Apple intern's thesis leaks secret project to port Mac OS X to ARM processors | Hacker News Reader

Apple intern's thesis leaks secret project to port Mac OS X to ARM processors

appleinsider.com

47 pointsUnited85716 comments

Apple has had Darwin on ARM processors for quite some time. It's called iOS. The cited project was porting Darwin to a certain ARMv5 chip. I don't know a lot about the ARM architecture, but from what I understand the original iPhone was already run by an ARMv6 and ARMv5 is pretty obsolete. So this can hardly be interpreted as part of a serious effort to port OS X to ARM.

it was a false flag/fake project..to see if the engineer was trust-worthy..do think the thesis would have been published outside if Apple was serious about the project? Its secretive Apple after all..

It's an open secret Apple keeps its options open. I just 'knew' they would have such a project after I read they bought P.A. Semi years ago.

http://news.ycombinator.com/item?id=2258693

Soon. :)

> Soon. :)

I'd still expect you to be wrong, iOS apps are unworkable on OSX and ARM processors, being significantly less powerful (though more efficient) than x86, would have a hard time translating x86 instructions to ARMv7 (à la Rosetta) with any kind of acceptable performances. ARMv7 also has very little support for 64b, which Lion mandates (dropped support for x86 and requires x86-64 instead) and ARMv8 is nowhere near ready for production chains.

There would be very little to be gained from putting an ARM chip in an Air, and pretty much everything to be lost. Asking for that feels like pointless wankery.

I'd expect that any such machine would be released as an extended iPad line, with OS X running a compat layer for iOS apps. Apps could theoretically ship as fat binaries (/again/) to work on both, but that much seems like quite a jump --- people can buy laptops now, with great battery life. It's the simplicity of iOS that drives its appeal.

In fact, the iOS Simulator used by developers today is pretty much just that, with the iOS app itself + the Cocoa Touch frameworks running within a OS X process. The only difference is that simulator app builds are specially compiled for X86 instead of ARM (thus, it's not a "emulator" in the strictest sense of the word).

With ARM, it'd be conceivably possible for the same iOS binary to run on both the device and the desktop simulator.

Emulation would indeed kill the advantage of ARM, but I can see Apple slashing the ecosystem (again) and allowing only vetted apps that go through their OSX app store and get compiled natively.

> I can see Apple slashing the ecosystem (again)

As far as I know, they have never done that without a pretty long transition period so far:

* The 68k -> PPC transition had a low-level 68k emulator in PPC for a pretty long time

* The OS9 -> OSX transition had Classic for emulation (until Tiger) and Carbon for recompilation (which still worked in Snow Leopard, not sure about Lion)

* The PPC -> Intel transition provided Rosetta from Tiger to Snow Leopard

* The 32b -> 64b transition took 4 years from 64b Cocoa introduction in Leopard (Tiger already had support for 64b headless processes, but that does not really count), and I believe Lion still has no problem running x86/32 binaries (even though 32b CPUs can't run Lion)

iOS is a pretty big chunk of OS X running on ARM already I understand.

I think this is very unsurprising.

Given the switch from PowerPC to Intel its unsurprising that Apple is keeping itself options open. I suspect that if Intel doesn't deliver the power consumption/performance ratio that Apple needs they would happily switch to a processor that does.

Given that at one point NeXTStep supported 4 different processor architectures (Motorola, Intel, Sparc, HP) the underlying system is probably one of the "easier" operating systems to port to new processor architectures.

This is a lot less of a "secret project leak" than someone discussing the ideal/advanced fast track to really learning how to kick ass in his new job.

Q:"This new kid is bright; what should we give him to cut his teeth on the weird stuff?"

A:"Break it at a low level and let him fix it."

I wonder if Apple is considering switching the AirPort/Time Capsule family to OS X; they're based on Marvell processors AFAIK.

I wasn't aware this was a secret. I thought Apple were planning an ARM build of OS X for a while, or had at least been discussing it.

And what with iOS evolving as it is, they probably think a full port of OS X to ARM is kind of pointless now.

This is the beginning of the end for Intel. It would've been anyway, but Apple has such a strong influence in the tech industry (considering everyone wants to copy them), that now every manufacturer will start gradually preferring the ARM architecture over x86 for "ultrathin" laptops with very long battery life and "good enough" performance, that are also less expensive than the Intel ones.

It's going to be a long time before desktop OS's running on ARM are as accepted as they are on x86, simply because of software compatibility. Apple's probably in better shape on that front (having already moved architectures once, and so having more of a system to include a compatibility layer), but users are going to be questioning why 95% of their software doesn't run, or runs very slowly, on their new and expensive laptop.

The middle ground is something like a Chromebook, which gets around the problem by not running any legacy apps. However, despite what Google thinks there's a lot of utility in a laptop outside of the browser, and people who have software that they need to run will not be moving over to ARM for a long time.

If people were willing to buy netbooks, they will be willing to buy something like this, too, which should even work better than Windows 7 on a single core Atom.

I agree that Chromebook should've done this a long time ago. But I wonder if they think Cortex A9 was subpar, and they're just waiting for Cortex A15 to do it. A $200 Tegra 3 Chromebook would've still been more appealing than a dual core Atom $400 one.

Reply on news.ycombinator.com