Welcome to the MacNN Forums.

Simon · Sep 13, 2004, 11:03 AM

OK, is this feasible and could this be a Tiger killer-feature?

This wired article claims software emulation speeds of up to 80% should be reached with this piece of software. They say it's the same code that was used to allow XBox development to be done on PowerMac G5s.

So, just guessing, could Apple license this software, build it into Tiger and basically do away with the 'this software doesn't exist on the Mac platform' argument?

Millennium · Sep 13, 2004, 11:16 AM

Highly unlikely. Frankly, I believe that this emulator from Transitive is nothing more than vaporware.

Besides this, simply emulating the x86 chip is not enough. The entire Win32 API would need to be emulated. Apple got away with emulating the 680x0 series on the early Power Macs back in the OS9 days, but this is because they still used the same API and the actual machine architectures were still mostly the same (major changes, such as PCI slots, would not come until much later, and when these came things started to break).

There are projects attempting to do exactly what you suggest, such as Darwine. I have serious doubts about the technological feasibility of these projects, however. Darwine has ported winelib over to OSX, but this doesn't do what you want because it still requires that apps be recompiled. I don't think that they will ever succeed in implementing the emulation aspect the way they want to, because the architectures are simply too different.

eyadams · Sep 13, 2004, 11:39 AM

They say it's the same code that was used to allow XBox development to be done on PowerMac G5s.

That's not what they say. They say their software will allow the next-generation XBox (which will use PowerPC chips, just like the Mac) to play games written for the first generation XBox (which uses Intel chips - a PIII, if I remember correctly). There were some rumors in the Mac community a few weeks ago that initial development for the next generation XBox was being done on Mac G5, but frankly that only makes sense - you want to develop on a machine as close to your target as possible.

Frankly, this isn't all that impressive to me - if the next generation XBox is sufficiently powerful, it should be able to emulate just about any other kind of previous generation console, include XBox, Playstation 1, Atari 2600, etc.

Mrjinglesusa · 02:01 PM

As someone in the lounge posted, the catch is that they are emulating processors not operating systems. This is MUCH easier. Notice they didn't give an example of Windows software running on OS X. It was Linux on OS X, Linux on Windows, etc. Even the XBox example is processor emulation because the underlying OS in XBox 1 and 2 is essentially the same.

Horsepoo!!! · 12:31 PM

Look no further than my name to describe the reality of this product.

First of all, it's extremely improbable that this emulator can emulate all platforms perfectly...and near impossible that it can emulate them really fast.

If they're going to tell me it can emulate a graphics card, a G5 CPU + Altivec at close to even 40% of the of the target computer's speed on say an x86 box, they're lying through their teeth. There is no ****in' way an x86 processor can emulate the G5, the very complex Altivec chip and a graphics card at anything close even 20% let alone the 80% they claim.

They then claim that all programs work with 100% functionality...HORSESHIT AGAIN! There are apps that rely on very specific hardware that probably isn't emulated by the product.

This is a hoax, folks...and probably border-line illegal...nothing to see here.

Horsepoo!!! · 01:00 PM

Originally posted by Mrjinglesusa:
As someone in the lounge posted, the catch is that they are emulating [i]processors[i/] not operating systems. This is MUCH easier. Notice they didn't give an example of Windows software running on OS X. It was Linux on OS X, Linux on Windows, etc. Even the XBox example is processor emulation because the underlying OS in XBox 1 and 2 is essentially the same.

You forget that consoles don't just have one processor. Even computers now don't have just one processor...and the other processors can't necessarly do general computations...only the CPU can. Emulating many different hardware CPUs and chips on one CPU is not easy and certainly not fast.

Heck, if the 8-bit 6502 NMOS (1.79MHz) NES console needs 800MHz G4 computers to run at cycle-accuracy with all the other emulated hardware components, an entire computer platform will *not* run at 80% of the host computer's speed. They must have super hero geek coding powers and magic wands at Transitive...or something.

cambro · Sep 13, 2004, 02:47 PM

I agree with horeepoo, but more generally....

