profiling .net c# code called via managed c++

We have a c++ application calling c# .net code via a unmanaged/managed "firewall" layer. I.o.w, the app is still build without /clr but uses a dedicated project with /clr where wrappers live, statically linked. I have tried the Ants profiler for the first time. It launched my .exe file ok, but all I get in the profiler are timings of managed and unmanaged context switches. Can I get it to profile the .net assemblies beeing called?
The Ants profiler is obviously affecting the application as the time it takes for the .net code to initialize has greatly increased so I'd say some "patching" has occured.
Regards,
conrad

Comments

  • Can I just check if you have selected "Wall Clock" timings or "CPU timings" (default). There's a drop down list on the results page (on the right). When profiling using "Wall Clock" timings you usually see transitions dominating. Have you been using the interactive timeline too? Using this, you can select data from only the code path you are really interested in- this helps to filter out "noise" often.
  • CPU/Wall Clock did not make any difference, but "Top-down (any method) did!. The Display defaults to Top-down (methods with source). Our solution does not contain the actual .net source although I am able to trace into it the source from the devstudio debugger.
    So, I suppose the next questions is: is there a way to tell Ants where to look for the .net c# source?

    Thanks for the prompt reply, you got me going!
    Conrad
Sign In or Register to comment.