On Apple Announcing the ARM Mac Transition at WWDC This Month
daringfireball.net
On Apple Announcing the ARM Mac Transition at WWDC This Month
1–10 of 22 posts
Re: On Apple Announcing the ARM Mac Transition at WWDC This Month
#2I'm excited about this. Can't wait to see what performance looks like. As for the effect on sales, it's probably a wash. For people who know and care about these things, I'd imagine a 50/50 split between people racing to get the last rev of Intel macs to wait out the transition until things stabilize and the other half waiting for the new stuff. macOS app devs won't really have a choice, but iOS devs will.
Re: On Apple Announcing the ARM Mac Transition at WWDC This Month
#3Why not just require apps to be distributed as LLVM bitcode, as is done for the Apple watch? That would give Apple the flexibility to change architectures--or even to create its own instruction set, now that it's making its own chips.
Re: On Apple Announcing the ARM Mac Transition at WWDC This Month
#4I really believe that the lack of x86 is going to be a big buying problem for a lot of people who need to run software on Windows. Without Boot Camp or x86 virtual machines its going to be a problem to justify the machine with a large group of people.
Re: On Apple Announcing the ARM Mac Transition at WWDC This Month
#5Why not just require apps to be distributed as LLVM bitcode, as is done for the Apple watch? That would give Apple the flexibility to change architectures--or even to create its own instruction set, now that it's making its own chips.
LLVM bitcode is not that portable in reality. Just like Plan9 assembly intermediary bitcode files are pretty much single architecture, due to containing lots and lots of low-level assumptions about the target architecture, not to mention possible incompatibilities in extended intrinsics (for example for vector units).
Re: On Apple Announcing the ARM Mac Transition at WWDC This Month
#6Smart move...
How would you otherwise convince existing customers to throw away working systems and buy new ones?
I don't think that we should expect new ARM64 to be 10 times better than x64...
The only idea I can think of is if chip will have bult-in hardware-accelerated OpenVG or any other 2D API to replace severely outdated CoreGraphics ...
Re: On Apple Announcing the ARM Mac Transition at WWDC This Month
#7Can they source enough chips to do this in the near term? Meanwhile, another major fabless designer (AMD) is ramping up too.
Re: On Apple Announcing the ARM Mac Transition at WWDC This Month
#8Can they source enough chips to do this in the near term? Meanwhile, another major fabless designer (AMD) is ramping up too.
Apple ships close to 200 million iPhones a year all with their custom ARM CPUs, fabricated by TSMC. They ship about a tenth as many Macs. I'm pretty sure they can handle that growth, especially as one of TSMC's biggest customers.
Re: On Apple Announcing the ARM Mac Transition at WWDC This Month
#9Why not just require apps to be distributed as LLVM bitcode, as is done for the Apple watch? That would give Apple the flexibility to change architectures--or even to create its own instruction set, now that it's making its own chips.
LLVM bitcode is not that portable in reality. Just like Plan9 assembly intermediary bitcode files are pretty much single architecture, due to containing lots and lots of low-level assumptions about the target architecture, not to mention possible incompatibilities in extended intrinsics (for example for vector units).
Most of the portability problems for retargetting to another little endian 64-bit architecture aren't the easy ones of bit widths and integer types, but rather more subtle things like a more relaxed memory model and different performance characteristics.
It's not worth the overhead of attempting to ship bitcode. Apple has shipped products on 6502, 68000, PPC, x86, x86-64, ARM, and ARM64. The transitions have been annoying for the lost software, but that's not an unusual occurrence for an OSX upgrade either.
Re: On Apple Announcing the ARM Mac Transition at WWDC This Month
#10Earlier quoted context omitted.
LLVM bitcode is not that portable in reality. Just like Plan9 assembly intermediary bitcode files are pretty much single architecture, due to containing lots and lots of low-level assumptions about the target architecture, not to mention possible incompatibilities in extended intrinsics (for example for vector units).
Most of the portability problems for retargetting to another little endian 64-bit architecture aren't the easy ones of bit widths and integer types, but rather more subtle things like a more relaxed memory model and different performance characteristics. It's not worth the overhead of attempting to ship bitcode. Apple has shipped products on 6502, 68000, PPC, x86, x86-64, ARM, and ARM64. The transitions have been ann…
What was the Newton using ?