IBM System 360 Halve Instructions

Talking about the Halve instructions, they were an oddity as
originally
implemented.
   To be sure, they halved an FPN, but did not normalize afterwards.
Hence, an extra bit could be lost at the least-significant end.
To avoid further loss of least-significant bits in subsequent add/
subtract operations, it was necessary to normalize, presumably by
adding zero.

   The main advantage of the halve instructions was that they were
much faster than using a divide instruction.

   Why the Halve instructions were done like that is puzzling,
because it wouldn't have taken much hardware to normalize the result.
0
Robin
2/2/2013 2:30:05 PM
comp.lang.pl1 1694 articles. 0 followers. Post Follow

13 Replies
113 Views

Similar Articles

[PageSpeed] 40
On 2013-02-02 14:30:05 +0000, Robin Vowels said:

> Talking about the Halve instructions, they were an oddity as
> originally
> implemented.
>    To be sure, they halved an FPN, but did not normalize afterwards.
> Hence, an extra bit could be lost at the least-significant end.
> To avoid further loss of least-significant bits in subsequent add/
> subtract operations, it was necessary to normalize, presumably by
> adding zero.
> 
>    The main advantage of the halve instructions was that they were
> much faster than using a divide instruction.
> 
>    Why the Halve instructions were done like that is puzzling,
> because it wouldn't have taken much hardware to normalize the result.

Since the main use of the Halve instructions was in calculating square 
roots, I suppose it was thought unimportant, as the successive results 
still had to converge to the true square root, anyway, so that only one 
bit, at most, would be lost in the final result.

-- 
John W Kennedy
If Bill Gates believes in "intelligent design", why can't he apply it 
to Windows?

0
John
2/3/2013 5:25:56 AM
John W Kennedy <jwkenne@attglobal.net> wrote:

(snip)
>>    Why the Halve instructions were done like that is puzzling,
>> because it wouldn't have taken much hardware to normalize the result.

> Since the main use of the Halve instructions was in calculating square
> roots, I suppose it was thought unimportant, as the successive results
> still had to converge to the true square root, anyway, so that only one
> bit, at most, would be lost in the final result.

The usual S/360 square root does the last iteration as

   Y2=(Y1-X/Y1)/2+X/Y1

(Well, PL/I (F) does it that way.)

I believe that works fine even if halve doesn't have a guard digit.

More specifically, it does:

         DE    FA,OP00(DR)
         AU    FA,OP00(DR)
         HER   FA,FA               Y1=(Y0+X/Y0)/
         DER   FB,FA
         MVC   OP01(3,DR),ROUND
         AU    FA,OP00(DR)         ROUND OFF X/Y
         SER   FA,FB               Y2=(Y1-X/Y1)/2 + X/Y1 IN ORDER
         HER   FA,FA               TO GUARD LAST DIGIT
         AER   FA,FB

-- glen
0
glen
2/3/2013 9:34:35 AM
On Feb 3, 4:25=A0pm, John W Kennedy <jwke...@attglobal.net> wrote:
> On 2013-02-02 14:30:05 +0000, Robin Vowels said:
>
> > Talking about the Halve instructions, they were an oddity as
> > originally
> > implemented.
> > =A0 =A0To be sure, they halved an FPN, but did not normalize afterwards=
..
> > Hence, an extra bit could be lost at the least-significant end.
> > To avoid further loss of least-significant bits in subsequent add/
> > subtract operations, it was necessary to normalize, presumably by
> > adding zero.
>
> > =A0 =A0The main advantage of the halve instructions was that they were
> > much faster than using a divide instruction.
>
> > =A0 =A0Why the Halve instructions were done like that is puzzling,
> > because it wouldn't have taken much hardware to normalize the result.
>
> Since the main use of the Halve instructions was in calculating square
> roots,

One use.
I opened up IBM's Scientific Subroutine package at a random page:
Subroutine MEAT has two divisions by 2.
Elsewhere in the manual I found multiplications by 0.5.

Generally, in algebraic formulas, one sees division by 2.

> I suppose it was thought unimportant, as the successive results
> still had to converge to the true square root, anyway, so that only one
> bit, at most, would be lost in the final result.

One bit doesn't sound like much, but is significant in single
precision, considering that 24 mantissa bits are available,
and that four leading bits would be then zero.

