allow
a 64K address space.
Ken Young
David Dyer-Bennet
Mar 28, 2012, 3:35:06 PM3/28/12
to
expelled self-modifying code from my mind that the idea is
inconceivable).
> There's another trick using move with word marks instead of
> just move that can make tables really small but I'd have to
> look at the bottom of the pile of my old listings to describe it.
and the move faster, but I don't see what I'm thinking of making the
*table* smaller.
(Don't have any old code around. I did 1401 code for about two years,
1969 and 1970, roughly, haven't had access to one since. Didn't own the
documentation (not that that helped; I passed on my 1620 documentation
when I moved away, people could still use it there, but I sort of miss
not having it, although all the info is online basically).
Quadibloc
Mar 28, 2012, 3:35:24 PM3/28/12
to
On Mar 26, 11:20 am, ken...@cix.compulink.co.uk wrote:
> When was word length standardised? I remember the computer books I have
> dating from the early era expected people to be familiar with octal,
> binary and hex notation. I have come across references to early
> computers using something like a 40 bit word.
The IBM System/360 was announced on April 7, 1964.
That computer was the one which was influential in causing computers
using the 6-bit character - which was then the most common
representation of textual data, leading to computers with word lengths
of 12, 18, 24, 30, 36, 48 and 60 bits being present at the time (30
was rare, used in some Univac military computers, and so was 60,
mainly found in the Control Data 6600) - to begin to be viewed as old-
fashioned.
Thus, for example, before the System/360, DEC had the 12-bit PDP-5/8,
the 18-bit PDP-4/7/15 and the 36-bit PDP-6/10 architectures. After the
System/360, it introduced the 16-bit PDP-11, and concentrated its
marketing efforts on that architecture.
The early advertising literature for the MOS Technology 6502
microprocessor and, I think, also the Intel 8080, noted that these
chips were influenced by the PDP-11, as it was a much-praised
architecture. The Texas Instruments 9900 is _very_ similar to a
PDP-11, except that it's big-endian instead of little-endian.
John Savard
> When was word length standardised? I remember the computer books I have
> dating from the early era expected people to be familiar with octal,
> binary and hex notation. I have come across references to early
> computers using something like a 40 bit word.
The IBM System/360 was announced on April 7, 1964.
That computer was the one which was influential in causing computers
using the 6-bit character - which was then the most common
representation of textual data, leading to computers with word lengths
of 12, 18, 24, 30, 36, 48 and 60 bits being present at the time (30
was rare, used in some Univac military computers, and so was 60,
mainly found in the Control Data 6600) - to begin to be viewed as old-
fashioned.
Thus, for example, before the System/360, DEC had the 12-bit PDP-5/8,
the 18-bit PDP-4/7/15 and the 36-bit PDP-6/10 architectures. After the
System/360, it introduced the 16-bit PDP-11, and concentrated its
marketing efforts on that architecture.
The early advertising literature for the MOS Technology 6502
microprocessor and, I think, also the Intel 8080, noted that these
chips were influenced by the PDP-11, as it was a much-praised
architecture. The Texas Instruments 9900 is _very_ similar to a
PDP-11, except that it's big-endian instead of little-endian.
John Savard
Quadibloc
Mar 28, 2012, 3:38:48 PM3/28/12
to
On Mar 28, 6:16 am, Peter Flass <Peter_Fl...@Yahoo.com> wrote:
> I believe the CDC 6400, 660, etc. series, with a 60-bit word, also
> stored multiple instructions in a word.
Yes. Instructions could be 15 or 30 bits long, but a 30-bit
> I believe the CDC 6400, 660, etc. series, with a 60-bit word, also
> stored multiple instructions in a word.
instruction could not begin in the fourth 15 bits of a 60-bit word,
because they weren't allowed to cross word boundaries.
John Savard
David Dyer-Bennet
Mar 28, 2012, 3:39:08 PM3/28/12
to
Walter Bushell <pr...@panix.com> writes:
> Anyway the byte pointers did everything you could want (except for
> allowing bytes to overlap words, which would have been nice, except
> probably not worth paying for).
