Re: XT/370 microcode

[emanuel stiebler via cctalk]

On 2018-03-15 18:35, Robert Armstrong via cctalk wrote:
>> You might look up Nick Tredennick's book "Microprocessor Logic Design: 
>> The Flowchart Method" which is sold at Amazon for an obscene price
> 
>   FWIW, there are several copies on Abe Books ranging in price from "only" 
> $800 (a steal!) to almost $1200.
> 
>   I'd love to read it, but that's ludicrous.  Is this book bound in solid 
> gold or something???

It was mentioned few month ago in a newsgroup. So all "reasonable"
copies got sold, and the price increased. Just wait ...

But it is a nice reading, I don't always agree with N.Tredennick,
and explains pretty much, why this CPUs were/are so weird ...

RE: XT/370 microcode

[Robert Armstrong via cctalk]

>You might look up Nick Tredennick's book "Microprocessor Logic Design: 
>The Flowchart Method" which is sold at Amazon for an obscene price

  FWIW, there are several copies on Abe Books ranging in price from "only" $800 
(a steal!) to almost $1200.

  I'd love to read it, but that's ludicrous.  Is this book bound in solid gold 
or something???

Bob 

Re: XT/370 microcode

[Camiel Vanderhoeven via cctalk]

On 3/15/18, 9:29 AM, "cctech on behalf of Veit, Holger via cctech" 
 wrote:

You might look up Nick Tredennick's book "Microprocessor Logic Design: 
The Flowchart Method" which is sold at Amazon for an obscene price - but 
maybe some university library has a copy. It's focus is on a methodology 
for designing microcode, and it uses the design of the single chip 370 
to explain it. I suppose it was a PhD thesis.

The main point is that the 370 is NOT an 68000 with a different 
microcode; instead it is said that it implements the bus interface of 
the 68K in order to interface easily with existing peripherals (rather 
than reinventing the wheel). The internal data paths and register sets 
may be similar between the 68K and the 370 (actually it is quite 
possible that at Motorola they were aware of IBM mainframe 
architectures...) but that's all likely. The control unit design 
described in the book was completely redesigned for the purpose of 
describing the proposed methodology.

Holger

There are two distinct products (well, one product and one project that never 
made it onto a product) here that get confused. First, the XT/370. This card 
uses a 68000 with modified firmware (nick was one of the Motorola engineers 
helping implement this for IBM), as well as a modified 8087, and a stock 68000. 
Later, Nick left Motorola and came to work at IBM on the Micro/370. The 
Micro/370 is the 370 on a single chip, and it uses the 68000 bus, but is 
otherwise different from it. A good writeup of the story can be found here: 
http://www.cpushack.com/2013/03/22/cpu-of-the-day-ibm-micro-370/

Camiel


 

Re: XT/370 microcode

[emanuel stiebler via cctalk]

On 2018-03-15 09:29, Veit, Holger via cctalk wrote:
> You might look up Nick Tredennick's book "Microprocessor Logic Design:
> The Flowchart Method" which is sold at Amazon for an obscene price - but
> maybe some university library has a copy. It's focus is on a methodology
> for designing microcode, and it uses the design of the single chip 370
> to explain it. I suppose it was a PhD thesis.

He did the 68000 microcode first, and then moved to IBM ...

Re: XT/370 microcode

[Veit, Holger via cctalk]

You might look up Nick Tredennick's book "Microprocessor Logic Design: 
The Flowchart Method" which is sold at Amazon for an obscene price - but 
maybe some university library has a copy. It's focus is on a methodology 
for designing microcode, and it uses the design of the single chip 370 
to explain it. I suppose it was a PhD thesis.


The main point is that the 370 is NOT an 68000 with a different 
microcode; instead it is said that it implements the bus interface of 
the 68K in order to interface easily with existing peripherals (rather 
than reinventing the wheel). The internal data paths and register sets 
may be similar between the 68K and the 370 (actually it is quite 
possible that at Motorola they were aware of IBM mainframe 
architectures...) but that's all likely. The control unit design 
described in the book was completely redesigned for the purpose of 
describing the proposed methodology.


