[HECnet] VMS/RSX Guest accounts --> 2020 Unibus(ses) --> Tops20 V4/5

Wed Apr 29 11:55:20 PDT 2020

LingLing, as well as MRC's other 2020's made it out to Seattle (Living 
Computer Museum).  I had asked them whether they had been able to 
recover any software off them, but I have yet to hear anything.  I ought 
to check in; they did some interesting recovery work on a PDP-8 that I 
enjoyed reading about.

I don't have a series 5 monitor handy, but for 7, the symbols live in 
section 5 and appear to be taking up about 253 pages of space.  That 
number may not be completely representative; the bug strings might be in 
there, too.  At any rate, you can get a feel for what engineering must 
have been up against when they went with symbol hiding in 4 and what MRC 
must have had to do to re-hide them in 5 along with the new code.

I think once he re-hide them, he probably had plenty of space and that 1 
MW wouldn't have made any difference.  I presume that for two reasons:

 1. You can see what kind of space got freed up; he wouldn't have needed
    the additional memory.  The real work would have been moving the
    .PSECT's and storage around to fit in a single section as well as
    modifying the macros for intersection transfer.  One imagines that
    KDDT would have needed some retrofitting to get at the other address
    space.
 2. Move over, using additional sections in the monitor might have been
    'hard' because you are talking about switching address spaces to
    transfer execution; essentially having the monitor task switch
    itself into another part.

As of 6, you get CFS support, which is a bucket load of code.  I think 
maybe Phase IV, if that only runs over NI; I think that may be the case 
because I believe the DN20's remained at Phase III.  I seem to recall 
that we were actually thinking of removing them at one point; that was 
mostly because of the potential saving in cooling and power (which 
didn't turn out to be that much).

On 4/29/20 3:55 AM, R. Voorhorst wrote:
> L.S.
>
> By the way, I am pretty sure MRC attempted his variant of Tops20 (4.2? and 5.x?) on the 512 kB KS10 variant; the results might have turned out differently if he had a 1 MW variant available.
> Is this work preserved somewhere?
>
> R.V.
>
> -----Original Message-----
> From: owner-hecnet at Update.UU.SE [mailto:owner-hecnet at Update.UU.SE] On Behalf Of R. Voorhorst
> Sent: Wednesday, 29 April, 2020 08:51
> To: hecnet at Update.UU.SE
> Subject: RE: [HECnet] VMS/RSX Guest accounts --> 2020 Unibus(ses)
>
> L.S.
>
> Well, look in Tops10 monitor sources and in the engineering stuff: multiple Unibus adapters (up to 4, standard 2 provided) each which its own mapping hardware and registers.
> It can easily expanded to even handle 4 MW memory space (I run here some 2 MW variants), but the current Tops10 single section monitor cannot handle the needed page mapping in monitor mapping space very well.
> 1 MW works fine though. 1 MW variants surfaced somewhat later than the 512k ones.
>
>
> R.V.
>
> -----Original Message-----
> From: owner-hecnet at Update.UU.SE [mailto:owner-hecnet at Update.UU.SE] On Behalf Of Johnny Billquist
> Sent: Wednesday, 29 April, 2020 08:14
> To: hecnet at Update.UU.SE
> Subject: Re: [HECnet] VMS/RSX Guest accounts
>
> I should point out that I actually suspect the 2020 have a Unibus map, just as the big PDP-11s or any VAXen have.
> Which then remaps the Unibus address space to the larger address space of the machine, so that DMA can go anywhere.
>
>     Johnny
>
> On 2020-04-29 08:10, Johnny Billquist wrote:
>> Nope. Unibus can only address 128Kword, and in this context that is
>> 128K of 18-bit words. So you cannot even DMA into a full 256K of 36 bit words.
>>
>>     Johnny
>>
>> On 2020-04-29 05:51, Thomas DeBellis wrote:
>>> You know, I have been racking my brains from time to time and I can't
>>> remember a thing about the ADP modifications to the 2020.
>>>
>>> So, the 2020 came with a maximum of 512K words, 2**9.  An addition
>>> bit would have brought it up to a full megaword, 2**10, which is
>>> quite reasonable for the target audience (some kind of installation
>>> that didn't hold stock in the local power utility).
>>>
>>> I guess there may have been modifications to the 2020 build of the
>>> monitor to allow for the extra bit.  I don't know if the Unibus
>>> devices could do DMA into the full address space.
>>>
>>> Did you pick up any software with these little jewels?  The monitor
>>> changes might be interesting.
>>>
>>>> --------------------------------------------------------------------
>>>> ---- On 4/24/20 12:12 AM, Dave McGuire wrote:
>>>>     Nice!
>>>>
>>>>     One of our 2020s is in the brown color scheme.  You know what
>>>> that
>>>> means: ADP, and one more address bit.
>>>>> -------------------------------------------------------------------
>>>>> -----
>>>>>
>>>>> On 4/23/20 6:15 PM, Thomas DeBellis wrote:
>>>>>
>>>>> The solution for 4.1 was one of the finest hacks I have ever heard
>>>>> of; while the 2020 doesn't support extended addressing, it does
>>>>> support multiple address spaces, so what they did was move all the
>>>>> symbols into a separate address space.  This was called 'hiding'
>>>>> symbols and I thought it was great because it made them harder to
>>>>> smash.  However, all of that went out the window with 5.0, which
>>>>> fully supported extended addressing.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sonic.net/pipermail/hecnet-list/attachments/20200429/0a3c1d15/attachment.html>