Resistance to modernization

I used two versions of optimization TOMS algorithm. Modernized one was done by experienced scientist and Fortran 95 programmer Allan Miller, who followed aforementioned approach and used his tool to make modernization. Finally, i quickly discovered the old one in f77 worked perfectly while the new one failed giving hard to trace down error. I gave up after few days and used the original implementation.
The second example concerns potentially efficient whole array constructs like where. They were never been as efficient as explicit loops since the latest were easily recognized patterns for advanced automatic compiler optimizations. My take is: be careful with modernization

2 Likes