[Info-vax] OpenVMS I64 V8.1 "Evaluation Release"?

[Johnny Billquist]

On 2012-03-27 17.35, Bob Koehler wrote:
> In article<jkqlkp$6aa$1 at speranza.aioe.org>, glen herrmannsfeldt<gah at ugcs.caltech.edu>  writes:
>>
>> As someone else noted, VAX doesn't allow for all 32 bits in the
>> physical address, even at the architectural level.
>
>     The very first VAX, the 11/780, had a 30 bit SBI.  With only 30 bits
>     on the bus, and memories starting at 1M, it made a lot of sense to
>     use a 30 bit hardware architecure.

One could argue that the actual implementation of a bus (SBI in this 
case) is not the same as, nor a good basis for, the architecture 
definition as such.
The fact is that all VAXen until the NVAX have a PTE format that will 
only allow the formation of a 30-bit physical address.

I don't know if memory really started at 1M on the 11/780. Sounds weird, 
but maybe it was so. On the 86x0, memory starts from 0. All I/O is 
located above 0x2000 0000, which limits physical memory to at most 512 
MB. (In practice, the 86x0 are restricted to 260 MB because of existing 
memory cards.)

>  For decades VAXen ran on 30 bit
>     and narrower busses.  But later on there was an ECO to the architecture
>     to take advantage of a wider bus and IIRC, you could actually order
>     a VAX with 4GB RAM.  Of course, no one process could actually use all
>     4GB since system space never used up the full upper half.

It wasn't an ECO (ECO = Engineering Correction Order). The VAX 
architecture was expanded with a second mode for the PTE, which was 
implemented on the NVAX CPU, which allowed for a 34 bit physical 
address. The same architecture change also changed the layout of the 
process memory space by removing the S1 memory space, and reducing the 
size of S0, so you could get much closer to actually using 4 GB in a 
single process (although you still couldn't get all the way, of course...)

>     And I for one, would not like it if P1 got too close to P0.  I've
>     never run out of stack space on a VMS system and I don't want to
>     learn what the error message is.

:-)

	Johnny