Distributed programs and subroutines for multiprocessors
Erik P. DeBenedictis · 1988
There seems to be a consensus that multiprocessors are harder to program than conventional computers. To change this consensus, this paper develops distributed analogies to important and well known programming techniques found in conventional computers — namely modular programming and variable scoping. This paper shows that with these techniques, truly parallel versions of some languages are possible — such as Unix (TM) shell programming. High level parallel languages could also be constructed where the program planning activities would be the same as conventional languages, although lower-level methods would be different. Within the scope of the issues addressed, this can make multiprocessors as easy to program as conventional computers.