Should the operation following HER be add or subtract, the fact that
the value may not be normalized will cause the loss of up to four
low-order bits when the magnitude of the other operand is such that
its mantissa must be shifted down during pre-normalisation.

I suppose that for single-precision evaluation of a formula,
the result could have been formed in an FPR that is first cleared
with SDR r,r and when the halve operation is carried out,
HDR r,r is used rather than HER.  The result could then be
normalised with some relevant instruction.
0
Robin
2/3/2013 10:54:13 AM
Robin Vowels <robin.vowels@gmail.com> wrote:
> On Feb 3, 4:25 pm, John W Kennedy <jwke...@attglobal.net> wrote:
>> On 2013-02-02 14:30:05 +0000, Robin Vowels said:
>> 
>>> Talking about the Halve instructions, they were an oddity as
>>> originally
>>> implemented.
>>>    To be sure, they halved an FPN, but did not normalize afterwards.
>>> Hence, an extra bit could be lost at the least-significant end.
>>> To avoid further loss of least-significant bits in subsequent add/
>>> subtract operations, it was necessary to normalize, presumably by
>>> adding zero.
>> 
>>>    The main advantage of the halve instructions was that they were
>>> much faster than using a divide instruction.
>> 
>>>    Why the Halve instructions were done like that is puzzling,
>>> because it wouldn't have taken much hardware to normalize the result.
>> 
>> Since the main use of the Halve instructions was in calculating square
>> roots,
> 
> One use.
> I opened up IBM's Scientific Subroutine package at a random page:
> Subroutine MEAT has two divisions by 2.
> Elsewhere in the manual I found multiplications by 0.5.
> 
> Generally, in algebraic formulas, one sees division by 2.

And did any of these compile to Halve instructions on pre-1967 S/360
compilers?

Anyway, it is well known that the original S/360 floating point was not all
it should have been. That’s why the 1967 EC was done, and that’s why the
IEEE designed a whole new floating-point format from scratch.
0
John
2/3/2013 2:16:35 PM
On Feb 4, 1:16=A0am, John W Kennedy <jwke...@attglobal.neg> wrote:
> Robin Vowels <robin.vow...@gmail.com> wrote:
> > On Feb 3, 4:25 pm, John W Kennedy <jwke...@attglobal.net> wrote:
> >> On 2013-02-02 14:30:05 +0000, Robin Vowels said:
>
> >>> Talking about the Halve instructions, they were an oddity as
> >>> originally
> >>> implemented.
> >>> =A0 =A0To be sure, they halved an FPN, but did not normalize afterwar=
ds.
> >>> Hence, an extra bit could be lost at the least-significant end.
> >>> To avoid further loss of least-significant bits in subsequent add/
> >>> subtract operations, it was necessary to normalize, presumably by
> >>> adding zero.
>
> >>> =A0 =A0The main advantage of the halve instructions was that they wer=
e
> >>> much faster than using a divide instruction.
>
> >>> =A0 =A0Why the Halve instructions were done like that is puzzling,
> >>> because it wouldn't have taken much hardware to normalize the result.
>
> >> Since the main use of the Halve instructions was in calculating square
> >> roots,
>
> > One use.
> > I opened up IBM's Scientific Subroutine package at a random page:
> > Subroutine MEAT has two divisions by 2.
> > Elsewhere in the manual I found multiplications by 0.5.
>
> > Generally, in algebraic formulas, one sees division by 2.
>
> And did any of these compile to Halve instructions on pre-1967 S/360
> compilers?

Whether they did or not is irrelevant.
I dispute your claim that the main use of HxR
was for computing square root.

> Anyway, it is well known that the original S/360 floating point was not a=
ll
> it should have been. That=92s why the 1967 EC was done, and that=92s why =
the
> IEEE designed a whole new floating-point format from scratch.

There were various floating-point hardware implementations in
operation, having different numbers of mantissa bits and exponent
ranges.
0
Robin
2/3/2013 10:55:32 PM
John W Kennedy <jwkenne@attglobal.neg> wrote:

(snip, someone wrote)

>> Generally, in algebraic formulas, one sees division by 2.
 
> And did any of these compile to Halve instructions on pre-1967 S/360
> compilers?

Does it compile to HALVE on any compiler?
 
