Enable non-finite values with Intel compiler
The Intel compiler by default turns on an optimization that assumes that
all floating point values are finite. This breaks any ligitimate uses of
non-finite values including checking values with functions like isnan
and isinf
. Turn off this feature for the intel compiler.
Backport: release
Edited by Kenneth Moreland