Replies: 2 comments
-
Yeah, I've been looking at it for the past couple of days (i.e. right after it was published lol), it looks very impressive! I started drafting a basic implementation yesterday, so I'm hoping to have some simple demos running soon-ish. They are running it on the GPU, so I'm curious how well it works on the CPU. With our existing solver, my plan was to implement graph coloring (I have a working prototype locally already) and to then use wide SIMD to speed up the solver further, but I'm not yet sure how well SIMD works with AVBD since the algorithm is structured quite differently. I think you could maybe use SIMD for the dual loop, but I'm not sure about the primal loop... we'll see! I'm just experimenting for now, but once I have it working, I'll definitely make a bunch of test cases to compare it to our existing solver and share the results somewhere. If it ends up consistently outperforming our existing solver, who knows, maybe we'll switch :) |
Beta Was this translation helpful? Give feedback.
-
Interesting stuff for sure - is it possible for avian to run it on gpu? |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Check out Augmented Vertex Block Descent (AVBD) paper. It seems to be pretty stable. The ideas presented may help the future of avian physics.
Beta Was this translation helpful? Give feedback.
All reactions