So when loading an avatar from Oculus, OvrAvatar sends off a specification request, gets either assets or asset ids back (not sure which yet), then loads them using ovravatarsdkmanager, stores them in ovravatarsdkmanager's asset cache, and draws on that cache to build the avatar. Just a quick background.
On Android, there is a major CPU bottleneck C side while waiting for the Specification Request to return. The slower the internet, the longer this bottleneck (cuts my framerate to 30fps), the point being that the length of bottleneck is highly variable even at constant CPU frequency. To me, this seems like strange behavior. If it was something tangible happening, like converting texture formats or something, it would be a constant length bottleneck, but it's highly dependent on internet speed and so must not be doing anything like that. For reference, the slowdown can be anywhere from non-existant to 3-4 seconds.
Does anyone know what's actually going on when you send a spec request? And has anyone come upon a solution for eliminating what seems to be an unnecessary slowdown?