Could have been nice sometimes. You'd want a choice, when incrementing
> Anyway the byte pointers did everything you could want (except for
> allowing bytes to overlap words, which would have been nice, except
> probably not worth paying for).
a byte pointer, which way it worked. There's that one unused bit in the
local bp format, that could say whether bytes filled words or not.
Would have been expensive at execution time I think, and probably
expensive in KA hardware at least.
Quadibloc
Mar 28, 2012, 3:45:47 PM3/28/12
to
On Mar 27, 7:07 am, jmfbahciv <See.ab...@aol.com> wrote:
> I could set the byte size from 1-36. I don't remember what happens
> if the byte size was 0. None of the PDP platforms insisted on
> one flavor of byte nor did they insist on byte-only accessing.
Well, the PDP-10 happened to have a feature that allowed operations on
variable-sized characters. Because it had a 36-bit word, characters
that were 6 bits long (six character per word), 7 bits long (one bit
wasted, five characters per word) or 9 bits long (four characters per
word) were the most reasonable choices.
Other DEC computers did not have this feature. The PDP-11 was built
around the 8-bit byte, and all its addresses were byte addresses, even
if most operations required alignment on at least a 16-bit word. Other
DEC computers only gave you word addressing; the 12-bit word of a
PDP-8 or the 18-bit word of a PDP-4, 7, or 15. Routines that used
shift instructions and AND and OR were used to pack and unpack
characters (usually 6 bits long) into a word.
Of course, in order to conserve disk space, the PDP-11 used special
routines to pack extra characters in less space for its filesystem.
But there, there was no character size in *bits*... that was, instead,
good old Radix-40.
John Savard
> I could set the byte size from 1-36. I don't remember what happens
> if the byte size was 0. None of the PDP platforms insisted on
> one flavor of byte nor did they insist on byte-only accessing.
Well, the PDP-10 happened to have a feature that allowed operations on
variable-sized characters. Because it had a 36-bit word, characters
that were 6 bits long (six character per word), 7 bits long (one bit
wasted, five characters per word) or 9 bits long (four characters per
word) were the most reasonable choices.
Other DEC computers did not have this feature. The PDP-11 was built
around the 8-bit byte, and all its addresses were byte addresses, even
if most operations required alignment on at least a 16-bit word. Other
DEC computers only gave you word addressing; the 12-bit word of a
PDP-8 or the 18-bit word of a PDP-4, 7, or 15. Routines that used
shift instructions and AND and OR were used to pack and unpack
characters (usually 6 bits long) into a word.
Of course, in order to conserve disk space, the PDP-11 used special
routines to pack extra characters in less space for its filesystem.
But there, there was no character size in *bits*... that was, instead,
good old Radix-40.
John Savard
Quadibloc
Mar 28, 2012, 3:55:44 PM3/28/12
to
On Mar 27, 12:05 pm, Walter Banks <wal...@bytecraft.com> wrote:
> Add to that
A company called ASI made some general-purpose computers with rather
unusual word lengths.
The ASI 210 had a word length of 21 bits, as did the ASI 2100.
The ASI 420 had a word length of 42 bits.
On the other hand, the ASI 6020 had a nice conventional 48-bit word
length.
John Savard
> Add to that
A company called ASI made some general-purpose computers with rather
unusual word lengths.
The ASI 210 had a word length of 21 bits, as did the ASI 2100.
The ASI 420 had a word length of 42 bits.
On the other hand, the ASI 6020 had a nice conventional 48-bit word
length.
John Savard
Rod Speed
Mar 28, 2012, 4:17:10 PM3/28/12
to
Dan Espen wrote
> Charlie Gibbs <cgi...@kltpzyxm.invalid> wrote
>> (Why are people so infatuated with using bright blue lights?
> Some studies have shown that blue is the most common favorite color.
I've often wondered why blue jeans ended up blue.
