Universal Kernel?

[Rick Langschultz]

Hey everyone,

I have a feature suggestion for version 1.0 or later... I think it  
would be called the universal kernel. The kernel could be compiled  
and made to run on various architectures: ia_32, ia_64, PPC, XBOX,  
PPC64, etc. The computer would translate natively compiled files from  
other systems into ia_32 or ia_64 code. I know that Windows XP (64  
bit) uses WoW which allows 32 bit programs to access 64bit memory or  
something like that. This way reactos could be installed on a variety  
of systems with one installation disc instead of a flavor of that  
disc. Just a suggestion for the future...
Rick Langschultz
[e-mail] (Home)
[e-mail] (Work)
[e-mail] (School)

[David Johnson]

Sounds great to me.. Very cool.

On 9/11/05, Rick Langschultz [e-mail] wrote:
> Hey everyone,
>
> I have a feature suggestion for version 1.0 or later... I think it
> would be called the universal kernel. The kernel could be compiled
> and made to run on various architectures: ia_32, ia_64, PPC, XBOX,
> PPC64, etc. The computer would translate natively compiled files from
> other systems into ia_32 or ia_64 code. I know that Windows XP (64
> bit) uses WoW which allows 32 bit programs to access 64bit memory or
> something like that. This way reactos could be installed on a variety
> of systems with one installation disc instead of a flavor of that
> disc. Just a suggestion for the future...
> Rick Langschultz
> [e-mail] (Home)
> [e-mail] (Work)
> [e-mail] (School)
>
> Ros-dev mailing list
> [e-mail]
> http://reactos.com:8080/mailman/listinfo/ros-dev
-- 
David Johnson
http://www.davefilms.us

[Rick Langschultz]

I thought it would give reactos uniformity, and able to be able to  
translate PPC and other binaries with a slight hit on speed, but  
probably not much...
On Sep 12, 2005, at 12:35 AM, David Johnson wrote:
> Sounds great to me.. Very cool.
>
> On 9/11/05, Rick Langschultz [e-mail] wrote:
> > Hey everyone,
> >
> > I have a feature suggestion for version 1.0 or later... I think it
> > would be called the universal kernel. The kernel could be compiled
> > and made to run on various architectures: ia_32, ia_64, PPC, XBOX,
> > PPC64, etc. The computer would translate natively compiled files from
> > other systems into ia_32 or ia_64 code. I know that Windows XP (64
> > bit) uses WoW which allows 32 bit programs to access 64bit memory or
> > something like that. This way reactos could be installed on a variety
> > of systems with one installation disc instead of a flavor of that
> > disc. Just a suggestion for the future...
> > Rick Langschultz
> > [e-mail] (Home)
> > [e-mail] (Work)
> > [e-mail] (School)
> >
> > Ros-dev mailing list
> > [e-mail]
> > http://reactos.com:8080/mailman/listinfo/ros-dev
>
> -- 
> David Johnson
> http://www.davefilms.us
>
> Ros-dev mailing list
> [e-mail]
> http://reactos.com:8080/mailman/listinfo/ros-dev

Rick Langschultz
[e-mail] (Home)
[e-mail] (Work)
[e-mail] (School)

[Alex Ionescu]

Rick Langschultz wrote:
> I thought it would give reactos uniformity, and able to be able to  
> translate PPC and other binaries with a slight hit on speed, but  
> probably not much...
> On Sep 12, 2005, at 12:35 AM, David Johnson wrote:

This technology is called "QuickTransit" (it's used by MacOS X under the 
name "Rosetta").
We could probably license it...

Best regards,
Alex Ionescu

[David Johnson]

Lets do it.

