[Info-vax] [OT] PDP-11 address space, was: Re: HP adds OpenVMS Mature Product Support beyond the end of Standard Support

[David Froble]

Johnny Billquist wrote:
> On 2014-02-02 10:30, Bill Gunshannon wrote:
>> In article <lclbf5$hs1$1 at dont-email.me>,
>>     Simon Clubley <clubley at remove_me.eisner.decus.org-Earth.UFP> writes:
>>> On 2014-02-02, Johnny Billquist <bqt at softjar.se> wrote:
>>>> On 2014-02-01 07:41, Bill Gunshannon wrote:
>>>>>
>>>>> Of course, being as they are still being made, it would be real
>>>>> interesting to see a new PDP-11 with more address lines, too. :-)
>>>>> There I go dreaming again.  :-)  But, you have to admiit, it could
>>>>> be a lot of fun.
>>>>>
>>>>> Hmmm..   Let's see...  How many overlays would it take to get Open
>>>>> Office running on a PDP-11?   :-)
>>>>
>>>> Eh.. What? Any "modern" PDP-11 already have 22 address lines...
>>>> My PDP-11 at home have 4 megs of ram.
>>>> Have you never seen a PDP-11 with that? Under which rock have you been
>>>> the last 40 years, Bill? :-)
>>>>
>>>
>>> In fairness to Bill, I think he's talking about program space address
>>> lines, instead of machine space address lines (ie: he wants to be able
>>> to run larger programs without getting stuck into reading the overlay
>>> section of the TKB manual yet again :-)).
>>>
>>> Simon.
>>>
>>> PS: I absolutely do _not_ miss TKB. :-)
>>>
>>
>> I only did RSX once, for a project when I was with Martin Marietta.
>> But I was real good at doing overlays with Ultrix-11.  Especially
>> for larger kernels.
> 
> The design of overlays in Unix/Ultrix is rather primitive compared to 
> RSX, but doing overlays is a complicated business.
> 
>> But, yes, I would love to be able to do bigger programs on PDP-11s.
>> But if nothing else develops with the OSes, I guess it won't make
>> much difference except for Ultrix-11.
> 
> If we're talking about getting a larger virtual address space, when is 
> needed is not more address pins, but larger registers, a redesign of the 
> instruction set, and so on. That is pretty much what the VAX did.
> 
> And yes, sometimes I really wish I had a larger address space for my 
> programs, and so on. But in many cases the need for that become rather 
> small with the introduction of I/D space, and supervisor mode libraries, 
> which means I can do RMS with almost no memory space lost, and have 64K 
> of code and 64K of data at the same time. Really, I very seldom need to 
> go play with overlays nowadays.
> 
> Ultrix do not have that kind of fancy stuff, so Ultrix is in a worse 
> position, I guess.
> 
> TKB is slow as molass, but very capable. And if you have a fast machine, 
> and if you use a modern TKB, which also exploits I/D space and 
> supervisor mode libraries, it is much faster. So life nowadays in RSX is 
> really not that bad.
> The only thing that still really is black magic is when you still need 
> overlays. Getting your overlays in a nice shape in RSX is complicated.
> 
> And of course, overlays only help for large code. If you need large 
> amounts of data, you need to do other tricks.
> 
>     Johnny
> 

Where I had left overlays and TKB (and thankfully I did leave it) was 
that some data was required to always be available, and much of the 
processing was pretty much threaded.  So the data stayed in memory, and 
the processing was broken into little bits, and overlayed in what I'd 
call the "swap area" as needed to do the processing.  It got interesting 
when the context for some of the code needed to remain in memory as it 
called pieces under it.

And now I have a 10 aspirin headache.

Stop mentioning TKB !!!