ABSTRACT
Data parallel computers provide more dramatic speed improvements, operating on hundreds or thousands of pieces of data simultaneously. Data parallel machines can be classified as single instruction, multiple data or multiple instructions, multiple data (MIMD). Linda is a simple set of programming language extensions to support coarse-grained parallel computing of the MIMD variety. While parallel computing is clearly helping business, industry, and science to solve some old problems faster, healthy skepticism is needed in evaluating new commercial disciplines enabled by massively parallel processing. A number of retailers and direct marketers are using parallel computing to detect subtle trends and relationships in data bases storing information on millions of customers. Retailers are using detailed analysis of transactions to help determine what items are often bought together. The key to programming parallel computers is adapting existing algorithms or creating new ones to make the most efficient use of a given architecture.
