Running a pretty high res water and 'sand' (want to simulate matcha) mixing mpm sim. Have the default 10k max substeps on the solver but looking at the detail attribute seeing it's doing ~13k substeps. Is this normal? It's taking a bit too long per frame so curious as to why this is and if there's a way to optimize/clamp it.

Bringing it down to 5k does result in the attribute coming down to ~6k. So it's working but not a hard limit.

Other odd behaviour with the min/max substeps in MPM. Wanted to test some stuff on this same setup with a quicker sim so brought down the max substeps to 2500. The particle separation was lowered though for the test I wanted to do (0.0003).

Noticed the small particlesep with the lower max substeps eventually made the simulation unstable as it would crash at frame 3 (telling me in the error message to try messing with minsubsteps or cfl condition). But, and this what I find odd, putting the max substeps around 3333 (not sure of where exactly, 3333 works) it stabilizes the sim. When looking at the detail attribute of the resulting cache it's because it's doing around 12k substeps. The min/max setup is quite misleading or not being applied correctly...

There should be some warning that the substeps are being pushed way above the setted threshold to stabilize the sim..

What I found with stiff objects like rocks (stiff 1e+9), and having instability, is not matter what, raising like crazy the max subs will still make the sim crushing,
I found a better result decreasing the CFL which is indeed decreasing the maxSubstep used (I'm working with 100000)

• Didn't find any benefit working with the minsubstep...
  
• Didn't find any "release on pressure" adding global substed, where I was expecting to halve the error if the velocity/computation was clustered it out on different steps, but mpm is not quite working in that way
  
• MaterialCondition is changing to much the behaviour
  
• Particle collision with Velocity is more difficult to solve respect the moving particles option
  

What I'm playing with now, having different material is to have the most stiff objects with lower resolution/particle separation,

I tried MPM only once, but it left the Impression on me that it is hard to understand what the parameters do. Wouldnt it be possible by some clever math to boil them down to three more human friendly values like: stiffness, brittleness and moisture?

Cheers
CYTE