Cross-Platform Universal Emulation?

[GregA]

They demonstrated Linux-Quake running on a Mac and Linux-GIMP running on Windows.

That's great and wonderful. But how about running Microsoft Office apps (either the Windows or the MacOS version - I don't care which) on Linux?

Call me crazy, but it seems that this test is striking in its absence. If I was developing this product, I wouldn't want to issue any press releases until I could demonstrate MS Office running on a foreign platform, if for no reason other than publicity.

Yes. Notice also they don't say which processor the Linux binaries are made for.

Are they saying for example
Quake for Linux-PPC running on a Mac-PPC? (ie NO processor emulation)
... or
Quake for Linux-Intel running on a Mac-PPC?

Same goes for
GIMP for Linux-Intel running on Windows-Intel (ie NO processor emulation)
... or
GIMP for Linux-PPC running on Windows-Intel

Not much of a feat for an emulator if they're not emulating any chip. But for multiple PC companies to be interested, maybe they've done more...

....in fact, their website implies some of that. It could be quite a good thing to allow Linux-Intel binaries (which normally run on a Linux-Intel OS) to instead run directly on a Linux-PPC OS. Then there's no OS or GUI overhead to emulate, just the application code itself. This seems a far more achievable goal than the Windows stuff.

If the Mac improved it's Linux compatibility AND added this, the Mac could start running Linux-Intel binaries directly (and where possible run Linux-PPC recompiled versions).

Naturally... Windows programs are entirely another issue!

 

[locovaca]

So, after you realize that impossibility of this, you start to wonder how that good old video card developer Bit Boys is doing nowadays... All of you guys buying into this must know about them and their wild success in overthrowing Nvidia, 3dfx, Ati, S3, Kyro, etc. because they proposed the same vein of impossible notions as this company.

Listen, this would be awesome, but this just is not real. Let's look at the possible ways of doing this:

1. Emulating APIs/ OS System Calls (ala Wine on X86)- as it already has been mentioned, APIs aren't documented 100%, some OS's don't have matching API's or even things similiar enough to emulate the calls (and this IS emulation, no matter what they call it)- take a look at the Wine project. For as long as they have been working at it you still cannot run a great deal of applications under wine other than basic, no frills stuff. Or take a look at Cygwin on Windows- there's tons of software that can never be ported because the NT Kernel does not have the system calls *nix does and nothing anywhere close to even attempt to emulate them. Yes, bsd can emulate Linux for many things, but those two are very similiar OS's. If this is what they are proposing (*nix to *nix), big whoop- it's been done already.

2. Emulating processors/architectures- This is the option that seems most likely. However, you still have huge performance problems (especially with risc style -> x86, as pearpc has shown), and there are too many levels of abstraction to run an os on an os (again, as pearpc or VPC shows) at any kind of performance level they are claiming. And this seems to be even worse than the PPC/VPC, based on their claims. You get:

Virtual Host
--------------------
Transitive commands <--This layer does not exist with PPC/VPC
--------------------
Host Operating System


See that extra layer of abstraction? To do what they're saying (Hardware graphics acceleration, etc.) that's universal to any platform they would need to define a sort of intermediate language (similiar to how Java/.Net has their byte code) to convert from any processor to any processor. Despite the fact that they are selling this on a per host-platform basis, to save on development costs you don't start from scratch for each platform, but rather define a common format and reuse that code, and create only the interpreter for each host.

Plus, this brings up another point- what OS's does this run on? Or, is it its own OS, or does it take over BIOS/Firmware code completely? OS adds overhead, own OS makes installation more "permanent" (overwriting existing install), and BIOS/Firmware will most definitely cause many people to avoid it completely.

If it sounds too good to be true, it probably is.

 

[solafide]

you're right - Windows apps NOT supported

This also means that Windows applications will NOT run:
"QuickTransit supports operating system mapping between any two Unix/Linux-like operating systems, as well as mapping between mainframe and any Unix/Linux-like operating systems."
QUOTE]

You're right - with all the excitement, I missed this. They never say that this will allow the use of Windows apps on PPC hardware - only between Unix/Linux-like operating systems.

 

[GregA]

This is not an impossible thing. Infact... I had this idea myself and it is quite possilbe to do this, but as a student working full time i don't quite have the resources to pour into this thing.

For the any OS, and CPU thing to really work, one key point must be observed.

Any program that you want to work this way must not directly interact with hardware.

Most programs written for windows 2000+, or OS X satsify this condition. They interact with the hardware though an OS abstaction of the hardware. (ie. to open a file you don't call interupt what ever, request a DMA channel blah, blah blah, you just call open file (or whatever).<snip>

The thing is, they'd need to map all the Win32 calls to OSX calls. Wine has worked on this (for Linux) for a LONG time and may never get all the Win32 calls. There's a LOT of work to do that!!!!

