A Quick Guide on How to Use the Fortran-to-Python - NO
Contents
1. a copy of the module in every project you conduct 5 Where to Go Next As a starter try to implement your own modules which may also include multiple subroutines We would also like to encourage you to redo the example but this time with the source code written in C instead of Fortran This tutorial was meant to be as short as possible and while we do not necessarily claim to have succeeded on this point we have given a marginal working example but neglected much of F2PY s functionality and applications Hence we d like encourage you to take greater look in its details and documentation We suggest you to try to write a generic integration routine in fortran in which takes a general one dimensional function and integration boundaries as arguments and returns the integral Nested loops are known to be an Achilles heel for all scripting language You can also take a look at this numerical integration example which explicity shows how powerful a tool F2PY can be by reducing the total running time of a pure Python script by a factor of 300 by using F2PY If you have any comments suggestions or even corrections you wish to add please feel free to send me an email magnud at stud ntnu no or mhsd91 at gmail com Shttp folk ntnu no magnud Projects F2PY_integration html2. A Quick Guide on How to Use the Fortran to Python F2PY Module Magnus H S Dahle February 2015 Abstract If you re familiar with Python and Fortran and wish to minimize reading Linux users may jump to section 4 2 while Windows and MAC OS users to 4 1 You should in addition be aware that F2PY also works for wrapping C code 1 Introduction The choice of programming language can be extensively difficult especially considering the computational effectiveness against implementation time and effort While scripting languages like MATLAB and Python may provide intuitive code which is fast to implement compiled languages like C C and Fortran yield superior computational speed Within this tutorial we ll show that it is in fact possible to obtain the very best of both worlds At least to some extent In the following example we ll make use of the Fortran to Python package F2PY in which enables a Python script to directly call a readily compiled Fortran module This is a powerful tool giving us the possibility for developing easy written code in Python but leaving more computational demanding parts of our script to more capable Fortran routines 2 Installation One should first note that F2PY is actually part of the NUMPY Numerical Python package of Python Thus having successfully installed NUMPY should imply also having a functional F2PY module If your computer is running on a Linux distribution using F2PY is rather straigh
3. ate a script and import whatever modules you d like including your own self made Print your self made module s documentation to see if it was imported correctly gt gt gt import numpy gt gt gt import chosen_module_name gt gt gt print chosen_module_name __doc__ Which should result in This module chosen_module_name is auto generated with f2py version 2 Functions su fortran_sum a b gt gt gt NOTE that in the process of creating the python module from the Fortran code upper case letters are transformed to lower case This is not that much of a big deal however you should note that Fortran is not case sensitive That is variable function names like aAA aaa AAA are equivalent and non distinguishable This is however not the case in Python Nevertheless the function may now be called similar to any other python function routine gt gt gt a numpy linspace 0 1 10 gt gt gt A chosen_module_name fortran_sum 1 2 4 2 gt gt gt print A 5 4 gt gt gt Hot Insider Tip If you re planning on using F2PY quite a lot you might end up with a large amount of self made Python modules We therefore suggest that you create a permanent library directory to store all modules you create this way Thus you can make a static path thread to import all your glorious modules from whenever you need them This way you ll save a lot of space compared to having
4. hich is given below and uses numpy distutils to generate a new Python module from the Fortran code file 3http docs scipy org doc numpy dev f2py gt documentation 4Equivalent to NUMPY SciPy etc ANIow kwnr PRR WwnNnrRowo The content of build_f2py_modules py from numpy distutils core import Extension ext Extension name chosen_module_name sources fortran_file_name f90 if _name__ __main__ from numpy distutils core import setup setup name f2py_example ext_modules ext Running this Python script will result in the creation of the new directory build lib 1linux x86_64 3 4 where you will find the newly generated file chosen_module_name cpython 34m so This is a Python module to be imported equivalently like any other module within a python script e g NUMPY or SCIPY 4 2 Creating the Python Module from Fortran Code The Easy Life of the Linux User Having installed F2PY you can call it directly from the terminal working_directory f2py c m chosen_module_name fortran_file_name f90 which will create the file chosen module name so This is a Python module to be imported equivalently like any other module within a python script e g NUMPY or SCIPY 4 3 How to Use the Module You ve Just Created Move copy paste the Python module file you just created to a directory you wish to work in Start a live Python console or simply cre
5. ial command line as f2py such that this is ignored by Fortran but used by Python to distinguish in output variables Notice the use of small cased letters f2py intent in a f2py intent in b f2py intent out Su Su A B END SUBROUTINE NOTE that older versions of fortran e g 77 uses file extension f not 90 but even more crucial the comment character may be different For Fortran77 it is an upper case C while for more modern versions of Fortran e g 90 it is an exclamation mark This ultimately also affects the f2py gt Cf2py command You may witness examples of this if you search for more tutorials elsewhere Finally we want to compile this Fortran Subroutine and incorporate it into a python module to be imported in a python script e Create your Fortran file fortran_file_name f90 It may contain multiple subroutines e Choose a name for the Python module you are about to create chosen_module_name e Create a directory in which you want to work in and move copy paste the Fortran file there Linux users may skip the next section and go straight to section 4 2 in which explaines a somewhat easier Linux approach 4 1 Creating the Python Module from Fortran Code The General Method The following method uses runs nothing but a pre written Python script and is thus eligible for all users regardless of their Operating System One needs the arbitrarily named Python script build_f2py_modules py w
6. t forward and easy going as Python with NUMPY is almost always integrated in the OS Linux users may check if F2PY is already installed by just opening a terminal by pressing CTRL ALT T and then type the command f2py with small letters If it is installed a short user manual will be printed in the terminal If not the terminal will complain that there exists no such command For Linux users F2PY is installed by installing NUMPY from official repositories Evidently this is also valid for computers running Windows or MAC OS That is download and install the NUMPY package It is easy to find on the web but if you re having issues you may want to take a look at the solved problem given by the following footnote NOTE If one struggles to make F2PY run correctly on one s own computer there are plenty of machines at the university NTNU running Linux Ubuntu where NUMPY and F2PY are both installed 3 When should You Use F2PY This is perhaps the ultimate question and there is no definite answer A good rule of thumb however is to use F2PY or compiled languages in general when considering nested loops Possibly the most typical example would be operations on elements in multidimensional matrices Other good examples could be programs calculating integrals or conducting Monte Carlo Simulations You might wonder if anyone has already made F2PY modules before you and the answer is most likely yes Most of the functions and ro
7. utines found in NUMPY and SCIPY are actually compiled Fortran routines which provides highly efficient and fast solvers for multiple problems Thus we advice you to always check if one of these two modules already have a routine in which may be suitable for your problem If not then implement your solver in pure Python to see if efficiency really is an issue If it is then F2PY may possibly provide the best solution for your problem lor more precisely numpy distutils nttp scientificcomputingco blogspot no 2013 02 f2py on 64bit windows python27 htm1 OANow kwnr nt el eS ee ee OOMONDRWNF OO 4 Usage and Syntax by a Marginal Working Example In the following we present a short example on how to use the F2PY module For details and references we strongly encourage you to take a look at the module s official documentation The F2PY module effectively compiles and incorporates Fortran code into a new Python module readily importable to any Python script We will next give an working example and obviously we ll need some Fortran code Furthermore we ll apply Fortran 90 and for simplicity we ll use the following banal and trivial Fortran routine This is the content of the file fortran_file_name f90 SUBROUTINE Fortran_Sum A B Su real 8 A B Su Real double precicion variables The Fortran Compiler ignores all lines starting with 7 The f2py utilizes this fact by defining its own spec
Download Pdf Manuals
Related Search