Intel CPU Designed For Transitive Binary Translator

[asian1]

If this new Intel CPU is succeful, is it easier to port AmigaOS 4 to the new Intel CPU?

From ServerPipeline.com:

"Intel Corp. has opted to work with Transitive Corp., a pioneer of processor virtualization and emulation technology, to allow software code originally written for RISC processors to operate on the Itanium 2 and Xeon processors, Transitive said Tuesday (March 7, 2006). The deal, despite being non-exclusive, is being presented as a booster for the Itanium community and as an attempt by Intel to fight back against rival Advanced Micro Devices Inc. in the high-end computer and server markets. Intel’s next step could be to use Transitive technology to take communications software—frequently written and compiled for RISC processors such as PowerPC and MIPS—and let it run transparently on Intel processors. Ultimately, it could start designing processors to assist the Transitive software."

[bloodline]

Transitive are the guys who designed Rosetta, the PPC->x86 translator in MacOS X, it works rather well... But the PPC and the x86 are quite similar chips, the Itanium is a bit more difficult... code optimised for the PPC will be dog slow on the Itanium.

Anyway since, intel is now looking for someone to buy the Itanium project (that's right they want to get rid of it), I suggest this is just a little token gesture to make the Itanium look a bit more attractive potential buyers.

[asian1]

What will happen if Freescale/IBM/AMCC/Culturecom/Xilinx buy the license from Transitive and create a new PowerPC CPU with support for the binary translator?

Can this move help software developers in porting their applications to PowerPC platform?

[bloodline]

asian1 wrote:
What will happen if Freescale/IBM/AMCC/Culturecom/Xilinx buy the license from Transitive and create a new PowerPC CPU with support for the binary translator?

Can this move help software developers in porting their applications to PowerPC platform?

Why would they do that?

The days of the Instruction set architecture are over... When the PPC was developed the CPU actually ran the instructions the programmer wrote... now they just get translated down inside the CPU to some weird instruction set that would make little sense to a programmer.

[adolescent]

bloodline wrote:
code optimised for the PPC will be dog slow on the Itanium.

Why is that?  

Also, since Xeon was mentioned, this could be an extension of the Xeon EM64T line.  

[Speelgoedmannetje]

Sounds like the thing Transmeta did a couple of years ago...

[bloodline]

adolescent wrote:

bloodline wrote:
code optimised for the PPC will be dog slow on the Itanium.

Why is that?  

The EPIC architecture has been shown to be inefficient at emulation of complex traditional OOO general purpose instruction sets.

Also, since Xeon was mentioned, this could be an extension of the Xeon EM64T line.  

Transistive's software already runs on this architecture, Apple use it. It's not really news.

[boing]

>When the PPC was developed the CPU actually ran the
>instructions the programmer wrote... now they just get
>translated down inside the CPU to some weird instruction set
>that would make little sense to a programmer.

Are ye sure you wouldn't like to... rephrase that... laddie?

[bloodline]

boing wrote:
>When the PPC was developed the CPU actually ran the
>instructions the programmer wrote... now they just get
>translated down inside the CPU to some weird instruction set
>that would make little sense to a programmer.

Are ye sure you wouldn't like to... rephrase that... laddie?

Hmmm, yeah I could elaborate a little... I meant to say, that the CPU will take the instructions, break them down into a subset of what the Programmers sees. Reschedule them and perform complex register renaming... to the point that the original programmer would have no idea what he was looking at, and certainly wouldn't recognise it as his original program.

[Karlos]

bloodline wrote:

to the point that the original programmer would have no idea what he was looking at, and certainly wouldn't recognise it as his original program.

Such as this, for example...

Oh wait, did you say OOO?

My bad ;-)

[bloodline]

Ook rocks!!!

[Karlos]

It does. I'm thinking to make a PHP Ook! interpreter at work just for a giggle...