Increasing soft error rates make the protection of combinational logic against transient faults in future technologies a major issue for the fault tolerance community. Since not every transient fault leads to an error at application level, software level fault tolerance has been proposed by several authors as a better approach. In this paper, a new software level technique to detect and correct errors due to transient faults is proposed and compared to a classic one, and the costs of detection and correction for both approaches are compared and discussed. ©2010 IEEE.