The Future of CPUs: What's After Multi-Core?

Anne & Lynn Wheeler

unread,

Oct 29, 2006, 10:51:58 AM10/29/06

to

The Future of CPUs: What's After Multi-Core?
http://www.informit.com/articles/article.asp?p=663085&rl=1

from above (the new, 40yr old thing)

This rule was driven home to me when I attended a talk by an IBM
engineer about his company's new virtualization technology. He
commented that his company had an advantage over other people working
in the area: Whenever they were stuck, they could go along the hall to
the mainframe division and ask how they solved the same problem a
couple of decades ago.

... snip ...

semi-related thread from comp.arch
http://www.garlic.com/~lynn/2006t.html#23 threads versus task

jsa...@ecn.ab.ca

unread,

Oct 29, 2006, 2:35:15 PM10/29/06

to

Anne & Lynn Wheeler quoted, in part:

> This rule was driven home to me when I attended a talk by an IBM
> engineer about his company's new virtualization technology. He
> commented that his company had an advantage over other people working
> in the area: Whenever they were stuck, they could go along the hall to
> the mainframe division and ask how they solved the same problem a
> couple of decades ago.

LOL!

It certainly *is* true that today's micros have a lot in common with
yesterday's mainframes.

It seems like only yesterday people were excited about the new 16-bit
microcomputer chips, that would give microcomputers the power of minis.

And it was but a briefer span ago that the first Pentiums came out,
with cache *and* pipelines, putting the power of a top-of-the-line
mainframe such as an IBM 360/195 on one's desk.

(Yes, the original Pentium only had SRT division instead of
Goldschmidt, and instead of one 32 K cache, it had two 8 K caches...
but it came close.)

Today's micros don't *quite* yet have full-bore pipeline vector
instructions like a Cray-1; while the Itanium has population count, it
lacks bit matrix multiply (it has halfword binary matrix multiply, and
some fixed byte permutations, but no bit shuffling)... but we're
getting to the point where the IBM engineer could no longer go down the
hall, but would instead have to leave the building.

Multi-core or multi-chip is what comes after everything else - when you
can't make a CPU any faster or more powerful, and so the only economies
of scale left are those that come from having just one box and power
supply.

But while experience at Cray might prove to have *some* value for IBM
engineers working on the next generation of micros, IBM has plenty of
experience, with machines like Blue Gene, with massively parallel
arrays of machines to draw on as well.

IBM might well remember why the IBM PC was so successful in the first
place. Yes, the "IBM name" was a factor, but the product had its own
intrinsic merits. It was competing against Z-80 based computers; *they*
could be expanded to 64 K (or even 128 K with certain kludges) while
the IBM PC could be expanded to 640 K.

Getting a factor of 10 jump on today's consumer PCs without pricing
oneself out of the market might seem challenging (and, of course, they
are so powerful already that one might ask "what's the point?", but
that doesn't seem to have stopped Microsoft and Intel)... but if anyone
could pull it off, IBM is one of the candidates.

John Savard

Anne & Lynn Wheeler

unread,

Oct 29, 2006, 2:50:32 PM10/29/06

to

jsa...@ecn.ab.ca writes:
> IBM might well remember why the IBM PC was so successful in the first
> place. Yes, the "IBM name" was a factor, but the product had its own
> intrinsic merits. It was competing against Z-80 based computers; *they*
> could be expanded to 64 K (or even 128 K with certain kludges) while
> the IBM PC could be expanded to 640 K.

original ref:
http://www.garlic.com/~lynn/2006t.html#27 The Future of CPUs: What's After Multi-Core?

my frequent theme has been that one of big PC market penetrations was
into commercial ... where a PC was about the same price as a 3270 ...
could provide both 1) 3270 terminal emulation and 2) some local
desktop computing in a single screen/keyboard footprint.
http://www.garlic.com/~lynn/subnetwork.html#emulation

once that enormous market penetration had been obtained ... it was
difficult for anything else to compete (something of a snowball
effect, large install base attracted a lot of application programmers,
a lot of applications attracted bigger market). clones and
plug-compatible was one of the few remaining approaches.

and for other drift related to cluster scaleup
http://www.garlic.com/~lynn/95.html#13

and ha/cmp work
http://www.garlic.com/~lynn/subtopic.html#hacmp

jmfb...@aol.com

unread,

Oct 30, 2006, 8:32:12 AM10/30/06

to

In article <m3ac3fn...@lhwlinux.garlic.com>,

Anne & Lynn Wheeler <ly...@garlic.com> wrote:
>
>The Future of CPUs: What's After Multi-Core?
>http://www.informit.com/articles/article.asp?p=663085&rl=1
>
>from above (the new, 40yr old thing)
>
>This rule was driven home to me when I attended a talk by an IBM
>engineer about his company's new virtualization technology. He
>commented that his company had an advantage over other people working
>in the area: Whenever they were stuck, they could go along the hall to
>the mainframe division and ask how they solved the same problem a
>couple of decades ago.
>

>.... snip ...

>
>semi-related thread from comp.arch
>http://www.garlic.com/~lynn/2006t.html#23 threads versus task

And what are they going to do when those who have the knowledge
are dead? Since it has been a Good Thing to stamp out all that
expertise over the last decade, all the biz has left is one
OS that is getting developed on the old principle. And that
OS' biz still has not experienced the growing pain of the
absence of one bit god.

I'm not sure that the biz is going to have time to reinvent the
wheels; the reports I get from developers is that they have
a time window of 3 months for complete delvelopment cycle.
I don't know how you would get an OS with a drastic rewrite
done, funded or ever shipped.

/BAH

Anne & Lynn Wheeler

unread,

Oct 30, 2006, 2:56:05 PM10/30/06

to

jsa...@ecn.ab.ca writes:
> But while experience at Cray might prove to have *some* value for IBM
> engineers working on the next generation of micros, IBM has plenty of
> experience, with machines like Blue Gene, with massively parallel
> arrays of machines to draw on as well.

re:
http://www.garlic.com/~lynn/2006t.html#27 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#31 The Future of CPUs: What's After Multi-Core?

and for other drift than the previous mention related to working on
cluster scaleup while we were doing ibm's HA/CMP product
http://www.garlic.com/~lynn/subtopic.html#hacmp

sort of part of the earlier hsdt project
http://www.garlic.com/~lynn/subnetwork.html#hsdt

... a recent hsdt post or two
http://www.garlic.com/~lynn/2006t.html#6 Ranking of non-IBM mainframe builders?
http://www.garlic.com/~lynn/2006t.html#11 Ranking of non-IBM mainframe builders?
http://www.garlic.com/~lynn/2006t.html#12 Ranking of non-IBM mainframe builders?

... the original ibm mainframe tcp/ip product had been implemented in
pascal/vs ... however because of various issues ... it would consume
nearly a whole 3090 processor getting 44kbytes/sec sustained thruput.

i did a rfc 1044 driver implementation (that was eventually shipped in
the product) ... which in some tuning at cray research got 1mbyte/sec
sustained between a cray and 4341-clone ... using only a modest amount
of the 4341-clone.
http://www.garlic.com/~lynn/subnetwork.html#1044

trivia item was we were to leave on flight from sfo to Minneapolis for
dong some testing ... and was 20 mins or so late getting off the
ground. part way thru the flight, i noticed a lot of whispering back
in the galley and wandered back to see what it was all
about. apparently something like five minutes after leaving the ground
the earthquake had hit.

before the previously mentioned incident that drastically
changed their direction
http://www.garlic.com/~lynn/95.html#13

kingston had been providing some assistance to chen ... as well as
working on their own machine of somewhat similar design. for other
recent drift along this line.
http://www.garlic.com/~lynn/2006q.html#9 Is no one reading the article?

Rostyslaw J. Lewyckyj

unread,

Oct 30, 2006, 3:53:10 PM10/30/06

to

jmfb...@aol.com wrote:

Has the bit god set them a task deadline of 3 months,
and you're worried they won't complete the rewrite
before GWB IIs decreed start of Armageddon?

Steve O'Hara-Smith

unread,

Oct 30, 2006, 2:09:17 AM10/30/06

to

On 29 Oct 2006 11:35:15 -0800
jsa...@ecn.ab.ca wrote:

> IBM might well remember why the IBM PC was so successful in the first
> place. Yes, the "IBM name" was a factor, but the product had its own
> intrinsic merits. It was competing against Z-80 based computers; *they*
> could be expanded to 64 K (or even 128 K with certain kludges) while
> the IBM PC could be expanded to 640 K.

Ahem. The Newbrain could (in principle at least) be expanded to 3MB
of RAM and 4MB of EPROM. The stack of expansion boxes would have been a
little on the tall side and the power consumption somewhat unfunny, but it
was designed to support that much address space.

--
C:>WIN | Directable Mirror Arrays
The computer obeys and wins. | A better way to focus the sun
You lose and Bill collects. | licences available see
| http://www.sohara.org/

jmfb...@aol.com

unread,

Oct 31, 2006, 8:10:48 AM10/31/06

to

In article <DDt1h.78443$zF5....@bignews1.bellsouth.net>,

Sigh! You deliberately misread what I write.

/BAH

jsa...@ecn.ab.ca

unread,

Oct 31, 2006, 8:42:53 AM10/31/06

to

Anne & Lynn Wheeler wrote:
> my frequent theme has been that one of big PC market penetrations was
> into commercial ... where a PC was about the same price as a 3270 ...
> could provide both 1) 3270 terminal emulation and 2) some local
> desktop computing in a single screen/keyboard footprint.

*That* reminds me.

When IBM came out with the 103-key keyboard - the one that *finally*
(admittedly, the PCjr keyboard was first) had both shift keys, the
Enter key, and the backspace key *all* in the right places... one
dissenting voice was heard.

John C. Dvorak asked in InfoWorld why we needed yet another keyboard
layout, and noted that most people were satisfied with the AT layout.
But one other objection he made would have seemed reasonable.

Why did IBM add two function keys, F11 and F12? Since existing
computers didn't have them, who would ever write software to use them?

Now, that _would_ have been a reasonable objection. If you didn't know
that IBM made a line of terminals with either 12 or 24 function keys -
for which terminal emulation packages were already available on the PC
platform, but which were forced into strange keyboard arrangements
(using alt-1 for PF 1, up to alt-= for PF12) by the lack of F11 and F12
on the PC keyboard.

And the terminal? The 3270.

John Savard

jsa...@ecn.ab.ca

unread,

Oct 31, 2006, 8:52:01 AM10/31/06

to

Steve O'Hara-Smith wrote:
> On 29 Oct 2006 11:35:15 -0800
> jsa...@ecn.ab.ca wrote:

> > IBM might well remember why the IBM PC was so successful in the first
> > place. Yes, the "IBM name" was a factor, but the product had its own
> > intrinsic merits. It was competing against Z-80 based computers; *they*
> > could be expanded to 64 K (or even 128 K with certain kludges) while
> > the IBM PC could be expanded to 640 K.

> Ahem. The Newbrain could (in principle at least) be expanded to 3MB
> of RAM and 4MB of EPROM. The stack of expansion boxes would have been a
> little on the tall side and the power consumption somewhat unfunny, but it
> was designed to support that much address space.

That may be true, but it was Z-80 CP/M boxes, and not the Newbrain,
that the IBM PC was competing against - initially, the IBM PC was
sufficiently pricey that it was ignored by the home market, that stuck
with computers like the C-64, and then later moved on to the Amiga 1000
and the Atari ST.

Of course the IBM PC wasn't the *first* computer to break the 64 K
barrier. There were some expensive 68000 systems, for example. IBM's
name, and the market positioning and price of the computer also were
important factors.

Given that the NewBrain was a *British* computer, and it had a keyboard
of the same type that almost killed the IBM PCjr, it was hardly a
threat. Oh: and, according to Wikipedia, it was released in 1983. The
IBM PC, of course, dates from 1981.

John Savard

Anne & Lynn Wheeler

unread,

Oct 31, 2006, 11:02:32 AM10/31/06

to

jsa...@ecn.ab.ca writes:
> *That* reminds me.
>
> When IBM came out with the 103-key keyboard - the one that *finally*
> (admittedly, the PCjr keyboard was first) had both shift keys, the
> Enter key, and the backspace key *all* in the right places... one
> dissenting voice was heard.
>
> John C. Dvorak asked in InfoWorld why we needed yet another keyboard
> layout, and noted that most people were satisfied with the AT layout.
> But one other objection he made would have seemed reasonable.
>
> Why did IBM add two function keys, F11 and F12? Since existing
> computers didn't have them, who would ever write software to use them?
>
> Now, that _would_ have been a reasonable objection. If you didn't know
> that IBM made a line of terminals with either 12 or 24 function keys -
> for which terminal emulation packages were already available on the PC
> platform, but which were forced into strange keyboard arrangements
> (using alt-1 for PF 1, up to alt-= for PF12) by the lack of F11 and F12
> on the PC keyboard.
>
> And the terminal? The 3270.

we had several arguments with the introduction of the 3274 controller
and 3278 terminal in the late 70s, ... one was that they initially
eliminated the 12 pfkey pad on the right.

3277 keyboard layout (from your web page)
http://www.quadibloc.com/comp/kyb01.htm

the 3277 had a lot of electronics in the head and keyboard of the
terminal (and not all back in the 3272 controller). to cut down on
manufacturing costs ... for the 3278 ... a lot of the electronics were
moved back into 3274 controller (as well as cutting down on
keys). this contributed to the increased response time ... and it also
made it impossible to do some of the local hardware fixes on the
terminal.

recent post that includes timing comparison of 3272/3277 against
3274/3278 (both channel attach):
http://www.garlic.com/~lynn/2006s.html#42 Ranking of non-IBM mainframe builders?

we also had a "FIFO" box that could be placed inline between the 3277
keyboard cable and where it plugged into 3277 display head ... that
handled the half-duplex tendency to sporadically lock the keyboard and
block keystrokes. also were able to add resister inside the 3277
keyboard that adjusted the typamatic delay and typamatic repeat rate .. to
some reasonable value ... past post discussing both timing comparison
as well as being to do local fixes to 3277 terminal:
http://www.garlic.com/~lynn/2005r.html#15 Intel strikes back with a parallel x86 design

the original 3278 keyboard took the program function key position for
numeric keypad ... and made program function keys alternates ...
past post mentioning this
http://www.garlic.com/~lynn/2005e.html#33 Stop Me If You've Heard This One Before

when we started arguing about with the 3278 product group about
program function keys (and slow 3274 controller and other stuff)
... they came back and said that the 3278 terminal wasn't designed for
programmers and interactive computer use ... but for data entry
applications.

you later got 3278 keyboard option with the pf1-pf12 across the top
and pf13-pf24 on the right.

another of your web pages
http://www.quadibloc.com/comp/scan.htm

Rostyslaw J. Lewyckyj

unread,

Oct 31, 2006, 1:08:51 PM10/31/06

to

jmfb...@aol.com wrote:

But Maam, There is no proper way. :)

Eugene Miya

unread,

Oct 31, 2006, 12:28:16 PM10/31/06

to

In article <m3ac3fn...@lhwlinux.garlic.com>,
Anne & Lynn Wheeler <ly...@garlic.com> wrote:

>The Future of CPUs: What's After Multi-Core?
>http://www.informit.com/articles/article.asp?p=663085&rl=1
>from above (the new, 40yr old thing)

So so article. More of the same thing which will run into quantum
effects.

What you have to do is hold these journalists making predictions to the
same standards that earthquakes are held to: 3-dimensions, time and a
magnitude. Sure some of that is ammenable to Moore's law which is just
a corollary to Newton.

>IBM

>Whenever they were stuck, they could go along the hall to
>the mainframe division and ask how they solved the same problem a
>couple of decades ago.

That works some time.
The Web didn't exist, non-hierarchical networks inside IBM didn't exist
back, then and a slew of other problems. Sure Unix and Linux were/are
behind some of the hierarchical mainframe ideas, but their didn't cut it
in France.

It's fine if you want compatible to backwards mistakes.

The future will be like Alan Kay notes: it's what you will make it.
It's the best way to predict. The problem is that we are coming out of
a discrete component way of thinking. Building Omega networks to
interconnect memories and processors one transistor at a time isn't the
way to achieve parallelism and even C and Java haven't been called
Fortran yet. I do expect new advances. It's not your father's old IBM
these days.

--

Stan Barr

unread,

Oct 31, 2006, 2:09:43 PM10/31/06

to

On 31 Oct 2006 05:52:01 -0800, jsa...@ecn.ab.ca <jsa...@ecn.ab.ca> wrote:
>
>That may be true, but it was Z-80 CP/M boxes, and not the Newbrain,
>that the IBM PC was competing against - initially, the IBM PC was
>sufficiently pricey that it was ignored by the home market, that stuck
>with computers like the C-64, and then later moved on to the Amiga 1000
>and the Atari ST.

We've just been discussing this subject on the Clasic Computer mailing
list. Initially the price of the IBM PC in the UK was a severe dis-
incentive to non-corporate purchasers. You could buy two Apple IIs
with twin floppys and monitors for the price of one floppy-based PC
with monitor.

>
>Of course the IBM PC wasn't the *first* computer to break the 64 K
>barrier. There were some expensive 68000 systems, for example.

IIRC the TRS-80 Model 4 and 4P (the portable(!) one) supported 128K.

--
Cheers,
Stan Barr stanb .at. dial .dot. pipex .dot. com
(Remove any digits from the addresses when mailing me.)

The future was never like this!

Anne & Lynn Wheeler

unread,

Oct 31, 2006, 2:48:47 PM10/31/06

to

eug...@cse.ucsc.edu (Eugene Miya) writes:
> That works some time.
> The Web didn't exist, non-hierarchical networks inside IBM didn't exist
> back, then and a slew of other problems. Sure Unix and Linux were/are
> behind some of the hierarchical mainframe ideas, but their didn't cut it
> in France.

