ViewHolder Performance Testing

The goal is to test the performance of the ViewHolder pattern in the following scenarios:

1. Worst possible implementation; Always inflate the layout
2. Good implementation; Reuse the convertView when not null
3. Best implementation; Use the ViewHolder pattern

Build Configurations

Three build flavors are used which contain the different Adapter implementations:

1. Baddoge - Always inflates the row layout
2. Gooddoge - Recycles the view if convertView is not null
3. Bestdoge - Uses the ViewHolder pattern.

The two default built types are used:

1. Debug - Contains Debug.start/endTrace call in getView() to generate the Trace file
2. Release - No Tracing, what should be used if you want to run the monkey script to capture the fps of the ListView.

The test app is a ListFragment that contains 10,000 row items. The row item view contains seven total views.

Build and Run

./gradlew tasks to see what assemble and install tasks are available

Results

Test Description

All tests were performed on the a Galaxy S4 API Level 17 emulator using Genymotion. Each adapter implementation was profiled with Traceview. The Traceview files can be found in the traces folder. In addition to profiling with Traceview, a monkey script was used to calculate the frames per second of the ListView. The monkey script was ran 25 times for each adapter implementation which in turn was used to calculate an average fps.

Test Results

Results of running the test on a Galaxy S4 API Level 17 Genymotion emulator.

	Bad	Good	Best
Average FPS	46.455	47.145	49.501
Real Time/Call	8.328	0.832	0.715
*Real Time/Call in milliseconds