关于 Apple M1 的群众意见每日汇总 [码农方向][2020.11.19]

this thing is honestly fast Rosetta 2 is kind of amazing
many CLI tools only run under R2 for now, and that's fine
native makes a difference for heavy apps
code signing is a bit annoying but makes sense, and hopefully tooling will catch up
https://twitter.com/FiloSottile/status/1329128976929988615

《 Developing on Apple Silicon 》 https://docs.google.com/document/d/1iWUstb66v66tTVxQWNMZ1BehgNzEmykzqDCUp5l8ip8/edit

For developing in Go on Apple Silicon, there's two main issues to be aware of:

arm64 binaries are not codesigned automatically ( https://github.com/golang/go/issues/42684)
amd64 binaries need GODEBUG=asyncpreemptoff=1 ( https://github.com/golang/go/issues/42700) https://twitter.com/FiloSottile/status/1329131111360983041

IDA 在 M1 Mac 上运行良好，来自 Checkra1n 团队成员 @qwertyoruiopz

M1 is actually pretty good, can’t believe the base air can handle ida like this https://twitter.com/qwertyoruiopz/status/1329152951294844932

so far so good. there’s some issues overall (eg brew currently doesn’t work well on it), but these will be sorted out and Rosetta 2 is just magic.

i just copied over my install from mac pro, so this is under rosetta 2. i’ll check later if there’s anything from hexrays in that regard

Are 8gb enough even for ida?
seems to handle it fine, but i am using it just to look into small machos

Accelerating TensorFlow Performance on Mac

https://blog.tensorflow.org/2020/11/accelerating-tensorflow-performance-on-mac.html

Google releases Apple Silicon-optimized version of Chrome for Mac

https://9to5google.com/2020/11/18/chrome-mac-apple-silicon/

If you want to use Chrome on new Macs with Apple M1 chips, head to https://google.com/chrome/ and download the Apple Silicon version we released in M87 today! https://twitter.com/elvin_not_11/status/1328816326542905345

Does it ARM? A list of reported app support for Apple Silicon

https://doesitarm.com

https://github.com/ThatGuySam/doesitarm

M1 Mac 合盖后依然保持运行，来自 Yachen Liu

写了非常简单的一小段代码验证了一下早上的观点，确认即使在电池供电且合盖的状态下，M1 设备 CPU 依然会定期的保持活跃，和 iDevice 一样，而且这个活跃不是系统特权。

第二张图是旧的 MBA，合盖后几分钟内依然活跃，然后就彻底休眠了。（代码中 counter 忘 ++ 了，忽略…） https://twitter.com/Blankwonder/status/1328952885841960960

关于 Rosetta 2，来自 Hector Martin

Also, remember that Apple cheated with their control over the CPU for Rosetta 2. Getting R2 x86 performance on any other ARM is impossible, due to the memory model mismatch. You have to massively slow down all loads and stores.

So Apple straight up implemented the x86 consistency model on their cores. That's the kind of high-impact detail that makes or breaks emulation performance for a different arch. Did they do this for any other x86-isms? Nobody knows so far.

https://twitter.com/marcan42/status/1328940797174099970

Also, everyone please stop eating up Apple's marketing bullshit and calling Rosetta 2 an "Ahead-of-Time Translator, Not An Emulator" as if it is some Fundamental Difference that Makes It Fast.

It's a JIT emulator, with a cache, with opportunistic partial ahead of time warming.

https://twitter.com/marcan42/status/1328956667548082177

Rosetta 2 runs everything under the correct memory model

“Excited” to hear about all the data races now that people will actually execute all their code under ARMs memory model

woah wait huh i never considered whether rosetta 2 does the necessary synchronization of x86

https://twitter.com/gsnedders/status/1328710885028884481

34 条回复 • 2020-11-20 09:15:39 +08:00

zckevin

2020-11-19 09:29:10 +08:00

## ARM x86 关于 Memory model 的区别，来自 Bruce Dawson (Google Chrome Team)

mov register-> a
mov register-> b

On x86/x64 the store to "b" is guaranteed to not be visible to other processors before the store to "a". On ARM that is not guaranteed.

Many benign x86/x64 data races will becomes real bugs on ARM, and it appears many programmers don't know this

The emulator _must_ emulate the x86/x64 memory model. It is as important as properly emulating the CPU status flags and arithmetic instructions. So, yes. If Rosetta didn't emulate the memory model it would not be an x86/x64 emulator.

warminer

2020-11-19 09:31:40 +08:00 via iPhone

让人兴奋的硬件，大家都在玩，玩着玩着兼容性就好起来了

zckevin

2020-11-19 09:50:42 +08:00

## Native FP16 on Apple M1, 来自 Steve Canon (Apple Swift team)

A couple random fun numerics things to play with on M1 macs:

- native half-precision arithmetic
- ARMv8.3 complex arithmetic SIMD instructions

https://twitter.com/stephentyrone/status/1328861121382666240