SNA in the 70s ... somewhat in conjunction with fs
http://www.garlic.com/~lynn/subtopic.html#futuresys

and somewhat afterwards ... defined hierarchical communication
architecture for big terminal networks ... pretty much wrapped around
vtam/sscp and 3705/ncp (or actually vtam/sscp and 3705/ncp wrapped
around the sna architecture ... there were some number of people that
complained that it didn't matter what sna architecture specified
... if you were to interoperate with vtam/3705, it had to conform to
whatever vtam/3705 did ... and the two weren't always kept in total
sync).

in that same time-frame ... my wife worked on a competing peer-to-peer
architecture (AWP39) that lost out to SNA. she then did a stint in the
JES2 group and then was con'ed into taking a job in pok in charge
of loosely-coupled architecture (mainframe for cluster) ... where
she origianted peer-coupled shared data architecture
http://www.garlic.com/~lynn/subtopic.html#shareddata

and had quite a few battles with the sna organization ... sort of
resulting in a truce ... where non-SNA was allowed within the
datacenter walls, but SNA was required whenever the walls of the
datacenter was crossed.

this also gave us lots of headache in our high-speed data transport
project (HSDT) starting circa 1980s
http://www.garlic.com/~lynn/subnetwork.html#hsdt

and also contributed to the terminal emulation
http://www.garlic.com/~lynn/subnetwork.html#emulation

and 3-tier architecture (and SAA) wars
http://www.garlic.com/~lynn/subnetwork.html#3tier

also in approx. the same time that arpa was starting up ... the
internal network
http://www.garlic.com/~lynn/subnetwork.html#internalnet

was originating at the cambridge science center
http://www.garlic.com/~lynn/subtopic.html#545tech

which was non-hiearchical, non-sna, non-vtam, and frequently non-3705
(much more akin to the peer-to-peer network architecture, AWP39 that
my wife worked on in the mid-70s).

recent posts mentioning size of internal network passing 1000 nodes in
1983 (year that arpanet switched over to internetworking protection)
... also includes reference that there were approx. 250 hosts on the
arpanet that required major changes as part of the 1/1/83 upgrade to
internetworking
http://www.garlic.com/~lynn/2006k.html#3 Arpa address
http://www.garlic.com/~lynn/2006k.html#8 Arpa address
http://www.garlic.com/~lynn/2006k.html#9 Arpa address

note while the earlier reference indicate that there approximately 250
hosts on the arpanet that required major changes as part of the 1/1/83
upgrade to internetworking ... this ARPANET newsletter artile was
predicting that there would be 100 arpanet nodes by the 1/1/83
switchover to internetworking
http://www.garlic.com/~lynn/2006k.html#40 Arpa address

it is possible that the difference between 100 arpanet nodes and 250
hosts ... was that on arpanet, the actual networking was handled by
the outboard IMPs ... with hosts then running host-to-host protocol
and connecting to the IMPs for the actual networking support
(potentially allowing greater than one-to-one relationship between
hosts and nodes).

by comparison, the internal networking implementation ... all of the
networking support executed directly on each host (and had nothing to
do with vtam ... which was the incarnation of hierarchical sna). any
outboard telecommunication control unit ... purely provided the
physical line point-to-point support (i.e. things like line scanner
operation, translating between line signal rise/lower and bit or
non-bit).

reference to computer history museum item on the more than 300
nodes/hosts on internal network in the 70s being instrumental in the
development and evolution of rexx
http://www.garlic.com/~lynn/2006p.html#31 "25th Anniversary of the Personal Computer"
http://www.garlic.com/~lynn/2006r.html#7 Was FORTRAN buggy?

i.e.
http://www.computinghistorymuseum.org/ieee/af_forum/read.cfm?forum=10&id=21&thread=7

from above:

By far the most important influence on the development of Rexx was the
availability of the IBM electronic network, called VNET. In 1979,
more than three hundred of IBM's mainframe computers, mostly
running the Virtual Machine/370 (VM) operating system, were linked by
VNET. This store-and-forward network allowed very rapid exchange of
messages (chat) and e-mail, and reliable distribution of software. It
made it possible to design, develop, and distribute Rexx and its first
implementation from one country (the UK) even though most of its users
were five to eight time zones distant, in the USA.

... snip ...

other posts this year mentioning awp39 peer-to-peer networking architecture
effort:
http://www.garlic.com/~lynn/2006h.html#52 Need Help defining an AS400 with an IP address to the mainframe
http://www.garlic.com/~lynn/2006j.html#31 virtual memory
http://www.garlic.com/~lynn/2006k.html#21 Sending CONSOLE/SYSLOG To Off-Mainframe Server
http://www.garlic.com/~lynn/2006l.html#4 Google Architecture
http://www.garlic.com/~lynn/2006l.html#45 Mainframe Linux Mythbusting (Was: Using Java in batch on z/OS?)
http://www.garlic.com/~lynn/2006o.html#62 Greatest Software, System R
http://www.garlic.com/~lynn/2006r.html#4 Was FORTRAN buggy?
http://www.garlic.com/~lynn/2006r.html#9 Was FORTRAN buggy?

Eugene Miya

unread,

Oct 31, 2006, 5:01:43 PM10/31/06

to

In article <ei4uss$8qk...@s787.apx1.sbo.ma.dialup.rcn.com>,

<jmfb...@aol.com> wrote:
>And what are they going to do when those who have the knowledge
>are dead? Since it has been a Good Thing to stamp out all that
>expertise over the last decade, all the biz has left is one
>OS that is getting developed on the old principle. And that
>OS' biz still has not experienced the growing pain of the
>absence of one bit god.

The OS problem is known. This is a thread in c.o.r.
In general, which some of the written knowledge is fine, other aspects
of computer knowledge is poor and quite biased. Some of the stuff is
quite good dying. This is not the only field where loss of knowledge
might not be a bad thing. D. MacKenzie in Edinburgh is wondering about
this when it comes to nuclear testing and nuclear weapons.

>I'm not sure that the biz is going to have time to reinvent the
>wheels; the reports I get from developers is that they have
>a time window of 3 months for complete delvelopment cycle.
>I don't know how you would get an OS with a drastic rewrite
>done, funded or ever shipped.

You get code generators to spin up new versions of tools like file
systems, or process and thread structures (oh so slightly different).
Wheel reinvention isn't completely bad: we now have steel belted radials
and coming rubberless tires.

--

Eugene Miya

unread,

Oct 31, 2006, 5:11:00 PM10/31/06

to

In article <1162150515.4...@m7g2000cwm.googlegroups.com>,

<jsa...@ecn.ab.ca> wrote:
>It certainly *is* true that today's micros have a lot in common with
>yesterday's mainframes.

Well, they are all von Neumann architectures.

>It seems like only yesterday people were excited about the new 16-bit
>microcomputer chips, that would give microcomputers the power of minis.

That was over 2 decades ago. That was before Macs.

>And it was but a briefer span ago that the first Pentiums came out,
>with cache *and* pipelines, putting the power of a top-of-the-line
>mainframe such as an IBM 360/195 on one's desk.

The 195 was good? Have you used one?

>Today's micros don't *quite* yet have full-bore pipeline vector
>instructions like a Cray-1; while the Itanium has population count, it
>lacks bit matrix multiply (it has halfword binary matrix multiply, and
>some fixed byte permutations, but no bit shuffling)... but we're
>getting to the point where the IBM engineer could no longer go down the
>hall, but would instead have to leave the building.

Well they hire consultants like Ken Kennedy.

>Multi-core or multi-chip is what comes after everything else - when you
>can't make a CPU any faster or more powerful, and so the only economies
>of scale left are those that come from having just one box and power
>supply.
>
>But while experience at Cray might prove to have *some* value for IBM
>engineers working on the next generation of micros, IBM has plenty of
>experience, with machines like Blue Gene, with massively parallel
>arrays of machines to draw on as well.

And this is supposed to be good?
John have you ever programmed an MPP architecture?

>IBM might well remember why the IBM PC was so successful in the first
>place. Yes, the "IBM name" was a factor, but the product had its own
>intrinsic merits. It was competing against Z-80 based computers; *they*
>could be expanded to 64 K (or even 128 K with certain kludges) while
>the IBM PC could be expanded to 640 K.

Well, that was an amusing story: how a mainframe firm came to produce
something like the PC in a rather not traditional way: in no way
like its bureaucracy could fathom.

Its future was not assured. Just sample a PC Jr.

>Getting a factor of 10 jump on today's consumer PCs without pricing
>oneself out of the market might seem challenging (and, of course, they
>are so powerful already that one might ask "what's the point?", but
>that doesn't seem to have stopped Microsoft and Intel)... but if anyone
>could pull it off, IBM is one of the candidates.

Well you just more of the same.
Otherwise you get a guy ready to tell you that if IBM can't produce it,
you don't need it.

--

Walter Bushell

unread,

Oct 31, 2006, 11:29:20 PM10/31/06

to

In article <4547d804$1@darkstar>, eug...@cse.ucsc.edu (Eugene Miya)
wrote:

> The 195 was good? Have you used one?

I've worked in places where that was the most powerful computer, and we
had hundreds of programmers.

--
"The power of the Executive to cast a man into prison without formulating any
charge known to the law, and particularly to deny him the judgement of his
peers, is in the highest degree odious and is the foundation of all totali-
tarian government whether Nazi or Communist." -- W. Churchill, Nov 21, 1943

Steve O'Hara-Smith

unread,

Nov 1, 2006, 2:43:29 AM11/1/06

to

On 31 Oct 2006 05:52:01 -0800

jsa...@ecn.ab.ca wrote:

> Steve O'Hara-Smith wrote:
> > On 29 Oct 2006 11:35:15 -0800
> > jsa...@ecn.ab.ca wrote:
>
> > > IBM might well remember why the IBM PC was so successful in the first
> > > place. Yes, the "IBM name" was a factor, but the product had its own
> > > intrinsic merits. It was competing against Z-80 based computers;
> > > *they* could be expanded to 64 K (or even 128 K with certain kludges)
> > > while the IBM PC could be expanded to 640 K.
>
> > Ahem. The Newbrain could (in principle at least) be expanded to
> > 3MB of RAM and 4MB of EPROM. The stack of expansion boxes would have
> > been a little on the tall side and the power consumption somewhat
> > unfunny, but it was designed to support that much address space.
>
> That may be true, but it was Z-80 CP/M boxes, and not the Newbrain,

Yes I know - I just wanted to mention that there were Z80 machines
with rather more than 128K of memory support. There were also some MP/M
boxes that supported rather more than 128K and those *did* get used in
preference to the IBM in many places (not least because they had a
multiuser OS).

> Given that the NewBrain was a *British* computer, and it had a keyboard
> of the same type that almost killed the IBM PCjr, it was hardly a
> threat. Oh: and, according to Wikipedia, it was released in 1983. The
> IBM PC, of course, dates from 1981.

Yes the Newbrain was killed by being late - it should have shipped
around 1979/80 but Newbury Labs couldn't get working ULAs from Ferranti.
Grundy (who took it over) simply shipped the TTL based prototype, but by
then the BBC contract had been taken away and handed to Acorn and the
'brain was dead in the water.

Eugene Miya

unread,

Nov 1, 2006, 11:19:51 AM11/1/06

to

In article <4547d804$1@darkstar>, eug...@cse.ucsc.edu (Eugene Miya)
wrote:
>> The 195 was good? Have you used one?

I was asking John.

In article <proto-9C21E9....@reader2.panix.com>,

Walter Bushell <pr...@panix.com> wrote:
>I've worked in places where that was the most powerful computer, and we
>had hundreds of programmers.

Sure, but Amdahl knew it was a loss leader.
There were what, a few dozen (maybe 4 dozen) made.
My friend Gail worked on 2 for Bank of America in SF off Market St.

I'm trying to figure out what John sees in a kitchen sink historic
approach to architecture.

--

Eugene Miya

unread,

Nov 1, 2006, 4:34:11 PM11/1/06

to

In article <m3bqns7...@garlic.com>,

Anne & Lynn Wheeler <ly...@garlic.com> wrote:

>you later got 3278 keyboard option with the pf1-pf12 across the top
>and pf13-pf24 on the right.

Programmed function keys are stupid. They are well intended, but
their key sequences aren't standard across all keyboards (like arrow keys).
I used to like them, but they lock you into a particular piece of vendor
hardware (thank god (Ken Arnold and the Evans crew) for things like termcap).

I think the best implementation of PF keys were Evans &
Sutherland Picture System boxes which were dynamically labelled
(really programmed), but these things weren't cheap.
Xerox did the right thing for their keyboards leaving them off.

--

Anne & Lynn Wheeler

unread,

Nov 1, 2006, 5:39:09 PM11/1/06

to

Walter Bushell <pr...@panix.com> writes:
> I've worked in places where that was the most powerful computer, and we
> had hundreds of programmers.

part of 195 was that it required highly optimized application to reach
peak sustained thruput from its pipeline ... about 10mips. for most
normal codes, thruput was more like 5mips.

sjr ran a 370/195 (os/360) mvt service thru the late 70s ... and there
were were jobs that sat in the work queue for several weeks waiting
for execution. also because of its highly complicated hardware
implementation ... it wasn't practical to add virtual address
translation ... which eventually became available for all other 370s
(and with the advent of virtual memory on all models of 370s, all the
operating systems transitioned to virtual memory).

i got somewhat involved in project to add dual i-stream
(multi-threading) to 195; replicate instruction address, registers,
etc ... workload in the pipeline would have one-bit tag added
indicating which i-stream it belong to. the idea was that if most
convential code only kept the pipeline half-full ... then a pair of
i-streams could keep the pipeline full and operating at aggregate
thruput of 10mips. this never shipped (and it didn't address the issue
of retrofitting virtual memory to the machine).

one of the applications that wasn't getting exactly great turn around
on the sjr 195 was air-bearing simulation in support of designing the
3380 floating, thinfilm disk heads (possibly an hr every several
weeks).

as i've mentioned before ... I got involved in hardening an operating
system so that it could be used in disk engineering and product test
labs (bldg. 14 & 15). they had a number of processors that were used
in dedicated stand-alone mode for doing disk regression testing.
because of the high rate of faults from prototype and engineering
hardware, they were unable to operate with conventional operating
systems
http://www.garlic.com/~lynn/subtopic.html#disk

the labs would get early processor models for dedicated regression
testing (including early availability of 4341 and 3033 processors).
with the availability of operating system on the labs machine ... we
found that normal disk regression testing only required a few percent
of processor capacity ... the rest of the processors became available
for other types of application.

3033 was about 4.5mips (a little less than half 195 peak sustained).
however, the air-bearing simulation work could get several hrs every
day on the bldg. 15 3033 ... compared to maybe an hr every several
weeks on the bldg. 28 195.

Eugene Miya

unread,

Nov 1, 2006, 4:41:59 PM11/1/06

to

Going to mainframe people fo solutions to current problems.

eug...@cse.ucsc.edu (Eugene Miya) writes:
>> That works some time.
>> The Web didn't exist, non-hierarchical networks inside IBM didn't exist
>> back, then and a slew of other problems.

>> ... in France.
>
In article <m33b946...@garlic.com>,

Anne & Lynn Wheeler <ly...@garlic.com> wrote:

>SNA in the 70s ... somewhat in conjunction with fs

Keep those dinosaurs running Lynn.

>and somewhat afterwards ... defined hierarchical communication
>architecture for big terminal networks ... pretty much wrapped around

Loosely coupled link stuff clipped.

>recent posts mentioning size of internal network passing 1000 nodes in
>1983 (year that arpanet switched over to internetworking protection)

Protection?

There you go again....

>it is possible that the difference between 100 arpanet nodes and 250
>hosts ... was that on arpanet, the actual networking was handled by
>the outboard IMPs ... with hosts then running host-to-host protocol
>and connecting to the IMPs for the actual networking support
>(potentially allowing greater than one-to-one relationship between
>hosts and nodes).

Damning ARPA with faint praise?

>By far the most important influence on the development of Rexx was the
>availability of the IBM electronic network, called VNET. In 1979,
>more than three hundred of IBM's mainframe computers, mostly
>running the Virtual Machine/370 (VM) operating system, were linked by
>VNET. This store-and-forward network allowed very rapid exchange of
>messages (chat) and e-mail, and reliable distribution of software. It
>made it possible to design, develop, and distribute Rexx and its first
>implementation from one country (the UK) even though most of its users
>were five to eight time zones distant, in the USA.

And no mention of Ira's use of RSCS in the entire post.

--

Anne & Lynn Wheeler

unread,

Nov 1, 2006, 6:23:49 PM11/1/06

to

eug...@cse.ucsc.edu (Eugene Miya) writes:
> Programmed function keys are stupid. They are well intended, but
> their key sequences aren't standard across all keyboards (like arrow
> keys). I used to like them, but they lock you into a particular
> piece of vendor hardware (thank god (Ken Arnold and the Evans crew)
> for things like termcap).
>
> I think the best implementation of PF keys were Evans & Sutherland
> Picture System boxes which were dynamically labelled (really
> programmed), but these things weren't cheap. Xerox did the right
> thing for their keyboards leaving them off.

so all the PF keys were software programmable ... although specific
applications tended to have specific default operations ... i.e. PF3
always mapped to the same function.

so the "data entry" 3278 started out with PF keys as ALT across
the top row.

3278 then introduced new row of PF1-PF12 across the top ... somewhat
like PC keyboard.

then you could keyboard with row of PF1-PF12 across the top and
PF13-PF24 on the right-hand side ... somewhat where the 3277 PF1-PF12
keypad was located.

so although the PF keys were software programmable ... a lot of
applications from 3277 had created embedded assignments for specific
PF keys. So with the availability of PF keys on the side of 3278
... there were several requests for operating system layer to swap the
scancodes between PF1-PF12 and PF13-PF24.

There was a separate issue that started earlier with software
programmable 3277 PF keys with regard to application consistency.
Given that applications could embed specific associations with
specific keys. given that there was no controlled specification for
the PF keys ... like the PC "page up" and "page down" keys
... different applications bound specific functions to specific keys
(and for some reason it took quite a long time to add user
configurable profiles to these applications).

in any case, w/o explicit, well established convention (like physical
labels on the keys) ... there was extended period of application
inconsistent specification for PF key use (aggrevated by applications
that failed to provide user profile configurability).

There was almost infamous arguments in the late 70s and early 80s
about PF3 consistently being used for "END/QUIT/EXIT" function. There
were a variety of emerging full-screen applications, application menu
environments, email readers, editors, etc. There was some irritability
when each environment assigned END/QUIT function to a different PF key
(and user profile configurability was not generally available )

from somewhere long ago and far away

To: wheeler
Date: 12/11/78 09:02:16

how about a cp command that would reset pfkeys 13-24 to be the same as
pfkeys 1-12 ?????

then on my 3278 i could use the pfkeys in the normal position for any
application that used pfkeys 1-12

... snip ...

To: wheeler
Date: 08/11/82 12:44:20

Hi there. I have had a bug reported to me concerning the assignment of
PF13-24 with my VMSG mods. Unfortunately, we don't have many terminals with
more than 12 PF keys around here, so that code wasn't tested very well. Sorry
'bout that! Hopefully, I'll have an updated version available soon and I'll
be re-distributing the package. I'll also be adding some other updates which
people have requested (6 EPILOG/PROLOG lines, for example). If you have any
other requests, let me know. It seems that there are various versions of VMSG
running around and I'm trying to get a hold of the other enhancements to
incorporate into my version. A bientot,

IBM Europe Technical Support
23 Allee Maillasson
92100 Boulogne, France

... snip ...

