Floating-point units configured to perform fused multiply-accumulate operations on three 128-bit extended operands, their methods, programs, and systems.

摘要

PROBLEM TO BE SOLVED: To provide a unit for performing a fused product sum floating point operation on a 128-bit extended operand. A floating point unit (10) configured to perform a fused product sum operation on three 128-bit extended operands (100, 102, 104), comprising: (i) 113×113-bit multiplication. (14), (ii) left shifter (18), (iii) right shifter (20), (iv) selection circuit (24) including a 3 to 2 compressor (25), and (v) selection circuit A first feedback path (36) connecting an adder (26) connected to the data flow from (24) and (vi) a carry output (91) of the adder (26) to a selection circuit (24). And (vii) a second feedback path (38) connecting the output of the adder (26) to the shifter (18, 20) for passing the intermediate expansion result (86) through the shifter (18, 20). A floating point unit (10). [Selection diagram] Figure 1