Re: NetBSD on SIMH, but not microvax3900

Hi.

On 2024-03-08 12:24, Antonio Carlini wrote:
> But after sending that I read a bit further in the Hardware Handbook and 
> found:
> 
> "The 28-bit SBI longword address field allows access to up to 512 Mbytes of
> main memory. The hardware supports a maximum of 64 Mbytes of main
> memory. Physical memory operations are performed when bit 27 of the SBI
> addresses is zero; 1/0 operations occur when bit 27 is one.
> 
> So SBI could do 256MB of which half was for I/O."

Actually, the SBI have a 1G address space. Yes, there are only 28 
address bit, but SBI is always addressing by longwords, so the two low 
bits are not included, and are implicitly always 0. Your quote actually 
says exactly this. 28-bit longword address. Up to 512Mbyte of main memory.

But even looking in the 1977 VAX Hardware Handbook (page 13), it's clear 
that memory space is 512M and I/O space is 512M on SBI.
However, it also points out that at that time, the maximum possible 
memory was 8M, because that was the most you could get into the memory 
boxes (the limit was the size of individual memory chips). That 
obviously grew over time...

> So there seems to be something that caps physical memory at 64MB (plus 
> perhaps a bit for multiport memory, but that was always just a handful 
> of MB).

Well. Any memory box is just responding on the SBI, so that's where the 
real hard limit is.
Actual memory boxes that contained more than 64M probably never got 
made. But that's more a limit on the actual design of a specific box 
than any limit in the hardware itself. A different box would allow more 
memory. SBI is generic enough that it don't really put any futher 
limitations here.

   Johnny

-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: [email protected]             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol