Call ancestry and invocation counts are wrong

Questions about YourKit .NET Profiler
Post Reply
andulvar
Posts: 4
Joined: Tue Dec 14, 2010 7:35 pm

Call ancestry and invocation counts are wrong

Post by andulvar »

I tried profiling a Silverlight 4 application. I really like the output and the detail of the tool. However, I notice that call counts and ancestries are not reported correctly. I observe the following:

- I have methods that are called 60,000 times, but the "Invocation Count" shows only one invocation. Oddly, the callees of this method show up with more reasonable numbers of calls.

- I have methods that simply do not show up in the call tree at all. I know they are called - they are direct callees of methods that are in the call tree, and they are on the only control flow path.

- I have methods that are in the tree, but attributed to a different caller. That is, a method that I expect to be a callee of a "parent" is in fact attributed to the parent's parent method.

The first thing I thought was that I was somehow profiling an optimized application, but I confirmed that I am profiling the debug build of the app.
Anton Katilin
Posts: 6172
Joined: Wed Aug 11, 2004 8:37 am

Re: Call ancestry and invocation counts are wrong

Post by Anton Katilin »

andulvar
Posts: 4
Joined: Tue Dec 14, 2010 7:35 pm

Re: Call ancestry and invocation counts are wrong

Post by andulvar »

I understand the filters. I have left the filters at the defaults. I don't understand how filtering on System.*, etc. would cause the profiler to report the wrong call counts and wrong ancestry on the functions in my program, or why it would cause my own function calls to simply go unreported, as if they had not been called.
andulvar
Posts: 4
Joined: Tue Dec 14, 2010 7:35 pm

Re: Call ancestry and invocation counts are wrong

Post by andulvar »

I figured out the call counts, I think. I was seeing the call counts on one of three code paths to a particular method call. The sum of all code paths looks correct, so I think it was just me misinterpreting the result screen.

I still haven't found an explanation for the wrong parentage or the missing function calls. I'll try to find a more precise description of those.
Anton Katilin
Posts: 6172
Joined: Wed Aug 11, 2004 8:37 am

Re: Call ancestry and invocation counts are wrong

Post by Anton Katilin »

Could you please provide snapshot demonstrating the issue.

ftp://upload.yourkit.com/pub/
user ftp
empty password
Post Reply