> (note: this post is mostly about data-parallel programming, or other parallel optimizations for computational performance purposes. There are other reasons to

> (note: this post is mostly about data-parallel programming, or other parallel optimizations for computational performance purposes. There are other reasons to use concurrency, but I’m not sure my argument will hold in most of those cases.) Respectfully, I think the usual claim is that concurrent programming is hard, not that parallel programming is hard. So, while the article is interesting in its discussions of the performance trade-offs involved in implementing data-parallel algorithms, it's arguing against a strawman, I think.