[Patch, fortran] PR 37131, inline matmul

[Thomas Koenig]

Hi Mikael,

> However, it introduces regressions on matmul_bounds_{2,4,5}.
> It seems the "incorrect extent" runtime errors are completely optimized
> away (even at -O0).
> Any ideas?

This is seriously wierd.  It seems that the call to gfortran_error is
really optimized away, because the middle-end decides something strange.

I would assume the backend decl for gfortran_error is somehow wrong.

I will take a look, but this is an area that I don't really know a lot
about...

	Thomas