Anne & Lynn Wheeler

unread,

Nov 1, 2006, 7:21:17 PM11/1/06

to

eug...@cse.ucsc.edu (Eugene Miya) writes:
> And no mention of Ira's use of RSCS in the entire post.

lots of past collected posts mentioning bitnet and/or earn
http://www.garlic.com/~lynn/subnetwork.html#bitnet

... university based network using similar technology to
that used in the internal network.
http://www.garlic.com/~lynn/subnetwork.html#internalnet

while there the number of nodes on the internal network was larger
than the number of arpanet/internet nodes ... from just about the
beginning until possibly sometime mid-85 ... there have also been
claims that the number of bitnet/earn nodes (independent of the
internal network nodes) were larger than the number of
arpanet/internet nodes for at least part of the early 80s.

post including old email referencing the creation of EARN
http://www.garlic.com/~lynn/2001h.html#65

note that while possibly all the nodes on the internal network were
ibm mainframes ... there were some number of bitnet/earn network nodes
that had vnet/rscs emulation running on other kinds of processors.

misc. posts from this year mentioning bitnet &/or earn:
http://www.garlic.com/~lynn/2006b.html#8 Free to good home: IBM RT UNIX

http://www.garlic.com/~lynn/2006b.html#12 IBM 610 workstation computer

http://www.garlic.com/~lynn/2006e.html#7 About TLB in lower-level caches

http://www.garlic.com/~lynn/2006e.html#25 About TLB in lower-level caches
http://www.garlic.com/~lynn/2006e.html#27 X.509 and ssh

http://www.garlic.com/~lynn/2006e.html#37 The Pankian Metaphor
http://www.garlic.com/~lynn/2006i.html#31 virtual memory

http://www.garlic.com/~lynn/2006j.html#8 ALternatives to EMail
http://www.garlic.com/~lynn/2006j.html#23 virtual memory

http://www.garlic.com/~lynn/2006j.html#43 virtual memory

http://www.garlic.com/~lynn/2006m.html#9 An Out-of-the-Main Activity

http://www.garlic.com/~lynn/2006m.html#10 An Out-of-the-Main Activity
http://www.garlic.com/~lynn/2006o.html#60 Greatest Software?

http://www.garlic.com/~lynn/2006r.html#5 Was FORTRAN buggy?

http://www.garlic.com/~lynn/2006t.html#6 Ranking of non-IBM mainframe builders?

heavily edited email ...

Date: 12/12/81 11:25
...

.. is the guy that Fuchs (CUNY - BITNET) wrote to about connecting
BITNET to VNET.. XXXXX is driving the BITNET connection with YYYY and
Fuchs) but they need some resources.. Also, YYYY (works in ZZZZ) also
just recently initiated a contract with Univ. of Wisconsin for putting
Internet/TCP protocol on VM/370

... snip ...

then there was also NSF funded CSNET in the early 80s (independent of
the heavily corporate funded bitnet and earn):

CSNET (Computer Science NETwork) is funded by NSF, and is an attempt
to connect all computer science research institutions in the U.S. It
does not have a physical network of its own, but rather is a set of
common protocols used on top of the ARPANET (Department of Defense),
TeleNet (GTE), and PhoneNet (the regular phone system). The
lowest-cost entry is through PhoneNet, which only requires the
addition of a modem to an existing computer system. PhoneNet offers
only message transfer (off-line, queued, files). TeleNet and ARPANET
allow higher-speed connections and on-line network capabilities such
as remote file lookup and transfer on-line, and remote login.

... snip ...

Brian Inglis

unread,

Nov 1, 2006, 11:21:00 PM11/1/06

to

On 1 Nov 2006 14:34:11 -0700 in alt.folklore.computers,
eug...@cse.ucsc.edu (Eugene Miya) wrote:

You must never have used CMS, XEDIT (and file/reader/email utility
commands/macros), VTEDIT (TECO macros), (TPU/)EVE or other screen apps
that really allowed you to customize and drive a shell or editor using
a right hand function keypad like a finance wiz in a spreadsheet.

--
Thanks. Take care, Brian Inglis Calgary, Alberta, Canada

Brian....@CSi.com (Brian[dot]Inglis{at}SystematicSW[dot]ab[dot]ca)
fake address use address above to reply

Brian Inglis

unread,

Nov 1, 2006, 11:33:37 PM11/1/06

to

On 1 Nov 2006 14:41:59 -0700 in alt.folklore.computers,
eug...@cse.ucsc.edu (Eugene Miya) wrote:

>eug...@cse.ucsc.edu (Eugene Miya) writes:

>In article <m33b946...@garlic.com>,
>Anne & Lynn Wheeler <ly...@garlic.com> wrote:

>>SNA in the 70s ... somewhat in conjunction with fs
>
>Keep those dinosaurs running Lynn.

He seems to be working on keeping them safe to use, which is arguably
more useful nowadays, as our savings and entire economy depends on it.

>>By far the most important influence on the development of Rexx was the
>>availability of the IBM electronic network, called VNET. In 1979,
>>more than three hundred of IBM's mainframe computers, mostly
>>running the Virtual Machine/370 (VM) operating system, were linked by
>>VNET. This store-and-forward network allowed very rapid exchange of
>>messages (chat) and e-mail, and reliable distribution of software. It
>>made it possible to design, develop, and distribute Rexx and its first
>>implementation from one country (the UK) even though most of its users
>>were five to eight time zones distant, in the USA.
>
>And no mention of Ira's use of RSCS in the entire post.

RSCS was the external product name, VNET was the internal product;
there may have been some overlap at times, just as there was between
Lynn's CP and various commercial releases of VM products. ;^>
Had IBM shenanigans not prevented the release of the internal (dare I
say "research" a la DMR et al) product, who knows if MVS, Unix,
Windows would be around?
IMHO IBM's biggest mistakes have not been in their execution of their
business, but in not allowing their research fellows to productize the
hell out of their labours a la 3M.

Anne & Lynn Wheeler

unread,

Nov 2, 2006, 1:11:27 AM11/2/06

to

Brian Inglis <Brian....@SystematicSW.Invalid> writes:
> RSCS was the external product name, VNET was the internal product;
> there may have been some overlap at times, just as there was between
> Lynn's CP and various commercial releases of VM products. ;^>
> Had IBM shenanigans not prevented the release of the internal (dare I
> say "research" a la DMR et al) product, who knows if MVS, Unix,
> Windows would be around?
> IMHO IBM's biggest mistakes have not been in their execution of their
> business, but in not allowing their research fellows to productize the
> hell out of their labours a la 3M.

actually the external product name went thru a number of iterations
... back and forth between RSCS and VNET ... as referenced in
this old email from 1985 included in this post
http://www.garlic.com/~lynn/2006k.html#51 other cp/cms history

aka "the internal version of RSCS" being eventually released as the
VNET PRPQ.

at some point VNET was used to describe both the internal networking
software and the internal network itself
http://www.garlic.com/~lynn/subnetwork.html#internalnet

as in this post
http://www.garlic.com/~lynn/2006t.html#43 The Future of CPUs: What's After Multi-Core?

that includes part of an email from 1981 that references a request
from CUNY to interface BITNET to VNET (where bitnet is the university
network running RSCS software and VNET is both the internal software
and the internal network)

in the 1985 email referencing the release of the VNET PRPQ in the
mid-70s, there was enormous objectives to allowing the VNET PRPQ to be
announced and shipped. you would probably never believe the tortuous
process and logic that eventually led to the company allowing VNET
PRPQ to be shipped.

in any case, this subthread is something of a repeat of a very
similar thread in 2002
http://www.garlic.com/~lynn/2002k.html#20 Vnet : Unbelievable

including references to bitnet/earn
http://www.garlic.com/~lynn/subnetwork.html#bitnet

the above posting from 2002 contains an section from Melinda's vm history
about some of the early rscs/vnet (software) history
http://www.princeton.edu/~melinda/25paper.pdf

Anne & Lynn Wheeler

unread,

Nov 2, 2006, 2:04:05 AM11/2/06

to

Anne & Lynn Wheeler <ly...@garlic.com> writes:
> in any case, this subthread is something of a repeat of a very
> similar thread in 2002
> http://www.garlic.com/~lynn/2002k.html#20 Vnet : Unbelievable

and some other posts in the similar thread/exchange from 2002
http://www.garlic.com/2002k.html#18 Unbelievable
http://www.garlic.com/2002k.html#19 Vnet : Unbelievable
http://www.garlic.com/2002k.html#21 Vnet : Unbelievable
http://www.garlic.com/2002k.html#22 Vnet : Unbelievable
http://www.garlic.com/2002k.html#23 Vnet : Unbelievable
http://www.garlic.com/2002k.html#26 DEC eNet: was Vnet : Unbelievable

the above posts references a post by morris that includes:

A listing I have from 1985 -- the accuracy of which I cannot guarantee --
shows the following node counts:

BITNET 435
ARPAnet 1155
CSnet 104 (excluding ARPAnet overlap)
VNET 1650
EasyNet 4200
UUCP 6000
USENET 1150 (excluding UUCP nodes)

... snip ...

i believe the number of nodes on the internet passed the number of
vnet nodes sometime later in 1985.

majority of the following email went into lengthy detail about
encryption and DES ... but concludes with the ending sentence.

From: wheeler
Date: 06/25/85 19:59:03

re: hsdt des;

...

This encryption technique is attractive for the internal network with
the number of nodes approaching 2000 and the desirability of
end-to-end encryption.

... snip ...

there was extensive use of link encrypters for lines that left
corporate physical facilities (but not much use of end-to-end
encryption). i have some memory of a claim in this period that the
internal network had over half of all the link encrypters in the
world.

the following is an old posting to info.nets from 18feb89

> Does anyone have a current table of size estimates for the academic
> and research networks?
>
> Network as of count Description
> -------- -------- ----- -----------------------------------------------
> BITNET 01/18/85 435 University/nonprofit/research network
> Arpanet 01/22/85 1155 DoD related

The December 1988 BITNET nodes file contains 2691 entries.
This includes BITNET/NETNORTH/EARN nodes.

... snip ...

google net.mail posting from 11aug85 giving list of 694 bitnet nodes as of 12jul85
http://groups-beta.google.com/group/net.mail/browse_thread/thread/1d703f2904d9ace0/551c1d15a1ab71d4?lnk=st&q=&rnum=21&hl=en#551c1d15a1ab71d4

jsa...@ecn.ab.ca

unread,

Nov 2, 2006, 8:05:24 AM11/2/06

to

Eugene Miya wrote:
> In article <4547d804$1@darkstar>, eug...@cse.ucsc.edu (Eugene Miya)
> wrote:

> >> The 195 was good? Have you used one?

> I was asking John.

I used a System 360/67, and then an Amdahl 470/V6, as a student of the
U of A, along with many others.

I don't know if the 195 was fun to use - it lacked the paging hardware
the 67 had, so it was stuck with IBM's old-fashioned batch approach to
computers - but it did succeed in the market.

The IBM 360/91, its predecessor, failed in the market - because the
much cheaper 360/85, using cache memory, offered comparable performance
for a lower price. So the 360/195 combined cache memory with the
performance-enhancing features of the 360/91 arithmetic unit. It
basically offered to IBM customers the same set of performance features
that were later introduced to the micro world by the Pentium.

That _was_ good; but it didn't have much to do with the architecture.

My "kitchen-sink historic approach to architecture", exemplified at

http://www.quadibloc.com/arch/arcint.htm

didn't really have anything to do with this. Partly, it's there simply
to allow people to *understand* what was happening in history. I don't
_really_ think there is an urgent need for present-day microprocessors
to offer a Honeywell 800-style cosequence counter, if anyone was under
that illusion.

In fact, if you'll take a look at my site, I have recently prepared for
a ruthless pruning of the addressing modes down to a minimal set.

John Savard

Anne & Lynn Wheeler

unread,

Nov 2, 2006, 10:48:28 AM11/2/06

to

so just for the fun of it ... here is the thread from 2002
(with correction for finger slip from previous post)
http://www.garlic.com/~lynn/2002k.html#18 Unbelievable
http://www.garlic.com/~lynn/2002k.html#19 Vnet : Unbelievable
http://www.garlic.com/~lynn/2002k.html#20 Vnet : Unbelievable
http://www.garlic.com/~lynn/2002k.html#21 Vnet : Unbelievable
http://www.garlic.com/~lynn/2002k.html#22 Vnet : Unbelievable
http://www.garlic.com/~lynn/2002k.html#23 Vnet : Unbelievable
http://www.garlic.com/~lynn/2002k.html#26 DEC eNet: was Vnet : Unbelievable

and the thread from last spring:

http://www.garlic.com/~lynn/2006j.html#34 Arpa address
http://www.garlic.com/~lynn/2006j.html#43 virtual memory
http://www.garlic.com/~lynn/2006j.html#45 Arpa address

http://www.garlic.com/~lynn/2006j.html#46 Arpa address
http://www.garlic.com/~lynn/2006j.html#49 Arpa address
http://www.garlic.com/~lynn/2006j.html#50 Arpa address
http://www.garlic.com/~lynn/2006j.html#53 Arpa address

http://www.garlic.com/~lynn/2006k.html#1 Hey! Keep Your Hands Out Of My Abstraction Layer!

http://www.garlic.com/~lynn/2006k.html#3 Arpa address
http://www.garlic.com/~lynn/2006k.html#6 Hey! Keep Your Hands Out Of My Abstraction Layer!

http://www.garlic.com/~lynn/2006k.html#8 Arpa address
http://www.garlic.com/~lynn/2006k.html#9 Arpa address

http://www.garlic.com/~lynn/2006k.html#10 Arpa address
http://www.garlic.com/~lynn/2006k.html#31 PDP-1
http://www.garlic.com/~lynn/2006k.html#32 PDP-1
http://www.garlic.com/~lynn/2006k.html#40 Arpa address

http://www.garlic.com/~lynn/2006k.html#42 Arpa address
http://www.garlic.com/~lynn/2006k.html#43 Arpa address

http://www.garlic.com/~lynn/2006k.html#53 Hey! Keep Your Hands Out Of My Abstraction Layer!

http://www.garlic.com/~lynn/2006k.html#56 Hey! Keep Your Hands Out Of My Abstraction Layer!

and the current thread:
http://www.garlic.com/~lynn/2006t.html#27 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#31 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#32 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#34 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#36 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#41 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#42 The Future of CPUs: What's After Multi-Core?

http://www.garlic.com/~lynn/2006t.html#43 The Future of CPUs: What's After Multi-Core?

http://www.garlic.com/~lynn/2006t.html#49 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#50 The Future of CPUs: What's After Multi-Core?

Eugene Miya

unread,

Nov 2, 2006, 7:54:48 PM11/2/06

to

In article <m33b946...@garlic.com>,
Anne & Lynn Wheeler <ly...@garlic.com> wrote:
>>>SNA in the 70s ... somewhat in conjunction with fs

eug...@cse.ucsc.edu (Eugene Miya) writes:
>>Keep those dinosaurs running Lynn.

In article <cosik29ms4rl047dn...@4ax.com>,

Brian Inglis <Brian....@SystematicSW.ab.ca> wrote:
>He seems to be working on keeping them safe to use, which is arguably
>more useful nowadays, as our savings and entire economy depends on it.

Naw, Tandems are more important.

>>>availability of the IBM electronic network, called VNET. In 1979,
>>

>>And no mention of Ira's use of RSCS in the entire post.
>
>RSCS was the external product name, VNET was the internal product;

I had an early demo of the VNET when I moved to the Bay Area.
It had good and bad points. The real problem, I should check Wikipedia,
is the revisionist history of the Bitnet guys who make Bitnet sound so
great.

>Had IBM shenanigans not prevented the release of the internal (dare I
>say "research" a la DMR et al) product, who knows if MVS, Unix,
>Windows would be around?
>IMHO IBM's biggest mistakes have not been in their execution of their
>business, but in not allowing their research fellows to productize the
>hell out of their labours a la 3M.

Well that's well known.

No time. I have to get out of the office.

--

Eugene Miya

unread,

Nov 2, 2006, 7:59:38 PM11/2/06

to

>>Programmed function keys are stupid. They are well intended, but

>>I think the best implementation of PF keys were Evans &

>>Sutherland Picture System boxes which were dynamically labelled
>>(really programmed), but these things weren't cheap.
>>Xerox did the right thing for their keyboards leaving them off.

In article <57sik298hvi8tnna5...@4ax.com>,

Brian Inglis <Brian....@SystematicSW.ab.ca> wrote:
>You must never have used CMS, XEDIT (and file/reader/email utility
>commands/macros), VTEDIT (TECO macros), (TPU/)EVE or other screen apps
>that really allowed you to customize and drive a shell or editor using
>a right hand function keypad like a finance wiz in a spreadsheet.

No, I used the slew of VT terminals and 3270s. I initially liked SPF.

You guys have never had a $300K hardwire graphics system of the E&S caliber.
You guys look down and see what? F[0-9] etc. E&S gave you little tiny
labels (plasma?) with glowing words. No abstract labels on the screen
or symbology on the keyboard. The labels changed, you programmed them.
Goes with their dial boxes, too.

--

Anne & Lynn Wheeler

unread,

Nov 3, 2006, 11:29:03 AM11/3/06

to

eug...@cse.ucsc.edu (Eugene Miya) writes:
> Damning ARPA with faint praise?

re:
http:/www.garlic.com/~lynn/2006t.html#36 The Future of CPUs: What's After Multi-Core

possibly just slight historical interest at where the arpanet scalling
limits were. somebody had made satirical comments about arpanet in the
late 70s about IMPs requiring 56kbit interconnect because majority of
the bandwidth could be taken up by inter-IMP administrative chatter
(including all the stuff about figuring out what would be the optimal
path for each packet). reaching 100 nodes by jan83 ... would imply
that scalling limit was starting to be reached with a lot less than
100 nodes in the late 70s.

later the comment was that part of the transition off of IMPs and
arpanet protocol (with the switch-over to internetworking protocol on
1/1/83) was being able to get out from under various of the scalling
limitations.

it probably doesn't a whole lot of difference some 25-30 years later
whether the arpanet scalling limits were 100 nodes or 250 hosts.

past threads where the load of the inter-IMP administrative chatter
was mentioned
http://www.garlic.com/~lynn/2003c.html#42 diffence between itanium and alpha
http://www.garlic.com/~lynn/2003c.html#47 diffence between itanium and alpha
http://www.garlic.com/~lynn/2003d.html#60 Bitnet again was: unix
http://www.garlic.com/~lynn/2003d.html#62 ARPAnet again: Bitnet again was: unix
http://www.garlic.com/~lynn/2003h.html#17 Why did TCP become popular ?
http://www.garlic.com/~lynn/2004o.html#53 360 longevity, was RISCs too close to hardware?

http://www.garlic.com/~lynn/2006e.html#35 The Pankian Metaphor

http://www.garlic.com/~lynn/2006j.html#45 Arpa address

Eugene Miya

unread,

Nov 3, 2006, 11:07:07 AM11/3/06

to

>> Damning ARPA with faint praise?