> Anyway, it is well known that the original S/360 floating point was not all
> it should have been. That???s why the 1967 EC was done, and that???s why the
> IEEE designed a whole new floating-point format from scratch.

-- glen
0
glen
2/3/2013 11:08:45 PM
On 2013-02-03 22:55:32 +0000, Robin Vowels said:

> On Feb 4, 1:16 am, John W Kennedy <jwke...@attglobal.neg> wrote:
>> Robin Vowels <robin.vow...@gmail.com> wrote:
>>> On Feb 3, 4:25 pm, John W Kennedy <jwke...@attglobal.net> wrote:
>>>> On 2013-02-02 14:30:05 +0000, Robin Vowels said:
>> 
>>>>> Talking about the Halve instructions, they were an oddity as
>>>>> originally
>>>>> implemented.
>>>>>    To be sure, they halved an FPN, but did not normalize afterwards.
>>>>> Hence, an extra bit could be lost at the least-significant end.
>>>>> To avoid further loss of least-significant bits in subsequent add/
>>>>> subtract operations, it was necessary to normalize, presumably by
>>>>> adding zero.
>> 
>>>>>    The main advantage of the halve instructions was that they were
>>>>> much faster than using a divide instruction.
>> 
>>>>>    Why the Halve instructions were done like that is puzzling,
>>>>> because it wouldn't have taken much hardware to normalize the result.
>> 
>>>> Since the main use of the Halve instructions was in calculating square
>>>> roots,
>> 
>>> One use.
>>> I opened up IBM's Scientific Subroutine package at a random page:
>>> Subroutine MEAT has two divisions by 2.
>>> Elsewhere in the manual I found multiplications by 0.5.
>> 
>>> Generally, in algebraic formulas, one sees division by 2.
>> 
>> And did any of these compile to Halve instructions on pre-1967 S/360
>> compilers?
> 
> Whether they did or not is irrelevant.
> I dispute your claim that the main use of HxR
> was for computing square root.

Oh for God's sake, /think/!

>> Anyway, it is well known that the original S/360 floating point was not all
>> it should have been. That’s why the 1967 EC was done, and that’s why the
>> IEEE designed a whole new floating-point format from scratch.
> 
> There were various floating-point hardware implementations in
> operation, having different numbers of mantissa bits and exponent
> ranges.

What era are you talking about?

-- 
John W Kennedy
"I want everybody to be smart. As smart as they can be. A world of 
ignorant people is too dangerous to live in."
  -- Garson Kanin. "Born Yesterday"

0
John
2/4/2013 1:39:28 AM
On 2013-02-03 23:08:45 +0000, glen herrmannsfeldt said:

> John W Kennedy <jwkenne@attglobal.neg> wrote:
> 
> (snip, someone wrote)
> 
>>> Generally, in algebraic formulas, one sees division by 2.
> 
>> And did any of these compile to Halve instructions on pre-1967 S/360
>> compilers?
> 
> Does it compile to HALVE on any compiler?

It wouldn't surprise me if it did nowadays; modern compilers, pushed by 
advances in the state of the art and pulled by RISC and the modern 
distaste for assembly language, do the most amazing optimizations. But 
I'm fairly certain it didn't in 1965-1967. (I have a vague memory of 
actually trying the experiment, but I could be just imagining it.)

-- 
John W Kennedy
"Give up vows and dogmas, and fixed things, and you may grow like That. 
....you may come to think a blow bad, because it hurts, and not because 
it humiliates.  You may come to think murder wrong, because it is 
violent, and not because it is unjust."
  -- G. K. Chesterton.  "The Ball and the Cross"

