A new method requiring O(log N) operations for the factorization of a symmetric tridiagonal Toeplitz matrix is presented. The method is shown to require O(N) operations in O(log N) steps using N processors for forward and backward substitution, and to exhibit a high degree of parallelism. The algorithm is derived using a generalized form of Gauss elimination which is applicable, with other operation-count bounds, to the general sparse matrix problem.
|Original language||English (US)|
|Title of host publication||Conference Proceedings - Annual Phoenix Conference|
|Number of pages||5|
|State||Published - Dec 1 1986|
|Name||Conference Proceedings - Annual Phoenix Conference|
ASJC Scopus subject areas