Standard algebraic problems and bitwise parallelism

Considers there standard algebraic problems: solving sets of linear equations, computing eigenvalues and singular values, and approximating matrices. It is shown that they can be solved by using a class of high-throughput, bit-parallel architectures. To illustrate the concepts, three concrete designs are surveyed in which the complete trajectory from definition of a new algorithm to realization using high-level macros was fulfilled. In all three cases, it was possible to derive fully pipelineable algorithms of minimal complexity. Limitations of the method and an overview of a design trajectory and system which leads from behavioral specification to the realization of the system are outlined