Previous: The special case of block tridiagonality
Up: Block factorization methods
Next: Blocking over systems of partial differential equations
Previous Page: The special case of block tridiagonality
Next Page: Blocking over systems of partial differential equations

Two types of incomplete block factorizations

One reason that block methods are of interest is that they are potentially more suitable for vector computers and parallel architectures. Consider the block factorization

where is the block diagonal matrix of pivot blocks.

Making the transition to an incomplete factorization we can replace the diagonal of pivots by either the diagonal of incomplete factorization pivots , or the inverse of , the diagonal of approximations to the inverses of the pivots. In the first case we find for the incomplete factorization

and in the second case

We see that for factorizations of the first type (which covers all methods in Concus, Golub and Meurant [55]) solving a systems means solving smaller systems with the matrices. For the second type (which was discussed by Meurant [152], Axelsson and Polman [20] and Axelsson and Eijkhout [15]) solving a system with entails multiplying by the blocks. Therefore, the second type has a much higher potential for vectorizability.