I\'ve recently come to maintain a large amount of scientific calculation-intensive FORTRAN code. I\'m having difficulties getting a handle on all of the, say, nuances, of a
Well, in one sense, you're lucky, 'cause Fortran doesn't have much in the way of subtle flow-of-control constructs or inheritance or the like. On the other, it's got some truly amazing gotchas, like the arithmetically-calculated branch-to-numeric-label stuff, the implicitly-typed variables which don't require declaration, the lack of true keywords.
I don't know about the "performance enhancing improvements". I'd guess most of them are probably ineffective, as a couple of decades of compiler technology have made most hinting unnecessary. Unfortunately, you'll probably have to leave things the way they are, unless you're planning to do a massive rewrite.
Anyway, the core scientific calculation code should be fairly readable. Any programming language using infix arithmetic would be good preparation for reading Fortran's arithmetic and assignment code.