I'd be surprised if this kind of thing isn't part of Microsofts plan for VirtualPC for Mac. They'd still have an entire virtual PC, but could rewrite certain parts of the OS to speak directly to the Mac OS - for example reading and writing files might use Mac filesystem calls directly (instead of a virtual filesystem). MS could also use fat binaries to speed up different parts of Windows XP. Time will tell...

 

[space2go]

Anybody else concerned that one of the apps they chose to test on was the Gimp on Windows, when there already is a Windows Gimp?

That caught me too.

And in the other example the say they ran the Linux version of Quake III on a Powerbook. They did not say on OSX.

So both things can be true without their software actually working...

 

[Doctor Q]

My take on it....

Excellent explanation, KingOfPain. Thank you. Comments like yours are very helpful when announcements send us off speculating past what's really been announced.

 

[jaromski]

i think univeral emulators are marketing b$

i just wanted to chime in and say i think "universal emulation" has an analog in cortez's city of gold or de leon's fountain of youth. nice little fairy tales told at varying intervals to inspire faith in something that is complete bull*****.

writing an emulator is quite difficult. writing a good emulator is nearly impossible. an emulator is a compiler. a compiler, in simplest terms, maps one computer langauge to another. this could be C => x86 asm, perl => c, or even x86 asm => ppc asm. i have written a compiler, a "simple" compiler for a "c" language => sparc assembly (well a subset of c) and it was the most difficult program i have grappled with ...ever. now through a few more "requirements" (other than running w/o crapping out)... say speed for one. well speed isn't a primary objective. speed is an optimization. optimizations have _HUGE_ costs programatically. and they usually don't pay off in terms of performance for the amount of effort put into "speed up" things. i mean it would be incredibly, nearly impossible, to generalize some algorithm that maps x86 into the optimal ppc assembly, every time, and in every context-- especially since there isn't an exact one-to-one map. there are "function equivalents" which immediately pre-suppose a one-size-fits-all to your solution space. one-size-fits-all usually means crap in the optimization realm. optimization is about finding those special cases and tweaking the **** out of them. now honestly is this "universal emulator" going to take every case into consideration and know how to tweak it up for a mere 20% performance hit? possible? maybe. probable? not by a long shot.

i don't know if i am making any sense, i just want to reiterate that i am highly skeptical of this software alchemy they have conjured up. emulators aren't exactly a new discipline in computing; they have been around for a long time. and the quality of your average programmer hasn't dramatically increased recently, i mean we have more tools and more abstractions, but for something as low level as emulation/compilers i think the rapid dev tools aren't making the dramatic inroads that these guys are suggesting.


my take, and i could be wrong.

jaromski

 

[BWhaler]

I really don't understand this, so I hope someone can help me out.

People claim it does not allow OSX run on PC's, for example, but rather let's programs run on any operating system.

SO then how is it a hardware emulator? Don't programs make calls to the operating system, not the hardware layer? (I know there are exceptions, but generally speaking.

I guess I am just plum confused as to what this is. Any help would be appreciated.

As for more my general reaction, if this let's me run any Windows app on OSX with similar speeds, but does not let OSX run on WinTel, I think this would be a HUGE boost for Apple. Fingers crossed...

 

[BWhaler]

I am going to go out on a limb here and say that Apple is one of the 6 companies in line. My prediction is that they will incorperate it into OS X, so that they never have to worry about Microsoft's commitment to the platform ever again -- I bet the boys over at the MacBU are soiling themselves at this very moment.

If this "hardware virtualisation" software is all its's cracked up to be, and they use it in Tiger, it would be the ultimate switch incentive to beleagered businesses who feel they are too heavily invested in software for an insecure windows environment. Of course other *nixes would benefit as well.

My copy of VPC is about to get KNOCKED THE #@!% OUT!

I hope you are right. Nothing would be sweeter than for 10.4 to be able to run Windows apps at a decent speed. That would be great for us as customers, and beyond fantastic for Apple.

 

[GregA]

I really don't understand this, so I hope someone can help me out.

People claim it does not allow OSX run on PC's, for example, but rather let's programs run on any operating system.

SO then how is it a hardware emulator? Don't programs make calls to the operating system, not the hardware layer? (I know there are exceptions, but generally speaking.

When you compile a program, you compile it for the operating system AND the hardware it runs on. So you might have a program compiled for Linux-on-Intel. When your program puts commands directly to the chip, it gets all the speed increases the Intel chip (in this case) can give. It's also using specific Intel functionality, and is optimised to how the Intel chips handle their instructions.