Hordes of our 'workers' like truck drivers used to wear
dark blue singlets and nothing else on the top too.
Doesnt explain why the case lights are so damned bright now tho.
> http://en.wikipedia.org/wiki/Color_psychology
> The hues in the blue range have been shown
> to have the highest preference amongst people.
> There are some hazards unique to blue light:
> http://www.mdsupport.org/library/hazard.html#blue
> I know I find the blue indicators most visually appealing but I
> would expect a good blinken-lights display to be multi-colored.
Yeah, my latest case stupidly has blue for both the power led and the HDD led.
None of the others are that stupid.
> Charlie Gibbs <cgi...@kltpzyxm.invalid> wrote
>> (Why are people so infatuated with using bright blue lights?
> Some studies have shown that blue is the most common favorite color.
Hordes of our 'workers' like truck drivers used to wear
dark blue singlets and nothing else on the top too.
Doesnt explain why the case lights are so damned bright now tho.
> http://en.wikipedia.org/wiki/Color_psychology
> The hues in the blue range have been shown
> to have the highest preference amongst people.
> There are some hazards unique to blue light:
> http://www.mdsupport.org/library/hazard.html#blue
> I know I find the blue indicators most visually appealing but I
> would expect a good blinken-lights display to be multi-colored.
None of the others are that stupid.
Dan Espen
Mar 28, 2012, 5:37:18 PM3/28/12
to
David Dyer-Bennet <dd...@dd-b.net> writes:
> Dan Espen <des...@verizon.net> writes:
>
>> A table with a key and value simply needs a chained move.
>> The A register is always left pointing to the next lower
>> entry.
>
> Yes, I understood the idea immediately (I have not, sadly, so completely
> expelled self-modifying code from my mind that the idea is
> inconceivable).
Cool!
> Dan Espen <des...@verizon.net> writes:
>
>> A table with a key and value simply needs a chained move.
>> The A register is always left pointing to the next lower
>> entry.
>
> Yes, I understood the idea immediately (I have not, sadly, so completely
> expelled self-modifying code from my mind that the idea is
> inconceivable).
>> There's another trick using move with word marks instead of
>> just move that can make tables really small but I'd have to
>> look at the bottom of the pile of my old listings to describe it.
>
> I think I sometimes used something like that to make the code smaller
> and the move faster, but I don't see what I'm thinking of making the
> *table* smaller.
only contain changes from the previous entry.
As I remember, I had an 8K table that boiled down to 3K.
The statistician that gave me the specs mentioned that he didn't think
the computer could handle the computations he needed.
> (Don't have any old code around. I did 1401 code for about two years,
> 1969 and 1970, roughly, haven't had access to one since. Didn't own the
> documentation (not that that helped; I passed on my 1620 documentation
> when I moved away, people could still use it there, but I sort of miss
> not having it, although all the info is online basically).
Started in 1964, worked 2 years on real 14xx, then for many years on S/360
running 14xx emulation. Maybe to 1975.
--
Dan Espen
Jorgen Grahn
Mar 28, 2012, 6:16:02 PM3/28/12
to
On Wed, 2012-03-28, Charlie Gibbs wrote:
> In article <ick425k...@home.home>, des...@verizon.net (Dan Espen)
> writes:
>
>> David Dyer-Bennet <dd...@dd-b.net> writes:
...
> In article <ick425k...@home.home>, des...@verizon.net (Dan Espen)
> writes:
>
>> David Dyer-Bennet <dd...@dd-b.net> writes:
>>> I really miss front-panel lights. Though at todays's speeds they
>>> wouldn't show much any more I suppose.
>>
>> Agree. All I have now is a disk access indicator light.
>> Why this isn't viewed as a marketing opportunity, I don't know.
See above -- what would they show? (Not that I know what blinkenlights
>>> wouldn't show much any more I suppose.
>>
>> Agree. All I have now is a disk access indicator light.
