Python vs. C++

It is an often reiterated statement that

Interpreted code is always slower then compiled code. We need speed! That's why we're using C/C++ in our project.

This assumption is based on the correct observation that large loops like for a dot product of two vectors u, v, are faster in C,

double out = 0.0
for (int k=0; k < n; k++) {
   out += u[k] * v[k];
}

than in Python:

out = 0.0
for k in range(n):
    out[k] += u[k] * v[k]

If you care about speed, you wouldn't do either of the above loops, though. In Python, most everyone does

import numpy

out = numpy.dot(u, v)

anyway. For C/C++, Eigen, CBLAS, and Boost come to mind.

This repository contains a comparison of some common costly numerical operations between C++ and Python.

As always, comments and suggestions are welcome!

Dot product of two vectors

Matrix-matrix product

Sum entries in a vector

Add two vectors

License

The code in this respository is published under the MIT license.