| R.I.P. PDP-10? [message #258154] |
Fri, 27 June 2014 14:34  |
cb
Messages: 300
Registered: March 2012
Karma: 0
|
Senior Member |
|
|
In article <lokbf4$v1g$1@dont-email.me>,
gareth <no.spam@thank.you.invalid> wrote:
> The Raspberry PI that is sweeping the world seems to be a full
> 32-bit plus floating point processor with 256Mbyte of RAM, all
> in a space the size of a paperback book for only a few pounds.
The Raspberry Pi is quite nice - and it's nice that such a relatively
powerful(*) computer is so readily and inexpensively available!
> In my final year at Uni (1972, Essex) I remember the PDP-10
> occupying a whole room, with every DEC black cabinet with
> a panel of blinkenlights at the top*****
You can run a PDP-10 emulator on a Raspberry Pi, either with Simh
(<http://simh.trailing-edge.com/>) or with klh10 (see
<http://victor.se/bjorn/its/raspi.php>). This may not give you an
actual PDP-10 to play with, but you could buy or make yourself a Panda
Display Panel (which is supported by the Panda distribution of klh10,
see <http://www.sparetimegizmos.com/Hardware/Panda.htm>) and run
TOPS-10 or TOPS-20 or ITS.
[ snippage ]
// Christian
|
|
|
|
| Re: R.I.P. PDP-10? [message #258155 is a reply to message #258154] |
Fri, 27 June 2014 14:13  |
Al Kossow
Messages: 237
Registered: July 2012
Karma: 0
|
Senior Member |
|
|
On 6/27/14 10:59 AM, gareth wrote:
> I wonder if DEC had professional stylists on thier design teams
Corporations normally contract with industrial design firms when setting
a style for their products.
Harwood's "The Interface: IBM and the Transformation of Corporate Design, 1945-1976"
discusses this in detail
Eliot Noyes and Assoc was the firm contracted for the industrial design of S/360
for example.
There were probably only and handful of people who worked on industrial design
for DEC in the early days since their machines were built from cookie-cutter
components that were also sold as end-user components (racks, light panels).
The actual details of who did the work probably exists in the DEC archives at the
Computer History Musueum, but that archive is currently inacessable and has not
been cataloged.
|
|
|
|
| Re: R.I.P. PDP-10? [message #258156 is a reply to message #258155] |
Fri, 27 June 2014 14:17 |
Al Kossow
Messages: 237
Registered: July 2012
Karma: 0
|
Senior Member |
|
|
On 6/27/14 11:13 AM, Al Kossow wrote:
> There were probably only and handful of people who worked on industrial design
> for DEC in the early days since their machines were built from cookie-cutter
> components that were also sold as end-user components (racks, light panels).
>
The industrial design of the HP and DEC products in the early 70's is particularly
startling. DEC machines look like they were put together with stone knives and
bearskins compared to their HP contemporaries.
|
|
|
|
| R.I.P. PDP-10? [message #258157 is a reply to message #258154] |
Fri, 27 June 2014 13:59 |
gareth
Messages: 598
Registered: June 2012
Karma: 0
|
Senior Member |
|
|
The Raspberry PI that is sweeping the world seems to be a full
32-bit plus floating point processor with 256Mbyte of RAM, all
in a space the size of a paperback book for only a few pounds.
In my final year at Uni (1972, Essex) I remember the PDP-10
occupying a whole room, with every DEC black cabinet with
a panel of blinkenlights at the top*****
I wonder if DEC had professional stylists on thier design teams,
because the colour-co-ordinated front panels of their processors
still, to my mind, look futuristic today?
***** when I were a lad, we didn't have computers, we had to make
our own abaci from pebbles on the beach, etc, etc, etc.
|
|
|
|
| Re: R.I.P. PDP-10? [message #258174 is a reply to message #258157] |
Fri, 27 June 2014 16:30 |
D.J.
Messages: 821
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
On Fri, 27 Jun 2014 18:59:45 +0100, "gareth"
<no.spam@thank.you.invalid> wrote:
> The Raspberry PI that is sweeping the world seems to be a full
> 32-bit plus floating point processor with 256Mbyte of RAM, all
> in a space the size of a paperback book for only a few pounds.
It is smaller than most paperback books. 2.5 x 3 inches. I have 2 of
them, and about 5 arduinos.
JimP
--
"Brushing aside the thorns so I can see the stars." from 'Ghost in the Shell'
http://www.linuxgazette.net/ Linux Gazette
http://travellergame.drivein-jim.net/ February, 2014
|
|
|
|
| Re: R.I.P. PDP-10? [message #258186 is a reply to message #258157] |
Fri, 27 June 2014 18:40 |
Gerard Schildberger
Messages: 163
Registered: September 2012
Karma: 0
|
Senior Member |
|
|
On Friday, June 27, 2014 12:59:45 PM UTC-5, gareth wrote:
> The Raspberry PI that is sweeping the world seems to be a full
> 32-bit plus floating point processor with 256Mbyte of RAM, all
> in a space the size of a paperback book for only a few pounds.
>
> In my final year at Uni (1972, Essex) I remember the PDP-10
> occupying a whole room, with every DEC black cabinet with
> a panel of blinkenlights at the top*****
>
> I wonder if DEC had professional stylists on thier design teams,
> because the colour-co-ordinated front panels of their processors
> still, to my mind, look futuristic today?
>
> ***** when I were a lad, we didn't have computers, we had to make
> our own abaci from pebbles on the beach, etc, etc, etc.
You had a BEACH?
We had to dig (with our bare hands) to find rocks, and
break the rocks up with a smaller rock to make our own
pebbles. Hurrumph! Hurrumph!
________________________________________ Gerard Schildberger
|
|
|
|
| Re: R.I.P. PDP-10? [message #258720 is a reply to message #258157] |
Tue, 01 July 2014 10:11 |
greenaum
Messages: 57
Registered: February 2011
Karma: 0
|
Member |
|
|
On Fri, 27 Jun 2014 18:59:45 +0100, "gareth"
<no.spam@thank.you.invalid> sprachen:
> The Raspberry PI that is sweeping the world seems to be a full
> 32-bit plus floating point processor with 256Mbyte of RAM, all
> in a space the size of a paperback book for only a few pounds.
512MB RAM, and in the size of a credit card with a couple of bits of
candy stuck to the top, really. You could squeeze a good 5 or 6 into a
paperback book.
They're brilliant, too. First time I've used Linux in a while, the
whole apt-get system is genius. You just need to know the magic name
of what you're after, and it just appears! Download, compile, install,
everything!
While you're being impressed, consider the power demand of something
like a watt. With the 3D graphics and MPEG codec and clocking up to
950MHz without even taking a breath! With a few brains and a
pico-projector you could have something where a few guys with
wristwatches could re-create the control room from the Apollo program,
giant screens 'n' all.
