Initial state and adaptive grid problem

[TomGore]

When starting render in Load&Start mode by loading frame 1000 to begin with smoke already extended in space (from the initial sim). There is a problem if adaptive grid was on initial smoke simulation - it forces loaded frame into the frame 0000 adaptive size, which is much smaller since it only started to be emited. What you see then is a large long trail of smoke scaled down to small box of first frames of new, actual sim. Setting this 2nd sim grid size to the identical size of the "initial" 1000th frame size would help, but this size has different propagation center and mess occurs when you have to match it into a specific emission points on the mesh. In one line: Load and Start mode is useless when working with Adaptive grids now. Or I am doing something wrong...

[Svetlin.Nikolov]

Indeed, Load and Start deliberately stretches the loaded cache file to the starting simulator size, so you would be able to load any resolution or cache size, even from a 3rd party software.

Thus, it does not know anything about the adaptive grid and this is by design. The Load&Start mode is intended so you would load the grid with some initial content, but not as a continuation of a pre-roll. What you could do would be to Restore the sim from the 1000th frame onwards and leave the frames before that as a pre-roll. You could also get rid of all frames before the nearest Backup Frame before 1000 so they don't take up space. Or you could even rename the last backup frame to 0000 and just restore from 0 - would that work for your setup?

Cheers!

[TomGore]

This is exactly what I do, including deleting files on the way, but I just got used to a normal "initial state" from other software, including realflow and fumefx.
Also, when looping a sequence with blending, is Advection Origin still required in Precise mode? Or loop blending doesnt use Precise mode and only retiming does?

[Svetlin.Nikolov]

Ah, let me clarify a bit more:
- Load and start reads a cache from any software, any resolution, stretches it to the grid size and regarless if it has any velocity or not, starts tge simulation from this state.
- Restore resumes a Phoenix simulation from a backup frame that has all the required info for the simulation to continue smoothly from this point on. This is not guarranteed to happen using load and start and especially for 3rd party caches - it cannot happen at all, so it's not the inention of load and start in the first place.

[Svetlin.Nikolov]

Let me check the code on looping and retiming just to be sure - off the top of my head I would say that precuse tracing won't be any good for looping since its idea is to find the best intermediate frame between two adjacent caches, while looping blends distant caches.

[TomGore]

Also, could you check if it is possible to successfuly use Playback Rate when in loop mode at the same time? I have some issues with this - phoenix starts flicking - switching frames on each frame...

[Svetlin.Nikolov]

Negative, sorry - simultaneous retime and loop overlap are not implemented yet. Just checked about the method used - for loop overlap since the velocity and precise tracing do not make sense, blending falls back to 'interpolation'.

Adding one more vote to looping and retiming in the TODO list...

[joconnell]

While you're at load and restore things, I see that the backup frame every nth frame doesn't start from the first frame of your max scene either. Not a biggy but I noticed that since we start our scenes at frame 1000, we get our first backup frame at 1005 so it must be taking multiples of 15 from frame 0?

[Svetlin.Nikolov]

Ah, yes, it's just taking multiples. It's gonna be a small change, until now it was done this way in the name of simplicity.

[cb LLC]

I agree there us a bug. We want to use load and start frame 500 and start it at frame 600. Unfortunately frame 500 changes so much it does not even look the same even though grid didn't change

[Svetlin.Nikolov]

You just use Load when you should be using Restore. I also updated the docs so it would be more clear and one should not be mistaken for the other.

Cheers!