Pre-rendered Frames/Flip Queue Size?

[Guest]

On Windows, I played around with FQS in order to have no pre-rendered frames mainly to reduce input and visual latency as far as possible. I'm curious if something similar can be done on Linux?

[curaga]

The open drivers have no such queue. Patches welcome if you want one, but now it's all in sync with minimal latency.

The blob has one, but it's completely untunable, resulting in all the Source input latency complaints and ugly glFinish workarounds.