Subject: Re: Transmaeta VAX?
To: None <port-vax@NetBSD.ORG>
From: John Wilson <email@example.com>
Date: 07/02/2001 11:39:20
>From: Brian Chase <firstname.lastname@example.org>
>This message seemed to have gotten swallowed up in the general flow of the
>off-topic threads. Well, what are the rumors? I've often thought that
>the VAX would be a wonderful personality to port to the Transmeta chips.
>But I sort of got the impression that although the Transmeta CPUs have the
>capability to emulate other architectures, it doesn't seem like a trivial
>thing to implement (i.e. Transmeta wouldn't bother coding a VAX emulation
>layer because it would cost them more money than they would make from it.)
I talked to someone (not a TM employee but someone whose company works with
their stuff a lot) who said that although writing a whole new personality
for the Transmeta is a Big Deal, it might not be all that hard to convince
them (i.e. maybe all you need is a check with enough zeros on it) to add
a few special-purpose instructions to the 80x86 morpher, to significantly
streamline an emulator written in mostly-80x86 code. Neat idea... Of
course, you'd presumably need to write a new check every time they make an
incompatible change to the micro-architecture (or whatever you call what
the Crusoe has), to port your instructions over for you. I'm sure they have
absolutely no interest in producing an SDK for someone writing their own
morpher thingy, insulating themselves from sub-architecture changes is the
whole point of what they're doing...
I've heard they've done at least one other personality as a proof of
concept kind of thing, but I doubt they'd have any reason to support that
kind of thing, when they've already covered 90% of applications with the
80x86 thingy they've already got, so if it's true I doubt it will see the
light of day. Unless they start making Crusoe chips that are actually faster
than something else, then they'd be able to spread out and start competing
with anyone they felt like.