Message boards : News : Status/Plans, 7th August 2012
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 851
Credit: 1,612,796
RAC: 174
Message 24559 - Posted: 10 Aug 2012, 9:49:29 UTC - in response to Message 24556.  

No; these jobs are re-running some incomplete cases to finish the
current studies. More news when I start the intensity scan. Eric.
ID: 24559 · Report as offensive     Reply Quote
grumpy

Send message
Joined: 1 Sep 04
Posts: 54
Credit: 1,803,161
RAC: 13
Message 24560 - Posted: 10 Aug 2012, 21:37:49 UTC
Last modified: 10 Aug 2012, 21:44:19 UTC

Just love fpu's,alu's,Floating-point libraries etc.

My favorite number is 31.1

Pop these in your favorite spreed sheet,and see what you get.

((186.6/31.1)-6)

(((6*31.1)/31.1)-6)

(6*31.1/31.1)-6

or

cell expression
A1 6*31.1
A2 6
A3 A1/31.1
A4 1/(A3-A2)
A5 1/A3-A2
ID: 24560 · Report as offensive     Reply Quote
BigBrownBear

Send message
Joined: 17 Aug 05
Posts: 2
Credit: 72,321
RAC: 0
Message 24634 - Posted: 17 Aug 2012, 23:29:16 UTC - in response to Message 24527.  

For you to gets identical bit for bit 0 ULP different results after many gigaflops with 5 different FORTRAN compilers at different levels of optimization, then you must be one serious programmer, FORTRAN or not. Besides, I haven't seen, much less thought about, FORTRAN since my days @ Alabama State University days in the late 90's.

I wonder if the source code was converted to come other high level programing language (ie C, COBOL, Pascal, APL, BASIC (ewwww!)) would their be a really big difference (speed/compiled code size/results from WU/ect.)?
<div>&nbsp;<img src="http://www.boincstats.com/signature/user_170873.gif"></div>
ID: 24634 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 851
Credit: 1,612,796
RAC: 174
Message 24641 - Posted: 18 Aug 2012, 13:38:24 UTC - in response to Message 24634.  

For you to gets identical bit for bit 0 ULP different results after many gigaflops with 5 different FORTRAN compilers at different levels of optimization, then you must be one serious programmer, FORTRAN or not. Besides, I haven't seen, much less thought about, FORTRAN since my days @ Alabama State University days in the late 90's.

Well thankyou, but I have made so many programming errors over the last
50 years I think I might say I am experienced.
Seriously I am going to publish soonest as I
think this is a minor breakthrough (previously thought to be perhaps
impossible) and a triumph for standardisation and the users versus rampant
commercial interests.

I wonder if the source code was converted to come other high level programing language (ie C, COBOL, Pascal, APL, BASIC (ewwww!)) would their be a really big difference (speed/compiled code size/results from WU/ect.)?


I don't think there would be an enormous difference either way in C, I don't think the other languages are really relevant.
Incidentally my methodology will work for C programmes too (and on GPUs) but I haven't verified/tested that experimentally yet, but I will.

Thanks for your comment; this was ten years work of which the last five were
"wasted" sailing into adverse winds. Eric.
ID: 24641 · Report as offensive     Reply Quote
Profile MAGIC Quantum Mechanic
Avatar

Send message
Joined: 24 Oct 04
Posts: 840
Credit: 36,847,510
RAC: 34,455
Message 24644 - Posted: 18 Aug 2012, 22:23:22 UTC - in response to Message 24641.  



Thanks for your comment; this was ten years work of which the last five were
"wasted" sailing into adverse winds. Eric.



I agree Eric, especially since I have been here checking every day for 8 years now.


