john@ghostwheel.unm.edu (John Prentice) (12/04/90)
The following note appeared in sci.math.num-analysis today. > I am looking for a C version of the SLAP package of netlib. > I tried already the f2c translator for it but the result is too > machine dependent for my purpose (actually, it is not the translated code > itself but the f2c libraries that are neccessary for it that require an > individual installation on every machine). I thought this was interesting considering the suggestions recently in the Fortran versus C debate that f2c solves the problem for C programmers of the big math libraries being written in Fortran. I have not talked to the person who posted this, so I don't know the specifics. There may be unusual circumstances or perhaps there is a simple solution. What I wanted to illustrate however is that clearly there are numerical programmers out there who are not finding f2c suitable for translating the math libraries into C. That is not to dump on f2c! However, a little realism about its capabilities is perhaps in order. What can one really expect to be able to do with f2c in translating Fortran mathematical packages into C, what are its limitations, how hard is it for the non-C or non-Unix guru to use, and how efficient is the resulting implementation? Please post any responses to the newsgroup since there is a wide audience for this topic. John Prentice john@unmfys.unm.edu
york@altger.UUCP (york) (12/07/90)
I used f2c to translate a program of about 40000 lines of standard F77 code (numerical analysis, graphics) and had no problems with it. The package includes COMPLEX statements and COMMON blocks.