Introduction to Concurrency in Programming Languages

Historical Context and Evolution of Languages 139

In many of the languages discussed thus far, one of the key discrepancies

was related to language support for distributing data amongst the processing

elements where concurrently executing threads or processes would act on it.

Some parallel languages, such as Cilk (used in later chapters and described in

Appendix C), focus on parallelism from the task and control flow perspective

with less emphasis on data. It was the sole responsibility of the program-

mer to write the necessary code to, say, distribute portions of large matrices

amongst a set of processors. The choice of data decomposition and distri-

bution is important for performance reasons to exploit locality. To remove

the responsibility of