If this *could* be done it wouldn't it be the death of Mac OS X applications?

Why would developers work on a Mac OS X native application when they can develop it for Windows and have it "work" almost as well on Macs?

This is why I don't like VPC and hope VPC never really runs fast. It will break the back of Mac developers and leave mac users with little more than an OS with a bunch of non-native applications.

Horsepoo!!! · Sep 13, 2004, 03:11 PM

Originally posted by cambro:
I agree with horeepoo, but more generally....

If this *could* be done it wouldn't it be the death of Mac OS X applications?

Why would developers work on a Mac OS X native application when they can develop it for Windows and have it "work" almost as well on Macs?

This is why I don't like VPC and hope VPC never really runs fast. It will break the back of Mac developers and leave mac users with little more than an OS with a bunch of non-native applications.

Nah...if this product actually works (and I don't see how it would without slow emulation), how would it handle the GUI and framework/libraries apps use?

The GUI is part of the OS and so are the frameworks.

This product may have worked when apps were self-contained like back in the days...but today?

Mac OS X apps of today use frameworks like QuickTime and Rendezvous to name a couple, SIMD-units like Altivec, frameworks use GPUs like CoreImage/Video...I can't even imagine any of this being possible without loading the full OS for the GUI and frameworks and some kind of emulation for emulating Altivec and graphics card GPUs (which would be hilariously slow since it *has* to be emulated by the CPU.) A CPU emulating a SIMD-unit or GPU at any decent speed is near impossible...emulating both a SIMD-unit, GPU, and CPU simultaneously at decent speeds is impossible. At best, all 3 chips would share the CPU and run at 1/3 the speed...and we're not counting how complex it is to emulate a SIMD-unit or GPU...or even another CPU.

Even if the app were to be something more than vaporware, this product will in no way affect development on Macs because there's no way it would run most if not any Mac OS X apps on another platform.

The reverse is true...Windows apps use frameworks, MMX, graphics cards, etc...it's highly improbable they'd work with fidelity on another platform.

Simon · Sep 13, 2004, 03:44 PM

I tend to agree with you guys that everything sounds a bit too good to be true and since we haven't seen much real-world results vaporware seems rather likely.

But, hearing the rather harsh tone some of you chose to use with regard to this 'product', what is more troublesome for us: the fact that it could just be a vaporware or the fact that if it were true, for the first time really, Win users would be able to emulate Mac-only apps (Motion, iLife, etc. anyone?) with reasonable speed...?

Horsepoo!!! · Sep 13, 2004, 03:52 PM

Originally posted by Simon:
I tend to agree with you guys that everything sounds a bit too good to be true and since we haven't seen much real-world results vaporware seems rather likely.

But, hearing the rather harsh tone some of you chose to use with regard to this 'product', what is more troublesome for us: the fact that it could just be a vaporware or the fact that if it were true, for the first time really, Win users would be able to emulate Mac-only apps (Motion, iLife, etc. anyone?) with reasonable speed...?

I spent a whole post explaining why things would never run at reasonable speed or at all even if it were true.

MacGorilla · Sep 13, 2004, 08:15 PM

Apple wouldn't do it, even if they could. Why? That'd be saying "Our architecture isn't good enough, here's another". It would a profoundly stupid business move.

Simon · Sep 14, 2004, 02:36 AM

Originally posted by MacGorilla:
Apple wouldn't do it, even if they could. Why? That'd be saying "Our architecture isn't good enough, here's another". It would a profoundly stupid business move.

I don't think that's the message.

It's actually not anymore the message than offering the X11 environment to all OS X users. It's just reflecting a true Mac strength: To blend in everywhere seamlessly and make use of as much software possible. I don't find this to be a bad thing per se at all.

