Gleitkommarechner.

摘要

The floating point computation unit for providing a sign/magnitude subtraction operation, uses propagate/generate logic (10, 11, 12) responsive to the subtraction operands (A and B) to produce intermediate and final propagate and generate outputs (P and G). First carry computation logic (13) is responsive to the final propagate and generate outputs (P and G) and to a carry-in bit (C63) to produce final carry outputs (C15, C31, C47) and a carry-out bit (Co). The latter outputs are used (14, 15, 17) to produce a first subtraction result (R). Second carry computation logic (18) responds to the intermediate propagate and generate outputs (P and G) to produce second final carry outputs (C min 15, C min 31, C min 47). The latter outputs and selected intermediate propagate outputs (P) are used (19, 20, 21) to produce a second subtraction result (-R). The carry-out bit (Co) then selects (22) one of the two subtraction results as the final subtraction result (R or -R). The floating point unit can further normalize an exponent result in conjunction with a mantissa result which is formed as a plurality of data slices. The number of leading zeros in each data slice is determined and each number of leading zeros is subtracted from the exponent result to produce a plurality of subtracted results. One of the subtracted results is then selected as the normalized exponent result.