Virtual PC works by emulating a whole Intel machine (a virtual Intel PC). You could then install Linux on this emulated Intel machine (the Linux would not realise it was emulated, it'd see it as an Intel machine).
..... With this setup - you could run a Linux-Intel application, installed on your Linux on the virtual PC. All the stuff your Linux-Intel application ran would be converted by the emulator - which slows it down a bit of course. ALSO, when your Linux program had a command to open a file or whatever, the Linux-Intel operating system would do it's stuff, slowing it down even more as it's also converted by the emulator.

This new system seems to leave the operating system alone as much as it can, and just converts the hardware specific stuff. So it doesn't emulate a machine, rather it translates chip specific instructions. This means, for instance, we might see yellow-dog Linux (for PPC) selling their Linux with the ability to run any and all Linux (for Intel) binaries.

Also, if Transitive went to IBM, they could help IBM run their mainframe stuff on PPC. It'd have to be a partnership - get the mainframe's APIs (or whole OS) running on the PPC, and use Transitive's hardware translation to handle the hardware specific parts. Any PC manufacturer supporting more than 1 architecture could do the same (IBM, HP, Sun, MS(Xbox)... ) or anyone on a different architecture wanting Linux-Intel compatibility (Apple?). I'm sure some of these companies are already developing this functionality themselves!

I guess I am just plum confused as to what this is. Any help would be appreciated.

As for more my general reaction, if this let's me run any Windows app on OSX with similar speeds, but does not let OSX run on WinTel, I think this would be a HUGE boost for Apple. Fingers crossed...

Yeah, well I don't think this will do that, they don't have access to Windows source code. They do have access to Linux source code naturally, so I'd guess they're using that to get the Linux APIs running on Windows.

Hope that helps... and someone will correct (or improve) my explanation I'm sure. It's great how the multiple viewpoints on here lead us to a better understanding!

 

[iMan]

Too good?

Hmm... this sounds really too good to be true, but if... it will shake the world as we know it...

Interesting part is that this is obviously not something you buy and then run on your computer - like a virtual PC thing... it is already built into the system from the manufacturer... they say two major manufacturers have already plans to release something late 2004 or early 2005... Would anyone believe Apple is one and this to be coincidal with the release of Tiger...? How about that for a MWSF killer! What a boost it would be for Mac users to utilize every windows app ever written just like that... the final argument of the windroids (you know that game thing...) finally falling flat to the ground...

yeah...

 

[Hubertus]

point it out

- this is a shipping product

- if this product really does emulate all the software, we would not discuss about this on a rumor site anymore. it would be in all newspapers, screenshots everywhere and this product would be advertised very different on the companys web page.

right?

and in my opinion this company would not exist anymore. money does a lot.

and they speaker about the server business as the target market. this means alot, no??

bye,

hubertus

 

[locovaca]

They do have access to Linux source code naturally, so I'd guess they're using that to get the Linux APIs running on Windows.

But this has already been done, so what would the big deal be? http://www.cygwin.com

(And there are limitations of Windows that keep from ANY program being run on Cygwin, which is why you do not see KDE/Gnome/Etc. running on Windows)

 

[Sol]

Impossible promises

The only commercial emulator software that delivered on its promises was Virtual Game Station and that was retired early due to the politics of a little Mac developer undermining SONY. Imagine what Apple's reaction would be if anyone could run the iLife applications on x86 hardware. Imagine what Microsoft would have to say in court about people loading Windows software on G5 Macs, without buying Windows XP.

Not that anything like that is going to happen. This smells like vapourware timed to coincide with the arrival of Virtual PC 7.

 

[~kilroy~]

From a propagation standpoint, this is bad - from a local damage standpoint, the viruses may or may not be written to access the OS X system properly (e.g. a virus trying to write to c:\ won't have any effect on a Mac).

Macs are not imune to viruses today. The only reason there are so few viruses for the mac has been because of market share.
Don't let that make you complacent though. I fully expect Macs to start getting thier fair share of viruses within the next year to 18 months.

 

[wPod]

sounds too good to be true

 

[sushi]

Transparent Aluminum Is Here

Not really, at least by Star Trek IV definition.

Sushi

 

[encro]

Hmmm... this remind's me of TAO Group's Intent.

It is a very similar idea in hardware abstraction but geared towards multimedia content.

 

[shamino]

"One of the key breakthroughs is an "intermediate representation," a kind of lingua franca that gives the software the flexibility to translate from one platform to another."
I'd be really interested in that one, because I once analysed basically dozens of processor architecture trying to come up with just such a thing. After several months I dropped the idea, because it's just not practical. Even some of the most basic stuff (eg. integer division) is so diverse that you either have to include every possible implementation or have it represented in very simple steps.
In the first case the intermediate representation becomes a real monster that no one is willing to write a backend for another architecture, and in the second case the code quality will be really bad.

