Open amontoison opened 4 weeks ago
There are already in64 tests for the makefile system. Myaybe @jfowkes can extend them to meson?
@nimgould you would need to add an option to use 64bit integers to the runcutest
script first.
I will have a look
OK, this is a major undertaking. Firstly, sifdecode will have to be rewritten (extended) to generate elfuns (etc) with 64 bit integer arguments if required. This will be a breaking fix, as it will violate the presumption that these functions are standard fortran 77, as we would need to use the iso-fortran module. OK, we could argue that nobody uses f77 anyway, but it would break that requirement.
Secondly, we would now have to have an extra set of suffices for the generated elfuns (etc), i.e., elfuns_s_64. Again no big deal, except that every cutest function, and every interface, would need to adapt (and to be checked that lines didn't break the 132 char limit).
And finally, the scripts would need to allow for an -64 (or something) flag (trivial).
But now ... is this really useful? No cutest example currently reaches the 32 bit limit, and as nobody is generating new SIF files except for me, are we actually inventing a complication that is not needed? I am glad we moved to parameterised types, but is now the time to use them? Frankly I am exhausted by all of this software engineering, so if we do want to push this, perhaps someone else should lead?
Comments?
I personally see no need for this, the current 32bit integer CUTEst works just fine.
It's fine for me too.
@jfowkes @nimgould What can we do to test CUTEst with 64-bit integer? Can we add some tests here?