Abstract: In this paper the performance of scientific applications are discussed by using Python programming language. Firstly certain techniques and strategies are explained to improve the computational efficiency of serial Python codes. Then the basic programming techniques in Python for parallelizing scientific applications have been discussed. It is shown that an efficient implementation of array-related operations is essential for achieving better parallel [11] performance, as for the serial case. High performance can be achieved in serial and parallel computation by using a mixed language programming in array-related operations [11]. This has been proved by a collection of numerical experiments. Python [13] is also shown to be well suited for writing high-level parallel programs in less number of lines of codes.

Keywords: Numpy, Pypar, Scipy, F2py.