0
John
2/4/2013 1:53:42 AM
On Feb 4, 12:39=A0pm, John W Kennedy <jwke...@attglobal.net> wrote:
> On 2013-02-03 22:55:32 +0000, Robin Vowels said:
>
>
>
>
>
>
>
>
>
> > On Feb 4, 1:16=A0am, John W Kennedy <jwke...@attglobal.neg> wrote:
> >> Robin Vowels <robin.vow...@gmail.com> wrote:
> >>> On Feb 3, 4:25 pm, John W Kennedy <jwke...@attglobal.net> wrote:
> >>>> On 2013-02-02 14:30:05 +0000, Robin Vowels said:
>
> >>>>> Talking about the Halve instructions, they were an oddity as
> >>>>> originally
> >>>>> implemented.
> >>>>> =A0 =A0To be sure, they halved an FPN, but did not normalize afterw=
ards.
> >>>>> Hence, an extra bit could be lost at the least-significant end.
> >>>>> To avoid further loss of least-significant bits in subsequent add/
> >>>>> subtract operations, it was necessary to normalize, presumably by
> >>>>> adding zero.
>
> >>>>> =A0 =A0The main advantage of the halve instructions was that they w=
ere
> >>>>> much faster than using a divide instruction.
>
> >>>>> =A0 =A0Why the Halve instructions were done like that is puzzling,
> >>>>> because it wouldn't have taken much hardware to normalize the resul=
t.
>
> >>>> Since the main use of the Halve instructions was in calculating squa=
re
> >>>> roots,
>
> >>> One use.
> >>> I opened up IBM's Scientific Subroutine package at a random page:
> >>> Subroutine MEAT has two divisions by 2.
> >>> Elsewhere in the manual I found multiplications by 0.5.
>
> >>> Generally, in algebraic formulas, one sees division by 2.
>
> >> And did any of these compile to Halve instructions on pre-1967 S/360
> >> compilers?
>
> > Whether they did or not is irrelevant.
> > I dispute your claim that the main use of HxR
> > was for computing square root.
>
> >> Anyway, it is well known that the original S/360 floating point was no=
t all
> >> it should have been. That=92s why the 1967 EC was done, and that=92s w=
hy the
> >> IEEE designed a whole new floating-point format from scratch.
>
> > There were various floating-point hardware implementations in
> > operation, having different numbers of mantissa bits and exponent
> > ranges.
>
> What era are you talking about?

Before IEEE FP, obviously.
0
Robin
2/4/2013 2:50:24 AM
Apart from the obvious division by 2, some constants lend themselves
to Halve instructions (as they were originally).

Multiplying by 1.5, for example, which can be performed by
HER r2,r0
AER r0,r2
without having to worry about any lost bits; also 1.25.
Similarly, divide by 16.
0
Robin
2/4/2013 3:05:48 AM
On Feb 3, 6:28 am, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:
> Robin Vowels <robin.vow...@gmail.com> wrote:
> > Talking about the Halve instructions, they were an oddity as
> > originally implemented.
> >   To be sure, they halved an FPN, but did not normalize afterwards.
> > Hence, an extra bit could be lost at the least-significant end.
> > To avoid further loss of least-significant bits in subsequent add/
> > subtract operations, it was necessary to normalize, presumably
> > by adding zero.
> >   The main advantage of the halve instructions was that they were
> > much faster than using a divide instruction.
>
> Even so, HALVE has to be used carefully.

That's what I said.

> Multiplying by two and then HALVE can lose low bits,

No-body would want to do that.
If you multiply by 2 and then halve the result, you finish up with
what you had to start with (apart from possible changes at the
least-significant end).

> such that
> the last iteration of square root is done differently on non
> base two machines.

As on the S/360 before the advent of the guard digit.
But not on the S/360 and later with the guard digit.

