[Info-vax] OT: obscure PDP11 OSes (even more dinosaury)

[Bill Gunshannon]

In article <mm954o$766$1 at iltempo.update.uu.se>,
	Johnny Billquist <bqt at softjar.se> writes:
> On 2015-06-15 15:47, Bill Gunshannon wrote:
>> In article <2f70d$557ed38b$5ed4324a$44068 at news.ziggo.nl>,
>> 	Dirk Munk <munk at home.nl> writes:
>>> Because one VAX could do the work of several PDP-11's?
>>
>> Maybe in later times, but the first VAX I worked on was a real dog
>> and I worked on several PDP-11's that outperformed it.  And, just
>> like the debates here over what Alpha might have become, IMHO the
>> PDP-11 could easily have kept up with the VAX.  All that would
>> have been needed, really, was a bigger address bus and being as
>> they had expanded it in the past, I don't think that was undoable.
>> As time went on they could have made a 32bit PDP-11 that maintained
>> 16bit compatability at the hardware level.  Intel did it twice,
>> 8088 to 80386 and 80386 to x86-64.  Things like Floating Point and
>> CIS could have been incorporated into the base architecture without
>> breaking anything.
> 
> DEC sortof did that. It was called the VAX. The early VAXen did have the 
> ability to run PDP-11 code in userspace.

Yes, but with no new features.   To get new features you had to totally
rewrite your MACRO-11 program in MACRO-32.  Iw as merely saying that
rather than totally changing the architecture I see no reason why they
could not have continued the architecture with enhancements at each
step, just like other architectures did.

> 
> And no, you would not have been able to extend the PDP-11 to 32 bits 
> without making it incompatible with a PDP-11, except as a special mode, 
> which is exactly what the VAX did.

Not sure I agree with this.  I expect that if it was desired they
could have done incremental modifications while maintaining backwards
compatability.  But then, this is all academic, isn't it.

> 
> Adding a larger physical address space would have had very little value. 
> Extending the virtual address space was crucial, and it could not be 
> done without being incompatible. And if you go that way, you might as 
> well make some general improvements anyway, which is exactly what DEC 
> did with the VAX.

The original x86 architecture had 64K segments.  It later added a mode
with a large flat address space while maintaining compatabiltiy with
"Small Model" programs.  It could be done.  

> 
>> Sorry, as much as I like the VAX, I think the PDP-11 was better.
> 
> I prefer the PDP-11 too, but let's not try to deny the obvious.
> 
>> Using the same argument we have heard here, what do yo think the
>> capabilites of a PDP-11 with no new features but made using current
>> technology for shrinks and clocking would be?  One screaming CPU.  :-)
>> How many RSTS/E users do you think could be supported on a 2.5Ghz
>> PDP-11/93 with 16GB of memory.  :-)
> 
> You could easily, from a CPU performance point of view, support 64 users 
> on an 11/9x today. At that point, it becomes a question of memory again. 
> And it's not the physical memory that you start running out of first. 
> It's virtual memory. Even the kernel have limitations based on this. In 
> RSX, one crucial part is system pool. It all sits in the kernel virtual 
> memory space, and thus cannot grow much larger than it currently is, no 
> matter what you do. You need a larger virtual address space. Anything 
> else essentially just means you are trying to do magic with your knees, 
> and you spend more and more time running code to just try and move 
> around the address space limitations.

Or, a big flat address space.  :-)  

> 
>> And X-11, too.  :-)
> 
> That would *not* happen without a larger virtual address space.
> 

Exactly.  :-)

As I have stated in the past, one project I would love to do, just
for the fun of it, was port RSTS to other architectures, both large
and small.  I can envision RSTS/86 with X-11 and large dataspace so
things like Postgres and otehr cool toold cold be ported.  Does it
have some commercial value?  Even i seriously doubt it, but as one
who really liked RSTS back in  the day, it wold be a lot of fun
doing it.  And who know who else might have fun working on and
enhancing it.  Even Minix is still being worked on.

bill

-- 
Bill Gunshannon          |  de-moc-ra-cy (di mok' ra see) n.  Three wolves
billg999 at cs.scranton.edu |  and a sheep voting on what's for dinner.
University of Scranton   |
Scranton, Pennsylvania   |         #include <std.disclaimer.h>