Split from #224: gtkperf
can cause some severe x11 Damage
storms where so many requests come in at once that our code struggles to keep up and the system latency goes through the roof.
Removing the many levels of indirection may help a little as we currently have to go from the UI event loop to CompositeHelper
to WindowModel
to Server
to ServerSource
to WindowSource
(s)
But ultimately, I don't think the python code will be fast enough to deal with these storms reliably, we should therefore move the damage batching to the event loop so that we only fire the python code once we know that the damage event is likely to result in a real action rather than a lot of churn just to postpone until later. Or maybe #505 is enough to fix the problem?
Closing, I think we manage Damage
storms well enough, the real problem with gtkperf
is #505
Milestone renamed
Milestone renamed
this ticket has been moved to: https://github.com/Xpra-org/xpra/issues/232