Algorthm based fault tolerane or algorithmic fault tolerance is an efficient scheme to improvethe reliability of parallel architectures used for real-time digital signal processing environments and other computation-intensive tasks. In this paper

a novel generalized weighted checksum encoding scheme with high error-correcting capability is taken into consideration

furthermore. a fast error detecting/error correting algorithm is proposed. Finally

the application employing the new encoding scheme in fault-tolerant matrix arithmetic is introduced.