The other way around seems more troublesome though...

waffffffle · Sep 14, 2004, 03:18 AM

When OS X was brand new I was thinking that Connectix should try and create an emulation virtual machine environment that mimicked the Classic environment. Classic is truly an innovative virtual machine and it is integrated with OS X extremely well. I was hoping that Connectix would try something similar with Windows XP. Microsoft could probably do it better, but they probably won't.

What I'm talking about is the (almost) transparent integration between OS 9 and OS X. OS 9 doesn't run in a box. You can drag and drop and cut and paste content between them. I feel like the Windows interface is even better for this sort of task since there does not need to be a persistent system-wide menu bar. I was thinking maybe this would be the return of "red box" (classic is blue box, cocoa is yellow). It probably won't happen, but I think it would have been a real leap forward in emulation. WINE offers this sort of interface interaction with Windows and linux window managers. I think it can be done with OS X too.

Simon · Sep 14, 2004, 03:22 AM

waffffffle, indeed it would be very nice to get something in Tiger that feels like WINE, although the task would certainly be more difficult since it's not just APIs but a complete hardware layer that has to be emulated.

Graymalkin · Sep 14, 2004, 03:32 AM

The claims of this company are absurd. First, syscall translation is nothing new, it has been around for a very long time. For years now FreeBSD and Solaris have had Linux syscall mapping so they could both run unmodified Linux software. Second, the claims about native speed emulation are complete crap.

The only way they could provide high speed emulation is to pre-compile all of the binaries you're going to run rather than doing real-time emulation or JIT compilation. Even then you have to put a ton of work into optimizing for situations where you don't have enough registers on the host processor to handle emulated instructions. Intel's chips only have 8 general purpose registers, PowerPC chips have 32.

Even if you managed to write an awesome emulator and an equally awesome syscall translator you'd still need a virtual environment in which to run the emulated system. Operating systems expect to be running on hardware and have contro lof said hardware. In order to run one OS inside of another you need to abstract the hosted OS's hardware usage so it thinks it is running on real hardware.

This whole thing sounds to me like a pump and dump scam. The company is promising a technology everyone wants in order to generate hype. They then go public in order to secure funds for "development". They make a good show of things for a little while and then the principals cash out with their hype-inflated shares leaving what remains of the company to be sold at a firesale. I think we need to send our thanks to Wired for becoming the new Register when it comes to ridiculous technological claims and hype. I mean these guys demoed programs readily available for the platforms they were supposedly running their emulator on. I mean GIMP on Windows and Quake 3 on OSX? Please.

Wired needs to go through their writing staff and drop some of them like a prom dress at a drunken afterparty.

OpenStep · Sep 14, 2004, 01:03 PM

It would be OS/2 all over again

Millennium · Sep 14, 2004, 01:42 PM

Originally posted by waffffffle:
When OS X was brand new I was thinking that Connectix should try and create an emulation virtual machine environment that mimicked the Classic environment. Classic is truly an innovative virtual machine and it is integrated with OS X extremely well. I was hoping that Connectix would try something similar with Windows XP. Microsoft could probably do it better, but they probably won't.

What I'm talking about is the (almost) transparent integration between OS 9 and OS X. OS 9 doesn't run in a box. You can drag and drop and cut and paste content between them. I feel like the Windows interface is even better for this sort of task since there does not need to be a persistent system-wide menu bar. I was thinking maybe this would be the return of "red box" (classic is blue box, cocoa is yellow). It probably won't happen, but I think it would have been a real leap forward in emulation. WINE offers this sort of interface interaction with Windows and linux window managers. I think it can be done with OS X too.

The problem is that such virtual machines do not require an emulator (you'll note that WINE only runs on machines which would be capable of running Windows). When you have to tie an emulator into the mix, things get screwed up, to the point where they can become actually impossible if the architectures are different enough, In the case of your average Mac versus your average Wintel PC, this is the case; the architectures are simply too different for seamless integration to be possible.

waffffffle · Sep 14, 2004, 01:48 PM

Originally posted by Millennium:
The problem is that such virtual machines do not require an emulator (you'll note that WINE only runs on machines which would be capable of running Windows). When you have to tie an emulator into the mix, things get screwed up, to the point where they can become actually impossible if the architectures are different enough, In the case of your average Mac versus your average Wintel PC, this is the case; the architectures are simply too different for seamless integration to be possible.

