If an implementation extends this specification to support IEEE-754 flags or exceptions, then all built-in relational functions shall proceed without raising the invalid floating-point exception when one or more of the operands are NaNs.
The built-in relational functions are extended with cl_khr_fp16 to include versions that take half, and half{2|3|4|8|16} as arguments.