In article <m38xisw...@lhwlinux.garlic.com>,

Anne & Lynn Wheeler <ly...@garlic.com> wrote:

>possibly just slight historical interest at where the arpanet scalling
>limits were. somebody had made satirical comments about arpanet in the
>late 70s about IMPs requiring 56kbit interconnect because majority of
>the bandwidth could be taken up by inter-IMP administrative chatter

Oh all those RFCs Jake is preserving? Oh yeah...

>(including all the stuff about figuring out what would be the optimal
>path for each packet). reaching 100 nodes by jan83 ... would imply

You stay in the late 70s and early 80s
and not the late 60s and early 70s when IBM poopooed it.

>that scalling limit was starting to be reached with a lot less than
>100 nodes in the late 70s.
>
>later the comment was that part of the transition off of IMPs and
>arpanet protocol (with the switch-over to internetworking protocol on
>1/1/83) was being able to get out from under various of the scalling
>limitations.

Better than no network in the 60s.

>it probably doesn't a whole lot of difference some 25-30 years later

matter

>whether the arpanet scalling limits were 100 nodes or 250 hosts.

Yep. Ethernet addressing was initially only 8 bits, but it's a LAN.

--

Eugene Miya

unread,

Nov 3, 2006, 11:16:43 AM11/3/06

to

>> Programmed function keys are stupid.

...

>> I think the best implementation of PF keys were Evans & Sutherland
>> Picture System boxes which were dynamically labelled (really
>> programmed), but these things weren't cheap. Xerox did the right
>> thing for their keyboards leaving them off.

In article <m3irhyn...@lhwlinux.garlic.com>,

Anne & Lynn Wheeler <ly...@garlic.com> wrote:

>so all the PF keys were software programmable ... although specific
>applications tended to have specific default operations ... i.e. PF3
>always mapped to the same function.

Consistency (standardization) are good for advancing the world.
No one can easily dispute that. The issue isn't software
programmability. Both Unix and emacs understood the issue of keybindings.
You can tell the IBM people (Backspace) from the DEC people (DELETE).

>so the "data entry" 3278 started out with PF keys as ALT across
>the top row.

irrelevant stuff snipped

>so although the PF keys were software programmable ... a lot of

Consider hardware programmable.

A bunch of you guys liked blinken lights.
How about letters, works, etc.?
Not stupid lights which turn off and one behind fixed
plexiglas signs with templated letters.

>about PF3 consistently being used for "END/QUIT/EXIT" function. There

Ah there is hope for enlightening Lynn yet.

>were a variety of emerging full-screen applications, application menu
>environments, email readers, editors, etc. There was some irritability
>when each environment assigned END/QUIT function to a different PF key
>(and user profile configurability was not generally available )

/*

--

Anne & Lynn Wheeler

unread,

Nov 3, 2006, 12:29:08 PM11/3/06

to

eug...@cse.ucsc.edu (Eugene Miya) writes:
> You stay in the late 70s and early 80s
> and not the late 60s and early 70s when IBM poopooed it.

re:
re:
http://www.garlic.com/~lynn/2006t.html#36 The Future of CPUs: What's After Multi-Core
http://www.garlic.com/~lynn/2006u.html#0 The Future of CPUs: What's After Multi-Core
http://www.garlic.com/~lynn/2006u.html#6 The Future of CPUs: What's After Multi-Core

"IBM" ... especially SNA organization poo-poo'ed all sorts of
stuff.

as i've mentioned before we constantly battled with them.

and as i mentioned before ... one of the first "internal network" activities
http://www.garlic.com/~lynn/subnetwork.html#internalnet

was between the science center, on the 4th flr, 545 tech sq
http://www.garlic.com/~lynn/subtopic.html#545tech

and endicott. this was software development activity to add "370"
virtual machine support to the cp67 kernel (running on 360/67). this
was before 370s were available ... and their were numerous
architecture differences between virtual memory hardware on 360/67 and
370 virtual memory architecture. the project required putting a lot of
code into the cp67 kernel to recognize the 370 virtual machine
operation and a whole lot of simulating 370 hardware characteristics
that were different than 370/367.

and there were people that cambridge was working with that were doing
arpanet support. for something a little bit different, my rfc index

http://www.garlic.com/~lynn/rfcietff.htm

and select "Author" in the "RFCs listed by" section and use browser
find function to get "Winett J." Joel worked on cp67 out at Lincoln
... after cambridge got cp67 up and working, they installed cp67 out
at Lincoln in 1967 (the third cp67 installation was at the university
in jan68, where i was undergraduate).

Winett J.
466 393 183 167 147 110 109

clicking on any RFC number, brings up the RFC summary in the lower frame,
i.e.

109
Level III Server Protocol for the Lincoln Laboratory NIC 360/67
Host, Winett J., 1971/03/24 (12pp)

normally, clicking on the ".txt=nnn" field in the RFC summary will
retrieve the actual RFCs. However, for (old) RFCs that haven't yet
been brought online, there won't be a ".txt=nnn" field.

however, there is also

466
Telnet logger/server for host LL-67, Winett J., 1973/02/27 (8pp) (.txt=17595)

which does have a ".txt=nnn" field.

Eugene Miya

unread,

Nov 3, 2006, 11:38:31 AM11/3/06

to

>> >> The 195 was good? Have you used one?
>> I was asking John.
>
>I used a System 360/67, and then an Amdahl 470/V6, as a student of the
>U of A, along with many others.

I can't remember if I ever used a V6. I've had 580s. I have used a 67.

In article <1162472724.3...@m73g2000cwd.googlegroups.com>,

<jsa...@ecn.ab.ca> wrote:
>I don't know if the 195 was fun to use - it lacked the paging hardware
>the 67 had, so it was stuck with IBM's old-fashioned batch approach to
>computers - but it did succeed in the market.

I am not certain with your accessment of "succeed."

>The IBM 360/91, its predecessor, failed in the market - because the
>much cheaper 360/85, using cache memory, offered comparable performance
>for a lower price. So the 360/195 combined cache memory with the
>performance-enhancing features of the 360/91 arithmetic unit. It
>basically offered to IBM customers the same set of performance features
>that were later introduced to the micro world by the Pentium.

Hmmmmm. I didn't have to maintain a 91 and its water cooling.
At this point it depends on the nature of codes. 32-bit only barely cuts
it for real scientific codes (long running codes).
So were you running business codes?

>That _was_ good; but it didn't have much to do with the architecture.
>
>My "kitchen-sink historic approach to architecture", exemplified at
>
>http://www.quadibloc.com/arch/arcint.htm

I am not enough of an architect to say more than I see too much verbage
and not enough stuff on layout (implementation).

>didn't really have anything to do with this. Partly, it's there simply
>to allow people to *understand* what was happening in history. I don't
>_really_ think there is an urgent need for present-day microprocessors
>to offer a Honeywell 800-style cosequence counter, if anyone was under
>that illusion.
>
>In fact, if you'll take a look at my site, I have recently prepared for
>a ruthless pruning of the addressing modes down to a minimal set.

No idea.

--

Anne & Lynn Wheeler

unread,

Nov 3, 2006, 12:51:41 PM11/3/06

to

eug...@cse.ucsc.edu (Eugene Miya) writes:
> Consider hardware programmable.
>
> A bunch of you guys liked blinken lights.
> How about letters, works, etc.?
> Not stupid lights which turn off and one behind fixed
> plexiglas signs with templated letters.

re:
http://www.garlic.com/~lynn/2006t.html#42 The Future of CPUs: What's After Multi-Core?

circa 76-78? ... and inexpensive enuf that they were starting to
appear on everybody's desk. at the start of this period, there was
requirement that to get 3270 on person's desk required vice-president
sign-off. we put together a business analysis that the 3yr
amortized/depreciated cost of a 3270 terminal was about the same as a
business phone (which was available on everybody's desk as matter of
course).

are we talking stuff that was ubiquitously available that everybody in
a corporation of several hundred thousand people could have one?

and i've frequently contended
http://www.garlic.com/~lynn/subnetwork.html#emulation

it was later that the ibm/pc and 3270 terminal was about the same cost
... so it was a relatively straight-forward transaction to justify an
ibm/pc as a 3270 terminal replacement ... where it was possible to
have both 3270 terminal emulation and some local desktop computing in
a single physical footprint (and same cost) ... that significant
market segment uptake contributed to enormously to wide proliferation
of the the ibm/pc.

as you mentioned in your other post ... are we talking 70s or 80s?
http://www.garlic.com/~lynn/2006u.html#7 The Future of CPUs: What's After Multi-Core?

and is it a commodity thing that can be ubiquitously deployed
world-wide that is generally available for potentially everybody in a
large corporation.

Eugene Miya

unread,

Nov 3, 2006, 12:24:21 PM11/3/06

to

>> Consider hardware programmable PF keys.

In article <m3zmb8v...@lhwlinux.garlic.com>,

Anne & Lynn Wheeler <ly...@garlic.com> wrote:

>circa 76-78? ... and inexpensive enuf

..

>are we talking stuff that was ubiquitously available that everybody in
>a corporation of several hundred thousand people could have one?

No. I didn't say this stuff was cheap. I said PF keys are stupid.

>as you mentioned in your other post ... are we talking 70s or 80s?

PF keys in the late 70s.
Networks in the 60s-70s.

>and is it a commodity thing that can be ubiquitously deployed
>world-wide that is generally available for potentially everybody in a
>large corporation.

Of course it's not commodity.
Most computer people have never heard of Evans and Sutherland as a firm.

--

Anne & Lynn Wheeler

unread,

Nov 3, 2006, 1:42:27 PM11/3/06

to

eug...@cse.ucsc.edu (Eugene Miya) writes:
> Of course it's not commodity.
> Most computer people have never heard of Evans and Sutherland as a firm.

part of old "decnews" email (clipped after mention of evans and
sutherland) ...

also notice DEC going full steam ahead with federal gov's OSI
replacement for internet

To: wheeler
Date: 31 Mar 88 21:43:52 GMT
To: lynn@wheeler

Digital Information March 88

The quote of the month comes from Shaku Atre, president of Atre Interna-
tional Consultants, Inc. in Rye, N.Y. In discussing how DEC is trying to
make headway into the "mainframe" market; "People shouldn't be trying to
get into the mainframe market right now. What DEC should be doing is
looking to its micros since there hasn't been anything OVER THE RAIN-
BOW".

DEC - SPECIFIC
______________

y Since DEC continues to have problems with their RA70 disk drive,
they are now offering another option to perspective MicroVAX 3500
customers. Now a customer can buy a MicroVAX 3500 with two RD-54
disk drives instead of the RA-70s. DEC previously offered the option
of RA-81 drives, but those big drives are not practical in many of-
fice environments (Digital Review 3/21/88).

y The number of DEC-compatible manufacturers has dropped from 36 three
years ago to about 6 today. Fewer new firms are getting DEC cpu and
microprocessor chips. Three of the remaining six sell ruggedized
versions of MicroPDPs, MicroVAX I & 2 and PDP-11s primarily to the
military (Digital Review 1/25/88).

y DEC's "announced" DESKTOP STRATEGY is to connect "MS-DOS, VAX-based
UNIX and Macintosh systems to DEC VAX/VMS systems, and to extend
services of Decnet/OSI to IBM OS/2 desktops" (Notice the use of the
words "Decnet/OSI". Is there any doubt about DEC's committed resolve
to OSI?) (Computerworld 1/25/88).

y DEC's NAS (Network Application Support) strategy is designed to sup-
port connectivity of different systems in one environment, in effect
creating a common interface across then (i.e., positioning DEC as
being the central control point for the entire network). NAS will
support the following protocols; DAP (the Data Access Protocol in
Decnet); Microsoft's SMB (Server Messgae Block), Sun's NFS (Network
File System); Apple's AFP file sharing Protocol; Adobe Systems'
Postscript for desktop publishing; DDIF (Digital Document Inter-
change Format), DEC's document processing internal standard; and SQL
for DB use

y DEC now includes DECnet/PCSA (Personal Computer System Architecture)
client licenses with VAXmates free of charge while at the same time
reducing VAXmate prices as much as 37%. The VAXmate originally sold
for $4,250.00 with an additional $500.00 charge for DECnet/PCSA
(Digital Review 3/7/88).

DEC - GENERAL
_____________

y Evans & Sutherland now markets an option for the newly-announced
VAXStation 8000 workstations that make it possible to view 3-D im-
ages on the screen "in stereo". The technology includes a time-
multiplexed liquid crystal shutter fitting over the display,
polarizing the display for both the left and right eye. To see the
effect, one must wear a pair of (you guessed it) glasses with
polarized lenses. The option costs $11,500.00 (Digital Review
3/21/88).

Eugene Miya

unread,

Nov 3, 2006, 1:17:07 PM11/3/06

to

In article <m3slh0v...@lhwlinux.garlic.com>,

Anne & Lynn Wheeler <ly...@garlic.com> wrote:

>eug...@cse.ucsc.edu (Eugene Miya) writes:
>> Of course it's not commodity.
>> Most computer people have never heard of Evans and Sutherland as a firm.
>
>
>part of old "decnews" email (clipped after mention of evans and
>sutherland) ...

Naw way too late.

>also notice DEC going full steam ahead with federal gov's OSI
>replacement for internet

Yeah and OSI is dead, too.

>To: wheeler
>Date: 31 Mar 88 21:43:52 GMT
>To: lynn@wheeler
>
>Digital Information March 88

...

> DEC - GENERAL
> _____________
>
> y Evans & Sutherland now markets an option for the newly-announced
> VAXStation 8000 workstations that make it possible to view 3-D im-
> ages on the screen "in stereo". The technology includes a time-
> multiplexed liquid crystal shutter fitting over the display,
> polarizing the display for both the left and right eye. To see the
> effect, one must wear a pair of (you guessed it) glasses with
> polarized lenses. The option costs $11,500.00 (Digital Review
> 3/21/88).

This really says nothing of the firm Ivan started.
Sure at this time, this was about the time of the infamous Computer
Division (the one great hope to compete with Cray, I saved one, it's a
good argument with Gordon Bell). This says nothing of the main meat of
the firm which is high performance real-time graphics for flight
simulation, and other kinds of simulators (cars, tactical, etc.).

It's the difference between Microsoft flight simulators and the big boys
with motion bases which can kill or injure you for real if you don't
appreciate their realism.

Great calendars and great demo reels (approach to crashing cars up to but
excluding the moment of impact).

Anyways, Ivan's at SUN now.

--

Anne & Lynn Wheeler

unread,

Nov 3, 2006, 2:22:42 PM11/3/06

to

eug...@cse.ucsc.edu (Eugene Miya) writes:
> This really says nothing of the firm Ivan started. Sure at this
> time, this was about the time of the infamous Computer Division (the
> one great hope to compete with Cray, I saved one, it's a good
> argument with Gordon Bell). This says nothing of the main meat of
> the firm which is high performance real-time graphics for flight
> simulation, and other kinds of simulators (cars, tactical, etc.).

so the (real) flight simulator are probably a couple million? ... mentioned
on the e&s wiki page

with respect to the following, i got some followup emails commenting
about the "400MIP" number, the actual workstation being only 4-5mips
... and the 400MIP comes from aggregating all the special purpose
graphic chips. I would guess that the $100k is on the low-end of
prices for E&S machines?

the original email goes on quite a bit more about lots more details
from the original announcement.

To: wheeler
Date: 05 Feb 88 18:12:41 GMT

DEC formally announced the VAXstation 8000. Based on their VAX 8250
and MicroVAX II chip sets and the PS390 from Evans & Sutherland, DEC
has developed one of the fastest workstations (over 400 million 32-bit
arithmetic operations per second) with the highest vector drawing
speed (over 500,000 3D vectors/sec) and highest "apparent" resolution
(8192 x 6912) in the industry. This is DEC's first workstation to
offer hi-performance graphics with full 3D functions. In spite of
this, it is still a single-user workstation which lists at almost
$100,000.00, and is limited to 3 RD54 disk drives (total of 477 MB),
obviously expecting customers to use the VAXstation 8000 as a node in
a VAX Local-Area VAXCluster. I was asked why DEC would use 8250 chips
when the new CMOS MicroVAX 3600 chips are available. From a design
standpoint, the 8250 chip was already complete. From a manufacturing
standpoint, the 8250 chips were already manufactured and probably
sitting in a warehouse somewhere, since sales of 8250s have not been
setting the world on fire. And bottom line: I can see the DEC salesman
now - "We already have a 400 MIP workstation, we certainly do not need
any more horsepower at this point".

The following information was downline loaded from the Digital
Electronic Store about the announcement.

DESCRIPTION

High-Performance, 3D, Realtime Graphics Workstation

The VAXstation 8000 is the newest and most powerful member of
Digital's VAXstation family of high-performance graphic workstations.
Developed jointly by Digital Equipment Corporation and Evans &
Sutherland Corporation, the VAXstation 8000 is a high-performance,
full color workstation that can manipulate complex three-dimensional
graphic objects in realtime.

Evans & Sutherland is widely recognized as a leader in computer
graphics technology. The combination of their expertise in computer
graphics with Digital's system and workstation expertise has resulted
in an industry-leading product -- the VAXstation 8000.

The VAXstation 8000 is designed for applications requiring the highest
levels of computer graphics speed and clarity, such as molecular
modeling, fluid dynamics, mechanical computer-aided engineering and
design, manufacturing engineering, command and control, and computer
animation.

With the VAXstation 8000, Digital now extends the range of the
VAXstation family even further, from the low-cost desktop VAXstation
2000, through the VAXstation II/GPX and VAXstation 3200 and 3500, to
the state-of-the-art VAXstation 8000.

