30 Aug
2006
30 Aug
'06
6:19 p.m.
From: "Jay West" <jwest at classiccmp.org>
Sean wrote....
> I seem to recall a supercomputer made by Intel
that used Pentiums. I
> forget the name though, but it was quite a behemoth of a machine from what
> little I recall (many CPUs).
Wasn't that "The Connection Machine"?
Nope. Connection Machines were made by Thinking Machines, Inc., and they
don't use Pentiums. They use gobs (up to something like 16k) of proprietary
processors.
The ASCI Red supercomputer built by Intel used PPros; see
http://www.ai.mit.edu/projects/aries/course/notes/ascii_red.pdf. I think
the compute nodes were pretty off-the-shelf SMP PPro designs, so some might
consider it off-topic.
30 Aug
30 Aug
4:45 p.m.
New subject: Pentium for Non-PC (Was Statement & apology (was Re: 10 year whine fest)
It was thus said that the Great Ken Seefried once stated:
From: "Jay West" <jwest at
classiccmp.org>
Sean wrote....
>Wasn't that "The Connection Machine"?
Nope. I did a search and it was the Intel Paragon I was thinking of,
which actually used i860s and not Pentiums. Ah well ...
I seem
to recall a supercomputer made by Intel that used Pentiums. I
forget the name though, but it was quite a behemoth of a machine from
what
> little I recall (many CPUs). The ASCI Red supercomputer built by Intel used PPros;
see
http://www.ai.mit.edu/projects/aries/course/notes/ascii_red.pdf. I think
the compute nodes were pretty off-the-shelf SMP PPro designs, so some might
consider it off-topic.
Ah, so the PPro *was* used in a super computer.
-spc (Now I don't feel so bad ... )
8:57 p.m.
On Wednesday 30 August 2006 09:19 pm, Ken Seefried wrote:
From: "Jay West" <jwest at
classiccmp.org>
Sean wrote....
Nope. Connection Machines were made by Thinking Machines, Inc., and they
don't use Pentiums. They use gobs (up to something like 16k) of
proprietary processors.
I'm vaguely remembering something about what chip that was, but now I can't
remember what it was called. Ah, would that have been the "transputer",
maybe?
I also recall an article in Byte, way back when.
--
Member of the toughest, meanest, deadliest, most unrelenting -- and
ablest -- form of life in this section of space, a critter that can
be killed but can't be tamed. --Robert A. Heinlein, "The Puppet Masters"
-
Information is more dangerous than cannon to a society ruled by lies. --James
M Dakin
I seem
to recall a supercomputer made by Intel that used Pentiums. I
forget the name though, but it was quite a behemoth of a machine from
what little I recall (many CPUs).
Wasn't that "The Connection Machine"? 
31 Aug
31 Aug
11:56 a.m.
On Wed, 30 Aug 2006, Roy J. Tellason wrote:
I'm vaguely remembering something about what chip that was, but now I can't
remember what it was called. Ah, would that have been the "transputer",
maybe?
No, transputers are 16 or 32 bit RISC chips, some with 64 bit IEEE
floating point either in hardware or emulated via microcoded instructions.
Their best features are: 4 async 2-pair 5, 10 or 20MHz full duplex
communication links on die, a minimal ammount of RAM on die, and the
ability to be booted and debugged over any of the 4 links. Thus, you can
build a parallel "computing surface" with just the transputer chips plus
power and clock sources. They also have a fairly flexible RAM/ROM
interface, but it's entirely optional. Finally, the whole family is
machine language compatible (later members only extended the instruction
set). In theory this let you mix family members in a machine, but
unfortunately there was no "what family member am I running on"
instruction.
Alexey
Wasn't that "The Connection Machine"?
Nope. Connection Machines were made by Thinking Machines, Inc., and they
don't use Pentiums. They use gobs (up to something like 16k) of
proprietary processors.
4:17 p.m.
