Motorola DSP96002 User Manual page 781

4. Unsigned Long Word Integer: 64 bits wide with unsigned magnitude representation. This stor-
age format can only be used in long (L) data memory space.
Long type integers can be moved to and from the data ALU register file. However, long integers can not be
directly used as input operands to data ALU operations. Long integers can however be results of data ALU
operations.
D.2.3 Integer Storage Format in the Data ALU
There are thirty 32-bit registers in which can contain integer words. However, data ALU arithmetic opera-
tions use the low portion of the register files as word source and destination operands. Long word integers
are only generated as results of integer arithmetic operations and are never used as source operands.
D.2.4 Integer Arithmetic
The integer arithmetic operations use the same arithmetic units in the data ALU as the floating point oper-
ations. These units consist of:
1. Adder: The subtract unit in the adder/subtracter unit described above is used for integer add
and subtract operations. It accepts two 32-bit integer operands from the low portions of the data
ALU source registers and delivers a 32-bit result in the low portion of the destination register.
2. Multiplier: The multiplier in the multiply unit described above also performs the integer multi-
plications. It accepts two 32-bit operands in the low portion of the data ALU source registers,
and delivers a 64-bit result in the low and middle portions of the destination register. Both
signed and unsigned multiplications are supported.
3. Logic Unit: The logic unit is responsible for the logical operations AND, ANDC, OR, ORC,
EOR, NOT, ROR. In addition, it performs the bit field manipulation instructions SPLIT, SPLITB,
JOIN, JOINB, EXT, and EXTB. The logic unit operates on 32-bit operands located in the low
portions of the data ALU registers. Results are also stored in the low portion of the destination.
4. Barrel Shifter: The barrel shifter in the normalization unit used for mantissa alignment in float-
ing point additions is also available for performing multibit shifts on integer (fixed-point) data.
Both single and multibit arithmetic shifts left and right and logical shifts left and right are sup-
ported.
MOTOROLA DSP96002 USER'S MANUAL D-29