http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40539
--- Comment #15 from Tobias Burnus <burnus at gcc dot gnu.org> --- As follow up: Recent Intel Fortran Compilers have the fixed it - but one currently has to use -standard-semantics in order to get the compatible result. [I don't know which options Intel uses for the Math Kernel Library.] See also http://software.intel.com/en-us/blogs/2012/05/11/doctor-fortran-in-i-can-c-clearly-now-part-i
