--Conversion to 32-bit unsigned integer
(operator `ru'): This conversion incorrectly raises the inexact exception together with the invalid exception (in single precision and double precision) in round-to-zero mode.
We have designed the algorithm to allow the selection of the data type of the input array components, which can be integer, unsigned integer
, float, double, long long or unsigned long long.
Adding the 16-bit unsigned integer
numbers F000h (hexadecimal representation) and 3000h generates a 17-bit result, or 12000h.
Each block in Figure 1 represents a 31-bit unsigned integer
and [as = [2.sup.31]p + q].
Therefore, if addition is done as unsigned integers
, the sum 4p+q can always fit into a 64-bit register.
Apparently, all dates must be represented by unsigned integers
of 16 bits, the maximum falling in the middle of 2079.