The only production example of this concept (an intermediate assembly language) that I'm aware of is the GNU assembler (part of their binutils package). It's my understanding that the gcc compiler works by generating an intermediate meta-assembly language, which is then assembled into native code using this assembler.

I suppose you could apply similar techniques to compile one processor's assembly language into another. Some instructions (those without analogues) would be painful, requiring whole subroutines as emulation, but I suspect that the most common instructions (including everything generated by typical compilers) may be portable enough for this approach to bear useful results.

Of course, WRT Transitive's product, I'm still going to remain in the "I'll believe it when I see it" camp. Theoretical discussions are one thing - a product good enough to be useful as a general-purpose tool is quite another.

 

[MadMan]

Hmm...

According to this, they claim it can run OS X on x86?!?!

http://www.extremetech.com/article2/0,1558,1645408,00.asp

 

[må¥å]

Doesn't Mac OS check the onboard flash memory for the firmware to register if it is indeed Apple HW....that is the reason why Mac OS X own't run on the clones.

PC don't have this pre-check HW system.


So I am guesing that you can install and run WinOS and x86 application on a Mac and have it run faster than emulation.

I don't believe you can install any Mac OS and Apple Application on a PC even with this solution. Since the Mac OS will pre-check for a code, and I am pretty sure this thing cannot obtain that code unless Apple allows it, which it won't.


Anyhow reading this article all I can say is that Windows OS and applciations can run close to native on a Mac system that will stop those people who keep holding on to they PC boxes for M$ Access for work purposes.

This is Bad news for M$ however great for Apple, unless M$ buys this out then we are back to the start

 

[EvilMole]

Alright, some people are getting totally the wrong end of the stick about this. First of all, it helps if you read the description of the product and the press release:

"Availability and Pricing Model
Transitive’s QuickTransit products are currently available and shipping to major OEM customers. Pricing for QuickTransit products is based on a one-time technology license fee and a usage fee model that depends on the customer’s deployment strategy. "

In other words, this is not an end-user product. So what is it? It's a solution for ISV's, OEMs, and large corporations who want to get their product running on a different platform without porting.

Say you're a large corporation which uses some ancient mainframe code. However, replacing this ancient mainframe code is impossible: there's too much data locked up in it, too much investment in time and money, and as a system, it just works. Now, you can deploy QuickTransit and get this thing running on a nice new POWER or Opteron (or Itanium for the masochistic) system, at a fraction of the cost of porting or simply abandoning your old mainframe code.

Of course, if a company was dependent on one processor platform - like, say, a certain Apple Computer Inc - QuickTransit would be of great interest to them, as it would allow them to transition their OS to another processor platform much more easily... most importantly of all, while retaining full backwards compatibility with PowerPC applications. Interesting idea, eh?

 

[iChan]

That's what happens when Slashdot picks up a story.

why?

 

[40167]

Finally I can start using OS X on my Dull I got from work!

Uhm that technoology already exists dude... www.pearpc.net

Edit: Forgot to note one thing... I use that for my main operating system, (obviously windows is behind it though) so its pretty spiffy; even moved the pc to a g3 case (even though its emulating a g4 system)

And for apple.. since the lisence says "apple branded computer required" well.. I got an apple case... so the system is apple branded, lol.

 

[Dr. Dastardly]

Alright, some people are getting totally the wrong end of the stick about this. First of all, it helps if you read the description of the product and the press release:

"Availability and Pricing Model
Transitive’s QuickTransit products are currently available and shipping to major OEM customers. Pricing for QuickTransit products is based on a one-time technology license fee and a usage fee model that depends on the customer’s deployment strategy. "

In other words, this is not an end-user product. So what is it? It's a solution for ISV's, OEMs, and large corporations who want to get their product running on a different platform without porting.

Say you're a large corporation which uses some ancient mainframe code. However, replacing this ancient mainframe code is impossible: there's too much data locked up in it, too much investment in time and money, and as a system, it just works. Now, you can deploy QuickTransit and get this thing running on a nice new POWER or Opteron (or Itanium for the masochistic) system, at a fraction of the cost of porting or simply abandoning your old mainframe code.

Of course, if a company was dependent on one processor platform - like, say, a certain Apple Computer Inc - QuickTransit would be of great interest to them, as it would allow them to transition their OS to another processor platform much more easily... most importantly of all, while retaining full backwards compatibility with PowerPC applications. Interesting idea, eh?

If i'm reading this correctly then this would be much more plausible and still a huge boon for PPC computers. You would see more and more applications coming out for BOTH windows and OSX. This is even better IMHO.