We have a large code for High Energy Physics, written in Fortran 2003/2008, which works flawlessly with gfortran 5.4+, NAG 6.2+, ifort 19+. With nvfortran it compiles, but already performing our test suites leads to many errors. Our code can be found here (together with instructions to build and run it): https://whizard.hepforge.org
(for many tests to be meaningful it is favorable to have the OCaml compiler at hand).
In our test suite, 58/136 unit tests are failing, and frightening 279/316 functional tests are failing. My question is how to report them: one by one, or - as I consider many of them connected - you are going to look into the full test suite?
Hi Jeurgen,
I sent a note to our engineering managers on what they want to do, but am encouraging them to update the version of Whizard that we run in our daily regression testing. We currently test 2.1.2, which works correctly, but obviously is not capturing these issues.
-Mat