Question

%%file fib.h

double fib(int n);

Writing fib.h

%%file fib.c

double fib(int n) {
    double a = 0, b = 1;
    for (int i=0; i<n; i++) {
        double tmp = b;
        b = a;
        a += tmp;
     }
    return a;
}

Writing fib.c

Using bitey and clang

This is perhaps the simplest method, but it only works with the clang compiler and does not geenrate highly optimized code.

import bitey

!clang -O3 -emit-llvm -c fib.c -o fib1.o

import fib1

fib1.fib(100)

354224848179261997056.0000

%timeit fib1.fib(100)

1000000 loops, best of 3: 941 ns per loop

Using Cython

I recomment using Cython for all your C/C++ interface needs as it is highly optimized and can do boht C \(\rightarrow\) Python and Python \(\rightarrow\) C. It is a littel more involved, but the steps always follow the same template.

Define functions to be imported from C

%%file fib.pxd

cdef extern from "fib.h":
    double fib(int n)

Writing fib.pxd

Define wrapper for calling function from Python

%%file fib2.pyx

cimport fib

def fib(n):
    return fib.fib(n)

Writing fib2.pyx

Use distutils to compile shared library for Python

This is the standard way all Python modules are compiled for distribution, and results in a build that is portable over different platforms.

%%file setup.py
from distutils.core import setup, Extension
from Cython.Build import cythonize

ext = Extension("fib2",
              sources=["fib2.pyx", "fib.c"])

setup(name = "cython_fib",
      ext_modules = cythonize(ext))

Overwriting setup.py

! python setup.py build_ext -i &> /dev/null

import fib2

fib2.fib(100)

354224848179261997056.0000

%timeit fib2.fib(100)

1000000 loops, best of 3: 224 ns per loop