I think I am missing something. I have not managed to fine tutorials covering this.
Here is what is happening. I have a simple scene. I load up a alembic. It is about 300 megs at about 200 frames. I then do a few very simple things to each frame of this alembic. Convert, add atributes and so on. Nothing over the top. Then I want to dump it on drive as bgeo. Great.
In theory this can very easily be multi threaded with tops. From what I understand all I need to do is dump down a “ROP geomatry out” inside tops. Point it to the last node in the network before I write out the cache and cook.
Yes this technically works. How ever it seems to want to load the entire 300 megs of the alembic up for each instance tops is running. So running 10 instances my memory spikes. When the instance is done with that frame it starts the next and reloaded the entire 300 megs again for this new instance.
I am sure I am doing something here. It cant work like this by default. What should be the way to set this up?