On Thursday 31 August 2006 02:56 pm, Alexey Toptygin wrote:
On Wed, 30 Aug 2006, Roy J. Tellason wrote:
I'm vaguely remembering something about what chip that was, but now I
can't remember what it was called. Ah, would that have been the
"transputer", maybe?
No, transputers are 16 or 32 bit RISC chips, some with 64 bit IEEE
floating point either in hardware or emulated via microcoded instructions.
Their best features are: 4 async 2-pair 5, 10 or 20MHz full duplex
communication links on die, a minimal ammount of RAM on die, and the
ability to be booted and debugged over any of the 4 links. Thus, you can
build a parallel "computing surface" with just the transputer chips plus
power and clock sources. They also have a fairly flexible RAM/ROM
interface, but it's entirely optional. Finally, the whole family is
machine language compatible (later members only extended the instruction
set). In theory this let you mix family members in a machine, but
unfortunately there was no "what family member am I running on"
instruction.
Which makes me wonder what would happen if you fed it an extended instruction
that it didn't know how to deal with.
I dunno, maybe they were just out in roughly the same time frame, or maybe
I've just got such a cluttered brain that it's all running together lately...
--
Member of the toughest, meanest, deadliest, most unrelenting -- and
ablest -- form of life in this section of space, a critter that can
be killed but can't be tamed. --Robert A. Heinlein, "The Puppet Masters"
-
Information is more dangerous than cannon to a society ruled by lies. --James
M Dakin
Wasn't that "The Connection Machine"?
Nope. Connection Machines were made by Thinking Machines, Inc., and
they don't use Pentiums. They use gobs (up to something like 16k) of
proprietary processors.
1 Sep
1 Sep
10:20 a.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
On Thu, 31 Aug 2006, Roy J. Tellason wrote:
I don't know, but finding out is one of the first things on my list of
things to do once I build some boards to plug mine into and find or make a
transputer link interface (anyone got a C012 they wouldn't mind selling?).
My wild-ass guess is they halt and wait for you to assert the reset or
analyze pins. What I really want to do is write a bug-compatible emulator
for them, so people could play with transputers without having to spend
hundreds of dollars and hours :-)
Alexey
No,
transputers are 16 or 32 bit RISC chips, some with 64 bit IEEE
floating point either in hardware or emulated via microcoded instructions.
Their best features are: 4 async 2-pair 5, 10 or 20MHz full duplex
communication links on die, a minimal ammount of RAM on die, and the
ability to be booted and debugged over any of the 4 links. Thus, you can
build a parallel "computing surface" with just the transputer chips plus
power and clock sources. They also have a fairly flexible RAM/ROM
interface, but it's entirely optional. Finally, the whole family is
machine language compatible (later members only extended the instruction
set). In theory this let you mix family members in a machine, but
unfortunately there was no "what family member am I running on"
instruction.
Which makes me wonder what would happen if you fed it an extended instruction
that it didn't know how to deal with. 
10:45 a.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
________________________________
From: cctalk-bounces at classiccmp.org on behalf of Alexey Toptygin
Sent: Fri 9/1/2006 1:20 PM
To: General Discussion: On-Topic and Off-Topic Posts
Subject: transputer internals (was Re: Pentium for Non-PC)
On Thu, 31 Aug 2006, Roy J. Tellason wrote:
That is not entirely true. There are instructions that gives it back on later versions of
the transputer. Can't remember off the top of my head right now though....
Ram
unfortunately
there was no "what family member am I running on"
instruction.
11 a.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
On Fri, 1 Sep 2006, Ram Meenakshisundaram wrote:
On Thu, 31 Aug 2006, Alexey Toptygin wrote:
Really? I've read my copy of "Transputer Instruction Set: A Compiler
Writer's Guide" cover to cover, and I don't recall such an instruction.
I think there's one that lets you load the address of the bottom of
on-chip RAM, but that won't distinguish the other features...
Alexey
unfortunately there was no "what family
member am I running on"
instruction.
That is not entirely true. There are instructions that gives it back on
later versions of the transputer. Can't remember off the top of my head
right now though....