HIGH PERFORMANCE GRAPHICS WITH FULL 3D FEATURES

The VAXstation 8000 is a single-user VAXstation based on the VAX 8250
CPU and a high performance graphics subsystem. It is housed in an
compact, desk-side system enclosure. Digital's first system to offer
high-performance graphics hardware with full 3D functions, the
VAXstation 8000 offers the fastest vector drawing speed in the
industry -- over 500,000 3D vectors per second.

By using unique anti-aliasing technologies, the VAXstation 8000
provides an apparent resolution exceeding 8000 by 6000 pixels.

Unlike other very high performance workstations, which require special
programming and application interfaces, the VAXstation 8000 supports
the X Window System version 11 (on both the VMS and ULTRIX operating
systems) and PHIGS standards. These high-level interfaces are
standards in the workstation market and provide a fully compatible
link to the other members of Digital's VAXstation family. VAXstation
8000 workstation windowing software also includes an extensive 3D
graphics library in addition to the X Window System.

VAX PHIGS, Digital's new hierarchically-oriented 3D and 2D graphics
software language, enables application programmers to take advantage
of the power and speed of the VAXstation 8000 by using this standard
high-level graphics programming language.

... snip ... and resume ...

VAXSTATION FAMILY COMPARISON CHART

---------------------------------------------------------------------------
| |VAXstation| VAXstation | VAXstation|VAXstation|VAXstation|
| |2000 | II/GPX | 3200 |3500 |8000 |
---------------------------------------------------------------------------
| CPU/FPU | MicroVAX | MicroVAX | CVAX | CVAX | VAX 8250 |
| | | | | | |
| # Planes | 1 or 4 | 4 or 8 | 4 or 8 | 4 or 8 | 58 |
| | | | | | |
| Monitors | 15 & 19" | 19" | 19" | 19" | 19" |
| | Mono & | Mono & | Mono & | Mono & |Color Only|
| | Color | Color | Color | Color | |
| | | | | | |
| Resolution | 1024x864 | 1024x864 | 1024x864 | 1024x864 | 1024x864 |
| | | | | |Effective |
| | | | | 8192 x 6912|
| | | | | | |
| Backplane | None | Q-bus | Q-bus | Q-bus | BI bus |
| # slots | | 8/12 | 8 | 12 | 6 |
| | | | | | |
|Expansion slots| None | 2/6 | 0/1 | 4/6 | 1 |
| (min/max) | | | | | |
| Ethernet | Thinwire | thick wire | thick wire|thick wire|thick wire|
| Controller | Standard | Standard | Standard |Standard |Standard |
| | | | | | |
| Max Memory | 6 MB | 16 MB | 16 MB | 32 MB | 32 MB |
| | | | | | |
| Disk Cavities | 2 | 1 or 3 | 1 | 2 | 3 |
| | | | | | |
| Min/Max Hard | 42 Mb/ | 71Mb/ | 71Mb/ | 159Mb/ | 159Mb/ |
| Disk Support | 318 Mb | 477 Mb | 318 Mb | 560 Mb | 477Mb |
| | | | | | |
| System | One year | One year | One year | One year | One year |
| Warranty | | | | | |
| | | | | | |
---------------------------------------------------------------------------

... snip ...

Eugene Miya

unread,

Nov 3, 2006, 5:10:05 PM11/3/06

to

>> Evans & Sutherland

In article <m3lkmsu...@lhwlinux.garlic.com>,

Anne & Lynn Wheeler <ly...@garlic.com> wrote:

>so the (real) flight simulator are probably a couple million? ... mentioned
>on the e&s wiki page

Well depending on the level/requirement for realism they have gome from
those little "toy" link simulators formerly made in the Fry's Santa Clara
which look like miniature planes for a few thousand to full on motion
bases with "infinity optics" and the full affordances of cockpits, real
joy sticks, etc. E&S graphics hardware depending on the amount of
memory in the pipeline could go $10M. And that's just the graphics.
The most expensive one I know is $100M with multiple cabs (the Shuttle
being one). The distant control tower looks like it's in the distance,
not at your nose.

>with respect to the following, i got some followup emails commenting
>about the "400MIP" number, the actual workstation being only 4-5mips
>... and the 400MIP comes from aggregating all the special purpose
>graphic chips. I would guess that the $100k is on the low-end of
>prices for E&S machines?

You are confusing 2 different lines.
The Picture System workstation is a completely different line.
They are attached processors to old minis and micros not for stand along
use. Vs. the flt. sims. or the ESCD (across the street from the Museum
and 1/2 block from the Googleplex on Shoreline).

>To: wheeler
>Date: 05 Feb 88 18:12:41 GMT

They had been in existence over a decade by this time.

>DEC formally announced the VAXstation 8000. Based on their VAX 8250
>and MicroVAX II chip sets and the PS390 from Evans & Sutherland, DEC

PS390? Hell I worked in PS2, and we have a PS 1.

>has developed one of the fastest workstations (over 400 million 32-bit
>arithmetic operations per second) with the highest vector drawing
>speed (over 500,000 3D vectors/sec) and highest "apparent" resolution

You like that metric?
At least it's not polygons.

>(8192 x 6912) in the industry. This is DEC's first workstation to
>offer hi-performance graphics with full 3D functions. In spite of
>this, it is still a single-user workstation which lists at almost
>$100,000.00, and is limited to 3 RD54 disk drives (total of 477 MB),

If one wants the best....

>obviously expecting customers to use the VAXstation 8000 as a node in
>a VAX Local-Area VAXCluster. I was asked why DEC would use 8250 chips
>when the new CMOS MicroVAX 3600 chips are available. From a design

...

>The following information was downline loaded from the Digital
>Electronic Store about the announcement.

^^^^^

>DESCRIPTION
>
>High-Performance, 3D, Realtime Graphics Workstation
>
>The VAXstation 8000 is the newest and most powerful member of

....

>Evans & Sutherland is widely recognized as a leader in computer
>graphics technology.

...

>HIGH PERFORMANCE GRAPHICS WITH FULL 3D FEATURES
>
>

>Unlike other very high performance workstations, which require special
>programming and application interfaces, the VAXstation 8000 supports
>the X Window System version 11 (on both the VMS and ULTRIX operating

done by people who didn't know graphics (but well meaning)

>systems) and PHIGS standards. These high-level interfaces are

dead standard.

>standards in the workstation market and provide a fully compatible
>link to the other members of Digital's VAXstation family. VAXstation
>8000 workstation windowing software also includes an extensive 3D
>graphics library in addition to the X Window System.
>
>VAX PHIGS, Digital's new hierarchically-oriented 3D and 2D graphics
>software language, enables application programmers to take advantage
>of the power and speed of the VAXstation 8000 by using this standard
>high-level graphics programming language.

Library, not language (DEC marketting's problem).

--

jsa...@ecn.ab.ca

unread,

Nov 3, 2006, 9:49:41 PM11/3/06

to

Eugene Miya wrote:
> I am not enough of an architect to say more than I see too much verbage
> and not enough stuff on layout (implementation).

Well, I do discuss implementation a little bit at

http://www.quadibloc.com/comp/cp0202.htm

John Savard

Brian Inglis

unread,

Nov 4, 2006, 12:37:05 AM11/4/06

to

On 3 Nov 2006 15:10:05 -0700 in alt.folklore.computers,
eug...@cse.ucsc.edu (Eugene Miya) wrote:

>>> Evans & Sutherland
>
>In article <m3lkmsu...@lhwlinux.garlic.com>,
>Anne & Lynn Wheeler <ly...@garlic.com> wrote:
>>so the (real) flight simulator are probably a couple million? ... mentioned
>>on the e&s wiki page
>
>Well depending on the level/requirement for realism they have gome from
>those little "toy" link simulators formerly made in the Fry's Santa Clara
>which look like miniature planes for a few thousand to full on motion
>bases with "infinity optics" and the full affordances of cockpits, real
>joy sticks, etc. E&S graphics hardware depending on the amount of
>memory in the pipeline could go $10M. And that's just the graphics.
>The most expensive one I know is $100M with multiple cabs (the Shuttle
>being one). The distant control tower looks like it's in the distance,
>not at your nose.
>
>>with respect to the following, i got some followup emails commenting
>>about the "400MIP" number, the actual workstation being only 4-5mips
>>... and the 400MIP comes from aggregating all the special purpose
>>graphic chips. I would guess that the $100k is on the low-end of
>>prices for E&S machines?

>They had been in existence over a decade by this time.

>
>>DEC formally announced the VAXstation 8000. Based on their VAX 8250
>>and MicroVAX II chip sets and the PS390 from Evans & Sutherland, DEC
> PS390? Hell I worked in PS2, and we have a PS 1.
>>has developed one of the fastest workstations (over 400 million 32-bit
>>arithmetic operations per second) with the highest vector drawing

adds?

>>speed (over 500,000 3D vectors/sec) and highest "apparent" resolution
> You like that metric?
> At least it's not polygons.

Nothing to do with Digital.

>>(8192 x 6912) in the industry. This is DEC's first workstation to
>>offer hi-performance graphics with full 3D functions. In spite of
>>this, it is still a single-user workstation which lists at almost
>>$100,000.00, and is limited to 3 RD54 disk drives (total of 477 MB),
> If one wants the best....

...one bought SGI in that timeframe, likely even more expensive, but a
lot more functional.

About 5 years ago, a workstation from Sun with dual graphics and
monitors, max memory, and extra 9GB still cost about $60K. Seismic
display/processing/analysis software probably doubled that cost.
Annual replacement for higher performance was not even an issue.
SGI was non-competitive in that market by then.
Sun is probably getting to the same stage, if not already there.

Still some SGI workstations around here, mainly to create fast
high-res 3D movies (with soundtracks) of seismic plays for investor
presentations.

>>Unlike other very high performance workstations, which require special
>>programming and application interfaces, the VAXstation 8000 supports
>>the X Window System version 11 (on both the VMS and ULTRIX operating
> done by people who didn't know graphics (but well meaning)
>>systems) and PHIGS standards. These high-level interfaces are
> dead standard.

Already an official standard and so hopelessly out of date given
advances...unlike OpenGL from SGI, still a de facto standard.

>Library, not language (DEC marketting's problem).

ISTR most of the 1980s graphics APIs being discussed and each
considered inadequate for their *intended* area of application, so
basically pretty useless outside those areas, except for trivial work.

Eugene Miya

unread,

Nov 6, 2006, 12:05:07 PM11/6/06

to

In article <1162608581....@m73g2000cwd.googlegroups.com>,

Architecture, not arithmetic.

You want a real serious architect to review.

--

Eugene Miya

unread,

Nov 6, 2006, 12:29:12 PM11/6/06

to

>>>> graphics

>>>speed (over 500,000 3D vectors/sec) and highest "apparent" resolution
>> You like that metric?
>> At least it's not polygons.

In article <598ok2prju3a86kck...@4ax.com>,

Brian Inglis <Brian....@SystematicSW.ab.ca> wrote:
>Nothing to do with Digital.

Well, some of the guys at DECWRL and DECSRC did on occasion sit in on
meetings (at places like SUN and SGI no less).

>>>$100,000.00, and is limited to 3 RD54 disk drives (total of 477 MB),
>> If one wants the best....
>...one bought SGI in that timeframe, likely even more expensive, but a
>lot more functional.

Sort of.
SGI make a bunch of early bad decisions: SV (they thought seemed good
idea), XNS, etc. These did not bode well for network integration, OS
issues, etc.

>About 5 years ago, a workstation from Sun with dual graphics and
>monitors, max memory, and extra 9GB still cost about $60K. Seismic
>display/processing/analysis software probably doubled that cost.
>Annual replacement for higher performance was not even an issue.
>SGI was non-competitive in that market by then.
>Sun is probably getting to the same stage, if not already there.
>
>Still some SGI workstations around here, mainly to create fast
>high-res 3D movies (with soundtracks) of seismic plays for investor
>presentations.

Yeah, we still have quite a few.
I was given a SUN (which I had initially more for the Java), but I spent
more times on PCs and Macs, but I'm posting from a SUN). FAQs go out
on the SUN.

>>>systems) and PHIGS standards. These high-level interfaces are
>> dead standard.
>Already an official standard and so hopelessly out of date given
>advances...unlike OpenGL from SGI, still a de facto standard.

Not my problem anymore. Like I told the guys in my old chapter:
I was more interested in the computer and less the graphics.

>>Library, not language (DEC marketting's problem).
>
>ISTR most of the 1980s graphics APIs being discussed and each
>considered inadequate for their *intended* area of application, so
>basically pretty useless outside those areas, except for trivial work.

The real problem was a flood of people who only thought about 2-D.
And all these languages aren't well made for a 2-D vs. 3-d dichotomy
(like how many indices or how big are arrays you plane to pass?).
Now with more LISP oriented guys, they think they can know that they can
handle this. It's partially true, but not the complete problem
(a spaid of MIT papers reinventing work done a decade before them in
other languages in the 80s).

But it shows how impressive a guy like Ivan Sutherland is.

--

Brian Inglis

unread,

Nov 7, 2006, 3:30:32 AM11/7/06

to

On 6 Nov 2006 10:29:12 -0700 in alt.folklore.computers,
eug...@cse.ucsc.edu (Eugene Miya) wrote:

>But it shows how impressive a guy like Ivan Sutherland is.

...how to clip lines using only a nybble!

Eugene Miya

unread,

Nov 7, 2006, 2:04:57 PM11/7/06

to

In article <s2d0l2lvigt85k6g2...@4ax.com>,

Brian Inglis <Brian....@SystematicSW.ab.ca> wrote:
>On 6 Nov 2006 10:29:12 -0700 in alt.folklore.computers,
>eug...@cse.ucsc.edu (Eugene Miya) wrote:
>>But it shows how impressive a guy like Ivan Sutherland is.
>
>...how to clip lines using only a nybble!

What's interesting about Ivan is that when you talk to graphics people,
they have no idea of his influence on computer networks and architecture.
When you talk to his DARPA buddies from those areas, they really knew
nothing of his PhD thesis Sketchpad and his graphics work which keeps
getting reinvented.

--

Jeff Jonas

unread,

Nov 16, 2006, 6:42:44 AM11/16/06

to

>> When IBM came out with the 103-key keyboard - the one that *finally*
>> (admittedly, the PCjr keyboard was first) had both shift keys, the
>> Enter key, and the backspace key *all* in the right places... one
>> dissenting voice was heard.
...

My keyboards are in storage but I bought some surplus keyboards
in the 80s that I never fully understood.

One was from the Herbach & Rademan catalogue: an IBM console
with a red, white and blue keyboard. Each key was a hall effect chip,
and it was EBCDIC encoded. But it was still a keypunch arrangement,
with shift for the numeric keypad on the right hand.
But the 13 red keys were NOT encoded at all: there were 13 lines for them
in addition to the parallel data and strobe.
There were tiny illuminated pushbuttons and numeric displays on the top.
And a ONE CHARACTER 5x7 LED that was also EBCDIC.
I never knew to what it attached.

I have similar Honeywell units: keypunch-layout fully electronic keyboards.
I wonder if that was intended for migrating keypunch operators
to data-entry as all-electronic systems started deployment?

--

-- mejeep deMeep ferret!

Jeff Jonas

unread,

Nov 16, 2006, 6:56:35 AM11/16/06

to

>Programmed function keys are stupid. They are well intended, but

>their key sequences aren't standard across all keyboards (like arrow keys).
>I used to like them, but they lock you into a particular piece of vendor
>hardware (thank god (Ken Arnold and the Evans crew) for things like termcap).

Thank you for speaking up.
When I saw keyboards with more than 12 special function keys,
I wondered "who can actually MEMORIZE so many actions,
regardless of the repetition?"
It's like the cash registers with a button for every possible item
(such as fast-food joints): it reduces everyone to hunt-and-peck
(well, except for really frequently-used-buttons along the edge).

And ergonomics? Even with my regular keyboard it's a long long reach
to get to a function key. With the "stadium" of 24 function keys
on some terminals, I can't fathom anyone really doing that well
unless there's no other data entry (no using the regular keyboard).

How far will we go with yet-another-shift key?
Shift and control keys were reasonable since that allows
generating most of the ASCII characters.
But EMACS and other folks went to meta-shift stuff
with combinations of shifts.
Then the ALT shift/modifier was added.
And the Windows key, or the open-Apple/closed-Apple shift.
On the one hand, each modifier essentially adds a keyboard-of-function-keys,
but the combinations of that with mouse buttons drives me mad!

Oh yea, and what about the ever-increasing number of mouse buttons
from Apple's 1 and M$'s 2 "that's enough for anyone"
to 3 (well, X-terms always used that well).
But with the scroll wheel that tilts left/right and clicks down,
I count 7-8 buttons on today's well-endowed mouse!

>I think the best implementation of PF keys were Evans &
>Sutherland Picture System boxes which were dynamically labelled
>(really programmed), but these things weren't cheap.
>Xerox did the right thing for their keyboards leaving them off.

I never got to use CAD/CAM systems but I loved the illuminated keypad
where buttons lit up only when they did something!

I've found some keyboards with a LCD across the top
that could be used for dynamic labelling the top row of keys,
and there's a proposed keyboard where every key is a tiny LCD.

stre...@rohan.sdsu.edu

unread,

Nov 16, 2006, 12:28:02 PM11/16/06

to

begin quoting Jeff Jonas <je...@panix.com> :
[attribution lost]

>>Programmed function keys are stupid. They are well intended, but
>>their key sequences aren't standard across all keyboards (like arrow keys).
>>I used to like them, but they lock you into a particular piece of vendor
>>hardware (thank god (Ken Arnold and the Evans crew) for things like termcap).

Applications shouldn't "predefine" the function keys -- they should be
dead keys. Let the user define what they should do.

These days, I generally have the window manager intercept function key
keystrokes and do things like "switch desktops".

> Thank you for speaking up.
> When I saw keyboards with more than 12 special function keys,
> I wondered "who can actually MEMORIZE so many actions,
> regardless of the repetition?"

Eventually, anyone /could/. Whether or not that would be a good
idea is another thing altogether.

> It's like the cash registers with a button for every possible item
> (such as fast-food joints): it reduces everyone to hunt-and-peck
> (well, except for really frequently-used-buttons along the edge).

...and god help you if you want an option that the designer didn't
think of. ("What do you mean, 'no pickles'? You're allergic to
cucumbers? I don't have an option for that. No, I can't just tell
the cook not to put on pickles. Yes, I know he's right over there.
I don't have an option for 'no pickles'...")

> And ergonomics? Even with my regular keyboard it's a long long reach
> to get to a function key. With the "stadium" of 24 function keys
> on some terminals, I can't fathom anyone really doing that well
> unless there's no other data entry (no using the regular keyboard).

