Parallel programming consists of four distinct phases: finding the parallelism, writing the code, debugging the code, and optimizing the code. In my opinion, frameworks like the Apple’s Grand Central Dispatch and Intel’s TBB only help with writing the code; they do not help with finding parallelism, or debugging, or optimizations (e.g., false-sharing, thread waiting, etc). I think that the difficulty in finding the parallelism , which can be an insurmountable barrier for many inexperienced parallel programmers, is often underestimated. In this post, I try to explain this challenge using a couple of parallel programming puzzles.
Continue reading “Parallel Programming: Why new frameworks only solve a part of the problem?” »