--
------------------------------------------------------------ --------------------
"There's nothing like eating hay when you're faint," the White King remarked to Alice, as he munched away.
"I should think throwing cold water over you would be better," Alice suggested: "--or some sal-volatile."
"I didn't say there was nothing better," the King replied. "I said there was nothing like it."
Which Alice did not venture to deny.
|
|
|
|
| Re: R.I.P. PDP-10? [message #258733 is a reply to message #258720] |
Tue, 01 July 2014 10:56 |
despen
Messages: 162
Registered: March 2005
Karma: 0
|
Senior Member |
|
|
greenaum@gmail.com (greenaum) writes:
> On Fri, 27 Jun 2014 18:59:45 +0100, "gareth"
> <no.spam@thank.you.invalid> sprachen:
>
>> The Raspberry PI that is sweeping the world seems to be a full
>> 32-bit plus floating point processor with 256Mbyte of RAM, all
>> in a space the size of a paperback book for only a few pounds.
>
> 512MB RAM, and in the size of a credit card with a couple of bits of
> candy stuck to the top, really. You could squeeze a good 5 or 6 into a
> paperback book.
>
> They're brilliant, too. First time I've used Linux in a while, the
> whole apt-get system is genius. You just need to know the magic name
> of what you're after, and it just appears! Download, compile, install,
> everything!
Agree. I've been using Fedora which is Yum based.
I'm sure there's very little difference.
You can find the magic package name with Google.
If you know the executable name, yum lets you find the package name with
yum provides *bin/executablename.
So if you know any file name the package provides, you're there.
Then there are a few GUIs that organize packages into categories so you
can just scan a list of package names.
Makes finding and installing a breeze.
Yet you still find people online complaining about having to compile
packages for Linux. Just shows, the ignorant just keep spouting.
--
Dan Espen
|
|
|
|
| Re: R.I.P. PDP-10? [message #258736 is a reply to message #258720] |
Tue, 01 July 2014 11:11 |
Morten Reistad
Messages: 2108
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
In article <53b2c063.4704453@news.eternal-september.org>,
greenaum <greenaum@gmail.com> wrote:
> On Fri, 27 Jun 2014 18:59:45 +0100, "gareth"
> <no.spam@thank.you.invalid> sprachen:
>
>> The Raspberry PI that is sweeping the world seems to be a full
>> 32-bit plus floating point processor with 256Mbyte of RAM, all
>> in a space the size of a paperback book for only a few pounds.
>
> 512MB RAM, and in the size of a credit card with a couple of bits of
> candy stuck to the top, really. You could squeeze a good 5 or 6 into a
> paperback book.
>
> They're brilliant, too. First time I've used Linux in a while, the
> whole apt-get system is genius. You just need to know the magic name
> of what you're after, and it just appears! Download, compile, install,
> everything!
>
> While you're being impressed, consider the power demand of something
> like a watt. With the 3D graphics and MPEG codec and clocking up to
> 950MHz without even taking a breath! With a few brains and a
> pico-projector you could have something where a few guys with
> wristwatches could re-create the control room from the Apollo program,
> giant screens 'n' all.
And it does not stop there.
You have the cubieboard, the hardkernel odroids, which give up to 8
processors at 1.2/1.6 GHz, 2G ram, and a gigabit ethernet port, and
some with SATA disk interface and USB3.0. And all of them have glorious
graphics and decent audio, and simple preinstalled ubuntu Linux. Or
images downloadable to SD-cards by simple DD.
I may be a little biased, using and programming small ARM processors
for work for a decade now, but the udroid U3 has become my primary screen,
driving a 1920x1280 physical and 4096x4096 virtual screen. All using
about 4 watts of power running full tilt, about 1700 mW idling.
(hardkernel.com has links to youtube videos where they measure this
while running various stuff.)
The backlight of the screen uses about 15.
With Xinerama and a dozen of these you can fill a wall, for the expense
of the screens plus about $80 per server, power and SD-card included.
And the chinese have made a "banana pi", with about 3 times the
performance. It seems a little dodgy project, though.
I find the raspberry a little too small, but the odroids perform as
a middle of the road laptop, running browser, video etc. Java is the
only sore point, it runs very slowly on arm.
-- mrr
|
|
|
|
| Re: R.I.P. PDP-10? [message #258864 is a reply to message #258720] |
Tue, 01 July 2014 16:05 |
Jon Elson
Messages: 646
Registered: April 2013
Karma: 0
|
Senior Member |
|
|
greenaum wrote:
> On Fri, 27 Jun 2014 18:59:45 +0100, "gareth"
> <no.spam@thank.you.invalid> sprachen:
>
>> The Raspberry PI that is sweeping the world seems to be a full
>> 32-bit plus floating point processor with 256Mbyte of RAM, all
>> in a space the size of a paperback book for only a few pounds.
>
> 512MB RAM, and in the size of a credit card with a couple of bits of
> candy stuck to the top, really. You could squeeze a good 5 or 6 into a
> paperback book.
>
> They're brilliant, too. First time I've used Linux in a while, the
> whole apt-get system is genius. You just need to know the magic name
> of what you're after, and it just appears! Download, compile, install,
> everything!
>
> While you're being impressed, consider the power demand of something
> like a watt. With the 3D graphics and MPEG codec and clocking up to
> 950MHz without even taking a breath! With a few brains and a
> pico-projector you could have something where a few guys with
> wristwatches could re-create the control room from the Apollo program,
> giant screens 'n' all.
>
There's also the Beagle Bone, all that a Pi has (with the possible exception
of a high-end video decoder/3D renderer) but WAY more I/O. It has at least
60 GPIO pins, plus a pair of 32-bit 200 MIPS microcontrollers that are
VERY good at implementing little bit banger tasks. Also very low
power, somewhere between 1-3 W depending on what you are doing.
Jon
|
|
|
|
| Re: R.I.P. PDP-10? [message #258975 is a reply to message #258736] |
Wed, 02 July 2014 14:40 |
greenaum
Messages: 57
Registered: February 2011
Karma: 0
|
Member |
|
|
On Tue, 1 Jul 2014 15:11:28 +0000, Morten Reistad <first@last.name>
sprachen:
> I find the raspberry a little too small, but the odroids perform as
> a middle of the road laptop, running browser, video etc. Java is the
> only sore point, it runs very slowly on arm.
Shouldn't do, really, since all the modern ones have Java extensions
built into the instruction set. Of course the JVM still needs managing
etc.
I'm gonna look up this Odroid thing. I don't use laptops, but since
playing on my mate's Raspi, it looks like it can even do desktop
stuff, web browsing, really fast and well. I might start recommending
them to relatives to replace their netbooks. For a computer that
spends 95% of it's time on *£%^&!% Faecebook, it may be just as good,
and for a tiny fraction of the price.
Since upgrades are just a matter of the apt-get I'm apparently in love
with, and a 16GB SD card is more hard drive than they'll ever need, I
bet it'd be ideal. Might try it out myself first. Plus if you take
your SD card with you, you've taken the whole personality with you.