I am quite aware of this. This is why I was hoping that Connectix would combine their emulation technology with a virtual machine. There has long been talk of combining WINE with an x86 emulator, but so far nothing has come of this.

yaro · Sep 14, 2004, 06:14 PM

Originally posted by waffffffle:
I am quite aware of this. This is why I was hoping that Connectix would combine their emulation technology with a virtual machine. There has long been talk of combining WINE with an x86 emulator, but so far nothing has come of this.

They are working on it.

Jablabla · Sep 14, 2004, 08:23 PM

Well, its interesting they would demo a video game. Most of the processing would be offloaded to the gpu anyways. I'd be interested to see what video card they were using.

If it was x86 linux then we would have different machine code running on their demo mac. Thus, I think it would be emulation. What I want to know is was is a power pc linux box that the original linux code was compiled on.

But if it was x86 linux code running on a x86 windows box then I think I could agree that they might get 90% of the performance.

If this is the case, then I'd see the windows platform getting hit pretty hard. The reality is no one really uses power pc based linux.

LordJohnWhorfin · Sep 14, 2004, 10:04 PM

The review focuses on points that are completely irrelevant to an emulator. Running programs written for another OS but for the same processor are not going to require the overhead of a CPU emulator, and as far as games are concerned most of the load is being handled by the GPU so running a graphics intensive game under emulation is nothing spectacular if you are capable of supporting the hardware (VirtualPC already does that).

Anyway, I think QuickTransit is a very unfortunate name for this product. It would be more appropriate for a high fiber cereal, or a brand of prunes...

SoClose · Sep 14, 2004, 10:08 PM

A long, long time ago, DEC had a very interesting code translator for the Alpha chip. IIRC, if you ran Windows NT on Alpha and ran an x86 Windows binary, the x86 -> Alpha translator would, over time, dynamically recompile the binary into Alpha code and save that code into the binary file image (and thus the file size would grow over time). So each time the user launched and ran an application, it had the potential to get faster as they'd most likely use the portion of the code that had been translated to Alpha. This clearly isn't what Transitive is purporting to do, but it's an interesting take on high-speed emulation and translation given the same API set (Windows) on different instruction sets (x86 vs. Alpha). Can someone fill in the large gaps in my memories on this technology?

Update: here's an article on the Alpha code translator and it mentions Transitive! From almost a year ago...interesting!

leperkuhn · Sep 15, 2004, 12:48 AM

Originally posted by cambro:
I agree with horeepoo, but more generally....

If this *could* be done it wouldn't it be the death of Mac OS X applications?

Why would developers work on a Mac OS X native application when they can develop it for Windows and have it "work" almost as well on Macs?

This is why I don't like VPC and hope VPC never really runs fast. It will break the back of Mac developers and leave mac users with little more than an OS with a bunch of non-native applications.

or it would allow people to switch to the mac os and leave their dell behind.

kcarnews · Sep 15, 2004, 04:17 AM

I really can't speak to the technological doubts previous posters have raised about Transitive's software. I CAN tell you, ,though, that I read a brief news article about Transitive's software--called "Dynamite" at the time--about a year ago. I found the idea fascinating, went to the company's Web site, noodled around and found there (IIRC) this link:

http://www.macworld.com/news/2001/10/15/transitive/

Yeah, the article is dated 10/15/2001. It's so old that when I did a keyword search on Maccentral, I couldn't find it. Fortunately, I'd already posted the link to a response to Derrick Story's Weblog on

http://www.oreillynet.com/pub/wlg/3712

back in August 2003. (I'm the anonymous coward posting at 2003-08-28 20:57:24).

