Does not support Apple Silicon

[fxcoudert]

Compilation on ARM macOS (aka Apple Silicon) leads to a build failure:

/tmp/io-20221111-68531-j5uskf/io-2017.09.06/libs/coroutine/source/asm.S:143:2: error: unrecognized instruction mnemonic
 mflr r0
 ^
/tmp/io-20221111-68531-j5uskf/io-2017.09.06/libs/coroutine/source/asm.S:144:2: error: unrecognized instruction mnemonic
 mfcr r5
 ^
/tmp/io-20221111-68531-j5uskf/io-2017.09.06/libs/coroutine/source/asm.S:145:2: error: unrecognized instruction mnemonic
 mfctr r6
 ^
/tmp/io-20221111-68531-j5uskf/io-2017.09.06/libs/coroutine/source/asm.S:146:2: error: unrecognized instruction mnemonic
 mfxer r7
 ^
/tmp/io-20221111-68531-j5uskf/io-2017.09.06/libs/coroutine/source/asm.S:147:13: error: unexpected token in argument list
 stw r0, 0*4(r3)
            ^

[shayneoneill]

Any chance of getting this fixed? Intel macs are getting closer and closer to EOLed for support, and with that usually comes the end of Rosetta.

[PedroBatista]

I think Rosetta will be around for many years, the problem is ( seems to be ) the assembly code in asm.S. It also blocked me from running this with WASM.

Since assembly programming is a long lost art, my guess would be to tell ChatGPT to translate it ( unlikely ) or to implement #define NEEDARMCONTEXT 1
#define SET setmcontext
#define GET getmcontext

in C.

---

@fxcoudert @shayneoneill Just noticed in the original post the build is entering in the #ifdef NEEDPOWERCONTEXT instead of the #ifdef NEEDARMCONTEXT. If you can make it enter the ARM part perhaps there is no problem with the code itself since it's 32bit ARM, I don't know for sure but ARM64 should support 32bit registers?

[stevedekorte]

I thought Apple finally got ucontext working. If so, we can just set the ifdef for that.

[stevedekorte]

I think Rosetta will be around for many years, the problem is ( seems to be ) the assembly code in asm.S. It also blocked me from running this with WASM.

There is an IoWASM project: https://github.com/stevedekorte/ioWasm

And an online REPL demo that uses it: https://iolanguage.org/repl/index.html

Would be great to see some bindings into JS and the DOM.

[shayneoneill]

@PedroBatista Perhaps, but the original Rosetta (for running PowerPC apps on Intel macs) really only was around for about 4-5 years, and we're at year 3 with Rosetta2. So maybe a couple of more years? Its a bit of breathing room, but not much. Theres still a fair few legacy Intel macs out there, and Apple for all its faults, is likely not in a hurry to go kicking its legacy customers out of its software distribution pipelines. But apple does have form for this sort of deprecation too.

That said, Apple will likely give some deprecation notice a version or two of macos before yeeting it.

[stevedekorte]

I'd happy offer a bounty (say $200?) for anyone to do this. I'd do it myself but I'm busy with another project and haven't touched XCode in ages.

[hellohejinyu]

I have forgotten how I managed to run io on my Apple Chip computer, and I am now looking for how to completely remove it since I no longer need it.🤣

[PedroBatista]

Just an heads-up, I've managed to make changes to the code and compile it natively to Apple Silicon. But I've been very busy, hopefully I'll do a PR soon.

[stevedekorte]

Great! Did you get coroutines working?

[PedroBatista]

@stevedekorte Yes, I remember I run the existing tests ( iovm/tests/correctness/CoroTest.io - I think? It was a couple months ago ) without any errors and the actual performance looked very good. I also run the whole list of tests and the errors it reported were exactly the same as the x86 version. It was something unrelated to coroutines, some UTF8? I might be wrong.

I'll try to revisit this later this week or in the weekend and make a PR.

Thank for you work and language btw, it has been one of my "guilty pleasures" since the late 00's 😄