Creation of Vector-Topological Data Structures
Michael J. Mineter · 2020
This chapter explores the algorithms, data structures and software architecture which are required to support such operations on Multiple Instruction Multiple Data parallel computers. The data structures within a Worker must accommodate partial topology from all strips processed by the Worker, and information related to the strip borders to allow stitching and collation to occur. Pointers to the data structures corresponding to these objects can be grouped together within a transition data structure. The transition structures are used to identify the data along this moving border. Domain data structures do not duplicate data held in strips. The domain structures need to allow: Stitching is initiated by topology building, stitching and output within a Worker when no more important event requires processing, and when a new domain has been generated, either from processing of a new strip or from previous stitching. During strip processing this list is held associated with strip data; during stitching it is associated with domain data.