On 9/11/05, Alex Ionescu [e-mail] wrote:
> Rick Langschultz wrote:
> > I thought it would give reactos uniformity, and able to be able to
> > translate PPC and other binaries with a slight hit on speed, but
> > probably not much...
> > On Sep 12, 2005, at 12:35 AM, David Johnson wrote:
>
> This technology is called "QuickTransit" (it's used by MacOS X under the
> name "Rosetta").
> We could probably license it...
>
> Best regards,
> Alex Ionescu
>
> Ros-dev mailing list
> [e-mail]
> http://reactos.com:8080/mailman/listinfo/ros-dev
-- 
David Johnson
http://www.davefilms.us

[art yerkes]

David Johnson wrote:
> Lets do it.
>
> On 9/11/05, Alex Ionescu [e-mail] wrote:
> > Rick Langschultz wrote:
> > > I thought it would give reactos uniformity, and able to be able to
> > > translate PPC and other binaries with a slight hit on speed, but
> > > probably not much...
> > > On Sep 12, 2005, at 12:35 AM, David Johnson wrote:
> >
> > This technology is called "QuickTransit" (it's used by MacOS X under the
> > name "Rosetta").
> > We could probably license it...
> >
> > Best regards,
> > Alex Ionescu
> >
> > Ros-dev mailing list
> > [e-mail]
> > http://reactos.com:8080/mailman/listinfo/ros-dev

Let's work on something we won't have to spend millions of dollars
licensing.  If you're really interested, you might work on a bare metal
implementation icvm and an NT port to that "architecture".  It would
come very close to going the whole way to architecture neutrality.  With
some setup magic, you should be able to make it work on most PCI and
PCI-like systems.  You might start by working with kjk on the userland
port (ROX-U), because that will likely be able to run out of the box on
ICVM.

[Mike Swanson]

I'd say that this should be left to user-mode translators (eg,
something like QEMU but for Windows/ReactOS). Most Windows
applications are currently built only for IA-32, it's hard to find
things compiled for other platforms (Win64 is gaining popularity, but
still...).

Running ReactOS on a platform in which Windows as limited or no
support already, such as SPARC or PowerPC, does indeed leave the user
practically left in the dust in terms of compatibility. However, most
users will likely be running ReactOS on x86 or AMD64, so in-kernel
translations of binaries will just add bloat that most users will
never need. It's my opinion that if any project is made to run
non-native binaries (like what QEMU on Linux does), it really should
be left in the user-space, as an optional thing to run, but does not
add any bloat to the core system itself.
-- 
Mike

[Radovan Skolnik]

Nice idea. However there's one thing that you missed: AMD64 bit processors
enable to run x86 32bit code without modification (same goes for Intel).
That means in this scenario there is no translation involved - the CPU
executes the 32bit instructions directly and OS just supports this scenario
(it's similar to running 16bit applications on 32bit processors).

The truth is that on Mac OS X there is this translation technology being
able to run PPC binaries on x86 (the thingy is pretty complex probably).
However it doesn't work on all of them and I doubt it would be easily
licensable to project like this. But one never knows. There are already
things to base such work on: PearPC as PPC emulator and some others probably
exist for other HW platforms.

Best regards

Radovan Skolnik

-----Original Message-----
From: [e-mail] [mailto:[e-mail]] On
Behalf Of Rick Langschultz
Sent: Monday, September 12, 2005 5:43 AM
To: ReactOS Development List
Subject: [ros-dev] Universal Kernel?

Hey everyone,

I have a feature suggestion for version 1.0 or later... I think it would be
called the universal kernel. The kernel could be compiled and made to run on
various architectures: ia_32, ia_64, PPC, XBOX, PPC64, etc. The computer
would translate natively compiled files from other systems into ia_32 or
ia_64 code. I know that Windows XP (64
bit) uses WoW which allows 32 bit programs to access 64bit memory or
something like that. This way reactos could be installed on a variety of
systems with one installation disc instead of a flavor of that disc. Just a
suggestion for the future...
Rick Langschultz
[e-mail] (Home)
[e-mail] (Work)
[e-mail] (School)

_______________________________________________
Ros-dev mailing list
[e-mail]
http://reactos.com:8080/mailman/listinfo/ros-dev

[Rick Langschultz]