Nobody needs more than 16GB of software. My video files etc can still
live on an HDD. Which also interfaces no problem with a Pi! Done that
too!
I could get my grandmother into having one, she could take the whole
thing on holiday with her!
------------------------------------------------------------ ---------
"hey let's educate the brutes, we know we are superior to them anyway,
just through genetics, we are gentically superior to the working
class. They are a shaved monkey. If we educate them, they will be able
to read instructions, turn up on time and man the conveyor belts,
sorted." #
|
|
|
|
| Re: R.I.P. PDP-10? [message #258976 is a reply to message #258864] |
Wed, 02 July 2014 14:50 |
greenaum
Messages: 57
Registered: February 2011
Karma: 0
|
Member |
|
|
On Tue, 01 Jul 2014 15:05:17 -0500, Jon Elson <jmelson@wustl.edu>
sprachen:
> There's also the Beagle Bone, all that a Pi has (with the possible exception
> of a high-end video decoder/3D renderer) but WAY more I/O. It has at least
> 60 GPIO pins, plus a pair of 32-bit 200 MIPS microcontrollers that are
> VERY good at implementing little bit banger tasks. Also very low
> power, somewhere between 1-3 W depending on what you are doing.
Yup, I knew about that. I think the Pi has the advantage with it's
vast user base though. Although stuff might be cross-compatible (same
CPU and OS, if not identical hardware), it wouldn't be utterly
hassle-free.
That's something about the Pi. And really really not like the
Slackware I tried some point in the 1990s. Shit just works! Apt-get
being one example. Config files are ok too, with perhaps a bit of
trouble working out which one you need to actually change. But pretty
much, from a standing start you can get along very well with the
beginning stuff, and learning as you go makes a fair amount of sense
too.
Though I am grateful for my experience learning to program (and hack,
and generally mess about with) the Honeywell Bull System V Unix mini I
learned C and Unix on. Like many OSes, the architecture, the feel, the
philosophy of the "place" is key to understanding how it works, and
why it does it this way and not that.
Any idea what that system might've actually been btw? Honeywell Bull,
bought in the early 1990s, 12MB RAM and some sort of 68020-ish CPU
(maybe a '30, who knows?). Seemed to have real virtual memory and
addressing, presumably a real MMU. As in, accessing the same RAM
address in 2 different programs gave different results. Had a hard
drive and a 5 1/4" floppy. About 2 feet tall in a "chunky-mid-tower"
style case.
Connected through a couple of dozen real RS-232 ports on the back of
the thing. 25-pin at both ends I think.
I still check Ebay for old Wyse 60 character terminals too. They'd
work fine with Linux I'm sure. Ironically the few I find cost hundreds
of pounds, when I can buy "thin clients" (remember that fad from the
1990s?) running Windows CE or XP for 20 quid or so.
I've thought of hollowing out the thin clients and selling them as
entry-level PCs to beginners for Facebooking. But I don't think what's
left inside 'em is up to the task.
------------------------------------------------------------ ---------
"hey let's educate the brutes, we know we are superior to them anyway,
just through genetics, we are gentically superior to the working
class. They are a shaved monkey. If we educate them, they will be able
to read instructions, turn up on time and man the conveyor belts,
sorted." #
|
|
|
|
| Re: R.I.P. PDP-10? [message #258982 is a reply to message #258975] |
Wed, 02 July 2014 15:26 |
scott
Messages: 4239
Registered: February 2012
Karma: 0
|
Senior Member |
|
|
greenaum@gmail.com (greenaum) writes:
> On Tue, 1 Jul 2014 15:11:28 +0000, Morten Reistad <first@last.name>
> sprachen:
>
>> I find the raspberry a little too small, but the odroids perform as
>> a middle of the road laptop, running browser, video etc. Java is the
>> only sore point, it runs very slowly on arm.
>
> Shouldn't do, really, since all the modern ones have Java extensions
> built into the instruction set. Of course the JVM still needs managing
> etc.
Even ARM has realized that Jazelle doesn't add much. The ARM64
instruction doesn't include the Jazelle instructions.
|
|
|
|
| Re: R.I.P. PDP-10? [message #259013 is a reply to message #258976] |
Wed, 02 July 2014 17:44 |
Jon Elson
Messages: 646
Registered: April 2013
Karma: 0
|
Senior Member |
|
|
greenaum wrote:
> On Tue, 01 Jul 2014 15:05:17 -0500, Jon Elson <jmelson@wustl.edu>
> sprachen:
>
>> There's also the Beagle Bone, all that a Pi has (with the possible
>> exception
>> of a high-end video decoder/3D renderer) but WAY more I/O. It has at
>> least 60 GPIO pins, plus a pair of 32-bit 200 MIPS microcontrollers that
>> are
>> VERY good at implementing little bit banger tasks. Also very low
>> power, somewhere between 1-3 W depending on what you are doing.
>
> Yup, I knew about that. I think the Pi has the advantage with it's
> vast user base though. Although stuff might be cross-compatible (same
> CPU and OS, if not identical hardware), it wouldn't be utterly
> hassle-free.
They have sold 170K of the Beagle Bones (as well as a bunch of earlier
Beagle Board models). So, that isn't such a tiny user base, either.
I've done two projects with the earlier Boards, and now two with the
Bone version, the last one using the PRU component (the 200 MIPS
microcontroller) that was awesome to use to emulate an obsolete DMA
card on an ancient PC. The availability of massive numbers of
I/O pins on the Bone is a huge plus for hardware hacking projects.
>
> That's something about the Pi. And really really not like the
> Slackware I tried some point in the 1990s. Shit just works! Apt-get
> being one example. Config files are ok too, with perhaps a bit of
> trouble working out which one you need to actually change. But pretty
> much, from a standing start you can get along very well with the
> beginning stuff, and learning as you go makes a fair amount of sense
> too.
>
Since I use Ubuntu as my standard desktop (or laptop) OS these days,
I am quite familiar with all the config and setup stuff already, so
almost everything is transferable. The only thing that is really
different are the devicetree for the GPIO config, and the PRU, if
you need to use that.
My only real Unix experience was trying to use real Bell Labs Unix
on a PDP-11 (our lab wanted to evaluate whether our large FORTRAN
codebase could be used), and trying to use BSD in a graphics course.
We checked the FORTRAN compiler and found the computed goto was BACKWARDS.
While we didn't use the computed goto much, it gave a bad taste of how
poorly the compiler might have been tested. Then, I wrote a program
like :
DO 10 I=1,10
A=I
PRINT A
10 CONTINUE
(omitting the format line)
And discovered it took about 700 ms to format and print one real number!
This seemed like a really bad implementation, and we didn't pursue it
much farther.
As for the graphics course, I went into a room with some terminals
and tried to edit up some c code in vi. After about 15 minutes struggling
with it, I managed to halt the CPU and couldn't figure out how to reboot.
