Design of A Double-Precision Floating- Point Multiply-Add-Fused Unit with Consideration of Data Dependence

A floating-point multiply-add-fused unit with data dependence considered is proposed in this paper. It is implemented in three pipeline stages. The main improvement is that the intermediate computation results are fed back to the first stage and floating-point operations can be executed directly following their preceding floating-point operations without being stalled due to data dependence. The experiment results show that 20% performance increase can be attained at the cost of only 0.16ns time delay added to the critical path.