A parallel version of a fast algorithm for singular integral transforms (Q1978039)

From MaRDI portal
Revision as of 17:45, 19 February 2024 by RedirectionBot (talk | contribs) (‎Changed an Item)
scientific article
Language Label Description Also known as
English
A parallel version of a fast algorithm for singular integral transforms
scientific article

    Statements

    A parallel version of a fast algorithm for singular integral transforms (English)
    0 references
    0 references
    0 references
    7 June 2000
    0 references
    The authors discuss fast algorithms for the evolution of singular integral operators, in particular the singular integral transform \[ T_m h(\sigma)=-{1\over \pi} \iint_{B(0;1)} {h(\zeta)\over(\zeta- \sigma)^m} d\xi d\eta,\qquad \zeta= \xi+ i\eta \] which arises in solving Beltrami's equation. Here \(h\) is a complex valued function of \(\sigma\) defined in the open unit disk \(B(0;1)\). These fast algorithms are based on a fast Fourier transform and some recursive relations. The authors develop a parallel version for these fast algorithms by reformulating in a novel way the inherently sequential recurrences. The resulting method uses only a linear neighbor to neighbor communication path is therefore suitable to many distributed memory architectures. The algorithms are worked out in great detail, including data distribution issues, complexity computations and numerical tests.
    0 references
    singular integral transform
    0 references
    fast algorithm
    0 references
    parallel processing
    0 references
    distributed memory
    0 references
    pipelining algorithm
    0 references
    numerical examples
    0 references
    Beltrami's equation
    0 references
    fast Fourier transform
    0 references
    complexity
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references