Re: is Vax addressing sane today

Michael S <already5chosen@yahoo.com> writes:

On Tue, 1 Oct 2024 10:12 +0100 (BST)
jgd@cix.co.uk (John Dallman) wrote:

PowerPC did for a while, but the company interested in NT on PowerPC
was IBM, and their hardware prices were a /lot/ higher than x86
prices. They didn't see that as a problem, but all the potential
customers did.

The ideas of ARC (MIPS) and PowerPC (which was not just IBM) was that
they would succeed the IA-32-based PC.  Given the assumed (and, around
1990, actual) performance superiority of RISCs over IA-32, this looked
plausible.  However, even with Alpha, which was often superior in
performance throughout the 1990s, and for which there were cheap
offerings (but without performance edge; e.g., I once was playing with
the idea of buying a 21164PC-based PC164SX system, where the CPU+board
(with 1MB L2 cache) cost ATS 6000 (~EUR 440) in 1998; but I went with
a K6-2, because I played some DOS games:-).  The cheap 164SX offer may have
been a clearance sale, however.

In any case, the performance advantage of the RISCs vanished during
the 1990s, the RISCs never had wide ISV support, and so WNT on RISCs
flopped.

Now I wonder what endiannes was used by PowerPC variant WinNT.
In theory, PPC/POWER could run in Little Endian mode, but before v3 of
POWER ISA it wasn't as full-featured as Big Endian mode. If I am not
mistaken, the difference was that in LE mode there was no support for
unaligned memory accesses.

Given that MIPS and Alpha require natural alignment, little-endian
PowerPC at the time was as full-featured as the other RISCs.

Alignment issues may have been a problem with the RISC ports, though.

- anton
--
'Anyone trying for "industrial quality" ISA should avoid undefined behavior.'
  Mitch Alsup, <c17fcd89-f024-40e7-a594-88a85ac10d20o@googlegroups.com>