Even though distributed memory parallel machines can usually have a much larger number of processors than shared memory machines, the only way to share information is by using messages. Because of these restrictions imposed by this communication model, some researchers developed special programming languages that support transparent data structure across processors and mimic the shared memory programming model in distributed memory systems. Linda [49], and Orca [23] are examples of such languages.
The approaches described below are solutions that do not use any special purpose emulation systems. The algorithms are hardcoded to the underlined architecture to achieve maximum performance.