Holger

Am 12.03.2018 um 12:13 schrieb Dave Wade via cctalk:

I don't suppose any one left in IBM has any knowledge of this. Perhaps no
one ever did and it was all done by Motorola. Wikipedia says there were/are
2x68000 CPU's..
.. I would ask on IBM Main...

http://www.cpushack.com/2013/03/22/cpu-of-the-day-ibm-micro-370/

seems to have some names..

Dave




-Original Message-
From: cctalk  On Behalf Of Lars Brinkhoff
via cctalk
Sent: 12 March 2018 07:02
To: cctalk@classiccmp.org
Subject: XT/370 microcode

Does someone have good connections with people inside IBM?  I'd like to

ask

about 68000 microcode for the XT/370 product.

Re: XT/370 microcode

[Steve Malikoff via cctalk]

ken said
>>> I am not at liberty to post the document.
>>
>>This?
> ...
>>Also:
> ...
>>If so, it might be this:
> ...
>
> There's a difference between "I am not at liberty..." and "what anyone
> can google...".

The Wayback Machine has it on archive.org
https://web.archive.org/web/20150608173633/http://maben.homeip.net/static/S100/IBM/5160%20588%20XT%20370/System%20370%20capability%20in%20a%20desktop%20computer.pdf

Whenever I remember, I try and submit pages like this to archive.org's crawler 
to catalog it for posterity, it only takes a moment. Often
they are already there. For instance I recently scanned and submitted an IBM 
Australia memo my dad had from TJ Watson congratulating IBM employees for
their part in Apollo 11's safe return:
https://archive.org/details/@galasphere347

Steve

Re: XT/370 microcode

[Ken Seefried via cctalk]

On Tue, Mar 13, 2018 at 6:40 PM, Fred Cisin  wrote:
> On Tue, 13 Mar 2018, Ken Seefried via cctalk wrote:

 I am not at liberty to post the document.
>>>
>>> This?
>>
>> ...
>>>
>>> Also:
>>
>> ...
>>>
>>> If so, it might be this:
>>
>> ...
>> There's a difference between "I am not at liberty..." and "what anyone
>> can google...".
>
>
> There do exist situations where somebody is NOT at liberty to post (NDA,
> etc), but the item in question is readily available.  The expiration of an
> NDA may be long after publication or leakage.
> Could always do a LMGTFY.com  link, . . . :-)

Yeah, Fred...I think that's exactly what I said without being enough
of an asshole to suggest LMGTFY.

Re: XT/370 microcode

[Fred Cisin via cctalk]

On Tue, 13 Mar 2018, Ken Seefried via cctalk wrote:

I am not at liberty to post the document.

This?

...

Also:

...

If so, it might be this:

...
There's a difference between "I am not at liberty..." and "what anyone
can google...".


There do exist situations where somebody is NOT at liberty to post (NDA, 
etc), but the item in question is readily available.  The expiration of an 
NDA may be long after publication or leakage.

Could always do a LMGTFY.com  link, . . . :-)

Re: XT/370 microcode

[Ken Seefried via cctalk]

>> I am not at liberty to post the document.
>
>This?
...
>Also:
...
>If so, it might be this:
...

There's a difference between "I am not at liberty..." and "what anyone
can google...".

RE: XT/370 microcode

[Paul Birkel via cctalk]

-Original Message-
From: cctalk [mailto:cctalk-boun...@classiccmp.org] On Behalf Of Paul Berger 
via cctalk
Sent: Monday, March 12, 2018 1:41 PM
To: Guy Sotomayor Jr via cctalk
Subject: Re: XT/370 microcode

The IBM Systems Journal article does not go into great detail but says 
that a complex of three separate processors are used, a modified 68000 
that executes "a large subset of 370 instructions", a standard 68000 to 
emulate the remainder odf the instructions and a modified 8087 to 
execute the floating point.  I am not at liberty to post the document.

Paul.
-

This one?  Seems to be freely available.

https://www.researchgate.net/publication/224102395_System370_capability_in_a_desktop_computer

-
paul

Re: XT/370 microcode

[Liam Proven via cctalk]

On 12 March 2018 at 18:41, Paul Berger via cctalk  wrote:
> I am not at liberty to post the document.

