On Dec 17, 2007, at 10:42 PM, Dave wrote:
>
> Hmm, the only difference I was aware of was the I/O
> module...keyboard/mouse interface or not.
>
> -Dave
For Onyx2 Deskside, the midplane is indeed different (has an XIO
connection to the graphics side, and a graphics side). Onyx2 rack uses
standard Origin2000 compute modules with the Xtown external XIO
interface linking the compute modules to the graphics module(s), which
in the case of a rack support 2 pipelines.
Deskside Onyx2 also comes with only null routers, so 4 procs is the
max. Some people have tried putting in full routers, but I don't recall
having heard any success. I don't think O2000 has this issue (except
for one model of IP31 (one of the 250MHz boards, I think) that has a
hard limit in the PROM about how many other nodes it will work with
before becoming sulky. Marketing department at work.
>
>Subject: Shift Registers as Delay Lines (was Delay lines in TV sets)
> From: "Rick Bensene" <rickb at bensene.com>
> Date: Mon, 17 Dec 2007 15:06:38 -0800
> To: "General Discussion: On-Topic and Off-Topic Posts" <cctalk at classiccmp.org>
>
>Tom Watson wrote:
>> If you decide to make up a delay line to hold CPU data, try a
>> prototype using shift registers. They are a bit easier to
>> make up, and often the chips are available. The problem is
>> that they come in weird (at times) sizes (132, 80, and the
>> like). I'll leave it to the reader to determine the
>> usability of odd sizes and their original use.
>>
Msot of those are later designs and tended to be for line printer,
CRT and other displey and printing systems. Common sizes for that
were line printer 132, CRT 80 and 72 (tty was 72!) and the true
oddball lengths were the 1024, 224, 192 and 56.
MOS shift registers of lengths greater than a few bits are very late
60s (after 67 or so). By early 70s parts 1024 long ere not uncommon.
What makes them interesting is you can use them word or digit parallel
and very deep by cascading them serially. For lesser registers older
4 and 8bit TTL devices in parallel can be very effecive in providing
a digit wide by 8 digits deep register in a small space. This is
truely getting into late 60s early 70s serial compurer design right
on the cusp of the days of the last serial drums or disks for local
store and the drums would then be for the larger program store.
Allison
>A lot of early MOS shift registers were developed specifically for use
>in electronic calculators, as solid-state replacements for magnetic core
>memory or magnetostrictive delay lines. Since most all electronic
>calculators in the mid-'60's through the late 70's operated in BCD or
>some alternate four (or sometimes five)-bit representation of decimal
>digits, the shift registers were usually made with a number of stages
>that was a multiple of four or five, with a few extra bits here and
>there for timing and synchronization. That's why many of these devices
>as an unusual number of stages. In some calculators from the late
>'60's, as IC logic had pretty much replaced discrete transistor designs,
>there were different versions of the same machine, earlier machines
>which used a magnetostrictive delay line, and "updated" versions which
>dispensed with the delay line, and replaced it with a number of MOS IC
>shift register devices. Functionally, the machines were identical. As
>far as the digital logic section of the machine went, also identical.
>The only real changes were the removal of the read amplifier and write
>driver for the delay line, and replacement with some simple
>level-shifting and power supply circuitry to properly drive the shift
>register chain.
>
>Rick Bensene
>The Old Calculator Museum
>http://oldcalculatormuseum.com
On Dec 16, 2007, at 11:19 PM, Richard wrote:
> OK, we talked earlier about how high this architecture could go and
> there was some mention of 256 and 512 processor systems. However, I
> just stumbled across a press release at LANL describing their
> installation of a massive 6144 processor system consisting of 48 sets
> of 128 nodes all arranged as a giant supercomputer:
>
Above 128 processors you need a special "Cray" router. Above either 512
or 1024 you'll probably need custom SGI setup, possibly a custom IRIX.
At any rate, 32 or 64 processors are pretty fast (remember as you scale
it's harder and harder to get parallel code to run across all
processors), and don't forget to factor in electricity and cooling.
Most hobbyists only run one or two racks.
The first LSI tester I ever worked on (manufactured by, I think, LSI
Testing Inc of Utah) used triple-66 bit MOS shift registers to store the
digital stimuli and compare patterns (3x64-bit registers plus 2 bits
used as control). Around 1970 or so at Hughes Microelectronics. Since we
manufactured MOS ICs (PMOS at the time) and the shift registers became
obsolete after a few years we designed our own as spares. The tester
itself used a PDP-8/L as the controller and was soon after replaced by a
bigger one which used a PDP8/I - my first contact with SEC PDP8
computers.
regards
Bob Adamson
> From: "Rick Bensene" <rickb at bensene.com>
>
> A lot of early MOS shift registers were developed specifically for use
> in electronic calculators, as solid-state replacements for magnetic
core
> memory or magnetostrictive delay lines. Since most all electronic
> calculators in the mid-'60's through the late 70's operated in BCD or
> some alternate four (or sometimes five)-bit representation of decimal
> digits, the shift registers were usually made with a number of stages
> that was a multiple of four or five, with a few extra bits here and
> there for timing and synchronization. That's why many of these
devices
> as an unusual number of stages. In some calculators from the late
> '60's, as IC logic had pretty much replaced discrete transistor
designs,
> there were different versions of the same machine, earlier machines
> which used a magnetostrictive delay line, and "updated" versions which
> dispensed with the delay line, and replaced it with a number of MOS IC
> shift register devices. Functionally, the machines were identical. As
> far as the digital logic section of the machine went, also identical.
