Re: Apple, Jobs, and the irony of it

[Sheldon Simms <sheldonsimms@yahoo.com>]

On Sat, 23 Aug 2003 21:58:35 -0400, Charlie wrote:

> "Supertimer" <supertimer@aol.com> wrote in message
> 20030823174449.00586.00000490@mb-m02.aol.com">news:20030823174449.00586.00000490@mb-m02.aol.com...
>> "Charlie" <charlied@NOSPAMbboard.com> wrote:
>>
>>> The Itaniums also have a 32 bit compatibility mode
>>
>> The Itanium runs 32-bit applications under emulation
> 
> How does the Opteron run 32-bit applications?  The AMD website and the Opteron
> data sheet are vague about how it is done.

Well it just runs them, since the Opteron instruction set is basically
the same as the Athlon x86 instruction set. It works pretty much exactly
like a 65816 running 8 bit applications.

Like the 65816, the Opteron has an "long" mode, that uses long (16/24 bit
for the 65816, 64-bit for the Opteron) addresses and operands with some
new instructions thrown in and a "compatibility" mode that works just
like a legacy CPU.

> 32-bit applications on an Itanium 2 run in "compatibility mode" which is a mode
> that is NATIVE to the processor.

Yes, but the Itanium's instruction set is completely, utterly different
than the x86 instruction set. The Itanium's compatibility mode is
really just a hardware-assisted software emulation of x86 in which
x86 instructions are translated to Itanium-native instructions on the
fly. This is better than pure software emulation, but it's not really
good enough to use for any real purpose (i.e., no one is going to buy
an Itanium to run primarily x86 software).

>> and it does so slower than a good 32-bit Athlon XP
>> or Intel's own 32-bit P4 for that matter.
> 
> True, but that still doesn't make the Opteron first <grin>.

There was/is x86 software emulation for 64-bit Alpha CPUs as well.

>> Note I said "a 64-bit processor that can run a
>> mainstream OS natively" with NATIVE being the
>> key.
> 
> Yes, and "native" is one of those words with various meanings when it comes to
> computers.  To me, native, in this case, means that the ability to run 32-bit
> software was built into the chip hardware.

Itanium does not natively execute x86 code in this sense, as
it requires low-level software to emulate an x86. The Itanium
processor provides hardware support for the emulation, but
that's not the same as native execution.

> Certainly, the Opteron was meant to be a transition chip (to make the
> transition from 32-bit to 64-bit easier) and from what I've read they have
> succeeded, but in my opinion it is not the first 64-bit processor to run a
> mainstream operating system.

People have been defining "mainstream operating system" to mean
whatever they want. I would call Windows NT mainstream for every
versions 4.0 and later. That would make Alpha the first 64-bit processor
to run a mainstream operating system.

> By the way, AMD says the Opteron is meant to be used in workstations and
> servers (whatever that means)

It means it's SMP certified and they charge too much for it.

> and they will have a 64-bit Athlon out in September, 2003.

Yes. The Athlon 64 and Opteron were originally scheduled to launch
much more closely together. I suspect the Athlon 64 was delayed because
the Opteron has some fabrication problems that AMD wanted to fix before
they did a mass-market launch. I expect the Athlon 64 will be a new
revision of the x86-64 architecture.

-Sheldon