[TICKET]: Simulation tends to be unstable and needs to be sped up

[yogwoggf]

Context

It's a long standing issue, the FleX simulation is extremely unstable for several reasons.

• The timestep must be sped up in order to appear "realtime"
• The parameters will cause constraint instability relatively quickly, burdening creative freedom
• Parameters have no definitive unit (cm, hu, ft, in?)

My proposal is to shrink the local space of the entire fluid simulation by some arbitrary factor, and then apply the Source Engine's ambient conditions in hammer units, which are then transformed by previously mentioned arbitrary factor.

This should work as the entire force applied upon the particles will be relatively smaller than what wouldve been applied if we kept the same unit dimensions. In fact, FleX really has only ever been known to work stably when the particle radius is about 0.01-0.5, several times smaller than the typical Gelly particle radius.

Scope

• Ensure all simulation inputs are scaled by the arbitrary factor
• Ensure all simulation outputs are put back into the original unit scale

Acceptance Criteria

• FleX no longer needs to be arbitrarily sped up

Simulation Change

Simulation Correctness

Other

No response

[yogwoggf]

I think this issue will heavily use our new trunk-based development--it's not necessarily high priority but could potentially bring far more pros than cons. Basically, it might be in low development for a while

[yogwoggf]

Was doing some ideation, fairly simple

• Create a new scaleFactor in the SimData object. Should be a divisor, not a coefficient, i.e: 100 instead of 0.01.
• Simulation internally divides all incoming inputs by scaleFactor.
• Renderer internally passes scaleFactor from SimData into the constant buffers
• Shaders multiply incoming vertices by scaleFactor (this is why we used a divisor, a mul instead of a div for each vertex is great)

That should just.. work. Will try later.