Using Intel® Parallel Advisor 2011 to model the correctness of an annotated program is a very thorough and detailed process, and due to this, it adds significant overhead to the run time of the application. This overhead is caused by several factors, which you’ll learn about in this tutorial
To view the tutorial, click here.
VERY GOOD ARTICLE
This is a regular pearl of wisdom. Minimizing the performance penalty when running instrumented code has been a problem since the 1940's but is less so today. We used to call it an "intrusion penalty" but that now refers to a jail term. These suggestions are useful but performance is relative and if it runs faster now, it will run even faster when you pull your nose out of it. I am going to invoke Werner Heisenberg and say that just looking at it makes it run slower. Many times however, stepping back and looking at how you're doing things will yeild more significant results. Parallel programming is an "algorthmic shift" and often it's not tactics (tweaking your code) that wins the day but adjusting your strategy - your algorithms. A lot of stuff doesn't have to run in any particular order as long as it all comes out right in the end.