[9fans] Plan9 VM for Windows/Linux?

[9fans] Plan9 VM for Windows/Linux?

[Joel Salomon]

On Tue Nov 26 16:24:52 2002 Jack Johnson wrote:
> Does anyone know of an inexpensive architecture that would virtualize
> notably easier than x86?  I don't think I've read of any attempts on
> PPC, Sparc or MIPS, though someone must have tried...

MMIX
http://www-cs-faculty.stanford.edu/~knuth/mmix.html

--Joel
______________________________________________________
Due to economic circumstances, the light at the end of
the tunnel has been turned off.

Re: [9fans] Plan9 VM for Windows/Linux?

[Robin KAY]

Doc Shipley wrote:

>   Only if you're a fan of x86's "backward-compatible-forever"
> philosophy.  I'm looking forward to seeing an AMD processor that isn't
> saddled with all of Intel's cruft.

But it is, the Hammer/Opteron will be fully compatible with the 32 bit Intel
Architecture. 64 bit addressing and the extended register set will be
enabled by a mode bit. The AA64 architecture is only a slight variation on
IA32 anyway.

--
Wishing you good fortune,
--Robin Kay-- (komadori)

Re: [9fans] Plan9 VM for Windows/Linux?

[Dan Cross]

>   Oh, wait.  I'm sorry, my mistake.  It won't run *Windows* in native
> mode.  Therefore, it *must* have been a stupid idea, right?

Uhh, no.

	- Dan C.

Re: [9fans] Plan9 VM for Windows/Linux?

[Doc Shipley]

On Wed, 27 Nov 2002, Dan Cross wrote:

> >   Well, the word here is that the Hammer series are "IA-64" *only* if
> > you want it to be.  By jumper or switch on the mainboard, you'll be
> > able to choose x86 mode, or native RISC.
>
> Great, yet another mode bit.  Only this one has to be physically
> manipulated!
>
> >   See Doc drool....
>
> Because it's a dumb idea, right?

  Only if you're a fan of x86's "backward-compatible-forever"
philosophy.  I'm looking forward to seeing an AMD processor that isn't
saddled with all of Intel's cruft.
  Just out of curiosity, what do you have against running a CPU with its
own native instruction set?
  Oh, wait.  I'm sorry, my mistake.  It won't run *Windows* in native
mode.  Therefore, it *must* have been a stupid idea, right?

	Doc

Re: [9fans] Plan9 VM for Windows/Linux?

[Dan Cross]

>   Well, the word here is that the Hammer series are "IA-64" *only* if
> you want it to be.  By jumper or switch on the mainboard, you'll be
> able to choose x86 mode, or native RISC.

Great, yet another mode bit.  Only this one has to be physically
manipulated!

>   See Doc drool....

Because it's a dumb idea, right?

	- Dan C.

Re: [9fans] Plan9 VM for Windows/Linux?

[Steven Taschuk]

Quoth Jack Johnson: [...]
> Does anyone know of an inexpensive architecture that would virtualize
> notably easier than x86?  [...]

The Java virtual machine, perhaps.

--
Steven Taschuk           |  o- @
staschuk@telusplanet.net | 7O   )
                         |  "  (    Hummingbird

Re: [9fans] Plan9 VM for Windows/Linux?

[Martin C.Atkins]

On 26 Nov 2002 20:47:25 -0000 Keith Nash <kjn9@citizenearth.com> wrote:
> > Does anyone know of an inexpensive architecture that would virtualize
> > notably easier than x86?  I don't think I've read of any attempts on
> > PPC, Sparc or MIPS, though someone must have tried...
> ...
>
> The 68020+ were good at virtualisation, but they topped out at 75MHz with the 68060.

It is my understanding the the PowerPC virtualizes easily. Hence
MacOnLinux, the system that lets you run MacOS in a window under
Linux on Mac Hardware.
The homepage is at:
	http://www.maconlinux.org

Does this count as an "inexpensive architecture"?

Martin
--
Martin C. Atkins	martin@mca-ltd.com
Mission Critical Applications Ltd, U.K.

Re: [9fans] Plan9 VM for Windows/Linux?

[Robin KAY]

Doc Shipley wrote:

> On 26 Nov 2002, Keith Nash wrote:
>
> > > Does anyone know of an inexpensive architecture that would virtualize
> > > notably easier than x86?  I don't think I've read of any attempts on
> > > PPC, Sparc or MIPS, though someone must have tried...
> >
> > I remember reading that AMD's IA-64/Hammer/Opteron platform fixes some of the x86's virtualisation problems.  Of course, you can't buy one yet.
>
>   Well, the word here is that the Hammer series are "IA-64" *only* if
> you want it to be.  By jumper or switch on the mainboard, you'll be
> able to choose x86 mode, or native RISC.

