/android-simpleperf_demo

android application projects using simpleperf

Primary LanguageC++

Examples of using simpleperf to profile android applications

https://android.googlesource.com/platform/system/extras/+/refs/heads/simpleperf-release/simpleperf/

simpleperf stat -e dTLB-loads,dTLB-load-misses,iTLB-loads,iTLB-load-misses,L1-icache-load-misses,page-faults -p pid

Table of Contents

Introduction

Simpleperf is a native profiler used on Android platform. It can be used to profile android applications. It's document is at here. This project is to show examples of using simpleperf to profile android applications. The meaning of each directory is as below:

simpleperf/                  -- contain simpleperf binaries and scripts.
SimpleperfExamplePureJava/   -- contains an android studio project using only java code.
SimpleperfExampleWithNative/ -- contains an android studio project using both java/c++ code.

Profiling java application

android studio project: SimpleExamplePureJava
test device: Android O (Google Pixel XL)
test device: Android N (Google Nexus 5X)

steps:

  1. Build and install app:
  # Open SimpleperfExamplesPureJava project with Android studio,
  # and build this project sucessfully, otherwise the `./gradlew` command below will fail.
  $cd SimpleperfExamplePureJava
  $./gradlew clean assemble
  $adb install -r app/build/outputs/apk/app-profiling.apk
  1. Record profiling data:
  $cd ../simpleperf
  $gvim app_profiler.config
    change app_package_name line to: app_package_name = "com.example.simpleperf.simpleperfexamplepurejava"
  $python app_profiler.py
    It runs the application and collects profiling data in perf.data, binaries on device in binary_cache/.
  1. Show profiling data:
  a. show call graph in txt mode
    $bin/linux/x86_64/simpleperf report -g --brief-callgraph | more
      If on other hosts, use corresponding simpleperf binary.
  b. show call graph in gui mode
    $python report.py -g
  c. show samples in source code
    $gvim annotate.config
      change source_dirs line to: source_dirs = ["../SimpleperfExamplePureJava"]
    $python annotate.py
    $gvim annotated_files/java/com/example/simpleperf/simpleperfexamplepurejava/MainActivity.java
      check the annoated source file MainActivity.java.

Profiling java/c++ application

android studio project: SimpleExampleWithNative
test device: Android O (Google Pixel XL)
test device: Android N (Google Nexus 5X)

steps:

  1. Build and install app:
  # Open SimpleperfExamplesPureJava project with Android studio,
  # and build this project sucessfully, otherwise the `./gradlew` command below will fail.
  $cd SimpleperfExampleWithNative
  $./gradlew clean assemble
  $adb install -r app/build/outputs/apk/app-profiling.apk
  1. Record profiling data:
  $cd ../simpleperf
  $gvim app_profiler.config
    change app_package_name line to: app_package_name = "com.example.simpleperf.simpleperfexamplewithnative"
  $python app_profiler.py
    It runs the application and collects profiling data in perf.data, binaries on device in binary_cache/.
  1. Show profiling data:
  a. show call graph in txt mode
    $bin/linux/x86_64/simpleperf report -g --brief-callgraph | more
      If on other hosts, use corresponding simpleperf binary.
  b. show call graph in gui mode
    $python report.py -g
  c. show samples in source code
    $gvim annotate.config
      change source_dirs line to: source_dirs = ["../SimpleperfExampleWithNative"]
    $python annotate.py
    $gvim annotated_files/home/yabinc/AndroidStudioProjects/simpleperf_demo/SimpleperfExampleWithNative/app/src/main/cpp/native-lib.cpp
      check the annoated source file native-lib.cpp.