Fortran and Python

P. de Buyl -- CPTG (U of T) & Phys. Complex Sys. (Brussels)

SNUG June

Background

Working in physics: statistical mechanics & nonlinear dynamics, mainly.
Lots of numerical work.
Mostly Fortran and Python.

Antithetical Languages

Python

An interpreted object-oriented programming language.
NumPy provides a powerful array syntax.

Fortran

A strongly typed compiled programming language.
Access to plenty of libraries.
A powerful array syntax built into the standard.

Tools

Python with NumPy -> http://www.scipy.org/
A Fortran compiler (the gcc based gfortran works, some commercial compilers also work).
All that is presented today runs on SciNet!

Introductory example

subroutine multiply(a,b,n,c)
  double precision, intent(in) :: a(n), b(n)
  integer, intent(in) :: n
  double precision, intent(out) :: c(n)
  integer :: i
  do i=1,n
     c(i)=a(i)*b(i)
  end do
end subroutine multiply

import numpy as np ; import my_f90mod
a = np.arange(10,dtype=np.float64) ; a /= len(a)
b = np.arange(10,dtype=np.float64) ; b /= len(b) ; b = -b + 1.
c = my_f90mod.multiply(a,b)
print a ; print b ; print c

A few remarks

The intent of Fortran arguments is always specified (else, it falls back on the default: intent(in)).
intent(out) arguments are returned by the Python function.
"Automatic" information, such as the size of the array, is made optional in the wrapping process.

Example

Computation

The computation of the velocity autocorrelation function is really slow in Python. \begin{equation*} < {\bf v}(\tau) \cdot {\bf v} > = \int_0^\infty {\bf v}(t+\tau) \cdot {\bf v}(t) \end{equation*}

Speed test

Some results for the speed test.

Acknowledgments

Prof. R. Kapral

Prof. P. Gaspard

P. de Buyl -- CPTG (U of T) & Phys. Complex Sys. (Brussels) - Fortran and Python