PDA

View Full Version : fortran 95




fluberman
Nov 23, 2004, 06:12 AM
I have been a pc user all the time, although I spend most of time running my PC under linux and now I'm considering to switch. I am also doing some programming using fortran 95 (intel fortran for linux, which is free). As far as I know, under mac os x I can run g95 or gfortran but they are still under heavy development. Does somebody here have any experience in running g95/gfortran and how about it's performance compared to a commercial compiler such as Absoft fortran ? I do really want to swich, but this issue is important to me. Thanks.



Hoef
Nov 23, 2004, 07:10 AM
Wow .... What do you use fortran for if I may ask?

fluberman
Nov 23, 2004, 07:23 AM
Wow .... What do you use fortran for if I may ask?

scientific computing, especially cfd

Mitthrawnuruodo
Nov 23, 2004, 08:09 AM
Maybe you'll find something you can use on Apple's own High Performance Computing (http://developer.apple.com/hardware/hpc/index.html) page.

I tried some simple MPI stuff for my parallell computing class, but as I don't have a cluster in the basement or even a dual processor Mac I can't say if really did any good... ;)

gwuMACaddict
Nov 23, 2004, 08:12 AM
Wow .... What do you use fortran for if I may ask?

still used quite frequently in finite element analysis and computational fluids

bousozoku
Nov 23, 2004, 11:12 AM
Absoft recently started selling IBM's XLF compiler, apparently, instead of their own. It's AltiVec-aware so it should provide good performance and it's G5/PPC970-aware so it can handle larger memory spaces.

jeremy.king
Nov 23, 2004, 03:31 PM
still used quite frequently in finite element analysis and computational fluids

:confused: what the heck is that?

Out of curiousity, what makes one language better than another in the field of computational fluids?

I got to go gooooooogling now. :)

visor
Dec 21, 2004, 08:37 PM
:confused: what the heck is that?

Out of curiousity, what makes one language better than another in the field of computational fluids?


Usually - the existing codebase. Somtimes better Compiler performance, however, if someone asks if there are compilers available - it's the existing codebase ,)

mwpeters8182
Jan 8, 2005, 10:49 PM
I took a class in Numerical methods during last semester (1st semester of grad school). We programmed exclusively in fortran, and I belive my prof. said g95 would work ok, though we used absoft's compiler.

MP

mkrishnan
Jan 8, 2005, 11:24 PM
:confused: what the heck is that?

Out of curiousity, what makes one language better than another in the field of computational fluids?

I got to go gooooooogling now. :)

The usual argument is that Fortran has the highest efficiency in compiling to machine code...that is, the ratio of the theoretical number of floating point and integer ops that the actual scientific problem takes, to the number of ops that the *computer program* takes is very favorable. So it's good for techniques that are essentially brute force, in the sense that they model a complex system by modelling relatively simple equations (such as basic fluid equations or Maxwell's electromagnetic equations) on a really fine mesh and/or lots of elements.

We used to use Fortran for modelling discrete particles (electrons and ions) in relativistic (but not QED) EM fields in plasma physics....

Oooh, btw, is there stuff you can't do in G77?

mwpeters8182
Jan 8, 2005, 11:51 PM
My guess is that if he had been programming in F90/95, G77 wouldn't work, as it adheres to the old (F77) standard IIRC, which requires strict formatting and all - F90 is MUCH easier to code in than F77. I'd give G95 and gfortran a try -- they're available here :

http://hpc.sourceforge.net/

MP