Automated testing of graphical applications is needed to ensure that applications meet specifications and perform properly. OpenGL applications typically query the viewing device as to the API extensions supported and implementation-dependent limits. The generated rendering stream is dependent on the results of the queries. The number of possible configurations is combinatorially daunting.
A simple example demonstrates: Ten OpenGL ES 2.0 extensions were reported from a query of a Nexus 7 (Tegra 3, Nvidia GPU) running 4.2.2 Android. On an Ubuntu Linux workstation with an Mesa Gallium OpenGL stack, the number of extensions reported was 30. Surprisingly, there were no extensions in common between these two systems! In addition, there are 32 different parameters defining the OpenGL ES 2.0 implementation-dependent limits which are different for each device.
Repeated testing for a large number of devices is difficult. However, with Ascender’s technology, it is possible to automatically run the application on a large number of device configurations and capture the rendering stream. This procedure can be performed in the cloud for hundreds of instances of device configurations. The resulting rendering streams can create a graphics transcription of the test execution and be tested for proper operation. The graphics can be replayed to see the visual state of the application when a test fails.