Think of those modern games... all those extra keys are, in theory,
useful.

> How far will we go with yet-another-shift key?
> Shift and control keys were reasonable since that allows
> generating most of the ASCII characters.
> But EMACS and other folks went to meta-shift stuff
> with combinations of shifts.
> Then the ALT shift/modifier was added.
> And the Windows key, or the open-Apple/closed-Apple shift.
> On the one hand, each modifier essentially adds a keyboard-of-function-keys,
> but the combinations of that with mouse buttons drives me mad!

I think your timeline is messed up. (For one, "windows key" is a
johnny-come-lately.)

Plus, when you move from a text terminal to a GUI console, adding another
modifier key makes sense. You still want the control key to go into the
existing terminal, but what key are you going to use for the (GUI) OS?

> Oh yea, and what about the ever-increasing number of mouse buttons
> from Apple's 1 and M$'s 2 "that's enough for anyone"
> to 3 (well, X-terms always used that well).
> But with the scroll wheel that tilts left/right and clicks down,
> I count 7-8 buttons on today's well-endowed mouse!

Mice started with three buttons, or had three within a short period
of time.

Personally, I think the scroll wheel makes for a *lousy* button. As
do those "little" buttons they sprinkle all over. :(

I think the additional buttons are also a side-effect from the gaming
folks[0], who do have a need for lots of different ways to initiate
in-game actions with a single button push.

[snip]

> I've found some keyboards with a LCD across the top
> that could be used for dynamic labelling the top row of keys,
> and there's a proposed keyboard where every key is a tiny LCD.

Oh, there are *all* sorts of amusing things you can do with that.

If they ever arrive.

[0] Twitch-gamers, that is, where there isn't a lot of time to type in
commands.
--
------------------------------------------------------------------------------
A diagnostic is someone who doesn't | Stewart John Stremler
know whether there are two gods. | stre...@rohan.sdsu.edu

Walter Bushell

unread,

Nov 16, 2006, 1:03:42 PM11/16/06

to

In article <ejhjlj$kfm$1...@panix5.panix.com>,
je...@panix.com (Jeff Jonas) wrote:

> Thank you for speaking up.
> When I saw keyboards with more than 12 special function keys,
> I wondered "who can actually MEMORIZE so many actions,
> regardless of the repetition?"

The people who use them, find it useful and the go into muscle memory.

--
"The power of the Executive to cast a man into prison without formulating any
charge known to the law, and particularly to deny him the judgement of his
peers, is in the highest degree odious and is the foundation of all totali-
tarian government whether Nazi or Communist." -- W. Churchill, Nov 21, 1943

Random832

unread,

Nov 16, 2006, 1:11:24 PM11/16/06

to

2006-11-16 <eji732$2fe$1...@gondor.sdsu.edu>,

stre...@rohan.sdsu.edu wrote:
> begin quoting Jeff Jonas <je...@panix.com> :

>> It's like the cash registers with a button for every possible item
>> (such as fast-food joints): it reduces everyone to hunt-and-peck
>> (well, except for really frequently-used-buttons along the edge).
>
> ...and god help you if you want an option that the designer didn't
> think of. ("What do you mean, 'no pickles'? You're allergic to
> cucumbers? I don't have an option for that. No, I can't just tell
> the cook not to put on pickles. Yes, I know he's right over there.
> I don't have an option for 'no pickles'...")

"I want you to hold 'em between your knees."

> I think your timeline is messed up. (For one, "windows key" is a
> johnny-come-lately.)

Plus, 90% of users don't know it's a modifier.

>> Oh yea, and what about the ever-increasing number of mouse buttons
>> from Apple's 1 and M$'s 2 "that's enough for anyone"
>> to 3 (well, X-terms always used that well).
>> But with the scroll wheel that tilts left/right and clicks down,
>> I count 7-8 buttons on today's well-endowed mouse!

You can't really count the scrolling directions as "buttons" - they're
mapped to buttons due to a deficiency in x11.

> I think the additional buttons are also a side-effect from the gaming
> folks[0], who do have a need for lots of different ways to initiate
> in-game actions with a single button push.

eh. what happened to using actual game controllers instead of the
mouse?

stre...@rohan.sdsu.edu

unread,

Nov 16, 2006, 2:28:52 PM11/16/06

to

begin quoting Random832 <ran...@random.yi.org> :

> 2006-11-16 <eji732$2fe$1...@gondor.sdsu.edu>,
> stre...@rohan.sdsu.edu wrote:

[chop]

>> I think the additional buttons are also a side-effect from the gaming
>> folks[0], who do have a need for lots of different ways to initiate
>> in-game actions with a single button push.
>
> eh. what happened to using actual game controllers instead of the
> mouse?

Too small, too few inputs / too many inputs too close together, and
joysticks aren't always as good as mice or trackballs.

--
----------------...@rohan.sdsu.edu----------------------------
It takes altogether too long to compose a reply if I have
to stop laughing first. -- Charlie Gibbs (July 2001)

Peter Flass

unread,

Nov 16, 2006, 5:37:23 PM11/16/06

to

Random832 wrote:

> 2006-11-16 <eji732$2fe$1...@gondor.sdsu.edu>,
> stre...@rohan.sdsu.edu wrote:
>>I think your timeline is messed up. (For one, "windows key" is a
>>johnny-come-lately.)
>
>
> Plus, 90% of users don't know it's a modifier.
>

Sure is, turns anything you type to cr@p.

Carl Pearson

unread,

Nov 16, 2006, 11:59:39 PM11/16/06

to

stre...@rohan.sdsu.edu wrote:

> Think of those modern games... all those extra keys are, in theory,
> useful.

Left, Right, Fire. Maybe a few others - Thrust/Up/Down - optional, if
required for simulated multi-dimensional movement.

What else would one need? ;)

Went to Best Buy today looking for a new VoIP box (turned out the old
router was going flaky). Long line of what looked like homeless folks
stretching down the exterior wall.

Figured it out, though. Been a bit chilly these past few days. Then I
snapped to the fact that most homeless folks don't have new looking
fold-up chairs & blankies, *clean* parkas or cell phones, separately or
in combination.

Not to mention there wasn't a single sign amongst the lot praising Ghu
or mentioning that "anything helps".

Yeah, PS3 starts selling here in the States tomorrow.

Interesting that gaming seems to drive the consumer demand for greater
horsepower (sure, you could argue for graphics, but nifty graphics are
the big draw in gaming, and there are a helluva lot more recreational
gamers than there are full-time graphic/cad employees), while the
average Joe never considers just how much big iron it takes to let any
credit card swipe on demand...

Still wonder what that game was that I played in the latter 70's in
Champaign, Illinois one weekend, a stand-up game sort of like Asteroids
but in a funky molded purple case, with square buttons and two enemy
ships that would come across the screen in parallel. It was before
Asteroids (at least, I saw it first), but similar.

grey...@gmaildo.tocom

unread,

Nov 17, 2006, 5:16:29 AM11/17/06

to

On Fri, 17 Nov 2006 04:59:39 GMT, Carl Pearson wrote:
> stre...@rohan.sdsu.edu wrote:
>
>> Think of those modern games... all those extra keys are, in theory,
>> useful.
>
> Left, Right, Fire. Maybe a few others - Thrust/Up/Down - optional, if
> required for simulated multi-dimensional movement.
>
> What else would one need? ;)
>
> Went to Best Buy today looking for a new VoIP box (turned out the old
> router was going flaky). Long line of what looked like homeless folks
> stretching down the exterior wall.
>
> Figured it out, though. Been a bit chilly these past few days. Then I
> snapped to the fact that most homeless folks don't have new looking
> fold-up chairs & blankies, *clean* parkas or cell phones, separately or
> in combination.
>
> Not to mention there wasn't a single sign amongst the lot praising Ghu
> or mentioning that "anything helps".
>
> Yeah, PS3 starts selling here in the States tomorrow.
>

And delayed in Yurop for a while, rumor hath it that Japanese ones a
re selling on eBay for 3xoriginal.

--
greymaus
Just Another Grumpy Old Man

Steve O'Hara-Smith

unread,

Nov 17, 2006, 2:11:05 AM11/17/06

to

On Fri, 17 Nov 2006 04:59:39 GMT

Carl Pearson <jman_spam...@gmail.com> wrote:

> stre...@rohan.sdsu.edu wrote:
>
> > Think of those modern games... all those extra keys are, in theory,
> > useful.
>
> Left, Right, Fire. Maybe a few others - Thrust/Up/Down - optional, if
> required for simulated multi-dimensional movement.
>
> What else would one need? ;)

Some of my son's games use every button on his Playstation controls
and combinations of them too, I've watched him going through the stages
where they are introduced and lost track really fast.

--
C:>WIN | Directable Mirror Arrays
The computer obeys and wins. | A better way to focus the sun
You lose and Bill collects. | licences available see
| http://www.sohara.org/

Charlie Gibbs

unread,

Nov 17, 2006, 11:06:40 AM11/17/06

to

In article <%kb7h.33975$rG.3...@tornado.texas.rr.com>,
jman_spam...@gmail.com (Carl Pearson) writes:

> Still wonder what that game was that I played in the latter 70's
> in Champaign, Illinois one weekend, a stand-up game sort of like
> Asteroids but in a funky molded purple case, with square buttons
> and two enemy ships that would come across the screen in parallel.
> It was before Asteroids (at least, I saw it first), but similar.

I think it's called "Computer Space". Next time you watch
Soylent Green, look for one in a corner of that futuristic
apartment with the "furniture girls".

--
/~\ 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!

Steve O'Hara-Smith

unread,

Nov 17, 2006, 11:52:31 AM11/17/06

to

On 17 Nov 2006 10:16:29 GMT
grey...@gmaildo.tocom wrote:

Ouch! Given the price that they were supposed to be selling for
there must be some people who are *really* keen to get one. When I saw the
signs up in Smyths saying they were taking deposits for advance orders I
asked what the selling price was going to be, and decided there and then
that my kids could do without one.

jsa...@ecn.ab.ca

unread,

Nov 20, 2006, 12:12:07 PM11/20/06

to

I wrote:
> but we're
> getting to the point where the IBM engineer could no longer go down the
> hall, but would instead have to leave the building.

I've been trying to make some sense out of the current explosion in
high performance architectures. Mainly, I'm trying to decide if it
would be worthwhile to buy a PlayStation 3 once it becomes actually
possible to find one in a store without camping out overnight in front
of it.

So, in the newsgroup comp.arch, I had compiled the following from some
web searching:

(first part)

The Grape-DR chip has 512 partial arithmetic units (i.e., adders or
multipliers), 300 million transistors, and measures 289 square
millimeters. It is manufactured with a 90nm process. It runs at 500
MHz, with 60 watts power consumption.

The CELL microprocessor has one main processor, and eight synergystic
processor units, each of which has AltiVec-style capability (so two
double-precision, four single-precision, ganged arithmetic units in
each) 234 million transistors, 221 square millimeters. It was made
using a 90nm process. The main processor runs at 3.2 GHz. As used in
the PlayStation 3, units with one failed SPU are fully acceptable.
Apparently its power consumption is 80 watts.

The original Itanium 2 had 3 megabytes of cache, 221 million
transistors, and was made using a 180nm process on a die of 421 square
millimeters, and consumed 130 watts of power.

A dual-core Itanium 2 with 1.72 billion transistors has just been
announced.

(less detailed stuff I added later)

The vector processor chip used in the NEC SX-6 supercomputer was
implemented with a 150nm process, and it contained 60 million
transistors. It provided 8 Gigaflop performance. This chip didn't use
any cache, but had a very fast bus to memory, so that the external RAM
had to be 2,048-way interleaved, thereby providing cache-like speeds,
at least when accessed sequentially - although accessed randomly, it
would be very nearly as good.

The SX-8 apparently uses a 90nm chip with 16 Gflop performance.

The CELL processor in a PlayStation 3 achieves 15 Gflop performance in
double precision - and 204 Gflops in single precision. But it doesn't
have a giant bandwidth to RAM - instead, the eight SPUs aren't even
ported to RAM. Getting full performance out of an SX-6 places _some_
restrictions on the type of program you have too, but the restrictions
with the CELL processor are, I would suspect, somewhat more severe.

And then there's the ClearSpeed CSX600, which is a 250 MHz chip, with
25 Gflop performance, and a power consumption of only 10 watts. There
seem to be many high performance alternatives to consider.

For comparison, a 3.6 GHz Pentium 4 is able to achieve over 3 Gflop
performance - and this is on benchmarks, while the other figures are
theoretical maximums.

John Savard

jsa...@ecn.ab.ca

unread,

Nov 20, 2006, 10:50:14 PM11/20/06

to

Anne & Lynn Wheeler wrote:
> The Future of CPUs: What's After Multi-Core?

> http://www.informit.com/articles/article.asp?p=663085&rl=1

As for the question that forms the name of that article, and the title
of the thread...

I'm not sure there's anything *architectural* after multi-core.

Certainly, there are other possibilities, like one main processor
handling a single thread controlling a large bank of slave ALUs or
simpler processors. And then, as we have even *more* gates available,
we could have multiple cores of *that* type.

But there *is* something after multi-core.

Gallium Arsenide.

Niobium.

Niobium Nitride.

Indium Phosphide and Indium Gallium Arsenide.

Diamond.

Silicon Carbide.

Right now, the other, faster materials haven't "caught up" with
silicon. But since the benefits of improvements in silicon technology
are now *limited*, so that we have to resort to multiple cores - when
the other technologies, still behind silicon in the number of gates
that can be placed on a chip, get to the point where a single processor
with hardware floating-point - and good quality floating-point
algorithms - can be used, then the latency issue will favor a *single*
core chip made of the new materials over even a 4,096-core chip using
silicon CMOS!

So that's what comes after multi-core. Single-core in some other
material from which faster transistors are made. And then the number of
transistors on a chip for that material will increase, and eventually
we go back to multi-core...

John Savard

jsa...@ecn.ab.ca

unread,

Nov 20, 2006, 11:24:21 PM11/20/06

to

I wrote:
> I'm not sure there's anything *architectural* after multi-core.
>
> Certainly, there are other possibilities, like one main processor
> handling a single thread controlling a large bank of slave ALUs or
> simpler processors. And then, as we have even *more* gates available,
> we could have multiple cores of *that* type.
>
> But there *is* something after multi-core.
>
> Gallium Arsenide.
>
> Niobium.
>
> Niobium Nitride.
>
> Indium Phosphide and Indium Gallium Arsenide.
>
> Diamond.
>
> Silicon Carbide.

Actually, there is at least *one* major advance after multi-core of an
architectural type that we are likely to see - if a need for increased
performance continues to be percieved, and if the wait for non-silicon
technologies is long enough - systems where the main memory is on the
same chip as the CPU.

After all, multi-core only exacerbates a major difficulty with
processors - one that large caches, mentioned in the article as a step
preceding multi-core tried to alleviate - the lack of memory bandwidth.

Put the memory on the chip, so you don't have to count pins, and memory
bandwidth can be colossal.

John Savard

jsa...@ecn.ab.ca

unread,

Nov 20, 2006, 11:38:53 PM11/20/06

to

I wrote:

> But there *is* something after multi-core.
>
> Gallium Arsenide.
>
> Niobium.
>
> Niobium Nitride.
>
> Indium Phosphide and Indium Gallium Arsenide.
>
> Diamond.
>
> Silicon Carbide.

I forgot Silicon Germanium.

John Savard

Gene Wirchenko

unread,

Nov 21, 2006, 12:19:17 AM11/21/06

to

jsa...@ecn.ab.ca wrote:

[snip]

>Actually, there is at least *one* major advance after multi-core of an
>architectural type that we are likely to see - if a need for increased
>performance continues to be percieved, and if the wait for non-silicon
>technologies is long enough - systems where the main memory is on the
>same chip as the CPU.

I have been wondering for *years* why this has not been done
already. Could you please explain?

[snip]

Sincerely,

Gene Wirchenko

Computerese Irregular Verb Conjugation:
I have preferences.
You have biases.
He/She has prejudices.

Brian Inglis

unread,

Nov 21, 2006, 1:36:26 AM11/21/06

to

On Mon, 20 Nov 2006 21:19:17 -0800 in alt.folklore.computers, Gene
Wirchenko <ge...@ocis.net> wrote:

>jsa...@ecn.ab.ca wrote:
>
>[snip]
>
>>Actually, there is at least *one* major advance after multi-core of an
>>architectural type that we are likely to see - if a need for increased
>>performance continues to be percieved, and if the wait for non-silicon
>>technologies is long enough - systems where the main memory is on the
>>same chip as the CPU.
>
> I have been wondering for *years* why this has not been done
>already. Could you please explain?

Has been done often: embedded processors.
How small a main memory can you tolerate running your workload in?
Or would those transistors be better used for a high speed cache in
front of main memory: answer always yes if using space for cache is
more effective than for main memory.
Then you have diminishing cash^Wcache returns, and time/space
(transistor budget) tradeoffs.
Control and data dependencies create the idle time, multi-threaded
CPUs take up some of the slack, but obviously not enough, otherwise
they'd add more threads to use up idle time, rather than more cores
using up space and transistors.
Remember Amdahl's law: 80% serial, 20% parallel, if lucky: obviously
not lucky enough, normally.

Andrew Swallow

unread,

Nov 21, 2006, 2:40:50 AM11/21/06

to

Brian Inglis wrote:
> On Mon, 20 Nov 2006 21:19:17 -0800 in alt.folklore.computers, Gene
> Wirchenko <ge...@ocis.net> wrote:
>
>> jsa...@ecn.ab.ca wrote:
>>
>> [snip]
>>
>>> Actually, there is at least *one* major advance after multi-core of an
>>> architectural type that we are likely to see - if a need for increased
>>> performance continues to be percieved, and if the wait for non-silicon
>>> technologies is long enough - systems where the main memory is on the
>>> same chip as the CPU.
>> I have been wondering for *years* why this has not been done
>> already. Could you please explain?
>
> Has been done often: embedded processors.
> How small a main memory can you tolerate running your workload in?
> Or would those transistors be better used for a high speed cache in
> front of main memory: answer always yes if using space for cache is
> more effective than for main memory.

The hardware has to cache pages rather than bytes. This requires a
new interface to the operating system. Each page has to be relocatable
so there has to be an index somewhere. Separate indexes for each
program, shared library routines and shared data area and a high
level index of indexes (pointers to programs). All have to be kept up
to date.

Capability machines can spend more time calculating and checking
addresses than actually working, although additional adders allows
address calculations and say multiplication to be performed in
parallel.