>> Why this isn't viewed as a marketing opportunity, I don't know.
used to show -- my beard is only partly gray.)
BTW, the disk activity LED is less useful than the /sound/ of the
disk I/O.
> It seems to be seen as an opportunity on everything _except_
> computers. Just about every gadget has far too many blinkenlights -
> most of them that irritating high-intensity blue - the purpose of
> which is to distract you from what's actually going on.
light. If I've been to the men's room, I want to see if anyone called
me without touching the damn phone.
> (Why are people so infatuated with using bright blue lights?
> Why does a dog lick his balls? Because he can.)
green twenty years ago, and some of the people in power now were young
then, and still think of them as expensive, thus classy?
/Jorgen
--
// Jorgen Grahn <grahn@ Oo o. . .
\X/ snipabacken.se> O o .
Message has been deleted
David Dyer-Bennet
Mar 28, 2012, 6:25:10 PM3/28/12
to
Dan Espen <des...@verizon.net> writes:
> David Dyer-Bennet <dd...@dd-b.net> writes:
>
>> Dan Espen <des...@verizon.net> writes:
>>> There's another trick using move with word marks instead of
>>> just move that can make tables really small but I'd have to
>>> look at the bottom of the pile of my old listings to describe it.
>>
>> I think I sometimes used something like that to make the code smaller
>> and the move faster, but I don't see what I'm thinking of making the
>> *table* smaller.
>
> You use word marks so that subsequent table entries
> only contain changes from the previous entry.
>
> As I remember, I had an 8K table that boiled down to 3K.
> The statistician that gave me the specs mentioned that he didn't think
> the computer could handle the computations he needed.
Useful, if you happen for example to have an 8k 1401!
> David Dyer-Bennet <dd...@dd-b.net> writes:
>
>> Dan Espen <des...@verizon.net> writes:
>>> There's another trick using move with word marks instead of
>>> just move that can make tables really small but I'd have to
>>> look at the bottom of the pile of my old listings to describe it.
>>
>> I think I sometimes used something like that to make the code smaller
>> and the move faster, but I don't see what I'm thinking of making the
>> *table* smaller.
>
> You use word marks so that subsequent table entries
> only contain changes from the previous entry.
>
> As I remember, I had an 8K table that boiled down to 3K.
> The statistician that gave me the specs mentioned that he didn't think
> the computer could handle the computations he needed.
My first paid programming project was to count up this five-dimensional
table relating SAT score, class rank, sex, and I forget what else (I
guess two other things) for the admission candidates at a college.
And then print it out on the line printer. (Sets of pages, each of
which contained a 2D grid, each cell of which contained two numbers, as
I recall).
I was a bit worried about testing this (since I was really quite willing
to believe I could screw up manual subscript calculations on a 5D table
in assembly language). I wrote a simpler, specialized program that
counted up the right number for *one* cell, simple enough to believe it
was true by inspection and simple testing, and used that to verify a
bunch of the results (not every single one). (No, not randomly chosen.)
I guess all that would have been right about the time I turned 15.
>> (Don't have any old code around. I did 1401 code for about two years,
>> 1969 and 1970, roughly, haven't had access to one since. Didn't own the
>> documentation (not that that helped; I passed on my 1620 documentation
>> when I moved away, people could still use it there, but I sort of miss
>> not having it, although all the info is online basically).
>
> I have some listings I haven't looked at in over 40 years.
>
> Started in 1964, worked 2 years on real 14xx, then for many years on S/360
> running 14xx emulation. Maybe to 1975.
but I was working on new software for the -11, rather than doing
anything with the old 1401 software.
Bill Leary
Mar 28, 2012, 6:35:43 PM3/28/12
to
"Peter Flass" wrote in message news:jkuuqe$mkn$1...@dont-email.me...
> operands had to be on word (32-bit) boundaries. It wasn't until the 370
> that this restriction was relaxed.
>
> I haven't commented on the PDP-11 because I'm not an authority, but didn't
> it have byte addressing but similar boundary restrictions?
