This may be something only one of the driver devs can answer, but here’s the scoop. I’m currently generating occlusion querys for meshes at initialization time, and starting/ending the queries at runtime. This results in a fairly large number of occlusion query objects being persistent for the run time of the application - is there a penalty for this, or a ‘break even’ point, of where generating the query objects during runtime may be faster?
Thanks!