Add osx-arm64 support? (Julia 1.8.0 now supports M1)

[jmuchovej]

Solution to issue cannot be found in the documentation.

• I checked the documentation.

Issue

Julia isn't installable on ARM-based MacOS. If helpful, I know that PyTorch has ARM support (this Azure Pipeline) and Python also has ARM support (this Azure Pipeline).

Installed packages

N/A

Environment info

N/A

[jmuchovej]

Whoops. I didn't see the open PRs:

• ARM OSX Migrator #219
• Adding osx-arm support #224

[mkitti]

Unfortunately it's a bit complicated. We are closer to this than we have been. The main complication is how to cross-compile the system image. I am contemplating a way to deliver Julia without the system image and then provide the user a way to build it themselves.

As a practical matter, I would consider using juliaup in the meantime: https://anaconda.org/conda-forge/juliaup

juliaup can be installed via conda-forge. juliaup can then install julia. You may want to consider stealing the activation and deactivation scripts from this repository.

Thank you for your interest @jmuchovej

[jmuchovej]

👋 @mkitti thanks for the quick response! I've never used juliaup, but will definitely do this, for now.

Maybe a question to post over there, but... would using juliaup in a conda environment be equivalent to using a local poetry environment (for Python) or regular npm install (instead of npm install -g)? Another way to think of this is... does juliaup add Julia version management atop what you "already get" by using conda environments?

Also – should we leave this open, since it's ongoing and folks might do like me and check open PRs as an afterthought? (Compared to checking issues.)

[mkitti]

To be clear, no Julia packages currently exist within conda-forge. Whether you use this feedstock or juliaup, conda-forge is not tracking Julia dependencies directly.

The closest we have come to directly managing a Julia package is SymbolicRegression.jl in conda-forge/pysr-feedstock#43