vfAnalyst starts where other profilers end, by providing you with a clear picture of where your performance bottlenecks are. These are the places where you want to explore opportunities for parallelism.

Using your test data, vfAnalyst runs dynamic analysis to collect a wealth of data, most of which is unavailable in any other tool. Using this information, you can see where your performance needs the most help from parallel execution.

You can view performance down to the loop level.

But performance analysis is just the start. During profiling, vfAnalyst also watches what happens as pointer usage results in memory writes and reads, and then identifies and classifies data dependencies as "memory" dependencies, "compute" dependencies, and "streaming" dependencies.