Algorithmic Fault Tolerance Using the Lanczos Method
Daniel L. Boley, Richard P. Brent, Gene Howard Golub, Franklin T. Luk · SIAM Journal on Matrix Analysis and Applications · 1992
This paper considers the problem of algorithm-based fault tolerance, and makes two major contributions. First, it shows how very general sequences of polynomials can be used to generate the checksums, so as to reduce the chance of numerical overflows. Second, it shows how the Lanczos process can be applied in the error location and correction steps, so as to save on the amount of work and to facilitate actual hardware implementation.