There’s that wonderful and dear to my heart bug again. All day I sit simulating something and catch myself thinking that somehow it became slow. Restarting Max and everything is nice and fast again.
Max 2022. I thought we beat this all a couple of versions ago.
P.S. Eight times slower. You open scene after closing - viewport speed is very fast (cache scrolling), you sim again - and boom it is slower and slower and slower. Like 35fps → 24 and so on. Also great looking single water particles, that are not merging with main water over time 
Looks very nice!
yacht_a2.zip (157 KB)
In the end performance is 0.05mvoxels. I wonder, why nobody complains.
Didn’t ring any bells at first, but just found a regression in the nightlies which appeared on 27 Apr. If you are using a nightly from this date and after that, get a latest one (30765) and try again?
Sure will do! Btw what that regression mean? Some weird old buggy code or what?
Yup, regression = something that worked but broke
Oh, this programmer stuff of yours! 
I have another question - a closed google group has closed. They promised to make a section on the forum instead - but time has passed and I don’t see it. Is there any news on this ?
P.S. After 50 frames - second sim went from 5.2mvoxels to 2.46. Todays build. Feels like phx stop using cores. Third sim 2.06. Fourth - 1.73
Can you attach a log file? It would show what is taking more time between frames.
As for the google group, there’s not much secret stuff going on in Phoenix - everything is in the nightlies, so there is no point in hidden groups right now.
Hm, where’s the log file for max 2022 ?
I was asking about vray group mainly…
Ah, sorry, don’t know about the V-Ray group. The log files should go into C:\ChaosPhoenix now (former C:\PhoenixFD)
Here we go. Something like 7 tries.
ChaosPhoenix 3ds Max 2022_sim.zip (41.1 KB)
Thanks! I can see something weird with the emitting from sources. Do you have anything exotic in the sources, like discharge modifiers, textures, animated emitters?
I love doing weird stuff, you know I think But the scene seems very simple and I don’t see anything exotic here. Water emission + foam emission directly from the source.
Hey Paul, so in today’s nightlies there is some more logging info that could help narrow down the issue. Can you run the scene again when you have time and send over the new log? Does not reproduce on my machine with the scene from the start of the thread, so we gotta do it the slow way…
Hey Paul , measuring a liquid simulation in vox/sec is kind of misleading , this is not a 100% grid based simulation, the particles count is important as well
@Ivaylo_Katev In general - yes, but in this particular case the log shows something else is going on though…
What difference does it make? If the phoenix gives results in its units, which decrease over time. Why does he do it
Does it make no sense ?
So the phoenix is doing some meaningless thing and you know it and don’t fix it ? OMG !
Uhm, not sure what you are talking about. Are you replying to Ivaylo or me? For sure Phoenix is not doing anything meaningless
Can you run the sim again with a latest nightly and send over the log file I asked you about? I hope this will say more about what the sources are doing.
Cheers!
I just don’t quite understand why counting voxels per second is a bad way to evaluate something. After all, it’s clear that when it goes down every time you press the simulate button, something weird happens.
And yes, I will definitely download a new build, computer is busy right now.
Ah, cool, so in short:
The fire/smoke solver is grid based. It will do the same amount of calculations no matter if the grid is full of smoke or completely empty, because it will calculate the velocity field. So if you gradually fill the grid with smoke, it won’t really change the simulation speed between frames. Technically it’s a bit different, but overall this is what’s up
The FLIP liquid solver used particles for the liquid and a grid for the velocities. So the grid takes some time and the particles take some time to calculate. You can have a huge grid full of particles and you can have the same grid with very few particles, and the few particles will be faster. E.g. if you have a grid that you fill up with liquid, the more liquid you have, the slower each frame will be.
HOWEVER, your log file shows that something entirely different happens to the source emission - it’s getting progressively slower on your pc, and it seems to get slower even after you stop the sim and start it again, which is very weird and smells like something is up with the interacting geometry or the scene in general. Does not seem related to the grid/FLIP stuff above.
Hope this makes sense.
I was just saying that same scene and same particles count and I think, that it is totally fine to measure performance. No ? I suppose, that there’s something with underwater foam emission.
