Intel Btrans: 68K to Itanium translator

[asian1]

Hello
Intel will use new strategy for porting old Intel X86 32 bit binary code to 64 bit Itanium binary code.
This Btrans approach use similar method with Digital / Compaq FX!32 for converting X86 to Alpha code.
The Btrans will do recursive / step-by-step binary translation from original 32 bit to 64 bit.

Btrans

If someone ported AmigaOS 4 to Itanium2, and use it on mass produced Supermicro Itanium 2 motherboard, is it possible to convert old Amiga 68K applications to Itanium2 using modified Btrans?
Is this idea have better price / performance compared with AmigaOne PowerPC G4?
 
Is this approach better than running WinUAE or LinuxUAE on top of Itanium2?

SuperMicro

[jd997uk]

Gaaaaaaaaaaaaaaaaaaaaaaaaah!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

A-bloody-nother Amiga-on-X86 thread.
Will you people get a life already? Unless you are prepared to fund it with real money by sponsoring a programming team to do it, as well as sponsoring application programmers to port important parts to native code and you are prepared to wait a long time  (ie never) for a return in your investment, it will never happen.
The easiest way is for Bernie Meyer to write a PPC to Itanium version of Amithlon. Since (with the situation being as it is) this about as likely  as Bill Gates porting Winxp to run on a Sinclair Spectrum, you may as well get used to the idea, that for the foreseeable future, OS4.x is PPC only.

-john

[asian1]

>A-bloody-nother Amiga-on-X86 thread.

Hello
Nope, Amiga on IA64 / Itanium 2.

>fund it with real money by sponsoring a programming team to do it
Yes, but I hope INTEL CAPITAL will support the effort, similar to porting BeOS to X86.
To convince INTEL and other investors, Hyperion and (if possible Amiga Inc) should have a good CFO and solid, realistic financial plan, schedule and proposal.

INTEL CAPITAL

>native code
According to Hyperion, most AmigaOS 4 codes are portable across different CPUs. I am sure the remaining codes can be ported from PowerPC to IA64 assembler.

At first run, old 68K and PowerPC codes are run using emulation (slow). At second run, parts of the program are converted to IA64 code. The process is repeated several times, until most or all of the code are in IA64 format.

[downix]

COnsidering that Itanium is the modern form of the PA-RISC core found
in Commodore's next-gen chipset design, Hombre, this is not such a bad
idea.

The new strategy isn't new at all.

They've moved from a hardware emulator (which runs all the x86 stuff really badly inside the Itanium, no prediction, no optimisations, just as it hits it) to a software one (JIT, branch optimisations, caching, whee!!)

Why would you need to use "this approach" when it's nothing more than a software emulation of one instruction set on top of another?

All you guys running Amithlon pretty much do this for 68k->x86, everyone with a Pegasos is running 68k->PPC on some stuff right now.

The best solution according to what you want: Amithlon III: Amitanium!

[DeQuevedo]

AMiGA and INTEL can´t be spoken in the very same TENSE!!!!!


AMiGA ROOLZ, 68K ROOLZ,PPC ROOLZ!!

[asian1]

>nothing more than a software emulation of one >instruction set on top of another?

No. At initial run, Btrans use pure emulator method, but later, this program will MODIFY and OPTIMIZE the original 68K program to native Itanium 2 application.
The converted program will NOT use the emulator anymore.

Perhaps Btrans can also substitute the parts of program that need Amiga chipsets with new codes.

In theory this approach is faster that JIT or other approach.

[asian1]

>Endian - From AmigaOne CPU thread.

Hello
One of the feature of Itanium 2 is "endian neutral". The CPU can read both big and little endian data.

Endian Neutral

>AROS X86 - Applications.
It is possible to port AROS to Itanium2.
(similar to current attempt to port AROS to PowerPC).

If someone re-compile the source code of Amiga 68K program to AROS X86 binary, the X86 binary can run on top of Itanium2 X86 emulator, or use a special Btrans code converter program.

[Floid]

downix wrote:
COnsidering that Itanium is the modern form of the PA-RISC core found
in Commodore's next-gen chipset design, Hombre, this is not such a bad
idea.

Yep, but before playing the heritage card, DH has said the PA-RISC was only planned as part of the GPU.  So better/worse/indifferent, it's more like basing a proverbial WindowsOne on a GeForce4000 than any particular 'logical extension' from 680x0.

What would serve us best would be, as always, a translator (and/or just a port)  to a VP-type system, making these issues moot.  But VP's featureset isn't quite there yet, and the alternatives suck worse for whatever reasons.

[mikeymike]

@ jd997uk

A-bloody-nother Amiga-on-X86 thread.

My feelings exactly :-)