I figure the reactos kernel could be ported to PowerPC; later. So  
when reactos is stable on powerpc or another system reactos could use  
a utility to allow reactos to run mac os x programs transparently  on  
top of ReactOS. Or even IA-32 programs on that system also. This  
would be reverable PPC-to-IA-32 and IA-32--to--PPC. I figure if  
PearPC contributed code to the project with XEN and others ReactOS  
could translate these binaries. Just a thought...
Rick Langschultz
[e-mail] (Home)
[e-mail] (Work)
[e-mail] (School)

[art yerkes]

Rick Langschultz wrote:
> I figure the reactos kernel could be ported to PowerPC; later. So 
> when reactos is stable on powerpc or another system reactos could use 
> a utility to allow reactos to run mac os x programs transparently  on 
> top of ReactOS. Or even IA-32 programs on that system also. This 
> would be reverable PPC-to-IA-32 and IA-32--to--PPC. I figure if 
> PearPC contributed code to the project with XEN and others ReactOS 
> could translate these binaries. Just a thought...
> Rick Langschultz
> [e-mail] (Home)
> [e-mail] (Work)
> [e-mail] (School)
>
> Ros-dev mailing list
> [e-mail]
> http://reactos.com:8080/mailman/listinfo/ros-dev

Good ideas, but let's see some code.  Also, many of these ideas apply
generally and don't need to be reactos projects.  Running OSX binaries
on windows would involve a colinux like darwin kernel userland port,
which is quite feasible and also quite outide the scope of the reactos
project.  I'd encourage somebody to finally work on a userland mach on
windows.  Heaven knows that the win32 api is rich enough to map
virtually every mach call to a handfull of api calls.  Once you had
that, i'd expect it to take only a month or two of tinkering to get
userland darwin running.

[TwoTailedFox]

Good Idea in principle, but coding it would be another matter.

I like the idea of a Rosetta-Style subsystem, that catches a binary,
and then translates it real-time to run on the host architecture.
Thing is, that's very taxing on hardware, and it'd need a Decent CPU,
and Hard Drive to work best (Consider the system used in the Mac OS x
x86 Developer PC's is a 3.6GHz Pentium 4, that gives you the basic
indication of how much juice may well be required).

While such a system is outside the general scope of ReactOS, this
would be where we borrow the "Distributions" idea of Linux. Sure, the
Core ReactOS Goal is to develop a quality Operating System that is
NT-Compatible, but as the system becomes more and more complete, I
don't see why this can't be undertaken by a seperate set of
developers, pursuing a different goal, in addition to the main one.

Just my 0.02c

On 9/12/05, art yerkes [e-mail] wrote:
> Rick Langschultz wrote:
> > I figure the reactos kernel could be ported to PowerPC; later. So
> > when reactos is stable on powerpc or another system reactos could use
> > a utility to allow reactos to run mac os x programs transparently  on
> > top of ReactOS. Or even IA-32 programs on that system also. This
> > would be reverable PPC-to-IA-32 and IA-32--to--PPC. I figure if
> > PearPC contributed code to the project with XEN and others ReactOS
> > could translate these binaries. Just a thought...
> > Rick Langschultz
> > [e-mail] (Home)
> > [e-mail] (Work)
> > [e-mail] (School)
> >
> > Ros-dev mailing list
> > [e-mail]
> > http://reactos.com:8080/mailman/listinfo/ros-dev
>
> Good ideas, but let's see some code.  Also, many of these ideas apply
> generally and don't need to be reactos projects.  Running OSX binaries
> on windows would involve a colinux like darwin kernel userland port,
> which is quite feasible and also quite outide the scope of the reactos
> project.  I'd encourage somebody to finally work on a userland mach on
> windows.  Heaven knows that the win32 api is rich enough to map
> virtually every mach call to a handfull of api calls.  Once you had
> that, i'd expect it to take only a month or two of tinkering to get
> userland darwin running.
> _______________________________________________
> Ros-dev mailing list
> [e-mail]
> http://reactos.com:8080/mailman/listinfo/ros-dev
-- 
"I had a handle on life, but then it broke"