We have a relatively large mesh that we author on the client side and write to Stream Cache on an in-memory device (not file system). The SC is then transported to the client side and rendered. We do the transport ourselves. What I am finding is that our performance on rendering has regressed from about 8 seconds (using OSG) to about 25 seconds using Hoops. Bulk of this 25 secs is spent in SC::Store::Model::PrepareStream and mainly in the compression. We are using CompressionStrategy::Fast.
Is there anything else we can do to speed things up?
For some perspective, the resulting SCS file (if written to file system) is about 50 MB.
You mentioned that you were using OpenSceneGraph (OSG) previously. To help us better understand and research the comparison, what is the analogue function in OSG to PrepareStream in Communicator?
OSG is not web platform based. So there is no need to generate Stream Cache kind of representation that you need to transport to the client side. The comparison is not fair because in the web world we have this additional step that is not needed on the OSG side. But from a users perspective, they observe a slow down in the time that rendering the scene takes and anything that can be done to speed up SC generation would help.