On my Mac (Fortran) testpostw90_si_geninterp has stopped passing.
The issue is that the location of the one of the kpts is calculated as 0.3923941084E-16 while in the benchmark it is 0.0000000. This gives a relative error of infinity (!).
This is happening because the compiler is calculating the coordinate from ab-ab. In one case it must evaluate this to zero - and on my machine to something very small. We are writing out with a G format specifier.
I don't think any of this is wrong.
I have just removed the test of the relative error from the k-point coordinate (so just the absolute error is used).
I will submit a PR shortly
On my Mac (Fortran) testpostw90_si_geninterp has stopped passing.
The issue is that the location of the one of the kpts is calculated as 0.3923941084E-16 while in the benchmark it is 0.0000000. This gives a relative error of infinity (!).
This is happening because the compiler is calculating the coordinate from ab-ab. In one case it must evaluate this to zero - and on my machine to something very small. We are writing out with a G format specifier.
I don't think any of this is wrong.
I have just removed the test of the relative error from the k-point coordinate (so just the absolute error is used).
I will submit a PR shortly