android: Cobalt APK comparison script

[sideb0ard]

Script to compare memory and CPU across 2 apks.

Requires installing pip modules 'pandas' and 'matplotlib' to run.

$ tools/cobalt_compare_apk_mem_and_cpu.py --help
usage: cobalt_compare_apk_mem_and_cpu.py [-h] [--apk1 APK1] [--apk2 APK2] [--duration DURATION] [--outdir OUTDIR] [--url URL]

Monitor and Compare Cobalt APKs.

options:
-h, --help show this help message and exit
--apk1 APK1 Path to the first Cobalt APK
--apk2 APK2 Path to the second Cobalt APK
--duration DURATION Duration to monitor each APK in seconds
--outdir OUTDIR Directory to save plots and CSV files
--url URL YouTube URL to launch

Bug: 483129786

[github-actions]

🤖 Gemini Suggested Commit Message

---

android: Add APK memory and CPU comparison script

Introduce a Python script to monitor and compare the memory and CPU
usage of two Cobalt Android APKs. This tool automates the process of
installing, launching, and collecting performance data via ADB.
It generates plots and CSV files to visualize differences, aiding in
performance regression detection and optimization efforts.

Bug: 483129786

---

💡 Pro Tips for a Better Commit Message:

1. Influence the Result: Want to change the output? You can write custom prompts or instructions directly in the Pull Request description. The model uses that text to generate the message.
2. Re-run the Generator: Post a comment with: /generate-commit-message

[gemini-code-assist]

Code Review

This pull request introduces a new Python script for comparing memory and CPU performance between two Cobalt APKs. The script is well-structured and provides a useful utility for performance analysis. For easier dependency management, consider adding a requirements.txt file listing pandas and matplotlib. My review focuses on improving robustness, maintainability, and adherence to Python style guides. Key suggestions include fixing a critical bug that would cause the script to crash, making the script more portable by removing hardcoded paths, improving the reliability of CPU monitoring, and addressing several style guide violations such as missing docstrings and excessive line lengths. I've also pointed out some unnecessary delays that could be removed to make the script run faster.

[gemini-code-assist]

Parsing the output of top is notoriously fragile as its format can vary across different Android versions and device manufacturers. For example, the CPU column header S[%CPU] might not be consistent. For a more robust solution, consider parsing the /proc/<pid>/stat file to calculate CPU usage. This involves reading CPU time values at two points in time and calculating the percentage against the total CPU time elapsed. While more complex, it's a standard and reliable method on Linux-based systems.

[gemini-code-assist]

There's a hardcoded 10-second sleep to wait for the app to start. The code that follows in run_test_on_apk already polls for the process PID to appear. This fixed-delay sleep is redundant with the polling mechanism and can be safely removed to make the script faster and more efficient. The polling loop will handle waiting for the app to be ready.