Using reserved words as variables

[PierU]

@RonShepard

It depends on the application, of course, but 20+ years ago I had that same kind of choice (between pointers and fixed arrays of max dimension) and I used pointers.

There was a 3rd choice, i.e. allocatable arrays not attached to a derived type. Of course codes sometimes were more complex/less structured because of this constraint, but performances were not hurt by the pointers.

Another solution was to have pointer components anyway, but pass them down to non pointer dummy arguments in the computing routines. Sometimes it was requiring an additional layer of calls, but at least performances were not hurted by the pointers.

Sure, it is now easier with the allocatable components, but it was possible nonetheless to have the same performances back then.

[FortranFan]

@RonShepard

… When allocatables were allowed (as dummy arguments and derived type components), I switched, picked up as much as a 2x speed improvement …

A huge problem in the Fortran ecosystem is the vagueness re: performance. Many of the claims are highly questionable and the actual practices in codes vary widely and myths too appear to drive choices.

It’s a sad state even now.