I tested the cpu-usage branch.. overall it is a really good improvement! :)
About loading time with models of 500+ tiles: cpu-branch as such does not really reduce tile loading time. But when I applied this kind of test, it reduced loading time by 30%!
var _requestTimer;
function callAfterRenderFunctions(scene) {
// Functions are queued up during primitive update and executed here in case
// the function modifies scene state that should remain constant over the frame.
var functions = scene._frameState.afterRender;
for (var i = 0, length = functions.length; i < length; ++i) {
functions[i]();
if (_requestTimer)
clearTimeout(_requestTimer);
_requestTimer = setTimeout(function() {
scene.requestRender();
}, 3000);
}
functions.length = 0;
}
I am wondering why is that afterRender request there in any case? Is it really needed? I also tried removing it and it speeded up loading even more and I did not see anything breaking..