3 ms·

Since Apple now controls the entire hardware stack for all their products (except the mac pro), they could conceivably design their own assembly language ('Appl

Since Apple now controls the entire hardware stack for all their products (except the mac pro), they could conceivably design their own assembly language ('Apple Assembly'?) optimized for, and only for, Apple silicon and bypass the problem.

lapcat 3y ago
Apple has already done this to some extent, though I don't have the references offhand. In any case, there's only so much that hardware can to do compensate for crappy software. It's not inherently a hardware problem.
- MichaelZuo 3y ago
  It's possible, if they were motivated to really invest in it, to introduce some innovations. For example, it could warn about badly designed software, major flaws, etc... at the assembler level. And since they mostly control software distribution, to then block anything that receives such warnings.
  - saagarjha 3y ago
    There’s no need to do something like that. They just don’t optimize for certain code patterns.
AceJohnny2 3y ago
Bypass the problem?? They'd only create more. There's millions of man-hours invested in optimizing software and hardware for the ARM64 ISA. Abandoning that for a nebulous "we can optimize for our stack" is not a worthwhile tradeoff.
- vaxman 3y ago
  Nebulous? When TSMC can no longer economically cram more transistors onto the dye, Apple will need other ways of increasing performance to keep the upgrade cash coming in. Apple previously co-developed PowerPC with Motorola (and IBM) for this very reason. At this time, Apple has all but kicked its third-party developers out of kernel space on its operating systems which will make for easy transition to future architectures. I would go so far as to speculate Apple is already designing proprietary RISC-V extensions and even their own FABs to produce them. One would expect such extensions to help Apple's systems execute Swift and ObjC better (faster and more securely) than general purpose processors and to optimize shared memory and fabric access between CPU, NPU and GPU cores.
saagarjha 3y ago
Yeah, that’s called arm64, except for the bit that other people can also use it.
- MichaelZuo 3y ago
  arm64 + special proprietary bits = 'Apple Assembly', or whatever marketing name, seems within the realm of possibility. It wouldn't be that much of an outlier in terms of naming convention.
  - alehlopeh 3y ago
    Not a single developer facing product is called Apple anything.
    - DerekL 3y ago
      Apple File System Apple events
- tinus_hn 3y ago
  I’m pretty sure there’s some Apple specific instructions that help with Rosetta 2.
  - saagarjha 3y ago
    Not for Rosetta 2.
    - tinus_hn 3y ago
      Here’s more information: https://dougallj.wordpress.com/2022/11/09/why-is-rosetta-2-fast/ https://dougallj.wordpress.com/2022/11/09/why-is-rosetta-2-f...
      - saagarjha 3y ago
        These are all standard ARM instructions. Apple tweaks the semantics of some of them in Rosetta mode to better support what they’re doing but it’s otherwise not involving any new encodings.

lapcat 3y ago

Apple has already done this to some extent, though I don't have the references offhand. In any case, there's only so much that hardware can to do compensate for crappy software. It's not inherently a hardware problem.

MichaelZuo 3y ago
It's possible, if they were motivated to really invest in it, to introduce some innovations. For example, it could warn about badly designed software, major flaws, etc... at the assembler level. And since they mostly control software distribution, to then block anything that receives such warnings.
- saagarjha 3y ago
  There’s no need to do something like that. They just don’t optimize for certain code patterns.

AceJohnny2 3y ago

Bypass the problem?? They'd only create more. There's millions of man-hours invested in optimizing software and hardware for the ARM64 ISA. Abandoning that for a nebulous "we can optimize for our stack" is not a worthwhile tradeoff.

vaxman 3y ago
Nebulous? When TSMC can no longer economically cram more transistors onto the dye, Apple will need other ways of increasing performance to keep the upgrade cash coming in. Apple previously co-developed PowerPC with Motorola (and IBM) for this very reason. At this time, Apple has all but kicked its third-party developers out of kernel space on its operating systems which will make for easy transition to future architectures. I would go so far as to speculate Apple is already designing proprietary RISC-V extensions and even their own FABs to produce them. One would expect such extensions to help Apple's systems execute Swift and ObjC better (faster and more securely) than general purpose processors and to optimize shared memory and fabric access between CPU, NPU and GPU cores.

saagarjha 3y ago

Yeah, that’s called arm64, except for the bit that other people can also use it.

MichaelZuo 3y ago
arm64 + special proprietary bits = 'Apple Assembly', or whatever marketing name, seems within the realm of possibility. It wouldn't be that much of an outlier in terms of naming convention.
- alehlopeh 3y ago
  Not a single developer facing product is called Apple anything.
  - DerekL 3y ago
    Apple File System Apple events
tinus_hn 3y ago
I’m pretty sure there’s some Apple specific instructions that help with Rosetta 2.
- saagarjha 3y ago
  Not for Rosetta 2.
  - tinus_hn 3y ago
    Here’s more information: https://dougallj.wordpress.com/2022/11/09/why-is-rosetta-2-fast/ https://dougallj.wordpress.com/2022/11/09/why-is-rosetta-2-f...
    - saagarjha 3y ago
      These are all standard ARM instructions. Apple tweaks the semantics of some of them in Rosetta mode to better support what they’re doing but it’s otherwise not involving any new encodings.