Processing sets on a SIMD machine
Alberto Baudino, Giancarlo Colla, Giuseppe A. Marino, Giancarlo Suci · 1995
Writing programs that really exploit the advantages of parallel computers is not easy.The programmer must know their architecture details and choose an algorithm that is suited to the kind of parallelism implemented by the machine.Impemtiue languages greatly feel the effects of the above mentioned problems.They lay complitely upon the programmer the responsibility of choosing the proper algorithm and of implementing it.Declarative languages does not define the operation sequence needed to solve a problem, but they define the problem itself.It is up to compiler and executor to divide job into different processes and to distribute data on machine processors.Therefore they seem to be well suited to be implemented on parallel architectures.For this reason we have chosen to implement a subset equational language (SEL) on the Connection Machine 2, because the set data structure has a high degree of implicit data parallelism.This paper describes in details the mechanism by which sets are processed by the Connection Machine 2.