Apple M1 Chip

[kcrisman]

Hi sage-devel,

The only references I could find to this so far are

https://ask.sagemath.org/question/54220/apple-silicon-m1-chip/

and 

https://groups.google.com/g/sage-devel/c/5yY3VmkT4kE/m/P21QdaMUBgAJ

where I asked more generally about ARM.

I am likely to be assigned such a laptop in the near future, but it would be nice not to be a total guinea pig (since I rely on having Sage for a number of different purposes in my teaching).  Though I am willing to be a partial guinea pig if I get the computer early enough in our semester break!

Anyway, in the intervening weeks since those posts, has anyone happened to hear more about this (either for Sage or for dependencies like BLAS/Numpy/MPIR)?  This will be the default for a lot of people pretty soon.

Thanks, 

kcrisman

[Zachary Scherr]

You may want to take a look at with https://github.com/Homebrew/brew/issues/7857.  I think the biggest thing at the moment is that gfortran won't be available for the M1 for some time.

[Isuru Fernando]

See https://conda-forge.org/blog/posts/2020-10-29-macos-arm64/

I added sage to the build of macos-arm64 packages and the progress can be seen at https://raw.githubusercontent.com/regro/cf-graph-countyfair/master/status/armosxaddition.svg?sanitize=true

Isuru

--
You received this message because you are subscribed to the Google Groups "sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/503ba9b5-637e-43cc-a448-35e196a9b643n%40googlegroups.com.

[Dima Pasechnik]

On Tue, Nov 24, 2020 at 3:51 PM Zachary Scherr <zsc...@gmail.com> wrote:
>
> You may want to take a look at with https://github.com/Homebrew/brew/issues/7857. I think the biggest thing at the moment is that gfortran won't be available for the M1 for some time.

it seems to me that one can still run the full Homebrew, including
gfortran, in their x86_64 emulator Rosetta.
https://soffes.blog/homebrew-on-apple-silicon (YMMV)

Anyhow, there is
https://www.nag.com/news/first-fortran-compiler-apple-silicon-macs
(and we have insiders in NAG, perhaps we can get these for free...)

>
> On Tuesday, November 24, 2020 at 10:47:42 AM UTC-5 kcrisman wrote:
>>
>> Hi sage-devel,
>>
>> The only references I could find to this so far are
>> https://ask.sagemath.org/question/54220/apple-silicon-m1-chip/
>> and
>> https://groups.google.com/g/sage-devel/c/5yY3VmkT4kE/m/P21QdaMUBgAJ
>> where I asked more generally about ARM.
>>
>> I am likely to be assigned such a laptop in the near future, but it would be nice not to be a total guinea pig (since I rely on having Sage for a number of different purposes in my teaching). Though I am willing to be a partial guinea pig if I get the computer early enough in our semester break!
>>
>> Anyway, in the intervening weeks since those posts, has anyone happened to hear more about this (either for Sage or for dependencies like BLAS/Numpy/MPIR)? This will be the default for a lot of people pretty soon.
>>
>> Thanks,
>> kcrisman
>

[François Bissey]

That would be a new fortran compiler. I did builds in the past with
the intel fortran compiler and the PG compiler but not NAG.
You’ll be in totally uncharted territory.

Fraçois

[Isuru Fernando]

At conda-forge, we are using a gfortran fork (https://github.com/iains/gcc-darwin-arm64) and it works great so far.

scipy, numpy test suites pass.

Isuru

To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/93A6A60B-D9BD-4869-AFB9-B4332245B196%40gmail.com.

[Dima Pasechnik]

On Tue, Nov 24, 2020 at 6:49 PM François Bissey <frp.b...@gmail.com> wrote:
>
> That would be a new fortran compiler. I did builds in the past with
> the intel fortran compiler and the PG compiler but not NAG.
> You’ll be in totally uncharted territory.

IMHO it's known to work with numpy, scipy, and R.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/93A6A60B-D9BD-4869-AFB9-B4332245B196%40gmail.com.

[kcrisman]

IMHO it's known to work with numpy, scipy, and R.

That's pretty much all we need fortran for, right?   Good news.   Still, the toolchain will be more annoying than on current Mac then. 

[François Bissey]

Also suitesparse.
Of course one of thing is actually blas/lapack (openblas)- but
with any luck we can use the accelerate framework from OS X.
I haven’t heard it is not available on macs with M1 chips so I am
assuming it is present.

> On 25/11/2020, at 3:46 PM, kcrisman <kcri...@gmail.com> wrote:
>
>
>
> IMHO it's known to work with numpy, scipy, and R.
>
>
> That's pretty much all we need fortran for, right? Good news. Still, the toolchain will be more annoying than on current Mac then.
>

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/fbdb111c-79a0-4319-9acb-b65e109a55d8n%40googlegroups.com.

[kcrisman]

Of course one of thing is actually blas/lapack (openblas)- but
with any luck we can use the accelerate framework from OS X.
I haven’t heard it is not available on macs with M1 chips so I am
assuming it is present.

I believe so, but it behaved differently sometimes on PPC (e.g. location) if I recall correctly from so long ago so perhaps at least some tweaks will be necessary.

See also https://developer.apple.com/forums/thread/649933