Andrew Swallow

Inter-program calls need very long names so that they can be unique.

jsa...@ecn.ab.ca

unread,

Nov 21, 2006, 7:53:03 AM11/21/06

to

Gene Wirchenko wrote:
> jsavard wrote:

> >Actually, there is at least *one* major advance after multi-core of an
> >architectural type that we are likely to see - if a need for increased
> >performance continues to be percieved, and if the wait for non-silicon
> >technologies is long enough - systems where the main memory is on the
> >same chip as the CPU.
>
> I have been wondering for *years* why this has not been done
> already. Could you please explain?

The first reason has to do with the economies of chip manufacture.

Wafers have bad spots on them. So, the bigger the chip you make, the
more likely you are to find a bad spot in your chip.

The second reason is historical.

It *has* been thought that with DRAM, as long as the bad spots aren't
in certain critical areas, one might well be able to route around the
bad sections. This has been attempted. It was called "wafer-scale
integration", and as it was tried before the technology was quite
ready, a lot of people lost a lot of money on it.

Investment capital, therefore, is hard to come by for anything vaguely
resembling that. It also doesn't help that since the idea was already
tried (when the technology was not ready) it isn't possible to patent
the idea *now* and tie it up.

The third reason has to do with how computers are marketed, or how big
chips are.

Another poster already noted that it has been "done already". For
example, there are one-chip programmable pocket calculators; the one
chip has a processor and ROM for its program and RAM for your program.
Such devices are called microcontrollers, for example.

If you make the processor and the RAM small enough, then you can put
them on a chip with good yields. But a small processor by *today's*
standards won't run *current* operating systems, which require a
processor of relatively recent vintage.

Eventually, when ideas like multi-core begin to *run out of steam*,
chip sizes will be big enough that a powerful CPU and an appreciable
amount of RAM can be put on one chip. Then, finally, putting the RAM on
the chip becomes a way to *increase* performance, instead of something
that is done for small-scale systems only.

John Savard

Stan Barr

unread,

Nov 21, 2006, 1:35:16 PM11/21/06

to

On Mon, 20 Nov 2006 21:19:17 -0800, Gene Wirchenko <ge...@ocis.net> wrote:
>jsa...@ecn.ab.ca wrote:
>
>[snip]
>
>>Actually, there is at least *one* major advance after multi-core of an
>>architectural type that we are likely to see - if a need for increased
>>performance continues to be percieved, and if the wait for non-silicon
>>technologies is long enough - systems where the main memory is on the
>>same chip as the CPU.
>
> I have been wondering for *years* why this has not been done
>already. Could you please explain?

When I asked this question (in connection with multiple processors per
chip) the answer I was given was that the fab processes for the processor
technology and for dynamic memory are different and incompatible.
Static memory of some sort can be done (viz caches) but takes up more
real estate.

I keep envisioning large arrays of simple processors with on-chip
memory and on-chip dedicated communications hardware.

--
Cheers,
Stan Barr stanb .at. dial .dot. pipex .dot. com
(Remove any digits from the addresses when mailing me.)

The future was never like this!

krw

unread,

Nov 21, 2006, 3:19:14 PM11/21/06

to

In article <slrnem5gdb....@citadel.metropolis.local>,
sta...@dial.pipex.com says...

> On Mon, 20 Nov 2006 21:19:17 -0800, Gene Wirchenko <ge...@ocis.net> wrote:
> >jsa...@ecn.ab.ca wrote:
> >
> >[snip]
> >
> >>Actually, there is at least *one* major advance after multi-core of an
> >>architectural type that we are likely to see - if a need for increased
> >>performance continues to be percieved, and if the wait for non-silicon
> >>technologies is long enough - systems where the main memory is on the
> >>same chip as the CPU.
> >
> > I have been wondering for *years* why this has not been done
> >already. Could you please explain?
>
> When I asked this question (in connection with multiple processors per
> chip) the answer I was given was that the fab processes for the processor
> technology and for dynamic memory are different and incompatible.

It normally isn't done, bit there is no reason it couldn't be
done (no reason to, so it's not). Embedded DRAM has been around for
a while.

> Static memory of some sort can be done (viz caches) but takes up more
> real estate.

About 6x. Maybe MRAM will change things.

> I keep envisioning large arrays of simple processors with on-chip
> memory and on-chip dedicated communications hardware.

I don't think you're going to see on-die main memory (in the high-
end processor market) for a while. Current needs are for far more
memory than one could reasonably integrate. Memory requirements
aren't going down.

--
Keith

Tim McCaffrey

unread,

Nov 22, 2006, 12:27:32 PM11/22/06

to

In article <MPG.1fcd25efd...@news.individual.net>, k...@att.bizzzz
says...

There is/was a company (Orbital?) that had an article in Electronic Products
several years ago describing a stacked DRAM package, they broke the
digital/analog interface into a separate chip, and layered the pure DRAM chips
on top of that. They claimed that they could stack up to 32 DRAM chips, and
were getting 4ns access time (and 250 Mhz transfer rate). Now, this was
before the Opteron came out, but the obvious next step is to stack the DRAM
interface/modules right on top of the CPU. 32 die with current technology
gets you 4 Gigabytes, and at 4ns (although I assume it would improve), the
access latency is very close to L2 cache. Remove the L2 cache and you could
probably add another CPU core (or graphics core like AMD/ATI is doing).

- Tim

NOT speaking for Unisys.

Charles Shannon Hendrix

unread,

Nov 22, 2006, 5:23:24 PM11/22/06

to

On 2006-11-16, Random832 <ran...@random.yi.org> wrote:

> eh. what happened to using actual game controllers instead of the
> mouse?

The mouse is far better than a game controller for a lot of games.

--
shannon "AT" widomaker.com -- ["An Irishman is never drunk as long as he
can hold onto one blade of grass and not fall off the face of the earth."]

Charles Shannon Hendrix

unread,

Nov 22, 2006, 5:43:36 PM11/22/06

to

On 2006-11-21, Brian Inglis <Brian....@SystematicSW.Invalid> wrote:

> Remember Amdahl's law: 80% serial, 20% parallel, if lucky: obviously
> not lucky enough, normally.

Amdahl's Law really only applies to single programs though, and that's
not how modern systems are working any more.

Just for example, systems like KDE. It is highly parallel because each
desktop function and all IPC is handled by a kio or other task/thread.
It's probably 50% parallel just sitting mostly idle.

Put a full load of applications on it, and your system is highly
parallel, so SMP is a win even on a desktop.

Also, we are finally starting to see more parallel applications.

Games and graphics applications are the primary proving ground for this
on the desktop.

Some things will never be parallel, but as desktops increase their
workloads, you can say that most of your overall load will be parallel.

Granted, in some ways that means we are running parallel bloat, but
that's a different argument... :)

stre...@rohan.sdsu.edu

unread,

Nov 23, 2006, 4:58:54 AM11/23/06

to

begin quoting Charles Shannon Hendrix <sha...@news.widomaker.com> :

> On 2006-11-16, Random832 <ran...@random.yi.org> wrote:

>> eh. what happened to using actual game controllers instead of the
>> mouse?

> The mouse is far better than a game controller for a lot of games.

Shanghai, Klondike, Lemmings, ...

--
--Stewart Stremler----------------...@rohan.sdsu.edu--
"I went to this computer conference down in Texas over the weekend. Those
computer guys are different. I mean, they told all these jokes, and they
*WEREN'T EVEN FUNNY!* -- Some Math Professor (according to Keith G. Murphy)

krw

unread,

Nov 23, 2006, 11:10:43 AM11/23/06

to

In article <ski2ke...@escape.goid.lan>,
sha...@news.widomaker.com says...

> On 2006-11-16, Random832 <ran...@random.yi.org> wrote:
>
> > eh. what happened to using actual game controllers instead of the
> > mouse?
>
> The mouse is far better than a game controller for a lot of games.

My son brought his Nintendo Wii over last night. That game
controller will beat any mouse!

--
Keith

krw

unread,

Nov 23, 2006, 11:12:34 AM11/23/06

to

In article <ek21a4$447$2...@trsvr.tr.unisys.com>,
t...@spamfilter.asns.tr.unisys.com says...

How do you get the heat out through all those DRAM chips?

--
KEith

Richard Steiner

unread,

Nov 23, 2006, 11:57:59 PM11/23/06

to

Here in alt.folklore.computers, stre...@rohan.sdsu.edu spake unto us, saying:

>begin quoting Charles Shannon Hendrix <sha...@news.widomaker.com> :
>

>> The mouse is far better than a game controller for a lot of games.
>
>Shanghai, Klondike, Lemmings, ...

Quake, Tribes, Unreal Tournament, ...

Some folks can do a 180-degree spin-and-fire so quickly with a mouse
(via wrist snap) that you can't react to it. :-) I used to have a
good 180 move in Doom, but I've since lost my edge...

--
-Rich Steiner >>>---> http://www.visi.com/~rsteiner >>>---> Mableton, GA USA
Mainframe/Unix bit twiddler by day, OS/2+Linux+DOS hobbyist by night.
WARNING: I've seen FIELDATA FORTRAN V and I know how to use it!
The Theorem Theorem: If If, Then Then.

jmfb...@aol.com

unread,

Nov 24, 2006, 9:28:24 AM11/24/06

to

In article <ski2ke...@escape.goid.lan>,

Charles Shannon Hendrix <sha...@news.widomaker.com> wrote:
>On 2006-11-16, Random832 <ran...@random.yi.org> wrote:
>
>> eh. what happened to using actual game controllers instead of the
>> mouse?
>
>The mouse is far better than a game controller for a lot of games.

[spluttering emoticon here] Are you nuts? [emoticon wiping
oatmeal off TTY]

The only game controller that is useful is the keyboard.

/BAH

>
>

jmfb...@aol.com

unread,

Nov 24, 2006, 9:30:11 AM11/24/06

to

In article <XvnZFpHp...@visi.com>,

rste...@visi.com (Richard Steiner) wrote:
>Here in alt.folklore.computers, stre...@rohan.sdsu.edu spake unto us,
saying:
>
>>begin quoting Charles Shannon Hendrix <sha...@news.widomaker.com> :
>>
>>> The mouse is far better than a game controller for a lot of games.
>>
>>Shanghai, Klondike, Lemmings, ...
>
>Quake, Tribes, Unreal Tournament, ...
>
>Some folks can do a 180-degree spin-and-fire so quickly with a mouse
>(via wrist snap) that you can't react to it. :-) I used to have a
>good 180 move in Doom, but I've since lost my edge...
>

What do they call injuries? Rodent Tunnel Syndrome?

/BAH

krw

unread,

Nov 24, 2006, 10:02:06 AM11/24/06

to

In article <ek6vlj$8qk...@s989.apx1.sbo.ma.dialup.rcn.com>,
jmfb...@aol.com says...

In a shoot-em-up? I think they're called "casualties". ;-)

--
Keith

krw

unread,

Nov 24, 2006, 10:08:57 AM11/24/06

to

In article <ek6vi8$8qk...@s989.apx1.sbo.ma.dialup.rcn.com>,
jmfb...@aol.com says...

If you';re ever near a game store, check out the Nintendo
controller. It's a blast! My son and I played baseball with a
couple of controllers. The pitcher winds up and throws the ball
and the batter swings the controller like a bat (they have a wrist
strap so one doesn't launch them through windows). I didn't play
much (too late) but the controller seems to know where it is in
6-space (likely dx/dt, dy/dt, and dz/dt, and roll, pitch and yaw)
and adjusts the pitch/swing accordingly. It would be hard to do
that on a keyboard.

The only problem we had was almost beaning each other while
swinging. ;-)

--
Keith

Richard Steiner

unread,

Nov 25, 2006, 12:38:25 AM11/25/06

to

Here in alt.folklore.computers, jmfb...@aol.com spake unto us, saying:

I don't know, actually. My wrist gets tired when I've played that type
of game for several hours straight (usually at a LAN party), but I've
never injured myself playing such games, and I find that my eyes tend
to dry out before my wrist becomes unusable. :-)

We've had folks play with keyboard only and with various other types of
controllers, though, and nothing seems to touch the mouse in games that
use a mouse+keyboard combination for movement and sight/weapon aiming.

jmfb...@aol.com

unread,

Nov 25, 2006, 10:38:59 AM11/25/06

to

In article <MPG.1fd0d518c...@news.individual.net>,

krw <k...@att.bizzzz> wrote:
>In article <ek6vi8$8qk...@s989.apx1.sbo.ma.dialup.rcn.com>,
>jmfb...@aol.com says...
>> In article <ski2ke...@escape.goid.lan>,
>> Charles Shannon Hendrix <sha...@news.widomaker.com> wrote:
>> >On 2006-11-16, Random832 <ran...@random.yi.org> wrote:
>> >
>> >> eh. what happened to using actual game controllers instead of the
>> >> mouse?
>> >
>> >The mouse is far better than a game controller for a lot of games.
>>
>> [spluttering emoticon here] Are you nuts? [emoticon wiping
>> oatmeal off TTY]
>>
>> The only game controller that is useful is the keyboard.
>
>If you';re ever near a game store, check out the Nintendo
>controller. It's a blast!

I always intend to. The systems are always dead.

> My son and I played baseball with a
>couple of controllers. The pitcher winds up and throws the ball
>and the batter swings the controller like a bat (they have a wrist
>strap so one doesn't launch them through windows).

Which windows?

> I didn't play
>much (too late) but the controller seems to know where it is in
>6-space (likely dx/dt, dy/dt, and dz/dt, and roll, pitch and yaw)
>and adjusts the pitch/swing accordingly. It would be hard to do
>that on a keyboard.

Slanging keyboards instead of slanging hash?

I wonder how much of that 6-space came out of those MIT projects?

>
>The only problem we had was almost beaning each other while
>swinging. ;-)

I think what I'll do is try to watch somebody else play.
Thanks for the pointer.

/BAH

jmfb...@aol.com

unread,

Nov 25, 2006, 10:40:38 AM11/25/06

to

In article <Rb9ZFpHp...@visi.com>,

rste...@visi.com (Richard Steiner) wrote:
>Here in alt.folklore.computers, jmfb...@aol.com spake unto us, saying:
>
>>In article <XvnZFpHp...@visi.com>,
>> rste...@visi.com (Richard Steiner) wrote:
>>
>>>Quake, Tribes, Unreal Tournament, ...
>>>
>>>Some folks can do a 180-degree spin-and-fire so quickly with a mouse
>>>(via wrist snap) that you can't react to it. :-) I used to have a
>>>good 180 move in Doom, but I've since lost my edge...
>>
>>What do they call injuries? Rodent Tunnel Syndrome?
>
>I don't know, actually. My wrist gets tired when I've played that type
>of game for several hours straight (usually at a LAN party), but I've
>never injured myself playing such games, and I find that my eyes tend
>to dry out before my wrist becomes unusable. :-)
>
>We've had folks play with keyboard only and with various other types of
>controllers, though, and nothing seems to touch the mouse in games that
>use a mouse+keyboard combination for movement and sight/weapon aiming.
>

Would you have to strap the keyboard to your leg?

/BAH

Richard Steiner

unread,

Nov 25, 2006, 6:21:23 PM11/25/06

to

Here in alt.folklore.computers, jmfb...@aol.com spake unto us, saying:

>In article <Rb9ZFpHp...@visi.com>,
> rste...@visi.com (Richard Steiner) wrote:
>
>>Here in alt.folklore.computers, jmfb...@aol.com spake unto us, saying:

>>We've had folks play with keyboard only and with various other types of
>>controllers, though, and nothing seems to touch the mouse in games that
>>use a mouse+keyboard combination for movement and sight/weapon aiming.
>
>Would you have to strap the keyboard to your leg?

Sorry, I don't understand the question. :-)

Normal layout is as follows:

The player's left hand typically uses the keyboard (placed in front and
centered) with fingers using the following map for movement:

Q W E Q = spin left W = move forward E = spin right
A S D A = strafe left S = move backward D = strafe right
X X = Crouch
SPACE SPACE = jump

Spin means rotate in place. Strafe means move left or right without
changing facing. The above is a fairly standard mapping. Some games
or players use different key mappings or variations of the above.

The mouse is used with the right hand to direct the weapon and fire it
independently of normal player movements done with the keyboard. It's
literally a point-and-shoot interface, and the crosshair indicating
weapon aiming will usually move as fast as you can physically move it.

The above combination will allow an experienced player to move/jump and
target/shoot independently, and it allows for very fast spin-and-fire
maneuvers if one has the required wrist snap down.

jmfb...@aol.com

unread,

Nov 26, 2006, 10:28:12 AM11/26/06

to

In article <z/MaFpHpv...@visi.com>,

Good grief. I always had problems playing the boogie-woogie
on the piano too.

/BAH

Richard Steiner

unread,

Nov 26, 2006, 7:28:00 PM11/26/06

to

Here in alt.folklore.computers, jmfb...@aol.com spake unto us, saying:

>rste...@visi.com (Richard Steiner) wrote:
>
>> [description of gaming wiht keyboard+mouse snipped for brevity]

>>
>>The above combination will allow an experienced player to move/jump and
>>target/shoot independently, and it allows for very fast spin-and-fire
>>maneuvers if one has the required wrist snap down.
>
>Good grief. I always had problems playing the boogie-woogie
>on the piano too.

It's less difficult than touch-typing. The left hand doesn't have to
move beyond a few keys, and the right hand just plays with the mouse.

Easier than chopsticks. :-)

Of course, doing it is one thing, but doing it WELL is quite another!

jmfb...@aol.com

unread,

Nov 27, 2006, 8:14:32 AM11/27/06

to

In article <QEjaFpHp...@visi.com>,

rste...@visi.com (Richard Steiner) wrote:
>Here in alt.folklore.computers, jmfb...@aol.com spake unto us, saying:
>
>>rste...@visi.com (Richard Steiner) wrote:
>>
>>> [description of gaming wiht keyboard+mouse snipped for brevity]
>>>
>>>The above combination will allow an experienced player to move/jump and
>>>target/shoot independently, and it allows for very fast spin-and-fire
>>>maneuvers if one has the required wrist snap down.
>>
>>Good grief. I always had problems playing the boogie-woogie
>>on the piano too.
>
>It's less difficult than touch-typing. The left hand doesn't have to
>move beyond a few keys, and the right hand just plays with the mouse.
>
>Easier than chopsticks. :-)

I can manage chopsticks with one hand. I could not manage
two pairs of chopsticks with each hand having unrelated beats
of motion.

>
>Of course, doing it is one thing, but doing it WELL is quite another!

