Hi,
This is a very performance related request so I totally understand if it’s not your top priority.
I have a very efficient job system render loop that use ALINE for large scale debugging, how large? Well…
At this point I realize we are bottlenecked not by the rendering, but by the SetVertexBufferParams
calls. By how much?
As you can see from the peaks, the more my view frustum move around, the more SetVertexBufferParams
cost. Beyond that, we pay a constant ~6ms for SetVertexBufferData
.
These calls can be jobified with MeshData API, my guess is you are aware of them but haven’t got time to investigate (I don’t like how these API work but at least they are fast).
So it’s something for your consideration