Without the guard digit, any F.P. computation that finished
with a division by 2, 4, 8, etc, via HxR, such as average of two
values, could lose a bit.
It's not something that was specific to square root.
0
Robin
2/4/2013 12:49:30 PM
Robin Vowels <robin.vowels@gmail.com> wrote:
> On Feb 4, 12:39 pm, John W Kennedy <jwke...@attglobal.net> wrote:
>> On 2013-02-03 22:55:32 +0000, Robin Vowels said:
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>>> On Feb 4, 1:16 am, John W Kennedy <jwke...@attglobal.neg> wrote:
>>>> Robin Vowels <robin.vow...@gmail.com> wrote:
>>>>> On Feb 3, 4:25 pm, John W Kennedy <jwke...@attglobal.net> wrote:
>>>>>> On 2013-02-02 14:30:05 +0000, Robin Vowels said:
>> 
>>>>>>> Talking about the Halve instructions, they were an oddity as
>>>>>>> originally
>>>>>>> implemented.
>>>>>>>    To be sure, they halved an FPN, but did not normalize afterwards.
>>>>>>> Hence, an extra bit could be lost at the least-significant end.
>>>>>>> To avoid further loss of least-significant bits in subsequent add/
>>>>>>> subtract operations, it was necessary to normalize, presumably by
>>>>>>> adding zero.
>> 
>>>>>>>    The main advantage of the halve instructions was that they were
>>>>>>> much faster than using a divide instruction.
>> 
>>>>>>>    Why the Halve instructions were done like that is puzzling,
>>>>>>> because it wouldn't have taken much hardware to normalize the result.
>> 
>>>>>> Since the main use of the Halve instructions was in calculating square
>>>>>> roots,
>> 
>>>>> One use.
>>>>> I opened up IBM's Scientific Subroutine package at a random page:
>>>>> Subroutine MEAT has two divisions by 2.
>>>>> Elsewhere in the manual I found multiplications by 0.5.
>> 
>>>>> Generally, in algebraic formulas, one sees division by 2.
>> 
>>>> And did any of these compile to Halve instructions on pre-1967 S/360
>>>> compilers?
>> 
>>> Whether they did or not is irrelevant.
>>> I dispute your claim that the main use of HxR
>>> was for computing square root.
>> 
>>>> Anyway, it is well known that the original S/360 floating point was not all
>>>> it should have been. That’s why the 1967 EC was done, and that’s why the
>>>> IEEE designed a whole new floating-point format from scratch.
>> 
>>> There were various floating-point hardware implementations in
>>> operation, having different numbers of mantissa bits and exponent
>>> ranges.
>> 
>> What era are you talking about?
> 
> Before IEEE FP, obviously.

Then they might have chosen 360 hexadecimal format. They didn’t.

In fact, I well remember numerical analysts decrying 360 hexadecimal in the
1967 era. Some of their complaints were addressed by the great EC, but some
simply couldn’t be.
0
John
2/4/2013 1:43:20 PM
Robin Vowels <robin.vowels@gmail.com> wrote:
> On Feb 3, 6:28 am, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:
>> Robin Vowels <robin.vow...@gmail.com> wrote:
>>> Talking about the Halve instructions, they were an oddity as
>>> originally implemented.
>>>   To be sure, they halved an FPN, but did not normalize afterwards.
>>> Hence, an extra bit could be lost at the least-significant end.
>>> To avoid further loss of least-significant bits in subsequent add/
>>> subtract operations, it was necessary to normalize, presumably
>>> by adding zero.
>>>   The main advantage of the halve instructions was that they were
>>> much faster than using a divide instruction.
>> 
>> Even so, HALVE has to be used carefully.
> 
> That's what I said.
> 
>> Multiplying by two and then HALVE can lose low bits,
> 
> No-body would want to do that.
> If you multiply by 2 and then halve the result, you finish up with
> what you had to start with (apart from possible changes at the
> least-significant end).
> 
>> such that
>> the last iteration of square root is done differently on non
>> base two machines.
> 
> As on the S/360 before the advent of the guard digit.

You mean on the 360 before the guard digit from 32-bit format was
reproduced on 64-bit format. 

> But not on the S/360 and later with the guard digit.
> 
> Without the guard digit, any F.P. computation that finished
> with a division by 2, 4, 8, etc, via HxR, such as average of two
> values, could lose a bit.
> It's not something that was specific to square root.
0
John
2/4/2013 1:43:20 PM
Reply:
Similar Artilces:

Auction: IBM 5360 Mainframe and IBM 5225 Printer
>http://cgi.ebay.com/IBM-5360-Mainframe-and-IBM-5225-Printer_W0QQitemZ130041937602QQihZ003QQcategoryZ162QQrdZ1QQcmdZViewItem< End time: Nov-04-06 10:44:14 PST Hutchinson, Kansas Folks, we are not in Kansas anymore... Has a pile of manuels, 3 monitors, three keyboards. Louis Ohland wrote: > Folks, we are not in Kansas anymore... Has a pile of manuels, 3 > monitors, three keyboards. As you would expect, each keyboard mates with one of the 5251 displays. It's easy to tell which display was the system console (the badly burned-in one on top) and which was th...