If you read the old Maccentral article, the company planned to demo an Athlon 1.4 Ghz machine running "PowerPC" applications at speeds equivalent to or better than a 1Ghz G4 chip. As I noted in my post to Derrick's Weblog, that must have been an impressive demo, 'cause single G4 chips were topping at 867 Mhz at the time. (I checked http://www.apple-history.com on this historical factoid--if I'm off, someone please let me know).

Another article from the Softwar Development Times

http://www.sdtimes.com/news/085/story10.htm back in Sept. 2003 gives a little more layman-level explanation of the technology.

I'd also point out to you all that Transitive scored $10 million in financing in March 2003 from two VC firms:

http://sanjose.bizjournals.com/sanjo...3/daily26.html

Finally, Alasdair Rawsthorne, Transitive's Founder and CTO, is a lecturer at Univ. of Manchester in England. From what I remember, he's been working on this software since '94. I have trouble believing that a lecturer (the equiv. of a tenured professor, I believe) at a respected British University would be involved in a swindle of such length and proportion. He would lose his position and his reputation in academia if this whole thing were a very long-playing scam--and he'd probably also wind up in jail.

Let's assume for the mo' that Transitive is for real. According to this piece

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

"Company executives said the technology could be used to run the Apple Macintosh OS on top of an X86 processor. However, the technology will most likely not be sold to individual users, so a company like Hewlett-Packard Corp. would have to decide to license the technology and include it with their PCs."

What would this mean for Apple? Would it hurt its hardware sales, its bread and butter? Wouldn't people buy a cheap Dell box and run OS X? Would Apple support such a mix of machine and software?

OR--perhaps folks would buy an Apple computer for the looks and reliability, knowing that they could run just about every damned Windows app out there. This is definitely a two-edged sword for Apple. It could hurt or help them. I can't imagine Steve refusing to include this in OS X at some point, though. The pressure from Apple users to build it in to the OS and support it will be tremendous.

Graymalkin · Sep 15, 2004, 05:59 AM

No matter how starry-eyed you are over the prospect of this technology, it is not going to meet the company's claims. Syscall translation as I tried to make clear is nothing new. A simple kernel service in OSX would allow PowerPC Linux binaries to run without issue on OSX. What is not nearly as easy is sticking an emulator in the mix to translate programs between processor architectures, executable formats, and ABIs. Let's not assume this software is real until they offer real demos of their technology. Quake 3 running on OSX is no big deal, running AutoCAD on OSX would be a real demo.

Jablabla · Sep 15, 2004, 10:48 AM

Though, I have to admit it could provide one positive thing for consumers.

It is very easy to port a x86 linux app to ppc linux. Then through the Transitive software get to os X. Though, porting code from linux to os x is not that difficult. Just run configure and your done. But using their software people could just bypass a deveopement build and test. This would save people money because you would not need to hire the developers. I'd say that 99.9% of the apps of the linux world have been developed for ppc linux. From this perspective I could see added value from the transitive software.

It would also be interesting to see all those iApps running on a linux ppc clone. That would be a problem for Apple because they initially developed the iApps to give value only to their hardware. I find it ironic now that microsoft with its new service pack is requiring their apps to be signed for security. This might also disable apps from running on other platforms.

Who knows, maybe the programs get converted to an intermediate type of code that is common to the transitive software. Lets call it Transitive byte code. Then it would be compiled on the first run on a new box. Rather than invent a new language like Java just make a translator for each platoform and do it all transparently. Now, that would be cool. You'd take the emulation hit on the first run and then cache the results for subsequent runs. Sort of a tangent to what a previous post was talking about with regard to the dec alpha.

Of course, if this all gets out of control, Apple pulls the rabbit out of the hat. They announce os X for x86 and then steal 30% of the windows market share. Hmm...

goMac · Sep 15, 2004, 11:12 AM

Transitive only works if you are running the same OS on both machines. You cannot run Windows software under Mac OS X for it. The Linux Quake 3 demo they did on a Powerbook ran under PPC Linux, not OS X.

