Flightgear stops loading scenery tiles only when cumulonimbus clouds are rendered in warm sector or thunderstorm scenario. Is it a memory issue since 3d clouds are demanding? This issue is present since 3.4.0. I don't know if its present in earlier versions.
FGFS version:2016.2.1 build 2220
OS:Windows 10 build 10586.164
GPU:GTX 650
Graphics driver 364.51
From email "One way to reproduce it is to start with ufo,select thunderstorm scenario and fly straight up to around 35000ft. "
If after going straight up it appears as though all terrain did load fly away from that area at high altitude (or even low altitude) and terrain will stop loading very soon.
Sorry, can't reproduce.
My FGFS settings. These are the settings I've changed.
Here is a video showcasing the issue.
This appears to be expected behavior, assuming that the terrain was not previously downloaded. (needs confirmation from user).
TerraSync can be slow when it must download all the necessary data for it to appear in realtime.
Based on the attached video, I believe you were not patient long enough to allow the scenery to download.
Many users use TerraMaster or TerraSync.py to prefetch the scenery files before launching FG.
@Mihajlo, can you confirm whether you feel this is still an issue?
I didn't use terrasync. The terrain was already downloaded via terramaster, and when there are no Cb clouds the terrain loads normally. I managed to fix the slowdown by adding these properties per Wkitty's reccomendation:
#--prop:/sim/rendering/multithreading-mode=AutomaticSelection
#--prop:/sim/rendering/multithreading-mode=CullDrawThreadPerContext
--prop:/sim/rendering/multithreading-mode=CullThreadPerCameraDrawThreadPerContext
This problem doesn't occur on every machine, so it's hard to reproduce.
For all I can see from the evidence, this is a jam in the model loading queue on weaker systems (the queue gets simultaneously requested by cloudlets, terrain, scenery objects etc.) I doubt it has to do with weather, so I removed the tag.
Thanks for explaining this.I wanted to write this as well, but forgot the suspected cause, and I couldn't find it on the mailing list.
Collecting notes for myself: we definitiely have some situations where the tile-loader gets stuck. I'm not sure it's the osgDB thread vs the tile scheduler we have one layer above. I should add some watchdogs to check for stuck tile loads / missed requests.
Unfortunately machine performance / memory consumption does seem to be a factor, which suggests it might be timing or race-condition-y, which would suck to track down.
Cannot reproduce on my current system (HP EliteBook 8470p; 16GB RAM, 8x 3.6GHz CPU, AMD Radeon 7750M) but I remember it being very annoying on my old system (Acer TravelMate 5360G; 4GB RAM + 8GB swap; 2x 1.7GHz CPU, NVidia GeForce 520M GPU). Both systems had an SSD.