I gave up and did the class exercises on a VMS system.
So, it was a long time before I went back to the Unix/Linux fold.
I did clone a National Semi 32016 CPU computer that ran Genix. It was
insanely slow, partially because of the 32016 CPU, and partially due to
slow memory I had grafted onto it. It worked, but was so glacially slow
I'd rather use an 8-bit Z80 CP/M system than the Genix system! Then I
got a uVAX at home, and used that for some time, until it became too
obsolete. And, then in 1997 I got dragged back in, with the outbreak
of Linux.
Jon
|
|
|
|
| Re: R.I.P. PDP-10? [message #259098 is a reply to message #259013] |
Thu, 03 July 2014 11:07 |
Quadibloc
Messages: 4399
Registered: June 2012
Karma: 0
|
Senior Member |
|
|
On Wednesday, July 2, 2014 3:44:00 PM UTC-6, Jon Elson wrote:
> Then, I wrote a program
> like :
> DO 10 I=1,10
> A=I
> PRINT A
> 10 CONTINUE
How horribly inefficient! It should have been:
A = 1.0
DO 7 I=1,10
PRINT A
A = A+1.0
7 CONTINUE
since a floating-point add is much faster than a conversion from integer to floating-point.
This assumes a variant of FORTRAN where PRINT A is used for format-free I/O, replacing WRITE(6,11) A with a FORMAT line numbered 11 omitted.
John Savard
|
|
|
|
| Re: R.I.P. PDP-10? [message #259205 is a reply to message #259098] |
Thu, 03 July 2014 18:33 |
Jon Elson
Messages: 646
Registered: April 2013
Karma: 0
|
Senior Member |
|
|
Quadibloc wrote:
> On Wednesday, July 2, 2014 3:44:00 PM UTC-6, Jon Elson wrote:
>> Then, I wrote a program
>> like :
>
>> DO 10 I=1,10
>> A=I
>> PRINT A
>> 10 CONTINUE
>
> How horribly inefficient! It should have been:
But, it was INTENTIONALLY done that way, to expose inefficiencies in the
Unix support libraries! I wanted it to be at least somewhat representative
of some of the code we were already using, which was not necessarily
efficient, but it is what would be transferred, if we went the Unix way.
After seeing how inefficient the compiler and/or libraries were, we went
from DOS-11/RT-11 to RSX-11M, and had good results with the same crummy
codes. They ran at least 10X faster, and probably closer to one HUNDRED
times faster on those OS'es. The guys in that lab were all FORTRAN users,
dyed in the wool, and continued on in the FORTRAN vein long after, with VAX
VMS. I used a fair bit of Pascal, especially when we moved to the VAXes.
> This assumes a variant of FORTRAN where PRINT A is used for format-free
> I/O, replacing WRITE(6,11) A with a FORMAT line numbered 11 omitted.
It was, almost certainly, a formatted print statement, but I have purged
my memory of how to code those up. All I remember was FORTRAN format
statements were a bit messy.
Jon
|
|
|
|
| Re: R.I.P. PDP-10? [message #259206 is a reply to message #259013] |
Thu, 03 July 2014 18:35 |
Gene Wirchenko
Messages: 1166
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
On Wed, 02 Jul 2014 16:44:00 -0500, Jon Elson <jmelson@wustl.edu>
wrote:
[snip]
> My only real Unix experience was trying to use real Bell Labs Unix
> on a PDP-11 (our lab wanted to evaluate whether our large FORTRAN
> codebase could be used), and trying to use BSD in a graphics course.
> We checked the FORTRAN compiler and found the computed goto was BACKWARDS.
COME FROM?
> While we didn't use the computed goto much, it gave a bad taste of how
> poorly the compiler might have been tested. Then, I wrote a program
> like :
> DO 10 I=1,10
> A=I
> PRINT A
> 10 CONTINUE
> (omitting the format line)
No FORMAT is needed with PRINT. I believe you should have
inserted a comma:
PRINT ,A
(I have no idea why it was defined that way.)
> And discovered it took about 700 ms to format and print one real number!
> This seemed like a really bad implementation, and we didn't pursue it
> much farther.
Are you sure that it was the formatting that took the time? I
have seen slow displays.
[snip]
Sincerely,
Gene Wirchenko
|
|
|
|
| Re: R.I.P. PDP-10? [message #259217 is a reply to message #259206] |
Thu, 03 July 2014 20:54 |
Andrew Swallow
Messages: 1705
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
On 03/07/2014 23:35, Gene Wirchenko wrote:
{snip}
> No FORMAT is needed with PRINT. I believe you should have
> inserted a comma:
> PRINT ,A
> (I have no idea why it was defined that way.)
Spaces are ignored in Fortran so the compiler would think PRINTA was a
variable.
Andrew Swallow
|
|
|
|
| Re: R.I.P. PDP-10? [message #259221 is a reply to message #259217] |
Thu, 03 July 2014 22:15 |
Walter Bushell
Messages: 1834
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
In article <qcednW0OgLrMZijOnZ2dnUVZ8gudnZ2d@bt.com>,
Andrew Swallow <am.swallow@btinternet.com> wrote:
> On 03/07/2014 23:35, Gene Wirchenko wrote:
> {snip}
>> No FORMAT is needed with PRINT. I believe you should have
>> inserted a comma:
>> PRINT ,A
>> (I have no idea why it was defined that way.)
>
> Spaces are ignored in Fortran so the compiler would think PRINTA was a
> variable.
>
> Andrew Swallow
Fortran also had no reserved words otherwise "PRINT A", could have
been parse, probably resulting in an error message "PRINT STATEMENT
IMPROPERLY FORMATTED" or such.
--
Never attribute to stupidity that which can be explained by greed. Me.
|
|
|
|
| Re: R.I.P. PDP-10? [message #259229 is a reply to message #259206] |
Fri, 04 July 2014 01:07 |
Jon Elson
Messages: 646
Registered: April 2013
Karma: 0
|
Senior Member |
|
|
Gene Wirchenko wrote:
> On Wed, 02 Jul 2014 16:44:00 -0500, Jon Elson <jmelson@wustl.edu>
> wrote:
>
> [snip]
>
>> My only real Unix experience was trying to use real Bell Labs Unix
>> on a PDP-11 (our lab wanted to evaluate whether our large FORTRAN
>> codebase could be used), and trying to use BSD in a graphics course.
>> We checked the FORTRAN compiler and found the computed goto was BACKWARDS.
>
> COME FROM?
No, computer goto in FORTRAN was something like
IF A (10,20,30)
it would go to 10 if A was minus, 20 if zero and 30 if positive and greater
than zero.
So, they had the + and minus reversed.
>
> Are you sure that it was the formatting that took the time? I
> have seen slow displays.