This?

http://ieeexplore.ieee.org/document/5387781/?reload=true

Also:

https://www.researchgate.net/publication/224102395_System370_capability_in_a_desktop_computer

If so, it might be this:

https://pdfs.semanticscholar.org/2007/dc1c4e24b021765707191eeef5930b29a69c.pdf


-- 
Liam Proven • Profile: https://about.me/liamproven
Email: lpro...@cix.co.uk • Google Mail/Hangouts/Plus: lpro...@gmail.com
Twitter/Facebook/Flickr: lproven • Skype/LinkedIn: liamproven
UK: +44 7939-087884 • ČR (+ WhatsApp/Telegram/Signal): +420 702 829 053

Re: XT/370 microcode

[Paul Berger via cctalk]

The IBM Systems Journal article does not go into great detail but says 
that a complex of three separate processors are used, a modified 68000 
that executes "a large subset of 370 instructions", a standard 68000 to 
emulate the remainder odf the instructions and a modified 8087 to 
execute the floating point.  I am not at liberty to post the document.


Paul.


On 2018-03-12 2:11 PM, Guy Sotomayor Jr via cctalk wrote:



On Mar 12, 2018, at 9:57 AM, Eric Smith via cctalk  
wrote:

On Mon, Mar 12, 2018 at 10:54 AM, emanuel stiebler via cctalk <
cctalk@classiccmp.org> wrote:


On 2018-03-12 15:49, Eric Smith via cctalk wrote:

As the most obvious example of the impedance mismatch between 370
architecture and 68000 microarchitecture, the 68000 is hardwired to have
eight each data and address registers, not sixteen GPRs, and microcode
can't easily paper over that.

I wouldn't bet on that ...


I'm fairly sure of it, based on microarchitectural details in the US
patents on the 68000 design.

I think there was also an article in the IBM Systems Journal when the XT/370
was announced that basically described how this was done and how everything
worked.

TTFN - Guy

Re: XT/370 microcode

[Guy Sotomayor Jr via cctalk]

> On Mar 12, 2018, at 9:57 AM, Eric Smith via cctalk  
> wrote:
> 
> On Mon, Mar 12, 2018 at 10:54 AM, emanuel stiebler via cctalk <
> cctalk@classiccmp.org> wrote:
> 
>> On 2018-03-12 15:49, Eric Smith via cctalk wrote:
>>> As the most obvious example of the impedance mismatch between 370
>>> architecture and 68000 microarchitecture, the 68000 is hardwired to have
>>> eight each data and address registers, not sixteen GPRs, and microcode
>>> can't easily paper over that.
>> 
>> I wouldn't bet on that ...
>> 
> 
> I'm fairly sure of it, based on microarchitectural details in the US
> patents on the 68000 design.

I think there was also an article in the IBM Systems Journal when the XT/370
was announced that basically described how this was done and how everything
worked.

TTFN - Guy

Re: XT/370 microcode

[Eric Smith via cctalk]

On Mon, Mar 12, 2018 at 10:54 AM, emanuel stiebler via cctalk <
cctalk@classiccmp.org> wrote:

> On 2018-03-12 15:49, Eric Smith via cctalk wrote:
> > As the most obvious example of the impedance mismatch between 370
> > architecture and 68000 microarchitecture, the 68000 is hardwired to have
> > eight each data and address registers, not sixteen GPRs, and microcode
> > can't easily paper over that.
>
> I wouldn't bet on that ...
>

I'm fairly sure of it, based on microarchitectural details in the US
patents on the 68000 design.

Re: XT/370 microcode

[emanuel stiebler via cctalk]

On 2018-03-12 15:49, Eric Smith via cctalk wrote:
> As the most obvious example of the impedance mismatch between 370
> architecture and 68000 microarchitecture, the 68000 is hardwired to have
> eight each data and address registers, not sixteen GPRs, and microcode
> can't easily paper over that.

