Performance question

[matrem]

Hi,

I have some questions regarding performances :

• When the documentation says :
  "tries not to block the main thread for longer than a certain time budget"
  It's only true for loading not instantiation ?
  So when calling InstantiateMainScene, we have a risk of render thread lag if object is to heavy?
• C# managed memory is used, so garbage collector can be called at any moment causing all threads to lag ?
  I experienced a big 3 sec unity freeze at gltf.load for a big model with many textures (for a total of 300MB), and the was during memory allocation/collection.

[atteneder]

Hi,

I have some questions regarding performances :

* When the documentation says :
  "tries not to block the main thread for longer than a certain time budget"
  It's only true for loading not instantiation ?
  So when calling InstantiateMainScene, we have a risk of render thread lag if object is to heavy?

That's true for the current version 4.x and before. In the main branch (or the upcoming 5.x version) instantiation is also async!

* C# managed memory is used, so garbage collector can be called at any moment causing all threads to lag ?
  I experienced a big 3 sec unity freeze at gltf.load for a big model with many textures (for a total of 300MB), and the was during memory allocation/collection.

I'm currently investigating and optimizing in this area wrt KTX textures and I'm on to something. What texture formats do you use and on which platform/hardware? Feel free to provide demo content. That always helps get a better understanding of user's needs.

[atteneder]

I've also documented the async instantiation progress here:

https://pixel.engineer/posts/gltfast-async-instantiation/

expect some improvement on complex scenes!