Seven Concurrency Models in Seven Weeks

Wrap-Up

For some reason, data parallelism seems to be ignored in many mainstream discussions of parallelism. As you can see, however, it’s an extremely powerful way to dramatically improve your code’s performance and one that all programmers should have in their repertoire.

Strengths

Data parallelism is ideal whenever you’re faced with a problem where large amounts of numerical data needs to be processed. It’s particularly appropriate for scientific and engineering computing and for simulation. Examples include fluid dynamics, finite element analysis, n-body simulation, simulated annealing, ant-colony optimization, neural networks, and so on.

GPUs are not only powerful data-parallel processors; they are also extremely efficient in ...