In the previous blog, I explained the rationale behind vectorizing with Array Notations inside of the Cilk ‘for’. There are lots of ways to vectorize inside a for loop once you’ve used the Cilk ‘for’, but using this colon/bracket style syntax has been the easiest way for me to get vectorization without having to think about using AVX128 vs. AVX256 at all. That’s simply much harder to wrap your head around than just changing your data structures to use a different, compiler-friendly style of expressing code:"Hi compiler, please vectorize this stuff inside my for loop. Thanks!"

So you’ve made the decision to go with a language extension and finish the parallel coding quickly. You don’t have to spend a lot of time thinking about your algorithm and problem with Cilk, and it's not meant for people that do want to spend a lot of time thinking about it. For this PBB model you can simply learn the syntax and usage of the Cilk ‘for’ and quickly augment your serial ‘for’ to get task parallelism across cores. After that you can protect against data races and add vectorization later (I will talk about those things in subsequent blogs/videos).

So you've used Intel Parallel Amplifier to discover that a sizable percentage of your processing time is being spent inside of a very hairy for loop. Task Parallel or Data parallel algorithm considerations aside, your first reaction is, how can I get this for loop, at the very least, parallelized across some or all of the cores (note: these considerations vary depending on data size - there are cases where parallelizing across all cores does not warrant the overhead in distributing across those cores)?

Welcome to Visualize this! the show where we talk about game development. I have an interesting lineup of guests for this year, and if you have guests you would like to hear from, send me note.

My guest today is Barry Tannenbaum Software Engineer and Cilk Developer at Intel. Barry and his team recently released the Cilk PLus SDK and he spoke with me on its value proposition to developers.