Absoft Corporation has announced the availability of Visual Numerics’ IMSL Fortran Numerical Library version 5.0 bundled with its new high performance MATFOR 3.
Absoft Corporation has announced the availability of Visual Numerics’ IMSL Fortran Numerical Library version 5.0 bundled with its new high performance MATFOR 3.
Good to hear some of the programming skils I acquired at my UNI (electronic engineering) are not obsolete. And yes, we used Fortran in our numerical methods course. That was hardcore stuff….
The only language is Fortran 95. I just spent two days trying to debug some C code that would have been a simple to spot error (or impossible to code wrongly) using Fortran. It was some wacky C array/pointer/derived data type structure that was causing it. AAA!
Give me native arrays and ultrarobust numerical libraries anyday. Sweet, sweet Fortran…
It was some wacky C array/pointer/derived data type structure that was causing it.
Unfortunately, standard F95 still has this problem for cases whey you absolutely need dynamic allocation within a derived type. There are two paths around this.
The first is use compilers that support the Technical Report (TR) for using “allocatable types” in derived types (I forget the TR number). Most newer (in the last year) Fortran 95 compilers support this TR (I know Intel ifort is supposed to). I think, but haven’t verified, that this Absoft compiler supports it as well.
Second, wait for a Fortran 2003 (the new standard, final draft due any time now) compiler. This is part of the F2003 standard. It will be some time before all F2003 functions work in compilers, but this should be available pretty early, especially since its available in most F95 compilers.
For those who don’t know, Fortran specified “allocatable” components as well as “pointers”. Pointers work a bit differently that in C, as they carry things like array bounds information, and arrays don’t have to be confined to contigous memory locations (they can have a predictable pattern which the pointer must supply when referenced). Allocatables don’t leak memory as they have a definite assigned symbol, instead of being “pointed to” by a symbol that can later point to something else. They also don’t have pointer-aliasing problems, helping to speed up Fortran programs that need dynamically allocated memory.
Historically, the authors Metcalf and Reid have a series of “Fortran Explained” books that are highly recommended as a Fortran reference.
The new edition _Fortran 95/2003 Explained_ is now available and will give you a good understanding of the new features availble in Fortran 2003: more object-oriented features, stream i/o, standardized interoperation with C, and the afore-mentioned new application of allocatables.
How do you expect it to die, there is so much out there written in it, it works great, and no one is going to re-write it unless they have to. Engineering models written in the 70s in fortran are still heavily used. I used programs in college my profs wrote in the 70s and are still using today.
Yes, very much so. Fortran is used for heavyweight numerical problems, mostly for Finite Element Methods (FEM). It seems to be particularly useful for fluid dynamics, stress computations, magnetic/electric field simulations, thermodynamics, weather predictions, astrononomy and celestial mechanics or what have you.