I know there are great reliable libraries on Fortran for Computational PDE/ODE, Computational Statistics and Numerical Linear Algebra.
Also almost all the legacy software packages of Civil Engineering, Mechanical and Aerospace Engineering, and Chemical Engineering are developed on Fortran. Newer versions of Fortran compilers are object oriented and incorporated many new features needed for Finite Element Mesh and Unstructured Mesh of Finite Volume, However, they still do not have some features of new languages as they have to be backward compatible.
I ask this question having to main point in mind:
1) Education and training of numerical methods.
2) Developing from the scratch or modifying the exciting in-house codes?
What are the Fortran limits? and what scientific and practical cons and pros you imagine for abandoning or sticking to Fortran?
Dear Kaveh Zamani
Under Abaqus, COMSOL, Fluent,...(software) you can introduce a free mesh parameter for Computational PDE. Those software are aimed to model many computational and numerical approach like fluid Mechanics, electronics, heat transport... for example under COMSOL Multiphysics, you can use the Finite element method (FEM) where you can choose the mesh and you can set all parameters separately.
See the link below
https://www.comsol.com/blogs/equation-based-modeling-in-comsol-4/
Best,
Houssem
Léonard de Izarra
Thanks for your suggestion. Just couple of practical points, I would like to know your idea on:
1- If you are not an advanced programmer (which usually the case for Mechanical, Civil and Chemical Engineers) you cannot tune data structure settings of C++ or C# such that it optimizes your computational performance. For Classic Fluid Mechanics (7 state variable x, y , z, U, V, W, Temp) and Stress Analysis Fortran is very well structured.
2- Fortran is easier to master and readable language compared to C++.
3- Fortran compiler on default is very picky and pre-check many issues which may reduce programming error, at the cost of difficulty for the CFD or FEM programmer.
What do you think?
It seems to me that these two programming languages have the same paradigms (OOP and generic) in their latest version. They both have good compilers (free or not), good libs and good performances at runtime.
For each one the ease of programming comes mainly from the habit ...
In fact the big difficulty currently lies in the fact of programming with objects (generic or not) for parallel computers. It is necessary to extract the data from the objects (to serialize them) before transmitting them to the other nodes of the computer, which is often difficult for the programmer and time consuming for the machine.
After, I admit that some libraries write in C ++ are really unreadable.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python