Special march Deal On DS10 systems
We are offering a blow out deal for Alphaserver DS10 These offers are NOT available to anyone but OpenVMS users and are NOT on our website Quote Promotion: DS10-466-SPCOV-M06 if ordering by phone or email/fax Alphaserver DS10 466Mhz EV6 512MB Memory 36GB U160 Disk SCSI ATI Radeon 7500 PCI Graphics Card Dual Channel U160 Controller Dual On Board 10/100 Ethernet Keyboard Mouse and Power Cord Only $2995 while stocks last This is WAY WAY below market pricing The pricing above does NOT include licenses Subject to availability Systems are reconditioned - look superb and carry Island's ...

ibm thinkpad 390x, sound card
i've just bought an ibm 390x old laptop. there is probably neomagic256av sound/graphic (?) card in it. i'm trying to configure soundcard but it is not easy. there are some threads about it in group archive, but i haven't found any complete solution. there are advices to start linux from msdos (i'm not sure if i like it) and some else which doesn't work for me. if i configure my card as ESS ES1969, i can hear only some noise when i try to play mp3. in linux kernel 2.6.0 soundcard support has been much changed. oss has become depreciated. the primary sound architecture is now...

US-TX-Austin: Sr. Systems Design Eng., Ph,D in Electrical and Computer Eng.; C- (45323057607)
US-TX-Austin: Sr. Systems Design Eng., Ph,D in Electrical and Computer Eng.; C- (45323057607) ============================================================================================== Position: Sr. Systems Design Eng. Reference: SMC01593 Location: Austin TX Duration: C-P Skills: Ph.D. in Electrical and Computer Engineering Scope: Research, design, and develop system level algorithms for data conversion and communication equipment mixed signal integrated circuits applications; apply computer and electric...

System restore thinks D: is system disk
I can not use system restore in XP Home Ed.. When I try, I get a message that claims drive D is the system disk. This might be related to the fact that my computer will not do a warm reboot. It says that boot files can not be found. If I turn off, then start, it works fine. This all started when I installed a new HD, transfering everything over, then keeping the original drive as D: drive. I sure hope that someone can help! Thanks, Rich The message <1116679523.049840.71700@g44g2000cwa.googlegroups.com> from thedells2@verizon.net contains these words: > I can not use syst...

Which is the best Ruby Wiki system like implemented so far?
I starting a Knowlege Manament project and would like to have as reference the best "Wiki" system implemented using Ruby. This project is intended to use a non linear Wiki system like TiddlyWiki (see http://www.tiddlywiki.com) and also a full cosmology implementation of the Personal Brain functionality (see http://www.thebrain.com). I have looked around about some, but it semeed to me not be able to know its current status e.g. instiki. So can anybody help me on that? Roberto Nogueira http://holoflux.wordpress.com On 9/4/06, zoat <enogrob@hotmail.com> wrote: > I startin...

IBM opens Linux Training Centre
http://www.reporter.gr/fulltext_eng.cfm?id=60324102743 ...

[News] Good Linux Systems Come to the Backbone
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Astaro Security Gateway boasts rich features ,----[ Quote ] | How important are flexibility and a rich feature set to you? If these | elements are your top considerations, then the Astaro Security Gateway should | be high on your short list. With roots in the Linux world, the Astaro is a | serious firewall with serious capabilities for a distributed enterprise UTM | box. `---- http://news.idg.no/cw/art.cfm?id=8343CC13-1A64-67EA-E460F07C187BFA98 NAS system houses 2.5-inch drives for up to 6TB ,----[ Quote ] | Qnap Systems announced a 2....

aberrant behavior of pthread_exit() on different POSIX-aware systems
Hello, Please consider following rather useless piece of code: #include <stdio.h> #include <stdlib.h> #include <pthread.h> #include <errno.h> #include <unistd.h> #define NUM_THREADS 5 void byebye(void); void print_data(void); void *workload(void *arg); struct thread_data { pthread_t tid; int sum; }; struct thread_data thread_data_array[NUM_THREADS]; void byebye(void) { printf("main thread [%d] rip\n", (int) pthread_self()); print_data(); } void print_data(void) { int t; ...

Hypergraph rewrite system?
Hi folks, (please post as my mailbox is filled to the brim with usenet spam) I am looking for a tool (or pointers to some methods) which I can do deduction over (coloured) hyper graphs. That is, I want to be able to draw out hyper graphs, and apply re-write rules to them (the rules themseleves are hypergraphs) to do deductions. The idea would be to look at the left hand side of the rule, do sub(hyper)graph isomorpism on your 'expression' and replace that found subgraph with the right hand side of the rule. That is something like: A: //The argument we will pass to the rewrite ...