11:03 a.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
Its undocumented...
Ram
________________________________
From: cctalk-bounces at classiccmp.org on behalf of Alexey Toptygin
Sent: Fri 9/1/2006 2:00 PM
To: General Discussion: On-Topic and Off-Topic Posts
Subject: RE: transputer internals (was Re: Pentium for Non-PC)
On Fri, 1 Sep 2006, Ram Meenakshisundaram wrote:
On Thu, 31 Aug 2006, Alexey Toptygin wrote:
Really? I've read my copy of "Transputer Instruction Set: A Compiler
Writer's Guide" cover to cover, and I don't recall such an instruction.
I think there's one that lets you load the address of the bottom of
on-chip RAM, but that won't distinguish the other features...
Alexey
unfortunately there was no "what family
member am I running on"
instruction.
That is not entirely true. There are instructions that gives it back on
later versions of the transputer. Can't remember off the top of my head
right now though....
11:11 a.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
Its lddevid. Look at the startup code (bootstrap) in the C compiler and/or OCCAM compiler
to see how it is used...
________________________________
From: cctalk-bounces at classiccmp.org on behalf of Ram Meenakshisundaram
Sent: Fri 9/1/2006 2:03 PM
To: General Discussion: On-Topic and Off-Topic Posts; General Discussion: On-Topic and
Off-Topic Posts
Subject: RE: transputer internals (was Re: Pentium for Non-PC)
Its undocumented...
Ram
________________________________
From: cctalk-bounces at classiccmp.org on behalf of Alexey Toptygin
Sent: Fri 9/1/2006 2:00 PM
To: General Discussion: On-Topic and Off-Topic Posts
Subject: RE: transputer internals (was Re: Pentium for Non-PC)
On Fri, 1 Sep 2006, Ram Meenakshisundaram wrote:
On Thu, 31 Aug 2006, Alexey Toptygin wrote:
Really? I've read my copy of "Transputer Instruction Set: A Compiler
Writer's Guide" cover to cover, and I don't recall such an instruction.
I think there's one that lets you load the address of the bottom of
on-chip RAM, but that won't distinguish the other features...
Alexey
unfortunately there was no "what family
member am I running on"
instruction.
That is not entirely true. There are instructions that gives it back on
later versions of the transputer. Can't remember off the top of my head
right now though....
12:10 p.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
On Fri, 1 Sep 2006, Ram Meenakshisundaram wrote:
On Fri, 1 Sep 2006, Alexey Toptygin wrote:
<groan> I hope there aren't a lot of undocumented instructions. You'd
think the instruction set reference would tell you about the instructions
used by the vendor's own compiler...
When you say C Compiler and OCCAM compiler, do you mean the "SGS-Thomson
Transputer ANSI C Toolset" and "SGS-Thomson Transputer Occam 2.1 Toolset"
you've got product info for on your page? I don't have either of those
(or any others for that matter), could I get a copy from you?
BTW, your mailer seems to add an extra copy of cctalk@ to the To: field
every time you reply.
Alexey
Really? I've read my copy of "Transputer
Instruction Set: A Compiler
Writer's Guide" cover to cover, and I don't recall such an instruction.
I think there's one that lets you load the address of the bottom of
on-chip RAM, but that won't distinguish the other features...
Its undocumented...
Its lddevid. Look at the startup code (bootstrap) in the C compiler
and/or OCCAM compiler to see how it is used...
12:15 p.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
I really shouldn't be doing this, but if you go to
http://www.classiccmp.org/transputer/software, you can find tons of goodies including the
aforementioned compilers. ST just really doesn't care about it though...
There are some other instruction sets that are out there, but I cant think of them at the
moment. That's strange, but thanks for pointing it out...
