Parallel programming

While SIMD is good at increasing the performance of a single thread running on one core, it doesn't work across multiple cores or processors and its applications are limited. Modern scaling means adding more CPUs, not simply making a single thread faster. We don't just want to parallelize our data as SIMD does; we should actually focus more on parallelizing our processing, as this can scale better.

There are various .NET technologies available to help with parallel processing so that you don't have to write your own threading code. Two such parallel extensions are Parallel LINQ (PLINQ), which extends the LINQ operations you're familiar with, and the Task Parallel Library (TPL).

Task Parallel Library

One of the main features of the TPL is to extend loops to run in parallel. However, you need to be careful with parallel processing, as it can actually make your software slower while doing simple tasks. The overheads involved with marshalling the multiple tasks can dwarf the...