{"id":518,"date":"2022-01-13T15:45:06","date_gmt":"2022-01-13T23:45:06","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/performance-diagnostics\/?p=518"},"modified":"2022-06-29T11:10:50","modified_gmt":"2022-06-29T18:10:50","slug":"new-android-logcat-support-available-in-microsoft-performance-tools-for-linux-android","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/performance-diagnostics\/new-android-logcat-support-available-in-microsoft-performance-tools-for-linux-android\/","title":{"rendered":"New Android logcat support available in Microsoft Performance Tools for Linux & Android"},"content":{"rendered":"

We recently added support for parsing Android logcat<\/a> logs to the OSS Microsoft-Performance-Tools-Linux-Android project<\/a>. Logcat is a text-based dump of system & app messages on Android.<\/p>\n

This is what the logcat log would look like when loaded in Windows Performance Analyzer (WPA)<\/a> \"Graphical<\/p>\n

Logcat for performance<\/h2>\n

Logcat is useful in general, but why do we find it useful for performance work? Don\u2019t we already have an official performance tooling available with a full-featured and system-level Perfetto tracing<\/a>?<\/p>\n

The answer is yes Perfetto is great for deep level system tracing, but one shortcoming is that Perfetto doesn\u2019t cover the boot scenario very well. There is no Linux or Android deep tracing<\/em> that I know of that covers kernel boot. The Perfetto tracing starts after kernel sometime in usermode, and only captures at best 1\/2 to 1\/3 of total boot<\/a>. So Perfetto is great for analyzing run-time perf after the system is booted, but currently doesn\u2019t cover boot very well, even on the upcoming Android 13.<\/p>\n

Logcat in tooling<\/h2>\n

Why is it useful to have this logcat log supported in the tooling?<\/p>\n

If you are just looking at diagnostics or errors, looking at the log with grep or your favorite text viewer should suffice. However, when analyzing performance, a bunch of useful features become available that would be cognitively difficult for a human analyst to easily perform. More features will be covered below but initially some of the benefits are:<\/p>\n

    \n
  1. Ability to automatically display & measure relative time differences between log statements. This is useful for quickly finding big jumps \/ gaps in the logs indicative of a long operation.<\/li>\n
  2. Calculation of duration information (how long something took). You have to find which operations are long, which is not trivial with a search pattern looking at thousands of entries with different log formats and time formats. You also want to get quickly be able to jump to where something started without doing the duration math.<\/li>\n
  3. A visual UI showing the longest duration operations and those gaps<\/li>\n
  4. The ability to correlate \/ show on the same timeline other events. For example, displaying Windows ETW events if Android is running in a VM with Windows Subsystem for Android (WSA)<\/a>.<\/li>\n<\/ol>\n

    Walkthrough using Windows Subsystem for Android (WSA)<\/h2>\n

    In our example we will be looking at the boot of a Windows Subsystem for Android (WSA) VM.<\/p>\n

    In this post we will use the following tools and will walk through each of them<\/p>\n