Unofficial AFNI macOS 26 clang build for Apple M2

Installation Issues
1

Hi AFNI community,

I used Codex to build an unofficial AFNI macOS 26 binaries for Apple M2. I am sharing it in case it helps anhone who want a optimized-built AFNI binary on newer Apple hardware.

This is not an official AFNI/NIMH build. Please treat it as a community build and do not use it in a production enviroment.

Build summary:

AFNI version: AFNI_25.2.00 'Gordian I'
Architecture: arm64
Build system: 2023 MacBook Pro, Apple M2 Max
Compiler: Apple clang
Optimization: -O3 -mcpu=apple-m2 -pipe
Package: macos_13_ARM.AFNI_25.2.00.m2clang.2.tgz

Dependency choices:

jpeg-turbo instead of system libjpeg
zlib-ng-compat instead of system zlib

Runtime Homebrew packages requirements:

brew install openmotif mesa-glu gsl glib libomp libxpm zlib-ng-compat

The package includes atlas/template/assets from https://afni.nimh.nih.gov/pub/dist/tgz/macos_13_ARM.AFNI_25.3.03.tgz and removes Python __pycache__ files.

Release:

https://github.com/cecilyen/MacM2_AFNI/releases/tag/v25.2.00-m2clang.2

README:

https://github.com/cecilyen/MacM2_AFNI/blob/main/README.md

I would appreciate any feedback as this is the first time Codex (AI agent) to help me building AFNI on macOS.

2

Very interesting idea. Hope it works well, and hope it was relatively easy.

3

I published an updated unofficial AFNI 26.2.01 build for Apple Silicon macOS:

  • Generic arm64 build using Apple Clang
  • Optimized with -O3 and ThinLTO
  • Uses jpeg-turbo and zlib-ng-compat
  • Includes AFNI/SUMA, atlases, templates, and support files
  • Tested with all 49 AFNI GUI plugins loading successfully
  • Excludes the optional, incompatible R_io.so

Download the release
Installation instructions

This is an unofficial community build. Please validate important results against an official AFNI installation.

4

Here are some of my .afnirc tweaks that may improve efficiency for advanced users.

.afnirc

AFNI_HINTS = KILL
AFNI_COMPRESSOR = PIGZ
AFNI_NOMMAP = YES
AFNI_NO_MCW_MALLOC = YES
AFNI_NOSPLASH = YES
AFNI_DEFAULT_IMSAVE = png
AFNI_VERSION_CHECK = NO
AFNI_MOTD_CHECK = NO
AFNI_IMSAVE_WARNINGS = NO
AFNI_SKIP_DYLD_WARNING = YES
AFNI_SESSION_SORT = MRITYPE
AFNI_START_SMALL = YES
AFNI_X11_REDECORATE = NO
AFNI_CLICK_MESSAGE = NO
AFNI_DISABLE_CURSORS = YES

Also, export DYLD_LIBRARY_PATH=/opt/X11/lib/flat_namespace is no longer needed in the lastest AFNI and macOS.

5

Hi, Cecil-

I have followed this thread as you have updated it. I just wanted to say thanks for going through these things, and we will look more deeply at this on our side here. (And I am surprised about the flat_namespace!)

Thanks,
pt