[TUHS] PG3 or Gen3.0?

[Warner Losh via TUHS]

On Fri, May 8, 2026 at 12:06 PM segaloco via TUHS <tuhs at tuhs.org> wrote:

> On Friday, May 8th, 2026 at 10:48, Adam Thornton via TUHS <tuhs at tuhs.org>
> wrote:
>
> > Generally, it's not that ARM is any faster than x86_64--but it can take
> > much less energy to give you the same amount of computation.
>
> With the advent of widespread mobile devices, this was crucial.  Necessity
> drove ARMs meteoric rise, and now priorities seem to be shifting once
> again, this time surrounding supply chains and fab redundancy, stuff like
> that.  My hunch is we're going to start seeing more interest in
> royalty-free designs like RISC-V and even moreso if there are such designs
> over in the realm of solid state memory.  No sociopolitical commentary but
> the TSMC situation is certainly on the mind.  Couple this with the hiring
> spree companies like Apple and nVidia had in the RISC-V realm several years
> ago and I suspect we'll start seeing the baby steps towards another
> tectonic shift possibly in the next decade.
>
> Granted this is now getting quite far afield from the original subject.
> To bring it back around, the modern pedagogical V6 port xv6 runs on RISC-V
> now, so that means there is also hope for bootstrapping other historic UNIX
> systems (like PG3) on RISC-V.


xv6 was written for 32-bit i386, so porting wasn't so bad.
The Unixes of this era (PG3 roughly V6) don't yet have the portability
fixes that even the 7th Edition has, so there's a lot of porting to be
done. None of the sizes of things are abstracted, so there's a lot of +2
baked into the code since it knows that's how big a pointer is..
Sure would be a cool project, though. The Miller papers on the Wollongong
Interdata port have many of the details...

Warner


> Given the nature of RISC-V as purely an ISA standard, not an
> implementation, this implies that UNIX's life has now been extended
> theoretically indefinitely, RISC-V is supposed to have future proofing
> baked-into the ISA design, offering an appealing interface to compute that
> isn't as likely to evaporate in a decade or two.
>
> But that's basicslly what they said about System/360 and VAX was it not?
> I guess time will tell if RISC-V really significantly moves the needle in
> the modern UNIX-like playfield.
>
> - Matt G.
>