The Motley Fool Discussion Boards

Previous Page

Miscellaneous / Climate Change

URL:  https://boards.fool.com/db2-i-imagine-it-will-be-difficult-to-find-a-30515527.aspx

Subject:  Re: FORTRAN??? Date:  1/30/2013  1:29 PM
Author:  LorenCobb Number:  40543 of 90939

DB2: I imagine it will be difficult to find a bright young researcher with expertise in FORTRAN.

Heh. I wish that were true.

In the real world, Fortran is the language of choice for scientific work on massively parallel multi-core computer architectures. To be more precise, it is the language of choice for writing the low- to mid-level numerical algorithms on which all higher-level code depends. Higher-level code can be written in anything -- it doesn't matter. But for low-level stuff, like finding the eigenvalues of million-dimensional matrices, conventional wisdom holds that nothing beats* the highly optimized Fortran compilers.

The state-of-the-art package of numerical algorithms is LAPACK, which is maintained and improved by a team of mathematicians in the next office down the hall from me. It is entirely written in Fortran 90, an updated object-oriented version of the old Fortran 77.

Fortran was already old when I learned it in 1967. I remember one wag in the Computer Science department saying, "I don't know what language I will be programming in when I retire, but I know it will be named 'Fortran'." That just about sums it up.

http://www.netlib.org/lapack/
http://en.wikipedia.org/wiki/LAPACK
http://arxiv.org/abs/0709.1272

Loren

* I have been having a running argument with the numerical jocks about this. I say, based on years of experience, that I can write hand-crafted assembler code that will beat anything produced by an optimizing Fortran compiler. The boys next door think I'm nuts, and swear by their compilers. Someday we will have to stage a contest to see who is right.
Copyright 1996-2020 trademark and the "Fool" logo is a trademark of The Motley Fool, Inc. Contact Us