[How To] Install Pyfhel in macOS

 

• Tested on Intel-based (x86_64) MacBook, not tested on M1/M2 (ARM)
• Apple Clang does not support -fopenmp and -libseal, use GCC or CLANG

1. % git clone --recursive https://github.com/ibarrond/Pyfhel

2.a) create requirements.txt file

  "setuptools<=60.9",
  "wheel",
  "cython>=3.0.0a9",
  "numpy>=1.20",
  "cmake>=3.15",
  "toml>=0.10"

2.b) % pip3 install -r requirements.txt

3. edit pyproject.toml

Line #111 [Afhel]

extra_compile_args

• {Darwin = ["-std=c++17","-O3","-fopenmp"]},

extra_link_args

• {Darwin = ["-fopenmp","-dynamiclib"]},

Repeat the above for [CYTHON EXTENSIONS]

4. edit Pyfhel/Afhel/Afhel.pxd

#from numpy cimport int64_t, uint64_t, uint8_t   

ctypedef long long int64_t
ctypedef unsigned long long uint64_t
ctypedef unsigned char uint8_t

5. edit Pyfhel/Pyfhel.pxd

#from numpy cimport int64_t, uint64_t

ctypedef long long int64_t
ctypedef unsigned long long uint64_t

6. edit setup.py

add these lines immediately after imports

os.environ["CC"] = "/usr/local/Cellar/gcc@12/12.1.0_1/bin/gcc-12"
os.environ["CXX"] = "/usr/local/Cellar/gcc@12/12.1.0_1/bin/g++-12"
os.environ["LDSHARED"] = "/usr/local/Cellar/gcc@12/12.1.0_1/bin/g++-12 -Wl,-x -dynamiclib -undefined dynamic_lookup"

Comment Line # 36-38 in setup.py:

#if platform_system == 'Darwin': # MacOS
    #raise SystemError("Pyfhel is not supported in MacOS (see issue #59)."
                      #"Please use a Linux VM or Docker.")

edit # Auxiliary methods

def get_lib_suffix(lib_type: str) -> str:
    if lib_type == 'static':
        if platform_system == 'Windows':
            return '.lib'
        else:
            return '.a'
    else:  # shared
        if platform_system == 'Windows':
            return '.dll'
        elif platform_system == 'Darwin':
        	return '.dylib'
        else:
            return '.so'

Build and install:

% python3 setup.py build && python3 setup.py install  (OR)
% pip3 install .
% cp build/lib.macosx-12-x86_64-3.9/Pyfhel/libAfhel.dylib /usr/local/lib/libAfhel.dylib

This guide installs Pyfhel, a Python wrapper for homomorphic encryption libraries, on an Intel based macOS machine. Because Apple Clang lacks some required flags, it uses an alternative toolchain and pulls the project with its submodules.

Homomorphic encryption allows computation directly on encrypted data. Pyfhel makes it approachable from Python, though the native build steps are the tricky part on macOS.