My understanding was that pipe->clear, blit, and resource_copy_region
should not increment query values (except for TIME_ELAPSED).
Currently, all radeon drivers implement this in the same way as this
patch does (it's done more efficiently though), but I've been meaning
to get rid of that and just t
I think I brought this up recently (perhaps this is even in response
to my comments about this)... my suggestion was to introduce a draw
flag that says whether the draw should be counted.
mipmap (and blit) should be using pipe->blit, which should never get
counted anyways, so all that's left are c
glClear, blitting, mipmap generation, etc. implemented with quad drawing
should not effect queries like occlusion count. To accomplish that, this
patch implements a simple pause/resume mechanism for active queries.
We now keep a list of active queries. When we "pause" queries we actually
end the