My experience with the PDP-11 was mostly porting code to Nova. I studied
the instruction set, but I'm not sure I ever actually used one. I don't
recall if 16-bit accesses had to be word aligned or not. I'm sure someone
who knows will chime in.
- Bill
> On 3/28/2012 2:37 AM, Bill Leary wrote:
>> ((...portions omitted..))
>> "David Dyer-Bennet" wrote in message news:ylfkzkb1...@dd-b.net...
>> ((...portions omitted..))
>> "David Dyer-Bennet" wrote in message news:ylfkzkb1...@dd-b.net...
>>> It used to mean the basic unit of memory access; but clearly that's
>>> the byte in nearly every case now.
>>
>> I'd have to disagree with that. The memory being byte addressable
>> doesn't mean it's word size is a byte.
>
> Truth. This slipped by me, but the 360 was byte-addressed, but word
>>> the byte in nearly every case now.
>>
>> I'd have to disagree with that. The memory being byte addressable
>> doesn't mean it's word size is a byte.
>
> operands had to be on word (32-bit) boundaries. It wasn't until the 370
> that this restriction was relaxed.
>
> I haven't commented on the PDP-11 because I'm not an authority, but didn't
> it have byte addressing but similar boundary restrictions?
My experience with the PDP-11 was mostly porting code to Nova. I studied
the instruction set, but I'm not sure I ever actually used one. I don't
recall if 16-bit accesses had to be word aligned or not. I'm sure someone
who knows will chime in.
- Bill
John Levine
Mar 28, 2012, 6:58:54 PM3/28/12
to
>I haven't commented on the PDP-11 because I'm not an authority, but
>didn't it have byte addressing but similar boundary restrictions?
Yes. Words had to be on an even boundary. For the models that had
>didn't it have byte addressing but similar boundary restrictions?
floating point, floats had to be on a word boundary but not a double
or quad word boundary.
R's,
John
--
Regards,
John Levine, jo...@iecc.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. http://jl.ly
Dan Espen
Mar 28, 2012, 7:28:04 PM3/28/12
to
Jorgen Grahn <grahn...@snipabacken.se> writes:
> On Wed, 2012-03-28, Charlie Gibbs wrote:
>> In article <ick425k...@home.home>, des...@verizon.net (Dan Espen)
>> writes:
>>
>>> David Dyer-Bennet <dd...@dd-b.net> writes:
> ...
>>>> I really miss front-panel lights. Though at todays's speeds they
>>>> wouldn't show much any more I suppose.
>>>
>>> Agree. All I have now is a disk access indicator light.
>>> Why this isn't viewed as a marketing opportunity, I don't know.
>
> See above -- what would they show? (Not that I know what blinkenlights
> used to show -- my beard is only partly gray.)
Often the lights were wired to the bits.
> On Wed, 2012-03-28, Charlie Gibbs wrote:
>> In article <ick425k...@home.home>, des...@verizon.net (Dan Espen)
>> writes:
>>
>>> David Dyer-Bennet <dd...@dd-b.net> writes:
> ...
>>>> I really miss front-panel lights. Though at todays's speeds they
>>>> wouldn't show much any more I suppose.
>>>
>>> Agree. All I have now is a disk access indicator light.
>>> Why this isn't viewed as a marketing opportunity, I don't know.
>
> See above -- what would they show? (Not that I know what blinkenlights
> used to show -- my beard is only partly gray.)
One would show the bits in the PSW.
> BTW, the disk activity LED is less useful than the /sound/ of the
> disk I/O.
Have the LEDs show disk arm position. You could have 4 for each disk
showing which quarter of the disk is being accessed.
Other lights for packets, temperature, fan speed, total power draw.
It really doesn't matter though as long as they blink impressively.
--
Dan Espen
Rod Speed
Mar 28, 2012, 8:42:01 PM3/28/12
to
Jorgen Grahn wrote
> Charlie Gibbs wrote
>> des...@verizon.net (Dan Espen) wrote
>>> David Dyer-Bennet <dd...@dd-b.net> wrote
> (Not that I know what blinkenlights used to show
Selectively, the various registers.
