Fortran is better for Array calculation and scientific computing in genral (built for that anyway), C is more practical for data handling (pointers, node count, meshes, coordinates...).

But now I think C++/F90 are really close to eachother to create quality CFD codes with high level of abstraction language (code as you speak, program as you write...).

jy

(1). I guess you are right. (2). If you don't want to become the endangered American buffalo, then you will sure come up with some new ideas.

i think this subject area is overworked now. the performance of a program has, in practice, not much to do with the programming language but the programmer. moreover there are so many other issues to consider when writing a code that the language of little importance. much of the time big codes written by more than one person can involve multiple programming languages. italian and english are just as good for communicating but when in Rome you speak italian etc

(1). I really would like to think in your way. (2). But it is likely that this C/C++ will become so powerful that Fortran will become the American buffalos in the Yellow Stone National Park. (3). After reading many books on C/C++, I think there is a big reason why C/C++ is becoming very popular. It is trying to model the real thing, something which has both the data and functions to operate on it. (4). Fortran used to be more on the calculator side, but this C/C++ can model the system, and CFD as a whole is a system. (5). I don't want to go further beyond this point. But I have to warn the reader that failure to recognize the difference between a calculator and CFD as a system and the difference between the old Fortran and the C/C++, will cost you a lot later on, just like the fate of the American buffalo. (6). But, that does not means that Fortran will disappear from the surface of the earth.

Although I'm very much in favour of C/C++ I have to surrender to some complaints about it on specialist compilers. Unfortunately alot of CFD still requires very large specialist machines which in turn have specialist compilers to exploit their features. It seems that C compilers have more difficulty in this exploitation compared with Fortran compilers.....

Ref. http://www.hpcf.cam.ac.uk/C_rant.html

I suspect we'll have to live with Fortran for a bit longer but this doesn't stop codes written in more than one language.

(1). Thank you very much for the information and the web site. I am getting a much better feeling about the Fortran VS C/C++. (2). About the statement "LINPACK in FORTRAN achieves 54 MFLOPS in FORTRAN and only 34 MFLOPS in C-significantly in FORTRAN's favour.", I feeling is that, there is no difference even there is a factor of two. I normally would upgrade my system when there is a clear gain in speed of at least four times. (3). The basic argument is: if we are going to retain the vector capability and performance in hardware and software, then the current FORTRAN is the best choice. (4). But if the users decided not to use the High Performance Computing Facility in favor of the new features available to C/C++, then I guess, at that point, the Facility will have to change its mind. (5). In many cases, people are curious about and interested in the new features and fields, where they can make some new contributions. This is how we make progress in engineering and sciences. For me, there is no difference between 54 MFLOPS and 34 MFLOPS. In the last ten years alone, we have move from a 25MH PC to a 800MH PC, there is a factor of 32 in clock speed gain. And the actual speed gain is many more times this value.