Volunteer Mad Scientist For Life
ID: 24644 · Report as offensive     Reply Quote
William C Wilson
Avatar

Send message
Joined: 11 Sep 08
Posts: 25
Credit: 384,225
RAC: 0
Message 24646 - Posted: 19 Aug 2012, 4:54:09 UTC - in response to Message 24634.  

ahh, I just had to answer this one! Fortran in the 90s, my grandson, I was programing Fortran in the 1969 and 1970 at NCAR where had a super large computer, CDC 7600 and then Star, for those days, 10s of millions of dollars - and my desktop now runs circles around it. Largest program I had was a space simulation for heat control, took over 7,000 cards, and I punched ever single one of them. And used 3 and half drawers to give it to the operator to read it in, and pray he would not drop it or get just one card out of sequence. Ahh, just to compile was fun in those days.

But I know what you experienced, difference results for same program, same data, different versions of compilers in those days. Well sounds like you got the bull by the horns.

And I loved those series of language. I started out with Fortran, then we built a PDP-11 ground station for digital data, paper tape etc., and started to program thousands of lines in code in machine language. Of course I had a great library of functions and routines built up by the time I went to DEC years later, 1973.

I am glad that you all are doing so much there in science, and computing science. You are the silent goal keepers that make this huge project work.

Keep up the good work, and the good humor.
William C Wilson
São Paulo Brazil
ID: 24646 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 851
Credit: 1,612,796
RAC: 174
Message 24647 - Posted: 19 Aug 2012, 6:47:46 UTC - in response to Message 24646.  

Well I just have to reply here, before dashing off to the pool and then to CERN.
Memories, memories. I shall send you my CV if I can't get it on the WWW here.
First collaborated with Berkeley back in the 60's on the CDC 6600 and also
visited NCAR later (common issues with data handling and magnetic tapes)
as well as LLL and Los Alamos.
Sadly even if Fortran 95 and especially Fortran 2003 are modern and just as good
as, or even better than C++ there is just not the market and hence money
for implementations, and everyone learns C.
I remember the punched cards, which were a great improvement over our 5 hole/
8 hole paper tape. Cards were too expensive for us! Surprised you didn't have
a sequence number in Cols 73-80! :-) The CDC operating system was on
binary cards, quick patches consisted of punching a few more holes by hand
or re-inserting some chips.
Finally I should publicise my article 50 years of Fortran (at CERN). Eric.
ID: 24647 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 851
Credit: 1,612,796
RAC: 174
Message 24648 - Posted: 19 Aug 2012, 6:51:28 UTC - in response to Message 24646.  

......and PS. It is hot and dirty in the Engine Room, and goodness
knows who is steering the ship, but if you can't stand the heat
don't go in the kitchen!
ID: 24648 · Report as offensive     Reply Quote
Profile Tom95134

Send message
Joined: 4 May 07
Posts: 250
Credit: 826,541
RAC: 0
Message 24656 - Posted: 19 Aug 2012, 20:54:07 UTC - in response to Message 24647.  

I remember the punched cards, which were a great improvement over our 5 hole/8 hole paper tape.


Do you remember having to key in the boot sequence using the front panel switches so it would read the paper tape?
ID: 24656 · Report as offensive     Reply Quote
Profile MAGIC Quantum Mechanic
Avatar

Send message
Joined: 24 Oct 04
Posts: 840
Credit: 36,847,510
RAC: 34,455
Message 24657 - Posted: 19 Aug 2012, 22:21:35 UTC - in response to Message 24647.  


Finally I should publicise my article 50 years of Fortran (at CERN). Eric.



Yeah I just happened to read one you wrote 5 years ago for the Cern Courier.

You have been hard at work for Cern for a long time Eric.

-Samson


Volunteer Mad Scientist For Life
ID: 24657 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 851
Credit: 1,612,796
RAC: 174
Message 24659 - Posted: 20 Aug 2012, 4:12:15 UTC - in response to Message 24656.  

Well I remember the CDC 6600 (and 6400/6500) dead start panel
with 10/12? 12-bit peripheral processor instructions to boot the
system. Was pretty good in octal in those days. Shame we switched
to hex!
ID: 24659 · Report as offensive     Reply Quote
Ruud van der Kroef

Send message
Joined: 16 Aug 05
Posts: 3
Credit: 2,234,903
RAC: 1
Message 24661 - Posted: 20 Aug 2012, 11:30:19 UTC - in response to Message 24659.  

