Virtual PC 64 bit: how fast could it be?

Hi,

Suppose (it won't happen too soon since MS owns VPC now) that VPC is
optimized for 64 bit architecture of G5. I am not talikng about
emulating 64 bit Intel CPUs (Itanium, I think). Suppse you just want to
emulate the 32 Pentium 4, but you write the code to use 64 bits on G5.

How fast could it be???

right now I would "guesstimate" (based more on feeling rather than
benchmarks) that VPC 5 runs at about 300 MHz on my 900 MHz G3 iBook.

Now would it be possible to get each G5 to emulate 2 P2 32-bit
intructions in one cycle, so you actually get 4 instruction if you have
a dual g5. How fast do you think VPC would be? Would you expect to
compare it to a 2 GHz P4???

Just wondering if you can emulate 2 32-bit instructions of P4 in one
cycle of G5...

Thanks

In article
<ch1b1tul-(E-Mail Removed)>,
chibitul <(E-Mail Removed)> wrote:

> Hi,
>
> Suppose (it won't happen too soon since MS owns VPC now) that VPC is
> optimized for 64 bit architecture of G5. I am not talikng about
> emulating 64 bit Intel CPUs (Itanium, I think). Suppse you just want to
> emulate the 32 Pentium 4, but you write the code to use 64 bits on G5.
>
> How fast could it be???
>
> right now I would "guesstimate" (based more on feeling rather than
> benchmarks) that VPC 5 runs at about 300 MHz on my 900 MHz G3 iBook.
>
My 17" PowerBook G4 running Win XP in VPC6 reports itself as a 1.33 GHz
Pentium, so I'm thinking 2 GHz should be a snap on a G5, even without
any optimization.


- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
"We really haven't done everything we could to protect our
customers ... Our products just aren't engineered for security."
-- Brian Valentine, Sr VP in charge of MS Windows' Dev Team

In article <ThomRosario-(E-Mail Removed)>,
Thom Rosario <(E-Mail Removed)> wrote:

> >
> > right now I would "guesstimate" (based more on feeling rather than
> > benchmarks) that VPC 5 runs at about 300 MHz on my 900 MHz G3 iBook.
> >
> My 17" PowerBook G4 running Win XP in VPC6 reports itself as a 1.33 GHz
> Pentium, so I'm thinking 2 GHz should be a snap on a G5, even without
> any optimization.

Well, does it "feel" like a 1.33 GHz Pentium???

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

At 2003-06-29T14:24:31Z, chibitul <(E-Mail Removed)> writes:

> Now would it be possible to get each G5 to emulate 2 P2 32-bit intructions
> in one cycle, so you actually get 4 instruction if you have a dual g5.

No. However, I *think* that the "G5" has more registers, so VPC's CPU
emulation core could be re-compiled to store more temporary information in
registers than RAM. Since registers are faster than even L1 cache and
aren't subject to being flushed, that may speed up emulation.

Out of curiosity, is VPC's CPU emulation multi-threaded? I haven't used it,
but I sort of doubt it - and I can't imagine that it would be even remotely
efficient it if did.
- --
Kirk Strauser
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)

iD8DBQE+/6l35sRg+Y0CpvERAhJcAKCDcWAbSVDGcyf4yZTF34cW1Lt9SQC fdEl8
XJ9pWuO28ROVfg9FNDFXuk4=
=ubD8
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

At 2003-06-29T15:07:12Z, Thom Rosario <(E-Mail Removed)> writes:

> My 17" PowerBook G4 running Win XP in VPC6 reports itself as a 1.33 GHz
> Pentium,

I used to run a VPC-alike program on my Amiga. I could configure it to
report as a Pentium 200 or a 386 16, but the speed was identical in either
case. The real question is: how fast does it truly run?

> so I'm thinking 2 GHz should be a snap on a G5, even without any
> optimization.

Not a chance.
- --
Kirk Strauser
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)

iD8DBQE+/6n75sRg+Y0CpvERAlI9AKCPWcS8e3k5uFbieILZ0qqtJJ7ORwC fa5aV
6Kp9kr9nsrpzaDwfjmR8wNQ=
=tJuD
-----END PGP SIGNATURE-----

In article <(E-Mail Removed)>,
Kirk Strauser <(E-Mail Removed)> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> At 2003-06-29T15:07:12Z, Thom Rosario <(E-Mail Removed)> writes:
>
> > My 17" PowerBook G4 running Win XP in VPC6 reports itself as a 1.33 GHz
> > Pentium,
>
> I used to run a VPC-alike program on my Amiga. I could configure it to
> report as a Pentium 200 or a 386 16, but the speed was identical in either
> case. The real question is: how fast does it truly run?
>
> > so I'm thinking 2 GHz should be a snap on a G5, even without any
> > optimization.
>
> Not a chance.

Have you used VPC on a G5? Or even a recent G4? It's quite impressive.


- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
"We really haven't done everything we could to protect our
customers ... Our products just aren't engineered for security."
-- Brian Valentine, Sr VP in charge of MS Windows' Dev Team

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

At 2003-06-30T13:19:04Z, Thom Rosario <(E-Mail Removed)> writes:

>>> so I'm thinking 2 GHz should be a snap on a G5, even without any
>>> optimization.

>> Not a chance.

> Have you used VPC on a G5? Or even a recent G4? It's quite impressive.

No, I really haven't. I'm sure that it is very nice. However, it would
have to be running at >100% efficiency to execute x86 code faster than a
current-model x86 processor. I have no doubt that it's very usable, but I
have severe doubts that it could outrun a 2 GHz Intel CPU when both are
running Intel code.
- --
Kirk Strauser
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)

iD8DBQE/AJl55sRg+Y0CpvERAkKkAJ0VNMFCDiFds0B+9dIy4Xz0n22jdg CgiGbQ
3tThAbbfvOkTZhZdgcRrmq8=
=fsUY
-----END PGP SIGNATURE-----

You could 'compile' the x86 code into PPC code. That is take a
segment of x86 code and produce a PPC shadow of the X86. That way
instead of have to interpret the x86 code everytime it is run, you
just run the x86 shadow.

I do not know if VPC does this or not. MS would be in a good position
to do things like this. To get really good performance out of VPC, MS
could compile critical portions of Windows for the PPC, but M$ is not
likely to do this.

Robert

(E-Mail Removed) (Robert) writes:

[x86->PPC code compilation]
>I do not know if VPC does this or not.

VPC does do a certain amount of JIT-style compilation on the x86 code,
and caches it as well, which improves overall performance if you can stuff
in more RAM and let it cache more of the system code.

>MS would be in a good position
>to do things like this. To get really good performance out of VPC, MS
>could compile critical portions of Windows for the PPC, but M$ is not
>likely to do this.

MS is not likely to optimize VPC for the Mac at all. They would be loathe
to encourage Windows growth on a platform they don't control well (look
at how quickly they dropped PPC WinNT, thank goodness).

--
Cameron Kaiser * (E-Mail Removed) * posting with a Commodore 128
personal page: http://www.armory.com/%7Espectre/
** Computer Workshops: games, productivity software and more for C64/128! **
** http://www.armory.com/%7Espectre/cwi/ **