In a similar vein to Build Performance Results this page aims to measure the performance of Xamarin.Android across different versions of Visual Studio. The goal is to get a higher-level picture than what we can achieve from profiling in MSBuild.

Using an internal Microsoft tool, we are able to get timing information out of Visual Studio. This timing gives a better overall picture of the devloop, as well as the exact experience developers get inside the IDE.

In each timing, I am merely pushing the "play" button (F5). We have timings for:

• Preparation Time: Build + Deploy, the time it took for the Xamarin.Android MSBuild targets
• Launch Time: app started + Mono debugging session started
• Total Time: Preparation Time + Launch Time

From the table, what are first/second/third?

• First: start a debugging session after an initial checkout, after nuget restore.
• Second: stop, new debugging session with no changes.
• Third: stop, modify a XAML file, start debugging again.

I used the same projects from Build Performance Results with a physical Google Pixel 3 device connected via USB.

This is a comparison between:

• Visual Studio 15.9.5, Xamarin.Android 9.1.5
• Visual Studio 16 Preview 2, Xamarin.Android 9.1.103.6

The results seem to support what we are seeing when profiling MSBuild command-line, such as Build Performance Results. There is an element of the timings being better than command-line, since MSBuild.exe is reused and other build-related performance caching you get in an IDE, but not command-line.

We think there might be a ~1 second performance regression in Launch Time in 16.0 P2. @jonathanpeppers will follow up with the IDE team to figure out what is happening there.