The IBM Multimedia collaboration
http://www.scientificsolutions.com/archives/archives.html In 1972 ... Dr. Martin Alpert starts Tecmar Scientific Solutions in Cleveland Ohio, the name "Tecmar" being derived from MARty's TEChnology....In the history of data acquisition, Tecmar and Scientific Solutions are actually the same company. Tecmar and Scientific Solutions - IBM MultiMedia collaboration In 1986, IBM approached Tecmar to design and manufacture a number of multimedia products. These products were to be designed and manufactured by Tecmar and marketed by IBM. Scientific Solutions provided the ...

Developers petition IBM to open source OS/2 [article]
"In the wake of IBM declaring that it is dropping support for it, over 11,000 people have signed a petition now up on the OS/2 World Web site asking the company to open source portions of it, if not the entire operating system. Supporters of the petition argue that in so doing the company's long-suffering user and development communities can continue to be well served." http://weblog.infoworld.com/techwatch/archives/001696.html -- Posted with OS/2 Warp 4.52 and IBM Web Browser v2.0.5 letoured@nospam.net wrote: > In <11e04m6jc5tcv00@corp.supernews.com>, on 07/21/...

Auction: IBM PC Computer Personal System PS/2 57 SX Type 8557 NR
8557SX (386SX-20) http://search.ebay.com/180213696883 (Warning: Lots of pictures! Plan your visit accordingly!) 1 bid of at least $0.99. Located in Pen Argyl, PA USA. eBay says the seller ships worldwide, shipping to the US is listed as $27.70. Ends Feb-13-2008, 08:59:55 PST (1 day, 21 hours). Good condition, comes with a dual async. Needs some cleaning and a set of screws to hold the case cover on. Nice box, think of it as a more expandable version of the 56SX/SLC/ SLC2/3 boxen. The processor can be upgraded, as IBM provided a "special socket" to do so. You can go ...

IBM
It frightens me that IBM is pushing for more Linux. They have gone to the point of releasing some of their software patents to help spark the growth of Linux. I for one shudder at the fact that some day companies like IBM may make a power play for the control of Linux software development. It will not be in ownership rights, because of the GPL. However, if the corporations of the world start "needing" open source, they will try to control and steer its development by offering large "donations" to get what they are depending on. If open source developers accept these &qu...

better way to fetch and install packages than just a bunch of system calls?
Howdy all! As part of a test script, I have to make sure my test machine (i.e. the machine that is running the script) has libraries zlib and hdf5. I do that with this code. Is there a better way? if ($opt{netcdf4} && ! -e $install_file) { copy("/upc/share/ed/downloads/zlib-$ {zlib_version}.tar.gz", $hdf5dir) || die; copy("/upc/share/ed/downloads/hdf5-$ {hdf5_version}.tar.gz", $hdf5dir) || die; chdir($hdf5dir) || die; system("gunzip -f zlib-${zlib_version}.tar.gz"); system("tar xf z...

Prometric & Pearson VUE Exam Center
Prometric & Pearson VUE Exam Center - Cisco,Microsoft,Sun,HP,IBM,ITIL,Websense Best Prometric and Pearson VUE Test center in Bangalore ( For the certification of=A0Microsoft, Sun, CISCO, Exin - ITIL, IBM,=A0HP, Citrix, etc.) =A0 Hi, =A0 Greetings from Everonn!!!! =A0 Are you looking for the best=A0Prometric and VUE exam center for your certification like Microsoft, Sun, CISCO, IBM, and HP, where you can take test whenever you want...? =A0 Then you are at the right place; =A0 Just come and take test whenever you are free..... =A0 We understand your difficulty while registering for your te...

(203) US-CT-Fairfield County: Senior Systems Analyst
********************************************************************** This job posting is brought to you by ISHUNTER.COM, offering computer jobs for programmers, engineers, systems analysts, network administrators and other high technology professionals. Search over 3,000 IS/IT jobs in the United States by clicking on http://www.ISHunter.com ********************************************************************** Title ...... Senior Systems Analyst - Systems Development Skills ..... program deploy develop Microsoft DNA .Net Java MSMQ SQL Server COM MTS Location ... Fa...