Do you turn your body or have you been able to train the torso
and other muscles not involved in the play to be still?

/BAH

Tim McCaffrey

unread,

Nov 27, 2006, 12:55:42 PM11/27/06

to

In article <MPG.1fcf927ed...@news.individual.net>, k...@att.bizzzz
says...
>

>
>How do you get the heat out through all those DRAM chips?
>

Well, the obvious answer is don't make it in the first place (e.g. ARM).
Since I'm not a chip packager, I don't really know other methods, although
some kind of sophisticated cooling socket comes to mind.... :)

krw

unread,

Nov 27, 2006, 1:14:22 PM11/27/06

to

In article <ekf8qt$4m3$1...@trsvr.tr.unisys.com>,
t...@spamfilter.asns.tr.unisys.com says...

Even an ARM produces heat. I guess that means we're not likely to
see it as a solution then.

--
Keith

Anne & Lynn Wheeler

unread,

Nov 27, 2006, 2:49:28 PM11/27/06

to

Anne & Lynn Wheeler <ly...@garlic.com> writes:
> The Future of CPUs: What's After Multi-Core?
> http://www.informit.com/articles/article.asp?p=663085&rl=1
>
> from above (the new, 40yr old thing)
>
> This rule was driven home to me when I attended a talk by an IBM
> engineer about his company's new virtualization technology. He
> commented that his company had an advantage over other people working
> in the area: Whenever they were stuck, they could go along the hall to
> the mainframe division and ask how they solved the same problem a
> couple of decades ago.
>
> ... snip ...
>
> semi-related thread from comp.arch
> http://www.garlic.com/~lynn/2006t.html#23 threads versus task

re:
http://www.garlic.com/~lynn/2006t.html#27 The Future of CPUs: What's After Multi-Core?

... "and the prospects for parallel-processing on the desktop? .... 20
plus years ago:

Date: 30 Jan 86 15:16:34 PST (Thursday)
Subject: Xerox PARC Forum Tuesday, Feb 11, Bill Joy: "Computer Workstation Architecture: 1982-1992"
From: Kluger.o...@xerox.com
To: ArpaColloquiums^.P...@xerox.com
cc: sun!bl...@ucbvax.berkeley.edu, Kluger.o...@xerox.com
Reply-to: Kluger.o...@xerox.com
Message-ID: <860130-165216-1935@Xerox>

Xerox Palo Alto Research Center Forum

Tuesday, February 11, 1986
2:00 pm, PARC Auditorium

Bill Joy

Vice President of Research and Development
Sun Microsystems

will speak on

Computer Workstation Architecture: 1982-1992

The talk will center on the architecture of computer workstations. We
have over a decade of experience with personal workstations, starting
with the Alto and the Dorado at Xerox. Over the last four years,
workstations have come into widespread commercial use driven by the
32-bit micros, driven largely by the Motorola 680x0 family.

The presentation will look at the architecture of workstations and the
supporting network environment, and speculate forward to 1992.
Questions to be addressed include the structure of a distributed
software environment, resource sharing among heterogeneous machines, the
impact of RISC technology, support for AI languages, and the prospects
for parallel-processing on the desktop.

... snip ...

other posts in the thread ...
http://www.garlic.com/~lynn/2006t.html#31 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#32 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#34 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#36 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#41 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#42 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#43 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#49 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006t.html#50 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006u.html#0 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006u.html#6 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006u.html#7 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006u.html#8 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006u.html#9 The Future of CPUs: What's After Multi-Core?
http://www.garlic.com/~lynn/2006u.html#10 The Future of CPUs: What's After Multi-Core?

jsa...@ecn.ab.ca

unread,

Nov 28, 2006, 4:43:59 PM11/28/06

to

Eugene Miya wrote:
> Architecture, not arithmetic.
>
> You want a real serious architect to review.

Well, I know that the architecture illustrated there - even if all the
excesses were removed - is not of the currently-fashionable RISC type.

I started out with what I thought was a basic instruction set
architecture that was relatively clean - similar to that of the IBM 360
or the Motorola 68000.

Basically, I start from the System/360 instruction, and then widen the
address field from 12 bits to 16 bits by reducing the number of
registers from 16 to 8 - that saves me three bits, one from the
destination register field, one from the index and base register fields
- and trimming the opcode field to 7 bits from 8.

I get some of what I lose back by:

using a zero base register field, rather than a different opcode, to
indicate register-to-register operations, and

by using my eight general registers also as index registers, but having
a separate set of eight registers for use as base registers (compare
the address registers of the Motorola 68000).

So I end up with an instruction format like that which was used by
Seymour Cray with the Cray-1 computer.

What has made this example architecture into a disaster area is...

the addition of types such as register packed decimal, compressed
decimal, simple floating (where the mantissa and exponent are contained
in two successive integers of equal size), decimal exponent floating
point,

hundreds of different variations on the ISA, aimed at shrinking the
size of programs, with one added after another as I fail to be
satisfied with the preceding one,

obscure features, like the ability to access main memory as if it were
made of 36-bit, 40-bit, 48-bit, or 60-bit words, the ability to select
your own floating-point format, the extended translate instructions
which allow one to set up a custom 'microprogram' for special string
operations.

I assumed that some operations (i.e. packed decimal arithmetic) are so
common that they shouldn't be relegated to a (slow) FPGA, but deserve
some silicon - and part of my design is aimed at having the ability to
*efficiently* emulate (even more efficiently at, say, the FORTRAN
source level, since then I can *vectorize* old 7090 decks) just about
every computer ever made.

Like the Honeywell 800, I even have two program counters!

If a chip having these kind of features were to become dominant, then
people with old programs from classic computers wouldn't have to
convert them to use them on the popular platform!

But I have to admit that it's a little late for that now. Everybody who
used to have, say, a Unisys 1110, or a DECSystem-20, has by now
converted any code that still needs to be used to x86, Itanium, or
PowerPC. And, unlike 360 emulation (addressed with general register
mode) and the emulation of architectures with 24-bit or 40-bit words, I
neglected what might be the *biggest* area where emulation is desired -
the IBM 1401.

John Savard

Eugene Miya

unread,

Nov 30, 2006, 1:44:40 PM11/30/06

to

In article <1164750239.6...@j72g2000cwa.googlegroups.com>,

<jsa...@ecn.ab.ca> wrote:
>Eugene Miya wrote:
>> Architecture, not arithmetic.
>>
>> You want a real serious architect to review.
>
>Well, I know that the architecture illustrated there - even if all the
>excesses were removed - is not of the currently-fashionable RISC type.

I don't see RISC has fashionable. Wintel is.

>I started out with what I thought was a basic instruction set
>architecture that was relatively clean - similar to that of the IBM 360
>or the Motorola 68000.

Clean? I think there are more CISCish examples (Hugh LaMaster isn't here
to defend the 205), but I would not call either clean.

>Basically, I start from the System/360 instruction, and then widen the
>address field from 12 bits to 16 bits by reducing the number of
>registers from 16 to 8 - that saves me three bits, one from the
>destination register field, one from the index and base register fields
>- and trimming the opcode field to 7 bits from 8.

Remember: more is not necessarily better.

>I get some of what I lose back by:
>
>using a zero base register field, rather than a different opcode, to
>indicate register-to-register operations, and
>
>by using my eight general registers also as index registers, but having
>a separate set of eight registers for use as base registers (compare
>the address registers of the Motorola 68000).
>
>So I end up with an instruction format like that which was used by
>Seymour Cray with the Cray-1 computer.

You mean by count?

>What has made this example architecture into a disaster area is...
>

>... types ...

>
>hundreds of different variations on the ISA, aimed at shrinking the

...
>obscure features,...

>
>I assumed that some operations (i.e. packed decimal arithmetic) are so
>common that they shouldn't be relegated to a (slow) FPGA, but deserve
>some silicon - and part of my design is aimed at having the ability to
>*efficiently* emulate (even more efficiently at, say, the FORTRAN
>source level, since then I can *vectorize* old 7090 decks) just about
>every computer ever made.

I've not seen what compiler technology you have used, but I know what
the Convex, IBM and Japanese compilers could do. And you certain don't
want cft from the mid-70s era, so I am unable to see how you can
vectorize old 7090 decks. Much less why you would want to do such
(better left dead except for historic study which even then (despite Al)
I think may be a questionable pursuit).

>Like the Honeywell 800, I even have two program counters!

Is that assuredly good?

>If a chip having these kind of features were to become dominant, then
>people with old programs from classic computers wouldn't have to
>convert them to use them on the popular platform!

Who do you know with 7090 source much less an environment to run on it?
If it were a Cray-1, excepting myself, I have yet to encounter anyone
with an unmodified source program much less any of the half dozen Cray
OSes in an unclassified or highly proprietary location. No
climatologist of today is interested in 20 year old climate models.
And that says nothing about lacking an OS for the I/O and or a compiler.

>But I have to admit that it's a little late for that now. Everybody who
>used to have, say, a Unisys 1110, or a DECSystem-20, has by now
>converted any code that still needs to be used to x86, Itanium, or
>PowerPC. And, unlike 360 emulation (addressed with general register
>mode) and the emulation of architectures with 24-bit or 40-bit words, I
>neglected what might be the *biggest* area where emulation is desired -
>the IBM 1401.

Huh? We (CHM, Al) has one of those. Running access.

--

Charles Shannon Hendrix

unread,

Nov 30, 2006, 1:51:39 PM11/30/06

to

On 2006-11-27, jmfb...@aol.com <jmfb...@aol.com> wrote:

> Do you turn your body or have you been able to train the torso
> and other muscles not involved in the play to be still?

That is personal preference... :)

However, people standing behind you watching have a tendency to go into
all kinds of convulsions, some even have siezures.

Something about watching 1st person without being in control really
messes some people's minds up.

--
shannon "AT" widomaker.com -- ["The trade of governing has always been
monopolized by the most ignorant and the most rascally individuals of
mankind. -- Thomas Paine"]

Charles Shannon Hendrix

unread,

Nov 30, 2006, 1:52:50 PM11/30/06

to

On 2006-11-24, jmfb...@aol.com <jmfb...@aol.com> wrote:
> In article <ski2ke...@escape.goid.lan>,
> Charles Shannon Hendrix <sha...@news.widomaker.com> wrote:
>>On 2006-11-16, Random832 <ran...@random.yi.org> wrote:
>>
>>> eh. what happened to using actual game controllers instead of the
>>> mouse?
>>
>>The mouse is far better than a game controller for a lot of games.
>
> [spluttering emoticon here] Are you nuts? [emoticon wiping
> oatmeal off TTY]

If you are spraying outmeal, I'd say you are nuts... :)

> The only game controller that is useful is the keyboard.

Yeah, right.

Personally, I use a combination of keyboard and mouse for most games.

I only like console controllers for games that were well designed for
them, and sometimes not even then.

--

jsa...@ecn.ab.ca

unread,

Nov 30, 2006, 10:59:44 PM11/30/06

to

Eugene Miya wrote:
> quoting me:

> >Well, I know that the architecture illustrated there - even if all the
> >excesses were removed - is not of the currently-fashionable RISC type.
.
> I don't see RISC has fashionable. Wintel is.

You can certainly argue that RISC has genuine merit, while the x86
architecture does not.

However, I don't think you can call the x86 architecture 'fashionable'.
Its popularity is not because people mistakenly think it's a good thing
for a 64-bit machine to have an architecture contorted by a history
going back to 8-bit processors. Nobody *likes* the x86 architecture.
They only like *the software that can run on it*.

If the x86 style of architecture were fashionable, other people
designing new chips which were not necessarily compatible would design
them to be similar.

> >I started out with what I thought was a basic instruction set
> >architecture that was relatively clean - similar to that of the IBM 360
> >or the Motorola 68000.
.
> Clean? I think there are more CISCish examples (Hugh LaMaster isn't here
> to defend the 205), but I would not call either clean.

Let's just say that the x86 suffers by comparison with either of them.
That may be faint praise indeed, of course.

> >Basically, I start from the System/360 instruction, and then widen the
> >address field from 12 bits to 16 bits by reducing the number of
> >registers from 16 to 8 - that saves me three bits, one from the
> >destination register field, one from the index and base register fields
> >- and trimming the opcode field to 7 bits from 8.
.
> Remember: more is not necessarily better.

I think that 64K-byte pages of contiguous memory *are* better than
4,096-byte pages of contiguous memory. Their cost is a limit on the
number of general registers, but eight general registers of each type
is not unreasonably low for a CISC architecture.

> >So I end up with an instruction format like that which was used by
> >Seymour Cray with the Cray-1 computer.
>
> You mean by count?

Only similar in the most superficial way - the widths of the various
fields. The Cray-1 was, of course, a RISC machine.

> I've not seen what compiler technology you have used, but I know what
> the Convex, IBM and Japanese compilers could do. And you certain don't
> want cft from the mid-70s era, so I am unable to see how you can
> vectorize old 7090 decks. Much less why you would want to do such
> (better left dead except for historic study which even then (despite Al)
> I think may be a questionable pursuit).

Basically, I've heard that FORTRAN compilers exist that can read old
code with DO loops in it, written for FORTRAN 77 (and, hence, maybe for
earlier FORTRANs too), and figure out which things could be done using
vector instructions, even though the code wasn't written in Fortran 95,
say.

There may, or may not, still be some programs out there which were
useful despite never having been fully converted.

When the 7090 gave way to the 360, many people complained that the
decrease in floating-point precision forced them to use
double-precision for scientific calculations for which 36-bit
floating-point was adequate, but 32-bit floating-point was not.

The IBM 360 floating-point format did have its deficiencies, and IEEE
754 squeezes somewhat more precision out of 32 bits. However, even so,
I think that it may well be beneficial to have a computer architecture
which allows more choices in the lengths of data items. If they don't
have to be any longer than necessary, then you can have more numbers in
RAM at once.

This is exclusive of any *direct* translations from computers of an
earlier era.

> >Like the Honeywell 800, I even have two program counters!
>
> Is that assuredly good?

It simply lets me explain how the Honeywell 800 worked, and what they
might have been thinking of when they included that feature.

> No
> climatologist of today is interested in 20 year old climate models.
> And that says nothing about lacking an OS for the I/O and or a compiler.

I figure something could be cobbled together to compile old 7090
FORTRAN; my assumption is that the *big* dependencies are on the data
formats used, not on the precise form of the PRINT statements.

And some code - say, lens design programs - has retained its relevance
better.

> >But I have to admit that it's a little late for that now. Everybody who
> >used to have, say, a Unisys 1110, or a DECSystem-20, has by now
> >converted any code that still needs to be used to x86, Itanium, or
> >PowerPC. And, unlike 360 emulation (addressed with general register
> >mode) and the emulation of architectures with 24-bit or 40-bit words, I
> >neglected what might be the *biggest* area where emulation is desired -
> >the IBM 1401.

> Huh? We (CHM, Al) has one of those. Running access.

As to the "Huh?", it is well known that some computing tasks got
translated fairly directly from tab equipment plugboards to 1401 RPG,
and then run on a 360 emulating a 1401... and, as long as they worked,
continued to be used an indecently long time. (Of course, now we are
after the year 1999, which means *some* of those programs have lost
their relevance...)

John Savard

jsa...@ecn.ab.ca

unread,

Dec 1, 2006, 6:47:20 AM12/1/06

to

Eugene Miya wrote:
> quoting me:

> >Like the Honeywell 800, I even have two program counters!

> Is that assuredly good?

Perhaps I should clarify matters further.

My design is perhaps "silly" on three different levels, some more
defensible than others.

1) It includes features that have appeared perhaps on one or two
computers in all history, and which never proved very useful.

- Two program counters, like the Honeywell 800. It probably helped with
some sorting routines, and acted as a cheap substitute for good
interrupt logic.

- A floating-point type where the mantissa and exponent are two
equal-sized integers, like the Recomp II. This illustrates how the
fixed-point arithmetic unit could be pressed into helping solve a
massive floating-point problem in a pinch. Anything to be a few extra
ways superscalar!

- A floating-point type where the mantissa is binary, but the exponent
is a power of 10, as done in software for JOSS.

These things are just there to illustrate computer concepts, and make
the pages educational.

The ability to access memory as if it were made up of 36-bit or 48-bit
words, even though it's really in the conventional 32/64 bit
organization, without wasting a bit of main memory, is, as far as I
know, novel - and it illustrates that it can be done, and how it can be
done.

2) It includes features that are genuinely useful, on some types of
computers.

It includes packed decimal arithmetic and character string
instructions. Those were genuinely useful on mainframes, which would
have to process tedious loops that dealt with 8 bits at a time, on an
architecture that could be operating on a whole 64-bit floating-point
number in one instruction.

It includes the option of flagging each instruction with a bit
indicating that it can or can't be executed in parallel with the
preceding instruction, like some popular DSP chips.

It includes explicit vector processing instructions, as found on many
supercomputers.

It includes parallel vector instructions, like MMX or AltiVec.

What's weird, though, is to have all these features on the *same*
computer. Anything with supercomputer-like vector instructions is
probably going to have a full set of pipeline interlocks.

And if you have an MMX type unit, with several parallel floating-point
units, wouldn't they be the ones that are pipelined, and shouldn't the
pipeline vector instructions work with them?

This gives me a chance to illustrate circuitry that lets a massive set
of parallel floating-point units behave like a single pipelined one,
only faster. (Basically, you want to ship a contiguous block of data to
successive ALUs in your choice of 32, 64, or 128 bit packets).

Why *not* have a family of parts - one supercomputer chip, one DSP
chip, one general-purpose microprocessor chip, one mainframe-like part
- that all share a common underlying instruction set? Of course some
implementations will favor some features, and omit others.

Where some features conflict, again, it gives the opportunity to
elucidate architectural concepts. Why do computers with packed decimal
data types tend to be big-endian?

3) It is out of harmony with current fashions.

Splitting up an 'add' instruction into a load operation and a later add
operation shrinks programs, and doesn't, in itself, add all *that* much
to instruction decoding logic.

Being big-endian makes the architecture less confusing, easier to
understand and explain.

Thus, while the architecture *as shown* is indeed indefensible, a very
trimmed-down subset, although still... different... might not be all
that bad.

John Savard

jsa...@ecn.ab.ca

unread,

Dec 1, 2006, 6:49:11 AM12/1/06

to

I wrote:
> Splitting up an 'add' instruction into a load operation and a later add
> operation

that is, in the instruction decoder, *instead* of doing so in the ISA,

> shrinks programs, and doesn't, in itself, add all *that* much
> to instruction decoding logic.

John Savard