>
Yes, as we ran RT-11 and later RSX-11M on the same hardware, it was
vastly faster, somewhere between 10 and 100 X faster on DEC OS and
compilers as on the Bell Unix. And, I'm sure Bell's C compiler did
fine, but we were not going to change our language of choice at the
time.
Jon
|
|
|
|
| Re: R.I.P. PDP-10? [message #259259 is a reply to message #259205] |
Fri, 04 July 2014 10:07 |
jmfbahciv
Messages: 6173
Registered: March 2012
Karma: 0
|
Senior Member |
|
|
Jon Elson wrote:
> Quadibloc wrote:
>
>> On Wednesday, July 2, 2014 3:44:00 PM UTC-6, Jon Elson wrote:
>>> Then, I wrote a program
>>> like :
>>
>>> DO 10 I=1,10
>>> A=I
>>> PRINT A
>>> 10 CONTINUE
>>
>> How horribly inefficient! It should have been:
>
> But, it was INTENTIONALLY done that way, to expose inefficiencies in the
> Unix support libraries! I wanted it to be at least somewhat representative
> of some of the code we were already using, which was not necessarily
> efficient, but it is what would be transferred, if we went the Unix way.
> After seeing how inefficient the compiler and/or libraries were, we went
> from DOS-11/RT-11 to RSX-11M, and had good results with the same crummy
> codes. They ran at least 10X faster, and probably closer to one HUNDRED
> times faster on those OS'es. The guys in that lab were all FORTRAN users,
> dyed in the wool, and continued on in the FORTRAN vein long after, with VAX
> VMS. I used a fair bit of Pascal, especially when we moved to the VAXes.
>
>> This assumes a variant of FORTRAN where PRINT A is used for format-free
>> I/O, replacing WRITE(6,11) A with a FORMAT line numbered 11 omitted.
> It was, almost certainly, a formatted print statement, but I have purged
> my memory of how to code those up. All I remember was FORTRAN format
> statements were a bit messy.
PRINT 1,A
1 FORMAT (1X,F2.0)
The compilers probably would probably output code which stored the
number into both locations when doing the indexing. The KA also
had float instruction so no code was needed to convert.
/BAH
|
|
|
|
| Re: R.I.P. PDP-10? [message #259285 is a reply to message #259229] |
Fri, 04 July 2014 14:55 |
Quadibloc
Messages: 4399
Registered: June 2012
Karma: 0
|
Senior Member |
|
|
On Thursday, July 3, 2014 11:07:59 PM UTC-6, Jon Elson wrote:
> No, computer goto in FORTRAN was something like
>
> IF A (10,20,30)
>
> it would go to 10 if A was minus, 20 if zero and 30 if positive and greater
> than zero.
> So, they had the + and minus reversed.
Actually, that's an IF statement;
IF (A) 10, 20, 30
And going to 10 if A is negative is correct. If it went to 30 then, that would indeed be backwards.
Computed GOTO is
GO TO (10, 20, 30), I
which goes to 10 if I .EQ. 1, 20 if I .EQ. 2, and 30 if I .EQ. 3. I guess if it was backwards, it would go to 10 if I was 3.
If they managed to get an assigned GOTO
GO TO I, (10, 20, 30)
to work backwards, they would have had to have expended some effort (such as implementing it as a backwards computed GOTO).
John Savard
|
|
|
|
| Re: R.I.P. PDP-10? [message #259318 is a reply to message #259229] |
Fri, 04 July 2014 18:31 |
Gene Wirchenko
Messages: 1166
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
On Fri, 04 Jul 2014 00:07:59 -0500, Jon Elson <elson@pico-systems.com>
wrote:
> Gene Wirchenko wrote:
>
>> On Wed, 02 Jul 2014 16:44:00 -0500, Jon Elson <jmelson@wustl.edu>
>> wrote:
>>
>> [snip]
>>
>>> My only real Unix experience was trying to use real Bell Labs Unix
>>> on a PDP-11 (our lab wanted to evaluate whether our large FORTRAN
>>> codebase could be used), and trying to use BSD in a graphics course.
>>> We checked the FORTRAN compiler and found the computed goto was BACKWARDS.
I think you are actually discussing an arithmetic IF. An example
of a computed GOTO is:
GOTO (10,20,30), A
and meant if a=1 then goto 10 else if a=2 then goto 20 else if a=3
then goto 30.
>> COME FROM?
> No, computer goto in FORTRAN was something like
> IF A (10,20,30)
I think
IF (A) 10,20,30
> it would go to 10 if A was minus, 20 if zero and 30 if positive and greater
> than zero.
>
> So, they had the + and minus reversed.
Yech!
Sincerely,
Gene Wirchenko
|
|
|
|
| Re: R.I.P. PDP-10? [message #259352 is a reply to message #259229] |
Fri, 04 July 2014 21:06 |
Charles Richmond
Messages: 2754
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
"Jon Elson" <elson@pico-systems.com> wrote in message
news:lrednYOW1YHTqyvOnZ2dnUVZ_o6dnZ2d@giganews.com...
> Gene Wirchenko wrote:
>
>> On Wed, 02 Jul 2014 16:44:00 -0500, Jon Elson <jmelson@wustl.edu>
>> wrote:
>>
>> [snip]
>>
>>> My only real Unix experience was trying to use real Bell Labs Unix
>>> on a PDP-11 (our lab wanted to evaluate whether our large FORTRAN
>>> codebase could be used), and trying to use BSD in a graphics course.
>>> We checked the FORTRAN compiler and found the computed goto was
>>> BACKWARDS.
>>
>> COME FROM?
> No, computer goto in FORTRAN was something like
> IF A (10,20,30)
> it would go to 10 if A was minus, 20 if zero and 30 if positive and
> greater
> than zero.
>
> So, they had the + and minus reversed.
>
It is called the "arithmetic if" statement. ISTR that this was the only
"if" in FORTRAN II. In FORTRAN IV you also had the "logical if" statement,
like: IF(X.GE.5.1). The "arithmetic if" was still supported.
--
numerist at aquaporin4 dot com
|
|
|
|
| Re: R.I.P. PDP-10? [message #259353 is a reply to message #259318] |
Fri, 04 July 2014 21:11 |
Charles Richmond
Messages: 2754
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
"Gene Wirchenko" <genew@telus.net> wrote in message
news:3haer9170vvgjupspff0tco9osu123kqtd@4ax.com...
> On Fri, 04 Jul 2014 00:07:59 -0500, Jon Elson <elson@pico-systems.com>
> wrote:
>
>> Gene Wirchenko wrote:
>>
>>> On Wed, 02 Jul 2014 16:44:00 -0500, Jon Elson <jmelson@wustl.edu>
>>> wrote:
>>>
>>> [snip]
>>>
>>>> My only real Unix experience was trying to use real Bell Labs Unix
>>>> on a PDP-11 (our lab wanted to evaluate whether our large FORTRAN
>>>> codebase could be used), and trying to use BSD in a graphics course.
