Multiplicity

One of the great truths of application development is that the exotic eventually becomes commonplace. Time was that multiprocessor machines were almost exclusively made by a company called Cray, and most mere mortals never laid eyes on one. Certainly the idea of running code on multiple pieces of hardware, (whether in separate rooms, on separate chips, or on the same sliver of silicon) was science fiction for the average user just a few short years ago. Today it's becoming the norm. And developers must keep pace.

Challenges arise from the growing prevalence of cluster environments (see Cluster Computing: Leaps, Bounds, and Bottlenecks), as well as the multicore revolution (see Multicore Meets Virtualization). C++ is changing to meet those challenges, but change comes slowly, and only after much deliberation. Which is good—no one wants hasty, ill-considered changes to the standard library, or (worse still) to the language itself.

Sure, C++ programmers already have workable solutions for multithreading. But they haven't been part of a standard, and have left developers with a jumble of platform incompatibilities, and no guarantees from their compilers that anything will work reliably everywhere. That is changing, however. Threading is directly in the crosshairs of the C++ standards committee, and threading-friendly changes to the language are expressly intended to be a part of C++0X. There are some great ideas on the table. For an overview of what's been said so far, check out the dirt from the latest ad-hoc meeting.

Posted by Kevin Carlson at 05:01 PM  Permalink