Study And Analysis Of The Leading-one Prediction Algorithm For Floating Point Operation

With the increasing requirement to the accuracy and speed of the floating point operation in all kinds of computer fields,the high performance floating point arithmetic gradually becomes an urgent demand.Lots of researches to the performance of floating-point arithmetic show that the leading-one-prediction(LOP)algorithm is always the key path in the floating point operation.Firstly,this paper studies and analyzes the traditional LOP algorithm which is aiming at the operation of two operands.Then,based on the research of traditional LOP algorithm,a LOP algorithm which can directly handle three operands is designed.This algorithm can apply to the situation which three operands are operated in the multiply-add component.What's more,this algorithm effectively shortens the circuit delay and reduces the circuit area.So,the overall operating efficiency is improved.Secondly,the research to the pre-encoding error correction algorithm of the three operands LOP algorithm is the innovation point of this paper.The rule of pre-encoding algorithm is the root to produce the one error.However,the three operands LOP algorithm designed can not judge the produce of the one error.Through the induction and summary of the pre-encoding rule,an error correction encoding rule is designed.Then,through the merge operations to the result of the designed error correction encoding rule,the judgment which shows whether the one error will be produced can be received.Finally,in this paper,the VHDL programming language is used to design the hardware circuit on the Quarters platform,and the simulation is carried out to prove the correctness of the proved algorithm.The analysis to the simulation results of the hardware circuit shows that,the algorithm in this paper can effectively shorten the delay and reduce the circuit area and power consumption.