Linux on VAX - Dead ?

[Jan-Benedict Glaw]

Hi!

On Thu, 2023-03-16 13:56:27 -0800, babydr DBA James W. Laferriere <babydr at baby-dragons.com> wrote:
> On Sun, 15 Jan 2023, Jan-Benedict Glaw wrote:
> > On Sun, 2023-01-15 01:18:27 +0000, Maciej W. Rozycki <macro at orcam.me.uk> wrote:
> ...snip...
> > While the Linux port is a sad story as it somewhat came to an end, at
> > least the system availability may get better. You know, I was long
> > searching for a room to setup my old system to be 100% operated
> > remotely. I'm actually negotiating something right now. :) So hope is
> > there that this may actually become a reality!
> > 
> >  The VAX port, as is, is pretty dead. I wouldn't try pulling in newer
> > Linux patches into our former tree, but use our arch stuff to add it
> > to current Linux. And glibc? What were the most "interesting" issues?
> > ISTR that it was threading and possibly a TLS ABI?
> 
> 	Are you saying the Vax-linux porting is dead as in never to be ?

I still would love to see Linux support VAX systems, with Linus's
upstream kernel having everything needed. The old/former port's
sources are there, but I think it'll be a non-trivial task moving this
source forward. So the (old) port is indeed dead.

  OTOH, Maciej did great work to improve/revamp GCC's VAX support.
These improvements didn't even yet arrive at NetBSD. (Indeed, I've
recently used NetBSD on SIMH VAX and it basically just works.) So with
compiler support getting _better_ for a long-dead ISA, at least one
operating system around that actively can use this, and a former Linux
port that can be used to start a _new_ Linux port.

  We learned a lot from the last attempt. I think the most important
part is that we need a working GNU libc quite early. Most of that is
quite easy (ie. just use old/existing code to do syscalls, define a
few macros, use all-standard C-coded library functions instead of
specifically coded VAX assembler variants), but some pieces are not so
trivial. We _need_ to define and implement a TLS ABI. And libm may or
may not be an issue, given that VAXen use a
different-to-what-is-usual-today representation of floating point
numbers. Basically, all of today's computers use the IEEE-754
representation of floating point numbers, along with this spec's
concept of infinity, rounding, ...  VAX is slightly different, which
in the past broke program's assumptions about how floating point
numbers work more than once.

  So... Linux on VAX is dead.  GCC for VAX is more vivid than ever
before. Linux on VAX is possible.

MfG, JBG

-- 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
URL: <http://lug-owl.de/pipermail/vax-linux/attachments/20230317/eb6727c4/attachment.sig>