C++ Versus F95

Set of Links that provide studies comparing C++ to F95

http://www.codecomments.com/archive271-2006-3-856235.html

http://www.amath.washington.edu/~lf/software/CompCPP_F90SciOOP.html

http://www.cs.rpi.edu/~szymansk/OOF90/performance.html

http://www.cts.com.au/compare.html (More of an education perspective.)

Scientific Computing: C++ Versus Fortran (The Dr Dobb article has a newer date, but has been edited very little and the page layout makes it much harder to read.)

Travis has also spoken with a colleague who has written several Adaptive Mesh Refinement (AMR) packages using C++. He is familiar with the SAMRAI package at Lawrence Livermore National Lab, which is a C++ AMR package with the lower level loops written in Fortran. Here are his comments:

"Regarding the C++/Fortran debate, templates per se do not degrade performance unless you are talking of expression templates which very few people use. They(regular templates) simply increase the size of your executable if you are not careful. Of course you have the option of not using templates at all. I use inheritance and compositional models in design and prefer it personally to using templates. SAMRAI uses both C++ and fortran as you probably know. The compute intensive tasks are always written in fortran for performance and it leverages C++ for flexibility, extendability, ease of use, and clean design. But this is an option you could try suggesting to people. Use C or fortran to do calculations and C++ for higher level operations. "