Also, from one of the developers on a Scott Manley Deep Space Hangout, there are supposedly some changes in the physics code to make it moderately less terrible, resulting in much better collision response resolution, which in turn means sooner early-exits, more numerical stability, and faster physics overall. And thus the ability to actually build stations a little larger.
Or in more technical detail, a change allowed them to specify which of 2 parts is 'dominant,' which should always be the heavier one but up until this point was just the first in the tree. Recent changes (I think in Unity) have allowed them to specify the heavy part as dominant. Basically, physics engines are like trying to estimate a highly complex interaction formula. They detect a collision is occurring through some means, then need to respond with an appropriate response/reaction. Since the problem itself isn't really solvable in a lot of cases, it ends up using iterations of a refining method, hopefully converging to a more accurate solution. By properly defining the 'dominant' part, these methods can converge more quickly. Since the methods themselves involve error calculations, they also include early-out "good enough" conditions that stops the algorithm when it has converged to an acceptable level. So converging more quickly means that more accurate solutions can be found in cases where the max number of iterations previously was insufficient, as well as reducing the computation time required for those which can be found quick enough for the "good enough" condition to trigger.
TBH, I'm more excited about bigger stations and such than the actual features. :>
Author
Topic: Kerbal Space Program: Now Hiring Optimistic Astronauts for Dangerous Munission (Read 1506178 times)