    <para>
      If an implementation extends this specification to support IEEE-754 flags or
      exceptions, then all built-in relational functions shall proceed without raising the
      <varname>invalid</varname> floating-point exception when one or more of the operands
      are NaNs.
    </para>

    <para>
      The built-in relational functions are extended with
      <citerefentry><refentrytitle>cl_khr_fp16</refentrytitle></citerefentry> to
      include appropriate versions of functions that take <type>half</type>, and
      <type>half{2|3|4|8|16}</type> as arguments and return values.
    </para>

 <!-- 13-Oct-2011 -->