Re: [Ql-Users] Stella

[Dave Park]

I was clued in by that whole no developments in twelve years aspect ;)

One of those times I'm happy to be spectacularly wrong. :D

Dave

On Tue, Mar 21, 2017 at 7:33 AM, Peter Graf  wrote:

> Hello Tony,
>
> it is good to read something from you.
>
> > 1) You need a system development environment that
> >  a) will support different processors - ruling out assembler
>
> That depends on where one looks for a niche market. Things have changed
> inasmuch small to medium sized FPGA have become large enough to contain
> a whole 32 bit CPU.
>
> In addition to that, patents on the original 68000 have expired.
>
> Because of the high code density, I found the 68K architecture best
> suited for use with the limited internal RAM resources of an FPGA. I
> have tried several 32 bit architectures, and 68K allows the smallest
> memory footprint.
>
> Within small to medium sized FPGA systems, having a real-time operating
> system would often advantageous, but existing systems are simply too large.
>
> For such targets, the best would be an assembler written OS, still
> allowing application code in C. Like SMSQ/E and C68, but capable of real
> time operation and oriented toward embedded systems instead of personal
> computing.
>
> > 2) You need to develop interface modules (drivers?) for a wide range of
> > "peripheral" devices - the device manufacturers will not develop them
> > for you.
>
> This also improves with FPGA - once there is a peripheral device in HDL
> (Hardware Definition Language) you can at least keep it th the same for
> the next product generation.
>
> > 4) You need to develop entirely different hardware architectures - this
> > century, architectures have become ever more tightly dedicated to UNIX
> > (Linux, Windows NT, ...)
>
> Maybe you can take the Q68 as a modest proof that a new hardware
> architecture can be close to 68K and the driver requirements of an
> assembler OS. I developed the Q68 as a hobby project, so is now 10 years
> old. Therefore it can not demonstrate the lastest chip performance, and
> of course the Q68 has a homecomputing flavour.
>
> But updated to latest chip generation, and adapted to an embdedded
> application, such a system suddenly could make a lot of sense with Stella.
>
> > 1) A first version of the Stella core was coded in MC68000 assembler and
> > benchmarked [...]
>
> So nothing would be better to continue development of the Stella than
> the 68K architecture.
>
> > 3) There is still no end in sight to the discovery of new "exploits" in
> > all sorts of Unix based devices.
>
> This is an important point. I have been joking with friends, that we
> might have to go back to 68K for internet use, simply because the
> architecture is too exotic to be exploited. Also, security requires
> lowest possible complexity.
>
> All the best
> Peter
>
> ___
> QL-Users Mailing List
>



-- 
Dave Park
Sandy Labs
d...@sinclairql.com
___
QL-Users Mailing List

Re: [Ql-Users] Stella

[Peter Graf]

Hello Tony,

it is good to read something from you.

> 1) You need a system development environment that
>  a) will support different processors - ruling out assembler

That depends on where one looks for a niche market. Things have changed
inasmuch small to medium sized FPGA have become large enough to contain
a whole 32 bit CPU.

In addition to that, patents on the original 68000 have expired.

Because of the high code density, I found the 68K architecture best
suited for use with the limited internal RAM resources of an FPGA. I
have tried several 32 bit architectures, and 68K allows the smallest
memory footprint.

Within small to medium sized FPGA systems, having a real-time operating
system would often advantageous, but existing systems are simply too large.

For such targets, the best would be an assembler written OS, still
allowing application code in C. Like SMSQ/E and C68, but capable of real
time operation and oriented toward embedded systems instead of personal
computing.

> 2) You need to develop interface modules (drivers?) for a wide range of 
> "peripheral" devices - the device manufacturers will not develop them 
> for you.

This also improves with FPGA - once there is a peripheral device in HDL
(Hardware Definition Language) you can at least keep it th the same for
the next product generation.

> 4) You need to develop entirely different hardware architectures - this 
> century, architectures have become ever more tightly dedicated to UNIX 
> (Linux, Windows NT, ...)

Maybe you can take the Q68 as a modest proof that a new hardware
architecture can be close to 68K and the driver requirements of an
assembler OS. I developed the Q68 as a hobby project, so is now 10 years
old. Therefore it can not demonstrate the lastest chip performance, and
of course the Q68 has a homecomputing flavour.

But updated to latest chip generation, and adapted to an embdedded
application, such a system suddenly could make a lot of sense with Stella.

> 1) A first version of the Stella core was coded in MC68000 assembler and 
> benchmarked [...] 

So nothing would be better to continue development of the Stella than
the 68K architecture.

> 3) There is still no end in sight to the discovery of new "exploits" in 
> all sorts of Unix based devices.

This is an important point. I have been joking with friends, that we
might have to go back to 68K for internet use, simply because the
architecture is too exotic to be exploited. Also, security requires
lowest possible complexity.

All the best
Peter

___
QL-Users Mailing List

Re: [Ql-Users] Stella

[Tony Tebby]

Dave Park thinks that

"So, fair to assume it's a dead project that will never see the light of 
day."


That seems quite reasonable!

The problem can be viewed in two ways

Negatively

1) You need a system development environment that
a) will support different processors - ruling out assembler
b) does not impose UNIX type structures - ruling out C, C++, etc

2) You need to develop interface modules (drivers?) for a wide range of 
"peripheral" devices - the device manufacturers will not develop them 
for you.


3) You need to develop an applications base for a new market

4) You need to develop entirely different hardware architectures - this 
century, architectures have become ever more tightly dedicated to UNIX 
(Linux, Windows NT, ...)


Therefore you need $.

Positively

1) A first version of the Stella core was coded in MC68000 assembler and 
benchmarked against Solaris 2 (UNIX SVR4) showing orders of magnitude 
(under unrealistic, best case, benchmark conditions) lower system 
overheads. Sun Microsystems rejected the idea of developing the system 
saying that it would not work on either SPARC systems or symmetric 
multiprocessor systems (false) and that benchmarks were not reliable 
indicators (true). Instead they bought Chorus, a UNIX variant, which 
disappeared without trace. Not everybody will always be so stupid


2) There are signs that the "you can always use a more powerful 
computer" is becoming less acceptable as an excuse for chronically 
inefficient software.


3) There is still no end in sight to the discovery of new "exploits" in 
all sorts of Unix based devices. Unix is fundamentally unsound as well 
as chronically inefficient.


4) Unix type systems will inevitably be wiped out by other systems. We 
could hope that, unlike Unix, these will be theoretically sound and fit 
for purpose. This may not happen in our lifetime.


5) The only real barrier is ignorance. The story of Android is 
illustrative. If, in 2005, Google had known that they could have 
developed an entirely new operating system and had it up and running 
sooner than trying to kludge something out of Linux, would they have 
bought Android Inc. for $50M? Of course not, but they did not know.


Tony






On 18/03/2017 12:49, Ralf Reköndt wrote:
I think, TT reads here, so he makes his own thougths about people, who 
still remember his projects and are still interested in these.


- Original Message - From: "Dave Park"

So, fair to assume it's a dead project that will never see the light 
of day.


Thanks all!

Dave

On Thu, Mar 16, 2017 at 4:10 PM, Norman Dunbar
wrote:


Sorry Dave,  I seem to have an inability to use a tablet! The url was
supposed to be:

https://web.archive.org/web/20050914114333/http://thgodef.
nerim.net/qdos/stella/intro.html

but I suspect it's Arnauld's page on Thierry's site which you 
mentioned.


Cheers,
Norm.
--
Sent from my Android device with K-9 Mail. Please excuse my brevity. 


___
QL-Users Mailing List




___
QL-Users Mailing List