Forgot the smiley:

;-) ;-) ;-)

Re: XT/370 microcode

[emanuel stiebler via cctalk]

On 2018-03-12 15:49, Eric Smith via cctalk wrote:
> As the most obvious example of the impedance mismatch between 370
> architecture and 68000 microarchitecture, the 68000 is hardwired to have
> eight each data and address registers, not sixteen GPRs, and microcode
> can't easily paper over that.

I wouldn't bet on that ...

RE: XT/370 microcode

[Dave Wade via cctalk]

Some of the code is now on bitsavers..

 

Dave

 

From: Eric Smith <space...@gmail.com> 
Sent: 12 March 2018 14:41
To: Dave Wade <dave.g4...@gmail.com>; General Discussion: On-Topic and 
Off-Topic Posts <cctalk@classiccmp.org>
Cc: Lars Brinkhoff <l...@nocrew.org>
Subject: Re: XT/370 microcode

 

On Mon, Mar 12, 2018, 05:13 Dave Wade via cctalk <cctalk@classiccmp.org 
<mailto:cctalk@classiccmp.org> > wrote:

. Wikipedia says there were/are
2x68000 CPU's..

 

One Motorola chip was the custom one, the other was normal (as indicated by 
mask code). There was also an Intel math co, presumably derived from 8087.

 

I used to have an AT/370, which had the same chipset, but I was never able to 
obtain the software.

 

I very strongly suspect the modified 68000 and 8087 have more than just 
microcode differences, and that full reverse-engineering of the die would be 
necessary to accomplish anything useful with the microcode. Neither chip was 
designed to be a general-purpose microcode engine; both were very heavily 
tailored for their exact visible architecture, and 370 architecture is enough 
different that it couldn't be implemented by microcode only changes with no 
data path changes; the microcode ROMs and PLAs just aren't big enough to work 
around the data path issues.

 

Re: XT/370 microcode

[Eric Smith via cctalk]

As the most obvious example of the impedance mismatch between 370
architecture and 68000 microarchitecture, the 68000 is hardwired to have
eight each data and address registers, not sixteen GPRs, and microcode
can't easily paper over that.

Similarly, the 8087 microarchitecture has hardwired support for binary
floating point normalization, and microcode can't efficiently force that to
do radix 16 normalization.

Both problems could be surmounted (inefficiently) with enough microcode,
but the chips were designed with no significant extra microcode ROM and PLA
capacity.

Re: XT/370 microcode

[Eric Smith via cctalk]

On Mon, Mar 12, 2018, 05:13 Dave Wade via cctalk 
wrote:

> . Wikipedia says there were/are
> 2x68000 CPU's..
>

One Motorola chip was the custom one, the other was normal (as indicated by
mask code). There was also an Intel math co, presumably derived from 8087.

I used to have an AT/370, which had the same chipset, but I was never able
to obtain the software.

I very strongly suspect the modified 68000 and 8087 have more than just
microcode differences, and that full reverse-engineering of the die would
be necessary to accomplish anything useful with the microcode. Neither chip
was designed to be a general-purpose microcode engine; both were very
heavily tailored for their exact visible architecture, and 370 architecture
is enough different that it couldn't be implemented by microcode only
changes with no data path changes; the microcode ROMs and PLAs just aren't
big enough to work around the data path issues.

Re: XT/370 microcode

[Lars Brinkhoff via cctalk]

Lars Brinkhoff wrote:
> Does someone have good connections with people inside IBM?  I'd like to
> ask about 68000 microcode for the XT/370 product.

Martin Ziskind says he has floppy disks with VM/PC, but no way to read
them.

RE: XT/370 microcode

[Dave Wade via cctalk]

I don't suppose any one left in IBM has any knowledge of this. Perhaps no
one ever did and it was all done by Motorola. Wikipedia says there were/are
2x68000 CPU's..
.. I would ask on IBM Main...

http://www.cpushack.com/2013/03/22/cpu-of-the-day-ibm-micro-370/

seems to have some names..

Dave



> -Original Message-
> From: cctalk  On Behalf Of Lars Brinkhoff
> via cctalk
> Sent: 12 March 2018 07:02
> To: cctalk@classiccmp.org
> Subject: XT/370 microcode
> 
> Does someone have good connections with people inside IBM?  I'd like to
ask
> about 68000 microcode for the XT/370 product.