>>>> We checked the FORTRAN compiler and found the computed goto was
>>>> BACKWARDS.
>
> I think you are actually discussing an arithmetic IF. An example
> of a computed GOTO is:
> GOTO (10,20,30), A
> and meant if a=1 then goto 10 else if a=2 then goto 20 else if a=3
> then goto 30.
>
There was also the "assigned goto". One used a statement like: ASSIGN 30
to IY ... where IY was an integer variable and 30 was a statement number
in the same subroutine or function or main program. Later a statement like:
GO TO IY (10,20,30) ... was executed, where 10, 20, and 30 were line
numbers. From what I read on the web, apparently the "list of line numbers"
can be omitted, but if the list is used, the variable IY *must* contain a
number from the list.
--
numerist at aquaporin4 dot com
|
|
|
|
| Re: R.I.P. PDP-10? [message #259361 is a reply to message #259353] |
Fri, 04 July 2014 22:37 |
Walter Bushell
Messages: 1834
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
In article <lp7jcp$6n0$1@dont-email.me>,
"Charles Richmond" <numerist@aquaporin4.com> wrote:
> There was also the "assigned goto". One used a statement like: ASSIGN 30
> to IY ... where IY was an integer variable and 30 was a statement number
> in the same subroutine or function or main program. Later a statement like:
> GO TO IY (10,20,30) ... was executed, where 10, 20, and 30 were line
> numbers. From what I read on the web, apparently the "list of line numbers"
> can be omitted, but if the list is used, the variable IY *must* contain a
> number from the list.
Without the list, one could write some real spaghetti code. Surprise!
--
Never attribute to stupidity that which can be explained by greed. Me.
|
|
|
|
| Re: R.I.P. PDP-10? [message #259362 is a reply to message #259361] |
Fri, 04 July 2014 22:50 |
Joe Pfeiffer
Messages: 764
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
Walter Bushell <proto@panix.com> writes:
> In article <lp7jcp$6n0$1@dont-email.me>,
> "Charles Richmond" <numerist@aquaporin4.com> wrote:
>
>> There was also the "assigned goto". One used a statement like: ASSIGN 30
>> to IY ... where IY was an integer variable and 30 was a statement number
>> in the same subroutine or function or main program. Later a statement like:
>> GO TO IY (10,20,30) ... was executed, where 10, 20, and 30 were line
>> numbers. From what I read on the web, apparently the "list of line numbers"
>> can be omitted, but if the list is used, the variable IY *must* contain a
>> number from the list.
>
> Without the list, one could write some real spaghetti code. Surprise!
*with* the list, one could write some real spaghetti code...
|
|
|
|
| Re: R.I.P. PDP-10? [message #259367 is a reply to message #259362] |
Sat, 05 July 2014 00:12 |
Andrew Swallow
Messages: 1705
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
On 05/07/2014 03:50, Joe Pfeiffer wrote:
> Walter Bushell <proto@panix.com> writes:
>
>> In article <lp7jcp$6n0$1@dont-email.me>,
>> "Charles Richmond" <numerist@aquaporin4.com> wrote:
>>
>>> There was also the "assigned goto". One used a statement like: ASSIGN 30
>>> to IY ... where IY was an integer variable and 30 was a statement number
>>> in the same subroutine or function or main program. Later a statement like:
>>> GO TO IY (10,20,30) ... was executed, where 10, 20, and 30 were line
>>> numbers. From what I read on the web, apparently the "list of line numbers"
>>> can be omitted, but if the list is used, the variable IY *must* contain a
>>> number from the list.
>>
>> Without the list, one could write some real spaghetti code. Surprise!
>
> *with* the list, one could write some real spaghetti code...
>
One of the few ways of implementing recursion in Fortran. The return
addresses have to be stored in an array.
Andrew Swallow
|
|
|
|
| Re: R.I.P. PDP-10? [message #259387 is a reply to message #259362] |
Sat, 05 July 2014 12:15 |
Walter Bushell
Messages: 1834
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
In article <1br420lds5.fsf@snowball.wb.pfeifferfamily.net>,
Joe Pfeiffer <pfeiffer@cs.nmsu.edu> wrote:
> Walter Bushell <proto@panix.com> writes:
>
>> In article <lp7jcp$6n0$1@dont-email.me>,
>> "Charles Richmond" <numerist@aquaporin4.com> wrote:
>>
>>> There was also the "assigned goto". One used a statement like: ASSIGN
>>> 30
>>> to IY ... where IY was an integer variable and 30 was a statement number
>>> in the same subroutine or function or main program. Later a statement
>>> like:
>>> GO TO IY (10,20,30) ... was executed, where 10, 20, and 30 were line
>>> numbers. From what I read on the web, apparently the "list of line
>>> numbers"
>>> can be omitted, but if the list is used, the variable IY *must* contain a
>>> number from the list.
>>
>> Without the list, one could write some real spaghetti code. Surprise!
>
> *with* the list, one could write some real spaghetti code...
Oh, sure. But at least you know the possibilities without it can go
anywhere. Perhaps reset in a subroutine masked by reference in a
COMMON or EQUIVALENCE statement. Which might be implemented so as to
jump into the middle of the subroutine. (Save a jump instruction to
the referenced label in the ASSIGNed variable.)
This could be as bad as a COMPUTED COME FROM.
--
Never attribute to stupidity that which can be explained by greed. Me.
|
|
|
|
| Re: R.I.P. PDP-10? [message #259418 is a reply to message #259387] |
Sat, 05 July 2014 18:14 |
Charles Richmond
Messages: 2754
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
"Walter Bushell" <proto@panix.com> wrote in message
news:proto-679665.12155805072014@news.panix.com...
> In article <1br420lds5.fsf@snowball.wb.pfeifferfamily.net>,
> Joe Pfeiffer <pfeiffer@cs.nmsu.edu> wrote:
>
>> Walter Bushell <proto@panix.com> writes:
>>
>>> In article <lp7jcp$6n0$1@dont-email.me>,
>>> "Charles Richmond" <numerist@aquaporin4.com> wrote:
>>>
>>>> There was also the "assigned goto". One used a statement like:
>>>> ASSIGN
>>>> 30
>>>> to IY ... where IY was an integer variable and 30 was a statement
>>>> number
>>>> in the same subroutine or function or main program. Later a statement
>>>> like:
>>>> GO TO IY (10,20,30) ... was executed, where 10, 20, and 30 were
>>>> line
>>>> numbers. From what I read on the web, apparently the "list of line
>>>> numbers"
>>>> can be omitted, but if the list is used, the variable IY *must*
>>>> contain a
>>>> number from the list.
>>>
>>> Without the list, one could write some real spaghetti code. Surprise!
>>
>> *with* the list, one could write some real spaghetti code...
