Performance Myth Busters (Android Dev Summit '19)

Let’s explore what myths we can bust about Android app performance! Are enums not OK? How many dex files are optimal? Is object pooling really a good thing? Can implementing Kotlin really impact app performance? Along with some best practices on performance profiling and benchmarking, we put these questions (and more!) to the test and find out which are true and which myths get…busted.

Presented by: Calin Juravle, Diana Wong, David Sehr, Eric Holk

Kotlin Sessions →…


  1. Why do you say there is not performance difference when the Kotlin version of Drive takes 0,002ms more to start up?

  2. If you have more than 400k LOC in Kotlin, then you will have more than 200% of compile time.

  3. Is there a white paper about those experiments? What are the percentiles and setup of those experiments?

  4. Don't think it was mentioned what min SDK was used for the multidex performance tests. My understanding is that legacy multidex (minSdk < 21) is much worse than ART's multidex, so would be interesting to see how this compares.

  5. 1. Using static allocated array is more efficient than using Pool.
    2. Dead code and other backwards capabilities most frustrated part of all Java-based languages when you come from C/C++ and realizes there is no macros at all

  6. This myths reminds me when as a teenager some one will came up with a weird think about a video game "Did you know you can make Shao Khan perform a fatality mid round?", where did kids come up with those? Where does devs come up with this?

Leave a Reply

Your email address will not be published. Required fields are marked *