|
Instant inlining of C/C++ code in Python
News
Release 0.9 (3/01/07): Port to Windows with mingw pluss some
added functionality and clean up.
Release 0.8 (5/10/06): Added support for NumPy
and Numarray.
Release 0.7 (12/09/06): Added functionality for the use of pkg-config files.
Release 0.6 (17/07/06): Cleanup of the user interface.
Release 0.5 (10/03/06): Instant uses SWIG directors for cross language inheritance.
Release 0.4 (03/01/05): Added md5sums to avoid unnecessary compilation.
Release 0.3 (27/11/05): Instant supports Numeric arrays.
Appetizer
Instant is a Python module that allows for instant inlining of C and
C++ code in Python. It is a small Python module built on top of SWIG and Distutils.
Example of use:
from Instant import inline
add_func = inline("double add(double a, double b){ return a+b; }")
print "The sum of 3 and 4.5 is ", add_func(3, 4.5)
Example with NumPy arrays converted to C double arrays:
from Instant import inline_with_numpy
c_code = """
double sum (int n1, double* array1){
double tmp = 0.0;
for (int i=0; i<n1; i++) {
tmp += array1[i];
}
return tmp;
}
"""
sum_func = inline_with_numpy(c_code, arrays = [['n1', 'array1']])
a = numpy.arange(10000000); a = numpy.sin(a)
sum1 = sum_func(a)
On my machine, the above example is about twice as fast as using the
sum function implemented in NumPy. This comparison
is of course not fair, since there are no safety checks in the
above example. The comparison is implemented in the file test9.py
that comes with the Instant package. The old package Numeric is about
as slow as NumPy (test2.py), while the same example with Numarray is
about as fast as with Instant (test10.py).
Requirement
Instant requires Python and SWIG. It has only been tested with
Python 2.3 and 2.4 and SWIG 1.3.24 or newer. It only runs on Unix systems.
Related Projects
The weave module
in SciPy
The PyInline module
The PyRex module
Magne Westlie
Kent-Andre Mardal
|