>
> Oh, sure. But at least you know the possibilities without it can go
> anywhere. Perhaps reset in a subroutine masked by reference in a
> COMMON or EQUIVALENCE statement. Which might be implemented so as to
> jump into the middle of the subroutine. (Save a jump instruction to
> the referenced label in the ASSIGNed variable.)
>
> This could be as bad as a COMPUTED COME FROM.
>
So you all must be glad that the latest FORTRAN standards have *eliminated*
the ASSIGNED GO TO.
--
numerist at aquaporin4 dot com
|
|
|
|
| Re: R.I.P. PDP-10? [message #259464 is a reply to message #259418] |
Sun, 06 July 2014 07:47 |
Walter Bushell
Messages: 1834
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
In article <lp9tc6$kus$1@dont-email.me>,
"Charles Richmond" <numerist@aquaporin4.com> wrote:
> "Walter Bushell" <proto@panix.com> wrote in message
> news:proto-679665.12155805072014@news.panix.com...
>> In article <1br420lds5.fsf@snowball.wb.pfeifferfamily.net>,
>> Joe Pfeiffer <pfeiffer@cs.nmsu.edu> wrote:
>>
>>> Walter Bushell <proto@panix.com> writes:
>>>
>>>> In article <lp7jcp$6n0$1@dont-email.me>,
>>>> "Charles Richmond" <numerist@aquaporin4.com> wrote:
>>>>
>>>> > There was also the "assigned goto". One used a statement like:
>>>> > ASSIGN
>>>> > 30
>>>> > to IY ... where IY was an integer variable and 30 was a statement
>>>> > number
>>>> > in the same subroutine or function or main program. Later a statement
>>>> > like:
>>>> > GO TO IY (10,20,30) ... was executed, where 10, 20, and 30 were
>>>> > line
>>>> > numbers. From what I read on the web, apparently the "list of line
>>>> > numbers"
>>>> > can be omitted, but if the list is used, the variable IY *must*
>>>> > contain a
>>>> > number from the list.
>>>>
>>>> Without the list, one could write some real spaghetti code. Surprise!
>>>
>>> *with* the list, one could write some real spaghetti code...
>>
>> Oh, sure. But at least you know the possibilities without it can go
>> anywhere. Perhaps reset in a subroutine masked by reference in a
>> COMMON or EQUIVALENCE statement. Which might be implemented so as to
>> jump into the middle of the subroutine. (Save a jump instruction to
>> the referenced label in the ASSIGNed variable.)
>>
>> This could be as bad as a COMPUTED COME FROM.
>>
>
> So you all must be glad that the latest FORTRAN standards have *eliminated*
> the ASSIGNED GO TO.
>
> --
>
> numerist at aquaporin4 dot com
Doesn't mean you don't have to maintain old programs. And is the
result still FORTRAN? OTOH, FORTRAN *was* designed to be a lean
language for numerical computation.
--
Never attribute to stupidity that which can be explained by greed. Me.
|
|
|
|
| Re: R.I.P. PDP-10? [message #259534 is a reply to message #259464] |
Sun, 06 July 2014 18:49 |
Charles Richmond
Messages: 2754
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
"Walter Bushell" <proto@panix.com> wrote in message
news:proto-125194.07471306072014@news.panix.com...
> In article <lp9tc6$kus$1@dont-email.me>,
> "Charles Richmond" <numerist@aquaporin4.com> wrote:
>
>> "Walter Bushell" <proto@panix.com> wrote in message
>> news:proto-679665.12155805072014@news.panix.com...
>>> In article <1br420lds5.fsf@snowball.wb.pfeifferfamily.net>,
>>> Joe Pfeiffer <pfeiffer@cs.nmsu.edu> wrote:
>>>
>>>> Walter Bushell <proto@panix.com> writes:
>>>>
>>>> > In article <lp7jcp$6n0$1@dont-email.me>,
>>>> > "Charles Richmond" <numerist@aquaporin4.com> wrote:
>>>> >
>>>> >> There was also the "assigned goto". One used a statement like:
>>>> >> ASSIGN
>>>> >> 30
>>>> >> to IY ... where IY was an integer variable and 30 was a statement
>>>> >> number
>>>> >> in the same subroutine or function or main program. Later a
>>>> >> statement
>>>> >> like:
>>>> >> GO TO IY (10,20,30) ... was executed, where 10, 20, and 30 were
>>>> >> line
>>>> >> numbers. From what I read on the web, apparently the "list of line
>>>> >> numbers"
>>>> >> can be omitted, but if the list is used, the variable IY *must*
>>>> >> contain a
>>>> >> number from the list.
>>>> >
>>>> > Without the list, one could write some real spaghetti code.
>>>> > Surprise!
>>>>
>>>> *with* the list, one could write some real spaghetti code...
>>>
>>> Oh, sure. But at least you know the possibilities without it can go
>>> anywhere. Perhaps reset in a subroutine masked by reference in a
>>> COMMON or EQUIVALENCE statement. Which might be implemented so as to
>>> jump into the middle of the subroutine. (Save a jump instruction to
>>> the referenced label in the ASSIGNed variable.)
>>>
>>> This could be as bad as a COMPUTED COME FROM.
>>>
>>
>> So you all must be glad that the latest FORTRAN standards have
>> *eliminated*
>> the ASSIGNED GO TO.
>>
>> --
>>
>> numerist at aquaporin4 dot com
>
> Doesn't mean you don't have to maintain old programs. And is the
> result still FORTRAN? OTOH, FORTRAN *was* designed to be a lean
> language for numerical computation.
>
ISTM that many older FORTRAN programmers will think that "modern" FORTRAN
has muddled up the language with a plethora of things *not* really necessary
for numerical computation. I have *not* written anything more modern than
FORTRAN77, but I have read about some of the changes made in more recent
FORTRAN standards.
--
numerist at aquaporin4 dot com
|
|
|
|
| Re: R.I.P. PDP-10? [message #259566 is a reply to message #259534] |
Mon, 07 July 2014 04:19 |
Quadibloc
Messages: 4399
Registered: June 2012
Karma: 0
|
Senior Member |
|
|
On Sunday, July 6, 2014 4:49:52 PM UTC-6, Charles Richmond wrote:
> ISTM that many older FORTRAN programmers will think that "modern" FORTRAN
> has muddled up the language with a plethora of things *not* really necessary
> for numerical computation.
But the *main* new features of FORTRAN 90 and thereafter are directly related to numerical computation - they give the power to directly specify vector operations. Which is very useful when programming machines like the Cray I, and presumably it even allows generating more efficient code on ordinary Intel processors, particularly with the option of using capabilities like SSE 2 and AVX.
John Savard
|
|
|
|
| Re: R.I.P. PDP-10? [message #259682 is a reply to message #259352] |
Mon, 07 July 2014 14:59 |
hancock4
Messages: 6746
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
On Friday, July 4, 2014 9:06:05 PM UTC-4, Charles Richmond wrote:
> It is called the "arithmetic if" statement. ISTR that this was the only
> "if" in FORTRAN II. In FORTRAN IV you also had the "logical if" statement,
> like: IF(X.GE.5.1). The "arithmetic if" was still supported.
