[Bf-committers] Blender developer meeting notes - 2020.1.6

[Jacob Merrill]

>
> what about accelerating picking, (BVHTREE stuff) - and accelerating
> writing data - (SIMD stuff)


 from what I understand
"lowest level tree" faces own sub faces, each subdivision level adds
another layer of depth to the quadtree.

'picking' the low level faces then automatically scrapes the next level and
beyond and puts it in a bucket with it's distance from the pick point
already computed.

Kdtree returns points in order of distance, I am not sure about bvhtree -
but having the quadtree do this could make it fast no?


when a 'stroke' happens it can roll through this data in order per "low
level face" and apply equations to the data (sculpt etc) and bail each time
the distance is too far
so initially we are not comparing the point vs every other point, instead
we are comparing the sculpt point only to the lowest level faces.