Successful PearPC/Mac OS X Install Documented

rocketjam writes "OS News has an article by a user who successfully installed Mac OS X using the 0.1 version of PearPC, the PPC emulator for x86 machines. He said it took 5 hours to run the first install CD but he did get it up and running on an AMD Athlon XP 1600+ with 512MB of RAM. The article has several screenshots of the Mac OS X install and new user set up running on his machine." See our previous story.

Comments

OS X Panther Here

[TravisWatkins]

I've gotten OS X Panther to install as well, you can see it here. Took about 7 hours on a Duron 1.6Ghz with 512MB SDRAM.

Re:I can see myself using this

[zgornz]

read the article

"Of course everything was not running very snappy; on their website they warn you: the emulated processor is about 40 times slower than the host processor. Still, I was amazed at what I saw: it worked!"

At 40 times slower than the host, you'd need one hell of a CPU to use this for as your primary environment.

Get a nice usb keyboard/mouse set, and a mac.

Re:I can see myself using this

[Twirlip of the Mists]

OS X treats the right button of a two-button mouse as a control-click, which seems logical enough..

Elaboration follows:

On a Mac, control-click sends the target a mouse-button-2 event. If you plug in a two-button mouse, the Mac automatically understands the second button as mouse-button-2. It's not that the Mac is remapping the second mouse click to some other kind of event; just the opposite.

Furthermore, a third mouse button works as well. Clicking the third button sends a mouse-button-3 event. Same with scroll wheels, and so on and so on.

Basically you can plug in just about any USB input device and it'll Just Work.

Legality

[BumpyCarrot]

It should be noted that this actually goes against the OSX EULA, which specifically states that the software cannot be used on anything other than Apple branded hardware, unfortunately :(

Re:I always wanted OSX on PC

[MBCook]

No. The number tossed around is at least 40x slower, and there are many reasons.

First is the obvious that if you can never emulate something the same speed that it would be if it was native. It will always be at least a hair slower.

In actuality, this is MUCH slower. There are a few reasons:

1. Registers - A PPC chip has something like DOUBLE the number of registers (on CPU memory that's used to hold variables while being worked on) as an Opteron chip. And Opteron has many more registers than a standard x86 chip. To make matters worse, while with PPCs and Opterons most registers are general purpose (can be used for anything), many operations in the x86 world require you to use a specific register, so they are less flexable. All this means lots of register swapping and other such trickery to make things work, and it costs speed. A version compiled/written for an Opteron should be faster, but it is still not the same. All these registers is one of the reasons why it's so easy to emulate a x86 on PPC but not vice-versa.
2. The second big reason is Alitvec. This is basically MMX/3DNow!/SSE, but I've heard it described as those things on steroids. It allows things to be done VERY fast that would take much longer without them. Matrix transforms, running the same instruction on a large table of data, etc. PearPC doesn't emulate Altivec right now. While OS X will run without it (G3s don't have it, IIRC), things would run much faster if Altivec operations could be mapped to SSE/MMX/etc. whenever possible. They are working on this.
3. Graphics - The graphics engine is all software (I think). If the graphics calls could be "pushed through" to the graphics card so that OS X's use of OpenGL in Quartz (to draw windows and do effects on them) could be done in hardware (instead of in software like on Macs that don't have good enough graphics cards) that would speed things up too.

Those are the main reasons. I think we'd all KILL for OS X on PCs, but I think we all know that realistically it's never going to happen.

Still, remember the software is only v0.1 so when they add things like Altivec and just do general optimisations, things should get faster.

Re:Emulator Scmemulator

[MBCook]

Macs have always been quite hard to emulate, at least that's my understanding. Thanks to Altivec, register starvation, and other things (see another post of mine in this topic) it's not easy.

It's mostly a problem of emulating the PPC chips themselves. There are emulators for the 68k based Macs (basillisk and executor to name two), and PPC based ones can be emulated too recently (SheepShaver has gotten this ability recently, I understand). Once you've got the chip emulated, the rest isn't that bad.

This is why there have always been "Mac on Mac" emulators (like Mac on Linux, or SheepShaver to run MacOS on PPC based BeOS and Linux machines). They don't have to deal with the whole processor issue, they just have to provide the right environment for the software.

So the ability to run OS X on Intel hardware is quite novel and interesting.

As for running Darwin, you can. Darwin is open source. The problem is that you can't run OS X on top of the x86 version because you can't get the source code to that. So you'd either have to rewrite ALL of the OS X libraries and then use emulation to run real Mac programs, or you'd have to use emulation to run the OS X libraries AND the software. Neither is easily done. Since they both require the CPU emulator, why not skip the middle man?

As for the "Virtual PC works well", see that post of mine I referenced above. It's MUCH easier to fake a x86 on a PPC than vice versa.

No one is claiming this is anywhere near usefull yet, but you never know what will come out if something like this.

Re:Bye Bye Mac Hardware

[wvitXpert]

I don't see where Apple hardware is really that much more epensive, especially when you consider the higher quality and better design of Apple's computers. Add the OS and iLife and I don't think there's any comparison. Maybe you haven't looked at Apple's hardware lately, thats the only reason I can see for your statement.
This is a comparison after a quick search on Dell.com and Apple.com...

eMac - $799 Dell Dimension 4600 - $746
1.25 GHz G4* 2.8GHz P4*
256MB RAM 256MB RAM
40GB HD 40GB HD
Combo Drive DVD-ROM Drive

12" PowerBook - $1599 Dell Inspiron 600m - $1368
1.33GHz G4 1.4 GHz Pentium M
256MB RAM 256MB RAM
60GB HD 40GB HD
64MB Graphics 32MB Graphics
Combo Drive Combo Drive

*note - regarding the eMac vs. the 4600 processor. I am writing this on a 2.66MHz Sony Vaio that seems for most things no faster than my 1GHz G4 PowerBook, so I don't think that comparing the two processors is too far off.

Re:I can see myself using this

[TravisWatkins]

Ask and ye shall see pain: OS X on WinXP on OS X

Re:Awesome...

[Anonymous Coward]

It's really not terribly bad on PearPC though. This has all been documented at Emaculation for about 3 days now.

The Jitc version of PearPC runs approximately 1/10-1/15 slower than a real mac. I successfully installed 10.2 on an Athlon64 3200+ and I can honestly say it's only a little slower than when I hacked 10.2 to run on a Powermac with a 603e procesor. The installation took about an hour and a half for a base install, and with the refresh set to around 40, it's quite usable. Were there a network bridge avaliable for Windows, I wouldn't mind doing basic functions on it.

Even the animation is bearable- again- only slightly slower than that 603e mac, which didn't have hardware acceleration either.

Also remember this is only the first release, 0.1. It's bound to increase in speed with subsequent releases. Just the fact that it works now is incredible in itself, given the architectural differences from x86 to ppc.