I think you'll find that that was a mistype of AA-64 (i.e. x86-64), the Hammer/Opteron cores are not capable of executing Itanic instruction
bundles.

--
Wishing you good fortune,
--Robin Kay-- (komadori)

Re: [9fans] Plan9 VM for Windows/Linux?

[Doc Shipley]

On 26 Nov 2002, Keith Nash wrote:

> > Does anyone know of an inexpensive architecture that would virtualize
> > notably easier than x86?  I don't think I've read of any attempts on
> > PPC, Sparc or MIPS, though someone must have tried...
>
> I remember reading that AMD's IA-64/Hammer/Opteron platform fixes some of the x86's virtualisation problems.  Of course, you can't buy one yet.

  Well, the word here is that the Hammer series are "IA-64" *only* if
you want it to be.  By jumper or switch on the mainboard, you'll be
able to choose x86 mode, or native RISC.

  See Doc drool....

	Doc

Re: [9fans] Plan9 VM for Windows/Linux?

[Keith Nash]

> Does anyone know of an inexpensive architecture that would virtualize
> notably easier than x86?  I don't think I've read of any attempts on
> PPC, Sparc or MIPS, though someone must have tried...

I remember reading that AMD's IA-64/Hammer/Opteron platform fixes some of the x86's virtualisation problems.  Of course, you can't buy one yet.

The 68020+ were good at virtualisation, but they topped out at 75MHz with the 68060.

Re: [9fans] Plan9 VM for Windows/Linux?

[suspect]

If a virtualisation of the hardware is what's wanted, then why
not go for Inferno ? (Aside from not being able to run your
C programs, but there are ways around that).

cheers,


On Tuesday, November 26, 2002, at 09:19  AM, Skip Tavakkolian wrote:

>>> As for a C compiler generating Dis, I believe that Dis
>>> cannot express all the things that one can do in C.
>>
>> It could be done, by mapping all C objects into a single
>> large Dis object and all code into a single module, for
>> example, also providing hooks to get at some of the non-C
>> features.  But why?
>
> The original thread was 'what if instead of drawterm there was a
> Dis-style VM and there was a C compiler for Dis so that you can
> run Plan9 term under other OS, without VMware'.
>

Re: [9fans] Plan9 VM for Windows/Linux?

[Jack Johnson]

On Mon, 2002-11-25 at 22:19, Geoff Collyer wrote:
> hard slogging, in part because the x86 processors are not as well
> suited to virtualisation as the IBM 370 was.  You'll note that they
> tell you up front that you can't run VMware recursively; VM/370 could

Does anyone know of an inexpensive architecture that would virtualize
notably easier than x86?  I don't think I've read of any attempts on
PPC, Sparc or MIPS, though someone must have tried...

-Jack

Re: [9fans] Plan9 VM for Windows/Linux?

[Skip Tavakkolian]

>> As for a C compiler generating Dis, I believe that Dis
>  > cannot express all the things that one can do in C.
>
> It could be done, by mapping all C objects into a single
> large Dis object and all code into a single module, for
> example, also providing hooks to get at some of the non-C
> features.  But why?

The original thread was 'what if instead of drawterm there was a
Dis-style VM and there was a C compiler for Dis so that you can
run Plan9 term under other OS, without VMware'.

Re: [9fans] Plan9 VM for Windows/Linux?

[Douglas A. Gwyn]

Geoff Collyer wrote:
> As for a C compiler generating Dis, I believe that Dis
 > cannot express all the things that one can do in C.

It could be done, by mapping all C objects into a single
large Dis object and all code into a single module, for
example, also providing hooks to get at some of the non-C
features.  But why?

Re: [9fans] Plan9 VM for Windows/Linux?

[Geoff Collyer]

If you're really talking about Plan 9 providing virtual CPUs, I think
it would take significant effort.  The VMware folks have done a lot of
hard slogging, in part because the x86 processors are not as well
suited to virtualisation as the IBM 370 was.  You'll note that they
tell you up front that you can't run VMware recursively; VM/370 could
be run recursively (e.g., to debug new versions of VM/370).  And of
course we have other processors to worry about than just the x86.

As for a C compiler generating Dis, I believe that Dis cannot express
all the things that one can do in C.  A free-standing limbo translator
for Plan 9 would be nice though.

[9fans] Plan9 VM for Windows/Linux?

[Skip Tavakkolian]

While we are discussing unusual (absurd? ☺) ideas, here is
another one.

Drawterm is great, but I wonder if it could be extended to include a
VM and a proper Plan9 kernel?  That is to say just like VMWare running
Plan9 term, except no VMware.  I keep wanting to think that a Plan9
kernel for Dis VM and a C compiler to generate Dis, would be nearly
there. Sounds too simple though.