In short, no one is going to be building fast Windows emulation any time soon.

Jablabla · Sep 15, 2004, 11:19 AM

About the syscall.

Todays new machines are so fast that even if I get to 90% performance the end user is not going to notice. And if they can get 90% performance that is going to be waay faster than java. Plus, any intensive graphics gets offloaded to a viedo card or altivec.

If they cannot run cross architecture then they probably are using syscall and not emulation.

scienceguy_ae · 02:31 PM

Originally posted by Simon:
OK, is this feasible and could this be a Tiger killer-feature?

This wired article claims software emulation speeds of up to 80% should be reached with this piece of software. They say it's the same code that was used to allow XBox development to be done on PowerMac G5s.

So, just guessing, could Apple license this software, build it into Tiger and basically do away with the 'this software doesn't exist on the Mac platform' argument?

I'm not sure either way whether this could be true or not, although it does seem a little to good to be...

By the way, they are claiming cross-architechture compatibility:

"Allows application binaries compiled for a MIPS�, x86 or mainframe processor to run on a POWER�/PowerPC� based computer." - http://www.transitive.com/products.htm

And this, [URL]http://www.linuxdevices.com/news/NS6404972696.html[URL], referring to a demonstration of dynamic, on-the-fly, CISC (x86 code) to RISC (MIPS architechture) translation. I gues they're saying that the same thing can be done with x86 to PowerPC.

That's the whole point, the reason this is raising so much debate.

If it does turn out to be a hoax (which would be difficult to produce considering so many demos have been given of the technology), what happens to the supporters like Pond Ventures and the University of Manchester? Guess they wouldn't be too happy...

Apparently, the University of Manchester has put alot of effort into making this work, including 8 grad students that are now working at Transitive.

http://www.mantechfund.com/dec2001news1.htm
http://www.pondventures.com/transitive.html

My 2 cents. Only time will tell I guess.

--
scienceguy_ae

msuper69 · Sep 15, 2004, 01:31 PM

Originally posted by kcarnews:
...

Let's assume for the mo' that Transitive is for real. According to this piece

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

"Company executives said the technology could be used to run the Apple Macintosh OS on top of an X86 processor. However, the technology will most likely not be sold to individual users, so a company like Hewlett-Packard Corp. would have to decide to license the technology and include it with their PCs."

What would this mean for Apple? Would it hurt its hardware sales, its bread and butter? Wouldn't people buy a cheap Dell box and run OS X? Would Apple support such a mix of machine and software?

..

Apple has included a little ditty in the EULA for OS X that prohibits running the OS on any hardware other than an Apple-branded Mac. This business of running OS X on X86 hardware will never go anywhere as long as Steve (and the Apple legal dept.) has anything to say about it.

Jablabla · Sep 15, 2004, 01:52 PM

You don't actually think a EULA would stop all the cheap x86 people from running OS X on their machines if they could?

msuper69 · Sep 15, 2004, 01:58 PM

Originally posted by Jablabla:
You don't actually think a EULA would stop all the cheap x86 people from running OS X on their machines if they could?

No, but the company producing the emulation software would stop after they get their butts handed to them in court.

Michel Fortin · Sep 15, 2004, 07:58 PM

Originally posted by Jablabla:
You don't actually think a EULA would stop all the cheap x86 people from running OS X on their machines if they could?

Well, they can.

macaddict0001 · Sep 16, 2004, 02:27 AM

It will no doubt be easier when intel/amd/whoever make's x86 arch's switche's to ppc because of the few major constraints with x86 mainly CISC as opposed to RISC. Of course there would therefore be ppc windows.

goMac · Sep 16, 2004, 02:31 AM

Originally posted by Michel Fortin:
Well, they can.

And legally Apple is within their rights to sue them at any time.

Just because people do it doesn't mean they can't get sued.