Inconsistent timing analysis

mikebk

I'm experimenting with using ANTS to improve performance. I've noticed that method times seem to usually reflect elapsed time (as opposed to processor time). Is this correct?

One of the areas that may allow for optimization is a procedure for reading in configuration. For this one, in the source line level column, the line invoking the method shows a number that's much more than shown in the method summaries. Why would this be? (The line where the method is invoked shows 77 seconds while the entry in the slowest methods summary shows .8 seconds. .8 second seems much closer to the sum of the line timings than 77. The invocation does show 77 seconds in the slowest line summary so there's some consistency.)

How do export the source code line timing? I'm not convinced that the .8 seconds is the sum of all the source code timings in the mthod.

Thanks

Mike

Brian Donahue

Hello Mike,

The source code data cannot be exported like the all methods data can. This is something we are looking at for the future. ANTS Profiler is a 'wall-clock' profiler, so it does not reflect CPU time or ticks. It's all actual user time, including sleeps.

There are also two categories that can help narrow things down -- the time and the time with children. The method time is how long the method had spent running when not calling other methods. Time with children includes these methods.