Brainstorm: Apple silicon support

[Edouard2laire]

This PR is here to use the maint branch of SPM when using an ARM Mac (which requires new Mex files) and fix a small bug

Note: this cannot be merged as is as brainstorm doesn't really like that SPM is in the folder called spm12-maint. But this works after adding SPM manually to the path: addpath('/Users/edelaire1/.brainstorm/plugins/spm12/spm12-maint')

Edit: actually would that be ok with you if we directly pull from the main branch of SPM: https://github.com/spm/spm ?

[rcassani]

Since the only difference between the main (the download link) and maint branches are the *.mexmaca64 files. It would make more sense to change the plugin to point to maint It should work by changing PlugDesc(end).Version from latest to github-maint and using the maint archive ULR.

• spm/spm12@c28485d
• https://neuroimage.usc.edu/forums/t/should-use-the-latest-version-of-spm12-for-better-compatible-for-apple-silicon/42703

[Edouard2laire]

Like that or would you change the version for everyone? I guess most users don't need the new mex.

actually shoudn't we remove the mex from https://github.com/brainstorm-tools/brainstorm3/tree/master/external/spm ?
as they are included in spm plugin ?

[rcassani]

The switch option should be ok. Not only saves the 6 MB for the Apple silicon binaries, but also saves the user of downloading from GitHub which can be slow, SPM is around 200 MB. I hope the zip in their page get updated at some point.

Regarding the duplicated binaries, the opposite approach will be done (later push by me also in this PR):

1. Add the to external also the Apple silicon ones
2. Remove them from the installed SPM as the binaries in external are the ones that add the In-Brainstorm MNI normalization (maff) method 😉

P.S. Adding the small bugfix for not existent plugin process directly in master

[Edouard2laire]

Thanks. let me know if i can help.

Maybe we should identify the toolbox that doesn't have new mex yet like iso2mesh, Mcxlab, and put a warning or something indicating that they cannot be used on ARM mac with matlab >2023b.

[rcassani]

These things are needed:

• A warning in the startup to indicate that Brainstorm is running on Apple silicon and some stuff may not work as expected, and recommend to use the Intel version.
• Add the Apple silicon binaries for external SPM
• Add warnings in the plugins that do not have support yet

Edit: It is quite a lot what is not supported yet:
https://www.mathworks.com/support/requirements/apple-silicon.html

[Edouard2laire]

Edit: It is quite a lot what is not supported yet:
https://www.mathworks.com/support/requirements/apple-silicon.html

it might seem like a lot but this is actually not really impacting us as its mainly packages for Simulink or specific decide like Arduino boards.

Here is an incomplete list of plugins that are not compatible yet

• iso2mesh Compile iso2mesh for ARM Mac fangq/iso2mesh#74
• mcxlab-cuda
• mcxlab-cl (Compilation on M1 Mac native fangq/mcxcl#48)

Maybe the generic warning is enough

[rcassani]

11-dec-2023: This is what I have checked so far. Parenthesis indicate not supported due to dependency

❓ 'adi-sdk'
✅ 'axion'
❓ 'bci2000'
✅ 'blackrock'
❌ 'brain2mesh' (iso2mesh)
✅ 'brainentropy'
✅ 'cat12'
❌ 'ct2mrireg'
✅ 'derivelfp'
❌ 'duneuro'
✅ 'easyh5'
✅ 'fastica'
❌ 'fieldtrip'
❌ 'iso2mesh'
✅ 'jsnirfy'
❓ 'kilosort'
❓ 'kilosort-wrapper'
❓ 'libsvm'
❌ 'mcxlab-cl'
❌ 'mcxlab-cuda'
❓ 'mff'
✅ 'mia'
✅ 'neuroelectrics'
✅ 'neuromaps'
✅ 'nirstorm'
✅ 'npy-matlab'
❓ 'nwb'
❌ 'openmeeg'
❓ 'phy' (?)
✅ 'picard'
❌ 'plexon'
❓ 'plotly'
❌ 'roast' (iso2mesh)
❌ 'simmeeg' (fieldtrip)
✅ 'spm12'
❓ 'tdt-sdk'
✅ 'ultramegasort2000'
❌ 'waveclus'
❓ 'xdf'

[Edouard2laire]

For brainentropy, if fminunc is available, then it's already working as part of the optimisation toolbox.

Edit; in case its needed, I recompiled it: multifunkim/best-brainstorm#15

[rcassani]

Added an error message for the Plugins that have been already tested.

As testing exhaustively all plugins may take a long time, due to lack of HW to test on my side. I think we should merge this PR for the moment and update the list (in f72d0a6) with future tests and reports from users. What do you think?

[Edouard2laire]

Also added an error when loading an already downloaded plugin. I'll try to compile them slowly first with iso2mesh and mxclab.

I agree that we can merge it like that.

[Edouard2laire]

update. I was able to compile iso2mesh today: fangq/iso2mesh#75

Some tests are needed but when this is merged; we should be able to check that from our todo list :)