[J3] [EXTERNAL] [BULK] GNU Ada Translator GNAT has physical units checking
Keith Bierman
khbkhb at gmail.com
Wed Jul 30 21:53:19 UTC 2025
> On Jul 30, 2025, at 3:19 PM, Van Snyder via J3 <j3 at mailman.j3-fortran.org> wrote:
>
>
>
> The examples show that the effort is enormous, and therefore probably wouldn't be used. The attribute system I proposed is much simpler, and more complete, and the attributes can easily be added to existing code.
>
Well, I am no longer working in a compiler writing organization (indeed, not formally working at all ;>), so this is somewhat speculative … but while I can imagine attributes, if part of the language itself, might manage to be handled with better performance during optimization phase(s) of the compilation system than anything that would result from templates (perhaps the C++ template system has gotten better optimized but it used to be an excellent way to help C++ underperform Fortran code.
Someone with direct skin the optimizing compiler writing game would be better positioned to provide input to the committee.
Physical units are, as Van has long observed, really critical in many Fortran centric applications. On the other hand, complicating the language has a wide impact on everyone, and not every application is apt to suffer from this class of error.
Keith Bierman
khbkhb at gmail.com
303 997 2749
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.j3-fortran.org/pipermail/j3/attachments/20250730/4539ae16/attachment.htm>
More information about the J3
mailing list