TWB: System 7.0 or System 7.1 floppies for Mac SE
I have a toaster Mac SE on which I wish to install Clarisworks 2.1 (which I already have). But Clarisworks 4.0 requires either System 7.0 or System 7.1 . Regrettably, the Mac SE has an early version of System 6.0. Does anyone out there have System 7.0 or System 7.1 for sale? Either originals or backups would suffice. BTW, I have never updated a Mac system. Do the disks have the self-installing feature, or do you need to move files into various folders, manually??? Please reply to me at: buggie@unm.edu Thanks, Stephen Buggie Steve Buggie | ...

IBM T30 BIOS MiniPCI
http://www.srcf.ucam.org/~mjg59/thinkpad/wireless.html Hello, My IBM T30 laptop requires a lil' hardware hacking (BIOS to be specific). Has anyone tried this hack? Any pointers or suggestions? Thanks, JiMbO J0nES ...

Your Xbox 360 Red Light ERROR ??
Get Easy Step By Step Guide to Fix. In 1 Hr Get Rid of Red Light Error http://xbox-redlight-review.123probiz.com Go XBox Mad :) Siswantoro Supervisor http://www.gogonai.info/?id=scumapro ...

how do i kerberos aunthetication in AIX system
dear team, how do i use kerberos client aunthetication in AIX server In article <1143364793.376303.246810@t31g2000cwb.googlegroups.com>, rohan <suddhasil@gmail.com> wrote: >dear team, >how do i use kerberos client aunthetication in AIX server With great difficulty. I eventually gave up, and disabled it. However, despite following the instructions, that broke it even worse. It is a pretty fair nightmare - even by Kerberos standards, and it is ALWAYS a nightmare :-( Regards, Nick Maclaren. ...

WTT: 32MB SIMMs
I have (6) sticks of SiliconMountain Memory (Parity, Gold leads, 70ns), product # SMM4342-9661. These SIMMs are supposed to be equivalent to IBM 92G7205. This memory works fine in my PC 750 MCA/PCI (6886) under Win98. But it does NOT work reliability under Windows NT 3.51 (My current OS). I would like to trade 192MB of SiliconMountain Memory for (2) sticks of IBM 92G7205 SIMMs. Please contact me via email at gorenbergh @ mags . net. Thanks, Steve (eTh1nk) ...

Auction: IBM Model 9595 Microchannel Mainboard w/32M IBM RAM
http://search.ebay.com/310005766659 End time: Dec-13-07 18:16:52 PST Single parallel / single serial planar, op panel and floppy cables, 32MB FPM. These folks are in contention for the coveted "Dollar Wrench" award from the Ardent Tool. See the price and make up your mind. Item: IBM Model 9595 Microchannel Mainboard w/32M IBM RAM Description: IBM Model 9595 Microchannel Mainboard w/32M IBM RAM - FRU #04G3863 - includes (8) 4MB parity SIMMS (32MB total) - includes device cables ...

Equivalent of system()?
Is there a Python equivalent of C++'s system()? TIA On 2008-02-23, Max <maxh.is.here@gmail.com> wrote: > Is there a Python equivalent of C++'s system()? The closest thing is probably system(). It's in the os module. -- Grant Max wrote: > Is there a Python equivalent of C++'s system()? More or less. You probably want subprocess.Popen: >>> import subprocess >>> subprocess.Popen("echo hello", shell=True) hello <subprocess.Popen object at 0x2ab8f3665d10> http://docs.python.org/lib/node533.html#CHILD_LINKS Thanks for the ...

IBM Using iPod to boot Linux on PCs
<http://apple.slashdot.org/apple/05/03/11/141246.shtml?tid=136&tid=126&tid=181&tid=201&tid=1&tid=106> Applejack writes "Looks like iPod fever has caught on to Big Blue. IBM has a yet unreleased iPod-based software for rescue, restore, and recovery of failed Windows PCs. I read this description of the software on Amit Singh's blog, whose group at IBM apparently created this stuff. If I understand this correctly (and I think I do), the iPod contains IBM's rescue software along with Linux. A crashed PC boots into Linux from the i...