-
Notifications
You must be signed in to change notification settings - Fork 3.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
RequestScheduler starves imagery requests for 3D Tiles #5954
Comments
Two ideas for fixing this:
|
Not sure if it's related, but the same thing seems to happen to terrain |
@hpinkos I would be really surprised if it weren't the same exact problem. |
Though the priority ordering could still be better and I have a possible fix coming. |
Terrain/Imagery is definitely waiting on a worker, but keep in mind that Chrome will only load so many files at a time, so the fact that JS files are being loaded is starving other requests too. So I definitely expect performance to be worse with the unbuilt version. This is one of the reasons we should always do profile/performance testing against the built version of Cesium; it's a huge difference. |
Hm so I guess the problem is that 3D Tiles requests are fighting with JS requests, which means it takes longer for imagery to begin. Is that something that we can fix? |
Not really (in code anyway). HTTP/2 fixes it because browsers will no longer limit to 6 requests at once. |
This was pre-streaming overhaul and I would consider it OBE. |
The NYC demo loads all 3D Tiles tiles before loading any imagery. This is really noticeable on slow connections or slower browsers (Firefox and Chrome throttles show the issue).
http://localhost:8080/Apps/Sandcastle/index.html?src=3D%20Tiles%20Feature%20Picking.html&label=All
Our theory is that since the camera starts at the location of the tileset, the position of the root imagery tile is so far away that it loads it last in the queue. We think something similar would happen with a large 3D tileset or terrain as well.
This makes some demos look really bad during load, but I'm not sure what the level of effort (or priority) for fixing this should be compared to other efforts.
CC @pjcozzi @lilleyse
The text was updated successfully, but these errors were encountered: