.NET Framework Cost Metrics

Rico Mariani included a very interesting file in his latest performance quiz post which shows all the members of the .NET Framework and their estimated performance signatures. These performance metrics are based on the number of allocations in the call tree of each method. (More info on this metric here)

While certainly not the only performance metric to consider it would be a nice starting point for identifying expensive framework calls in performance sensitive spots in your code.  As Krzysztof Cwalina points out these would make for great FXCop rules to to warn when you invoke an expensive framework member.

[ plug ] If you've used one of the early Red Five Labs private betas you might've noticed we provide a small analysis tool that provide you with similar performance measurements of your application code (as well as framework code) after program execution as opposed to doing a static analysis of the code. Whether this tool will be included in release is not decided though. [ /plug ]