IIRC, the IBM 1130 supported only arithmetic IF's. S/360 supported both.
With ANDs and ORs, one could code complex logic conditions that would be much more cumbersome using only arithmetic IFs.
FWIW, I had a compsci instructor who said logical IFs were inefficient and would not allow their use in our programs. I strongly suspect she was wrong, at least at the time we were using it (S/360-65).
|
|
|
|
| Re: R.I.P. PDP-10? [message #259683 is a reply to message #259418] |
Mon, 07 July 2014 15:01 |
hancock4
Messages: 6746
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
On Saturday, July 5, 2014 6:14:29 PM UTC-4, Charles Richmond wrote:
> So you all must be glad that the latest FORTRAN standards have *eliminated*
> the ASSIGNED GO TO.
As had COBOL eliminating the ALTER statement.
|
|
|
|
| Re: R.I.P. PDP-10? [message #259740 is a reply to message #259682] |
Mon, 07 July 2014 19:16 |
Walter Bushell
Messages: 1834
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
In article <be02cf49-19f4-416f-b9f7-e06fd86025f0@googlegroups.com>,
hancock4@bbs.cpcn.com wrote:
> On Friday, July 4, 2014 9:06:05 PM UTC-4, Charles Richmond wrote:
>
>> It is called the "arithmetic if" statement. ISTR that this was the only
>> "if" in FORTRAN II. In FORTRAN IV you also had the "logical if" statement,
>> like: IF(X.GE.5.1). The "arithmetic if" was still supported.
>
> IIRC, the IBM 1130 supported only arithmetic IF's. S/360 supported both.
>
> With ANDs and ORs, one could code complex logic conditions that would be much
> more cumbersome using only arithmetic IFs.
>
> FWIW, I had a compsci instructor who said logical IFs were inefficient and
> would not allow their use in our programs. I strongly suspect she was wrong,
> at least at the time we were using it (S/360-65).
For complex conditions LOGICAL IF's are better, you want the logic
spread over 5, 6 or more IF Statements?! Anyway breaking the pipeline
is far more time consuming than a logical statement, if there be any
loss of efficiency.
--
Never attribute to stupidity that which can be explained by greed. Me.
|
|
|
|
| Re: Fortran archaeology, was R.I.P. PDP-10? [message #259758 is a reply to message #259740] |
Mon, 07 July 2014 20:53 |
John Levine
Messages: 1405
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
>> FWIW, I had a compsci instructor who said logical IFs were inefficient and
>> would not allow their use in our programs. I strongly suspect she was wrong,
>> at least at the time we were using it (S/360-65).
She was wrong. In Fortran G the arithmetic IF probably generated
useless branches, in Fortran H it's have been optimized to be the
same.
> For complex conditions LOGICAL IF's are better, you want the logic
> spread over 5, 6 or more IF Statements?! Anyway breaking the pipeline
> is far more time consuming than a logical statement, if there be any
> loss of efficiency.
Pipeline? What's a pipeline? This was a 360/65. Next thing you know
you'll be talking about this mythical thing called a "cache".
--
Regards,
John Levine, johnl@iecc.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. http://jl.ly
|
|
|
|
| Re: R.I.P. PDP-10? [message #259775 is a reply to message #259682] |
Tue, 08 July 2014 04:34 |
Quadibloc
Messages: 4399
Registered: June 2012
Karma: 0
|
Senior Member |
|
|
On Monday, July 7, 2014 12:59:23 PM UTC-6, hanc...@bbs.cpcn.com wrote:
> FWIW, I had a compsci instructor who said logical IFs were inefficient and
> would not allow their use in our programs. I strongly suspect she was wrong,
> at least at the time we were using it (S/360-65).
Of course. But on the 7090, while the Fortran II compiler continued the tradition of the original Fortran compiler for the 704 in being highly optimized, the Fortran IV compiler was *not* optimized. And, of course, the Logical IF statement was introduced with Fortran IV.
So I can understand where she might have gotten that idea from.
John Savard
|
|
|
|
| Re: R.I.P. PDP-10? [message #259790 is a reply to message #259740] |
Tue, 08 July 2014 07:00 |
Peter Flass
Messages: 8375
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
Walter Bushell <proto@panix.com> wrote:
> In article <be02cf49-19f4-416f-b9f7-e06fd86025f0@googlegroups.com>,
> hancock4@bbs.cpcn.com wrote:
>
>> On Friday, July 4, 2014 9:06:05 PM UTC-4, Charles Richmond wrote:
>>
>>> It is called the "arithmetic if" statement. ISTR that this was the only
>>> "if" in FORTRAN II. In FORTRAN IV you also had the "logical if" statement,
>>> like: IF(X.GE.5.1). The "arithmetic if" was still supported.
>>
>> IIRC, the IBM 1130 supported only arithmetic IF's. S/360 supported both.
>>
>> With ANDs and ORs, one could code complex logic conditions that would be much
>> more cumbersome using only arithmetic IFs.
>>
>> FWIW, I had a compsci instructor who said logical IFs were inefficient and
>> would not allow their use in our programs. I strongly suspect she was wrong,
>> at least at the time we were using it (S/360-65).
>
> For complex conditions LOGICAL IF's are better, you want the logic
> spread over 5, 6 or more IF Statements?! Anyway breaking the pipeline
> is far more time consuming than a logical statement, if there be any
> loss of efficiency.
From what I understand (not much) it shouldn't matter much. At the
machine-language level either type of IF gets translated to a series of
compares and branches.
--
Pete
|
|
|
|
| Re: Fortran archaeology, was R.I.P. PDP-10? [message #259801 is a reply to message #259758] |
Tue, 08 July 2014 09:17 |
Walter Bushell
Messages: 1834
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
In article <lpfff2$1pst$3@miucha.iecc.com>,
John Levine <johnl@iecc.com> wrote:
>>> FWIW, I had a compsci instructor who said logical IFs were inefficient and
>>> would not allow their use in our programs. I strongly suspect she was
>>> wrong,
>>> at least at the time we were using it (S/360-65).
>
> She was wrong. In Fortran G the arithmetic IF probably generated
> useless branches, in Fortran H it's have been optimized to be the
> same.
>
>> For complex conditions LOGICAL IF's are better, you want the logic
>> spread over 5, 6 or more IF Statements?! Anyway breaking the pipeline
>> is far more time consuming than a logical statement, if there be any
>> loss of efficiency.
>
> Pipeline? What's a pipeline? This was a 360/65. Next thing you know
> you'll be talking about this mythical thing called a "cache".
Do people run very intensive calculation on the descendants of the
360, or have said systems cached up?
--
Never attribute to stupidity that which can be explained by greed. Me.
|
|
|
|
Members
Search
Help
Register
Login
Home
