I have a situation with flip sim where every time i write out / cache there is a change in particles positions in some areas of sim.
My scene particles have somewhat moderate velocities with wide variation in viscosity ranging between (0.000 to 1.000)*40 in few decimals places.
Although overall sim shape remains the same but there is a visible change in secondary shapes, fine tendrils it makes.
My suspicion is on the viscosity solver as i bypass viscosity calculations the error goes away. 64 bit float precision doesn't help the case either.
I am aware of rounding errors that are inherent in floating-point computation but have no idea of how is viscosity solved.
Has anyone experienced it behaving it non-deterministic ?
Thought to ask around before submitting as a bug.