On Tue, 2025-01-14 at 14:34 -0800, Fred Cisin via cctalk wrote: > >> A bad reason to pick Fortran or C is having been taught it at school > >> and then making no effort to update one's skills at any point in the > >> intervening decades. > > On Tue, 14 Jan 2025, Van Snyder via cctalk wrote: > > This assumes that your professors are teaching Fortran 66 instead of > > Fortran 2025. I'd be interested to know what you believe its defects to > > be. > > I first taught Fortran in community college in Fall 1983. > In the lab, we used the IBM/Micorosft Fortran, which worked very well for > THAT task, but had some serious deficiencies. A "Sieve of Erastothanes" > compiled in it ran slower than in BASICA. Bob Wallace (who was still at > Microsoft) warned me to avoid the run-time library.
Microsoft Fortran was a joke, but I liked Bob Allison, one of the developers. He retired at age 35. Ryan-MacFarland Fortran was promoted by IBM for the AT. I reported problems with it. I eventually got an invitation to buy the next version, but by then I had been using Lahey F77L for about a year. We found an error in F77L complex divide and expected the same sort of "reply" we had gotten from Ryan-MacFarland. The next day, Bruce Bush called and told me "type this into your fix file." I asked "what's a fix file?" He told me it's a small text file that the compiler reads and patches itself when it loads. We stuck with Lahey until Tom retired. By then, he wasn't selling his own compiler, but rather the Fujitsu Fortran 95 Windoze compiler that he had modified for Linux. Fujitsu didn't provide any meaningful support, never provided a 64-bit compiler, and never provided Fortran 2003 or anything newer. I use NAG and Intel now. gfortran has too many bugs and too many weird interpretations of "standard compliant." It refuses to compile some of my clearly-compliant modules. It gets occasional fatal internal errors and offers to send a message to the developers — from whom I never get a reply. Intel ifx is free and by far the best for run time performance, while NAG is by far the best for both compile-time and run-time diagnostics. ifx is free but nagfor isn't. > My FORTRAN experience was from fifteen years prior (PDQ FORTRAN, WATFOR, > and FORTRAN 4?), so, until I managed to catch up, I was teaching my > students to write FORTRAN programs in Fortran77. Damian Rouson teaches classes in parallel programming. About two weeks into the course he reveals to his students that they're using the coarray SPMD features of modern Fortran — far easier to use and understand, and generally more efficient, than MPI or PVM. > -- > Grumpy Ol' Fred ci...@xenosoft.com
