In order to solve the pipeline stall caused by the physical register source conflict in deep pipeline superscalar processor

a non-blocking instruction issue method by sharing one physical register to multi-instructions is presented.This method continues to distribute physical register even when the physical register source conflict happens by using the instruction issue queue to buffer the conflicting instructions and increasing the amount of physical registers that can be distributed.By using this method

the processor releases the issue window and improves the using parallelism of physical register.Experimental results show that compared with the traditional renaming method

to achieve the same performance

iteratively reusing method saves 27.3 % of the physical register resources.