Building PyFFTW for osx-arm conda envs

[garrettwrong]

Below are some notes regarding building the PyFFTW and FFTW dependencies for ASPIRE on Apple silicon (M1 and M2 chips).

The following needs to occur:

1. Make and configure a local install prefix for use with the libraries, bin, and incudes we're about to make.
2. Get FFTW source.
3. Patch FFTW
4. make, check, and install FFTW for doubles.
5. make, check, and install FFTW for singles.
6. create an empty conda env
7. configure magic environment variables
8. Get pyfftw source.
9. Install pyfftw from source
10. Install ASPIRE
11. Test ASPIRE install.

More details are below. This is just meant for reference and not recommended for getting started with ASPIRE. At some point PyFFTW will probably just work out of the box on osx-arm, until then the recommendation is to try and use the osx-64 platform which will emulate x86_64 on M1 and M2 chips.

[garrettwrong]

# setup a local install location
mkdir -p ~/work/opt
export DYLD_LIBRARY_PATH="$HOME/work/opt/lib"
export LDFLAGS="-L$HOME/work/opt/lib"
export CFLAGS="-I$HOME/work/opt/include"

cd ~/work/
wget https://www.fftw.org/fftw-3.3.10.tar.gz
tar -xvzf fftw-3.3.10.tar.gz
cd fftw-3.3.10

# patch the source
cat << EOF > hack.txt
3934c3934
<                 aarch64)
---
>                 arm)
EOF
patch configure hack.txt

# configure and build the software towards the arm target with threading (doubles default)
./configure --prefix=$HOME/work/opt --enable-threads --enable-neon --enable-armv8-cntvct-el0
make -j
make check
make install

# clean
make clean
make distclean

# configure and build the software towards the arm target with threading, this time with single precision
./configure --prefix=$HOME/work/opt --enable-threads --enable-neon --enable-armv8-cntvct-el0 --enable-float
make -j
make check
make install

[garrettwrong]

conda create -n test_fftw_src python=3.8
conda activate test_fftw_src

cd ~/work
git clone https://github.com/pyFFTW/pyFFTW.git. # this was at v0.13.1
cd pyFFTW

# not sure if we need these, but I set them anyway.
export PYFFTW_INCLUDE=$HOME/work/opt/include
export PYFFTW_LIB_DIR=$HOME/work/opt/lib

# build the local source, the verbose might help if I forgot a step or there is other issues.
pip install -e . -v

# Finally we're ready to install ASPIRE as per usual
cd ~/work
git clone https://github.com/ComputationalCryoEM/ASPIRE-Python.git  # this was at v0.11.1
cd ASPIRE-Python
pip install -e ".[dev]"

# All tests should pass
pytest