> The only real changes were the removal of the read amplifier and write
> driver for the delay line, and replacement with some simple
> level-shifting and power supply circuitry to properly drive the shift
> register chain.
>
> Rick Bensene
> The Old Calculator Museum
> http://oldcalculatormuseum.com
>
Hello !
Save some old computer books, before they go for paper recycling.
I still have most of the collection available.
The books vary from the years 1970-1990
The collection contains books about:
LISP, Data management, Artificial Intelligence, Automatic data
processing, Computer programming etc.
>From publishers like:
McGraw Hill, M&T Books, Wiley Press, Microsoft
They are all in English language and in good condition.
You can get them for FREE, only pay for post delivery. The books are shipped
>from Slovenia, Europe.
The link to the list is available at:
http://yang.mtveurope.org/books.html
If you are interested, contact me to my private email address.
Kind regards,
Jan Prunk
--
Jan Prunk <janprunk at gmail.com> http://www.prunk.si
GnuPG: 1024D/00E80E86 77C5 156E 29A4 EB6C 1C4A 5EBA 414A 29F5 00E8 0E86
Second call....
Anyone interested in an HP "LaserJet+" for the cost of merely coming and
picking it up from Birmingham?
Also available, one IBM "ProPrinter XL24". Same deal.
The HP works but the IBM is in unknown condition.
These both need to be picked up by the end of January or they *WILL* get
dumped....same goes for the rest of my computer collection (more "free for
collection" postings to follow soon).
Contact me off-list if interested. Thanks.
TTFN - Pete.
There is a lot of really cool source code here; a DXF vector library, math and graphics libraries.....
Randy
_________________________________________________________________
Get the power of Windows + Web with the new Windows Live.
http://www.windowslive.com?ocid=TXT_TAGHM_Wave2_powerofwindows_122007
>
>Subject: Re: Homebrew Drum Computer
> From: ard at p850ug1.demon.co.uk (Tony Duell)
> Date: Sun, 16 Dec 2007 22:20:38 +0000 (GMT)
> To: cctalk at classiccmp.org
>
>> >I didn't think the NTSC system used a delay line at the receiver, but I
>> >must admit I've never repaired a US TV set.
>>
>> Yes they did/do but is was not ultrasonic. The common ones were basically
>> the same as LONG (4.5us?) transmission lines.
>
>Is that a luminance deleay line, to compensate for the different
>bandwidths of the luminance and chromanance channels? PAL sets have those
>as well,
>
>The delay line I was sgggesting for use as a computer memory device is a
>glass block wioth untrasonic transducers on it. It's got a delay period
>of almost one complete line-time, and is used to store one of the
>chromanace signals as part of the PAL decoding process.
I don't know about PAL but one line for NTSC is only 63us and thats
not enough delay.
Computers that used delay lines were in the milliseconds range as they
needed to store a lot of bits/digit/words.
Allison
>
>-tony
Performance.
These are generally the best compilers around, with support for parallelism, distributed memory, loop unrolling, support for vector hardware.... Even using the GPU as a general purpose processor.
High performance parallel computing with Fortran 90 and HPF.
Real men code in FORTRAN
;)
> Date: Mon, 17 Dec 2007 13:05:14 -0800
> From: chrism3667 at yahoo.com
> To: cctalk at classiccmp.org
> Subject: Re: have you seen www.fortran.com?
>
> I realize that old doggies die hard, but who except
> for the scientific/engineering crowd would want to
> create new apps, and for what reason? I imagine there
> could be some good reasons to port old (iron?) code to
> peecees, but gcc can compile FORTRAN already (albeit
> only F77).
>
> --- Randy Dawson <rdawson16 at hotmail.com> wrote:
>
> >
> > There is a lot of really cool source code here; a
> > DXF vector library, math and graphics libraries.....
> >
> > Randy
> >
> >
> _________________________________________________________________
> > Get the power of Windows + Web with the new Windows
> > Live.
> >
> http://www.windowslive.com?ocid=TXT_TAGHM_Wave2_powerofwindows_122007
>
>
>
> ____________________________________________________________________________________
> Never miss a thing. Make Yahoo your home page.
> http://www.yahoo.com/r/hs
_________________________________________________________________
i?m is proud to present Cause Effect, a series about real people making a difference.
http://im.live.com/Messenger/IM/MTV/?source=text_Cause_Effect