problems with higher res FLIP sims

   7985   12   2
User Avatar
Member
143 posts
Joined: Feb. 2012
Offline
Im trying to do a large spash using FLIP.
Im getting pretty nice results at a particle separation of 0.06 (around 3.6 mill particles) but when I go to 0.04 in the particle separation (about 8.7 mill) I loose my splash completely! I understand that there should be some change, but it seems quite dramatic.
I have tried increasing the smoothness and changing the surface extrapolation to get back the splash but I cant seem to get a decent sim out of it.
Im using a fliptank setup with pretty much default settings. Using volume velocity with a velocity scale of 1.5, smoothing at about 0.2 with no viscosity, density or divergence.
Any tips of getting better hi res sims?
Thanks!

Attachments:
flip_test_001.jpg (39.1 KB)

User Avatar
Member
373 posts
Joined: March 2009
Offline
Any chance of posting a hip file so we can get a better look at what's going on under the hood? You may need to increase the max cells to Extrapolate on the Volume Motion / Solver tab of the flip solver as well maybe since your voxels are inherently smaller, you may need to increase this to compensate for fast objects. Also, I would take a look at some of the other fields, such as the collision field to see if everything is still coming in the same way between both resolutions… which they should be.
Ian Farnsworth
User Avatar
Member
143 posts
Joined: Feb. 2012
Offline
On further investigation it looks as if this is happening when i submit the higher res sim to the farm and use a pre-render python script to change the res there. Do I have to set the Initialize Simulation OPs checkbox??
Higher res seems to work fine locally… Ill do some more tests on the farm tonight
User Avatar
Member
373 posts
Joined: March 2009
Offline
You shouldn't need the initialize simulation checkbox since you are starting fresh on the farm, it's going to be initialized anyway. It's prob not a bad thing though, to check it. Also, do you have all of your collision geometry cached and accessible to the farm? I'd suggest it, as opposed to generating it on the fly. It can also narrow down any potential issues you might come across.
Ian Farnsworth
User Avatar
Member
143 posts
Joined: Feb. 2012
Offline
I have my collision geo cached to a bgeo.gz sequence and it is colliding fine, as I can get a lower res sim off the farm fine.
I have removed the pre render script and I am still getting sims with no splashes when simming 8.7 million particles on the farm. Memory hits about 4-5GB and the machines have 24GB. Maybe some houdini cache settings?
I have loaded the scene up on one of the farm machines and seems to sim ok in houdini, but running the sim through hython on deadline gives bad results. (I have modified the hython script for deadline to run rop.render once for the frame range, not once every frame which is how it came shipped)
User Avatar
Member
373 posts
Joined: March 2009
Offline
Sounds like you're on the right path for debugging at least. I haven't used deadline and Houdini, but I could see how that default set up of render each frame (when doing a sim) may cause issues for a sim's cache. I've set up hython / hscript for our own renderfarm software and haven't had any issues, setting it to render once should work fine.
Ian Farnsworth
User Avatar
Member
143 posts
Joined: Feb. 2012
Offline
Ok, it seems to be running the render command from python/hython gives totally wrong simulation results.
Im trying to replicate the scene with with a simple setup but im having trouble getting it to break.
Maybe I need to rebuild my scene
User Avatar
Member
143 posts
Joined: Feb. 2012
Offline
Im still having massive problems with this. Its like the cached out collision volume drops its velocity once the particle count gets over 7million.

Time to randomly change parameters till it works!!!
User Avatar
Staff
818 posts
Joined: July 2006
Offline
What do the Cache settings on the DOPnet look like? If you're using DOPs collisions (i.e. your collision object is a Static Object), then DOPs needs the previous frame in the cache to compute good collision velocities, especially while substepping the FLIP sim.

It's possible that if you have really small cache settings, then that one frame is not fitting in memory and is being cached to disk when running in the Houdini UI. But Cache to Disk defaults to Off for non-interactive sessions like hython or hbatch, so you won't get good velocities when simulating on the farm.

Just a theory. What do those parameters on the DOPnet look like?

Edit: another thing to try is to turn off Allow Caching on the Creation tab of the FLIP Object, since it's really only important that the collision objects are cached.
User Avatar
Member
143 posts
Joined: Feb. 2012
Offline
yesss! I had my cache set to 500meg, settings to 5Gig worked!
Thanks Johner!!!!

On that note, whats the best way to use the cache amount? Di I just crank it and allow enough headroom for the OS and the houdini scene? (so maybe set it to 15Gig on a 24Gig machine?) Is there an easy way to get feedback about it running out of ram or over allocating it (if it even does that!)

I tried disabling the cache an that didnt work.
Also, from some of my tests last night I found that adding a trail sop to the collision geo before it goes into dops and changing the collision velocity from volume to point worked also. I guess this is a better method if I have a constant point count on my collision geo? (faster, less ram, more accurate?)

Anyway, thanks for the help! now to crank the particle count!!! weeeeee!
User Avatar
Staff
818 posts
Joined: July 2006
Offline
Glad to hear things are working now.

On that note, whats the best way to use the cache amount?

I'd recommend turning off Allow Caching on the FLIP FluidObject, since then you'll never risk running out of cache due to particle count. I also usually turn off Allow Caching to Disk on the DOPNet, since 1) it can be slow once you hit the disk, and 2) you're not surprised when things work differently in the UI from hbatch or hython.

With those settings a 1Gb cache or so should usually be plenty, since you only care about caching the collision objects.
User Avatar
Member
143 posts
Joined: Feb. 2012
Offline
cool, thanks for the tips!
So is the cache just there for collision objects and not the actual sim?
User Avatar
Staff
818 posts
Joined: July 2006
Offline
JordanWalsh
cool, thanks for the tips!
So is the cache just there for collision objects and not the actual sim?

No, it's there for the sim, and very helpful with medium-sized simulations. But past a certain size simulation, there's not much chance that it will all fit in memory (even on a 24Gb machine), at which point it starts paging to disk and gets pretty slow.
  • Quick Links