Hello,
I am using the GL_QCOM_motion_estimation on a Meta Quest 2 and the performance seems way off.
My input images are 1440x1584 (Quest 2 base resolution), and I am measuring 10ms of time spent in the estimator for each. The Quest 2 performance tools also confirms this number.
Per Qualcomm's own blog post: Improving VR performance using motion estimation OpenGL Extensions (qualcomm.com)
> For typical VR frame resolutions, the motion vectors can often be generated in under 1ms for both eyes, allowing the vectors to be created on-demand.
The number I am seeing is 20 times off!
I reduced resolution to 720x792, and I now see 2ms in the estimator (so 4ms for both eyes), which is unfortunately quite low resolution and not really usable.
Am I doing something wrong to cause such low performance? Are the numbers in your blog post erroneous?
Thanks.