Cheers,
Ram
________________________________
From: cctalk-bounces at classiccmp.org on behalf of Alexey Toptygin
Sent: Fri 9/1/2006 3:10 PM
To: General Discussion: On-Topic and Off-Topic Posts
Subject: RE: transputer internals (was Re: Pentium for Non-PC)
On Fri, 1 Sep 2006, Ram Meenakshisundaram wrote:
On Fri, 1 Sep 2006, Alexey Toptygin wrote:
<groan> I hope there aren't a lot of undocumented instructions. You'd
think the instruction set reference would tell you about the instructions
used by the vendor's own compiler...
When you say C Compiler and OCCAM compiler, do you mean the "SGS-Thomson
Transputer ANSI C Toolset" and "SGS-Thomson Transputer Occam 2.1 Toolset"
you've got product info for on your page? I don't have either of those
(or any others for that matter), could I get a copy from you?
BTW, your mailer seems to add an extra copy of cctalk@ to the To: field
every time you reply.
Alexey
Really? I've read my copy of "Transputer
Instruction Set: A Compiler
Writer's Guide" cover to cover, and I don't recall such an instruction.
I think there's one that lets you load the address of the bottom of
on-chip RAM, but that won't distinguish the other features...
Its undocumented...
Its lddevid. Look at the startup code (bootstrap) in the C compiler
and/or OCCAM compiler to see how it is used...
12:41 p.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
BTW,
Did you pick up the AVX "SuperComputer"?
Cheers,
Ram
4:59 p.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
Which makes me
wonder what would happen if you fed it an extended instruction
that it didn't know how to deal with.
I don't know, but finding out is one of the first things on my list of
things to do once I build some boards to plug mine into and find or make a
transputer link interface (anyone got a C012 they wouldn't mind selling?). My wild-ass guess is they halt and wait for you to
assert the reset or
Assert Error?
analyze pins. What I really want to do is write a
bug-compatible emulator
for them, so people could play with transputers without having to spend
hundreds of dollars and hours :-)
Considering it took me a morning to go from 'open the Transputer
Databook' to 'Having a handwired bootable machine', I think 'hundreds of
hours is a little pesimistic. And are they really that hard to find or
expenssive that you have to spend hundreds of dollars to get one?
-tony
2 Sep
2 Sep
12:45 p.m.
New subject: transputer internals (was Re: Pentium for Non-PC)
Tony Duell wrote:
Are ISA transputer boards (B004, etc) that hard to find now?
I've only stumbled across one in the last four years, and that one was missing
all the actual transputer modules from it and was just the ISA host board :-(
Which makes me wonder what would happen if you fed it
an extended instruction
that it didn't know how to deal with.
I don't know, but finding out is one
of the first things on my list of
things to do once I build some boards to plug mine into and find or make a
transputer link interface (anyone got a C012 they wouldn't mind selling?). And are they really that hard to find or
expenssive that you have to spend hundreds of dollars to get one?
I'd say so. I suppose the "supercomputer" associations tend to push the
price
up or something. Annoyingly we've had an offer of a complete Meiko CS1 along
with the associated Microvax controlling host / graphical display, and nobody
seems to be interested apart from me :(
--
If you've ever wondered how you get triangles from a cow
You need buttermilk and cheese, and an equilateral chainsaw
31 Aug
31 Aug
11:30 a.m.
Ken Seefried wrote:
The ASCI Red supercomputer built by Intel used PPros;
see
http://www.ai.mit.edu/projects/aries/course/notes/ascii_red.pdf. I
think the compute nodes were pretty off-the-shelf SMP PPro designs, so
some might consider it off-topic.
Bit more info for the PDF-impaired:
http://www.pact.sscc.ru/hardware/computer/#IntelASCIRed
(plus two of the three 'see also' links still seem to work!)
cheers
J.
6456
days inactive
6458
days old
15 comments
7 participants
participants (7)
-
alexeyt＠freeshell.org -
ard＠p850ug1.demon.co.uk -
julesrichardsonuk＠yahoo.co.uk -
ken＠seefried.com -
RMeenaks＠olf.com -
rtellason＠verizon.net -
spc＠conman.org