Adrian (a.delete@this.acm.org) on September 24, 2021 1:21 am wrote:
> David Hess (davidwhess.delete@this.gmail.com) on September 23, 2021 6:00 pm wrote:
> > sr (nobody.delete@this.nowhere.com) on September 22, 2021 6:55 am wrote:
> > >
> > > 8086 segments were bad but those had nothing to do with MMU and privilege levels. 80286's MMU
> > > and specially its extended version on 80386 instead had some serious brainpower behind them.
> > > Remember they were designed in beginning of 80's and did have hardware limitation of what they
> > > could implement. I believe that I read that 386 was on reticle limit on it's first process.
> >
> > I remember reading that the 386's paged MMU was directly inspired by the DEC VAX. I would have noticed but
> > the VAX was just before my time. This would make the relationship between VMS and NT less surprising.
> >
>
> While it is likely that various other CPUs with paging existed earlier, DEC
> VAX is the one which made paging well known and popular, so almost all later
> architectures included paging more or less similar with that of DEC VAX.

Yep there was paging before then! Not just the Manchester Atlas but the IBM/360 67 and then other IBM/370 machines and I think they played a much bigger role in popularizing the idea.

> Intel 80386 was inspired by DEC VAX in many other details, which was natural, because during the
> eighties DEC VAX was the incumbent 32-bit architecture, so it was unavoidable for all the designers
> of 32-bit microprocessors from National Semiconductor, Motorola, Intel etc. (before the fashion moved
> to RISC designs) to have as their target to match the features of VAX at a lower price.
>
> The new addressing modes of Intel 80386 were also a subset of the VAX addressing modes,
> but this is one of the very few cases when the Intel designers were wiser than their competition
> because they included only the simpler of the VAX addressing modes, which was good because
> the more complex modes with memory indirection became obsolete later.
>
> Motorola 68020 had included more addressing modes, even with extra flexibility
> above VAX, but a decade later Motorola had to remove all the too complex addressing
> modes, giving up on backward compatibility with ColdFire.
>
> However, it might be that the good choice of addressing modes in 80386 was not due to the better
> foresight of the designers, but they might have thought to include the complete set of VAX addressing
> modes, and then they gave up, either because it would have increased too much the chip size or
> because they were unable to find a good way to encode more addressing modes without increasing
> the average instruction length and in a way not too different from the 8086 encoding.

Motrorola's complex addressing modes caused real problems with virtual store handling. To support virtual store Motorola needed to store intermediate states within instructions.