Yes, being a former CDC engineer, that brings up a lot of memories.
The dead start panel on the 6x00 had 12 rows of toggle switches, each row representing a 12-bit (PP) word. The later Cyber 180 (initialy Cyber 170/800) had 20 rows of 16 switches.
If I really think hard, I can come up with the dead start switch settings to dead start (IPL in IBM terms) the machine from punched cards / magnetic tape/ disk.
It's a long time since.

Ruud
ID: 24661 · Report as offensive     Reply Quote
tullio

Send message
Joined: 19 Feb 08
Posts: 586
Credit: 3,539,915
RAC: 2,845
Message 24662 - Posted: 20 Aug 2012, 14:42:37 UTC

I have completed a task on my new HP 635 laptop in 166 hours,and it was validated a week beyond its deadline.The laptop has an AMD E-450 CPU which uses only 18 W and seems to survive the heat wave we have now in Italy, while I had to shut down the SUN WS whose Opteron 1210 uses 75 W and has the fans going full speed. All this on Linux, SuSE SLES 11 sp1 on the laptop and OpenSUSE 12.1 on the SUN.
Tullio
ID: 24662 · Report as offensive     Reply Quote
William C Wilson
Avatar

Send message
Joined: 11 Sep 08
Posts: 25
Credit: 384,225
RAC: 0
Message 24676 - Posted: 21 Aug 2012, 16:36:09 UTC - in response to Message 24647.  

Go for 50 years of Fortran. Would love to have memories of that, well few more years to get 50 - LOL.

Yes, I had sequence holes punched, but problem, was when made them there fixes, you did not ever know if fixed or not, so skipped it or punched in the same number as continued. Even at NCAR, they used to look badly at you, if you decided to throw away a few drawers of cards. And those punches, like the O 29 as was better. And the card duplicators, had to hard wire in the last columns when ready to sequence them. Gee have come a long way, and I even used the Radio Shack TRS-80, with magnetic tape when came to Brazil. Bought one of the first ones, wow, 32 kb of memory.

Came thru customs here in Brazil, when moved here in 76, and they did not even know what it was so let me thru. Said it was a new type of calculator, which it really was basically.

Then we had at NCAR, the DD80 which we could make movies, frame by frame. To do clouds, we told the operator to defocus it slightly.

Want to be on TOP of your list for your book, go for it. New generation of programers would think it was from the stone age. LOL
William C Wilson
São Paulo Brazil
ID: 24676 · Report as offensive     Reply Quote
William C Wilson
Avatar

Send message
Joined: 11 Sep 08
Posts: 25
Credit: 384,225
RAC: 0
Message 24677 - Posted: 21 Aug 2012, 16:38:19 UTC - in response to Message 24656.  

Sounds like what I had to do when rebooted the DEC PDP-10 (DEC 10, models KA and KI). If was not for use, switch makers would have been out of business.
William C Wilson
São Paulo Brazil
ID: 24677 · Report as offensive     Reply Quote
Phil
Avatar

Send message
Joined: 26 Jul 05
Posts: 63
Credit: 4,083,755
RAC: 2
Message 24879 - Posted: 1 Oct 2012, 10:36:40 UTC - in response to Message 24647.  

Well I just have to reply here, before dashing off to the pool and then to CERN.
Memories, memories...
I remember the punched cards, which were a great improvement over our 5 hole/
8 hole paper tape. Cards were too expensive for us!


Well 5-hole tape was OK except for all those FIGS/LTRS and CR/LF conventions. "ZCZC" and "NNNN" are still in use to this day!
The problem with 8-hole tape was the Great DELete Conspiracy, where DEL on the Teletype110 meant "ignore the character I've just punched on this tape" but in ASCII meant "ignore the character preceding this one" which required that every tape we've punched over the last 5yrs had to be remade!
ID: 24879 · Report as offensive     Reply Quote
Previous · 1 · 2

Message boards : News : Status/Plans, 7th August 2012


©2019 CERN