> -- my beard is only partly gray.)
> BTW, the disk activity LED is less useful than the /sound/ of the disk I/O.
Not with modern hard drives, you cant hear any of mine.
>> It seems to be seen as an opportunity on everything _except_
>> computers. Just about every gadget has far too many blinkenlights -
>> most of them that irritating high-intensity blue - the purpose of
>> which is to distract you from what's actually going on.
> Except for cell phones, which just recently reacquired a "missed call" light.
Not just cellphones. Landline phones have had them for decades now.
> If I've been to the men's room, I want to see if
> anyone called me without touching the damn phone.
>> (Why are people so infatuated with using bright blue lights?
>> Why does a dog lick his balls? Because he can.)
> Perhaps because blue LEDs were much more expensive than red/amber/
> green twenty years ago, and some of the people in power now were young
> then, and still think of them as expensive, thus classy?
Doubt it. Bet its just because they now dont cost anymore and they prefer it.
> Charlie Gibbs wrote
>> des...@verizon.net (Dan Espen) wrote
>>> David Dyer-Bennet <dd...@dd-b.net> wrote
> ...
>>>> I really miss front-panel lights. Though at todays's
>>>> speeds they wouldn't show much any more I suppose.
>>> Agree. All I have now is a disk access indicator light.
>>> Why this isn't viewed as a marketing opportunity, I don't know.
> See above -- what would they show?
Indeed.
>>>> I really miss front-panel lights. Though at todays's
>>>> speeds they wouldn't show much any more I suppose.
>>> Agree. All I have now is a disk access indicator light.
>>> Why this isn't viewed as a marketing opportunity, I don't know.
> See above -- what would they show?
> (Not that I know what blinkenlights used to show
> -- my beard is only partly gray.)
> BTW, the disk activity LED is less useful than the /sound/ of the disk I/O.
>> It seems to be seen as an opportunity on everything _except_
>> computers. Just about every gadget has far too many blinkenlights -
>> most of them that irritating high-intensity blue - the purpose of
>> which is to distract you from what's actually going on.
> Except for cell phones, which just recently reacquired a "missed call" light.
> If I've been to the men's room, I want to see if
> anyone called me without touching the damn phone.
>> (Why are people so infatuated with using bright blue lights?
>> Why does a dog lick his balls? Because he can.)
> Perhaps because blue LEDs were much more expensive than red/amber/
> green twenty years ago, and some of the people in power now were young
> then, and still think of them as expensive, thus classy?
JW
Mar 29, 2012, 5:22:10 AM3/29/12
to
On Wed, 28 Mar 2012 17:50:35 +0100 Ahem A Rivet's Shot <ste...@eircom.net>
drives in the mid 80's that used AMD bit slice processors, although for
the life of me I can't remember the exact part number. It may have been
the 2901.
wrote in Message id: <20120328175035....@eircom.net>:
>On Wed, 28 Mar 2012 10:51:22 -0500
>ken...@cix.compulink.co.uk wrote:
>
>> By the way when was bit slicing depreciated?
>
> The TI-990 was bit slice mini (1973), the TMS9900 (1976) was a
>single chip CPU largely compatible with it, so I'd put the end of bitslice
>somewhere around the mid to late 70s.
There a word I haven't heard in ages. I remember repairing Digi-Data tape
>On Wed, 28 Mar 2012 10:51:22 -0500
>ken...@cix.compulink.co.uk wrote:
>
>> By the way when was bit slicing depreciated?
>
> The TI-990 was bit slice mini (1973), the TMS9900 (1976) was a
>single chip CPU largely compatible with it, so I'd put the end of bitslice
>somewhere around the mid to late 70s.
drives in the mid 80's that used AMD bit slice processors, although for
the life of me I can't remember the exact part number. It may have been
the 2901.
It is loading more messages.
0 new messages