I have downloaded the Intel* Parallel Studio beta and worked with it for parallelizing few test applications such as Delaunay triangulation and a demo Finite element analysis application. It works great and is extremely useful. I found the amplifier extremely useful for identifying hotspots and also for checking concurrency levels of the program.
The clarification that I need is about the concurrency index that is displayed in the summary window :
After my application was parallelized, I get a different concurrency index each time I run it. This is mainly because of the time I take to actual give the inputs. If I wait for some time before I give the inputs and start the computations, the total time for which the application was running increases and this bring the concurrency index down. In my opinion, this is quite misleading since the amount of concurrency during the computations is the same.
Can you please tell me how we can measure the concurrency only during the intensive computations ?
Also can you please tell me how I can get to know more about the Intel* Parallel Advisor. The objective of this tool interests me a lot since all hotspots need not be necessarily areas that need parallelization.