24 replies to this topic

[Slydog43]

your on fire mukuste! Thanks for all that you do!

[jimmyfingers]

@ mukuste - So, to confirm, does this mean that your separate branch physics changes would allow for VP based VSync to be used without the previously random inconsistencies noticed regarding the higher speed setting flippers?  Also, it seems that even without VSync that some considerable improvement in diminishing (or even removing) variance would be attained, so will it benefit both situations?  Do you have a test version / build you could PM me to check out?

[mukuste]

JF:

 

VP9 has quite a lot of variance in its physics routines, some of it depending on FPS, some on other factors. Most of it is due to sloppy coding and not taking the proper timestep into account. In the current VP9, that variance can be reduced quite a lot by simply forcing the physics timestep to always be constant (in fact, the FPS dependence can be completely taken out of the equation this way). This is a simple mod and shown by my earlier posts in this thread. The remaining variance, as I have found out, is practically completely due to an inherently buggy implementation of friction; by removing friction, I managed to get very consistent numbers. The final piece of the puzzle is upping the physics framerate from 100Hz to 400Hz (or cutting the timestep in 1/4th, if you prefer). This should help to combat any stutter that was introduced by the additional temporal aliasing that using only fixed timesteps brings. It should also increase the overall fidelity of the physics simulation.

 

So, yes, getting vsync to work smoother is hopefully one of the benefits. But everyone will hopefully get more consistent aiming, especially, but not only, the people who had framerates on the lower end of the spectrum.

 

As for a test version, I plan to start a new thread on the whole physics topic after things with DX9 settle down a bit (as they seemingly already begin to do). I'll again have public test versions there. So I'll have to ask you to be patient for a while yet; anyway not everything is completely worked out in this current version, re-adding friction in a more sound way being an important task left.

Edited by mukuste, 04 March 2014 - 06:55 PM.

[StevOz]

I'm somewhat confused, does the dampening friction parameter actually apply to the ball vs all walls or just the flippers exclusively, in accordance with the elasticity setting?

 

There is also contact friction which I assumed was for the ball vs the playfield, which should be virtually zero anyway.  

[gtxjoe]

Added ball velocity measurements and standard deviation.

 Flippertest3_desktop.vpt   424KB   6 downloads