Hi,
I have a heavy FLIP simulation, my machine has 64 GB RAM and it can barely finish the sim! Anyways, the sim result is stored as surface and vel VDB and packed geo for particles.I using these VDB and particles for whitewater simulation, however it's terribly slow (e.g. it takes couple seconds just to create source particles for whitewater sim).
My question : is there a way to simplify the results coming out from FLIP simulation? I've tried using VDB LOD but it's seems to be broken (it changes volumes names randomly and nothing works after), also I've tried VDB Resample, and it did lower the VDBs resolution, but I got the same particles count which still makes whitewater sim very slow??
Simplifying FLIP simulation
1899 3 1- catchyid
- Member
- 39 posts
- Joined: Feb. 2016
- Offline
- Enivob
- Member
- 2540 posts
- Joined: June 2008
- Offline
It might be better to control the voxel count before generation rather than after the fact.
On the FLIPObject create a new float channel and paste this expression into the field. The expression assumes default naming of nodes so adjust as needed for your setup.
This expression creates a scientific notation number based upon the particle separation and the size of the flip container. Values on the right side of the + symbol represent the millions of voxels multiplier. So 06 is times a million, 07 is ten million and 08 is hundreds of millions. If you are 09 or larger (billions), you may want to reduce the voxel count by increasing particle separation. I typically strive for 250 million for final but work lower when developing the effect.
This GIF reports the voxel estimate of 800,000 first. Then shows 8 million then 37 million. Being able to accurately dial in voxel counts helps control performance and memory usage a lot.
On the FLIPObject create a new float channel and paste this expression into the field. The expression assumes default naming of nodes so adjust as needed for your setup.
ch("../flipsolver1/limit_sizex")/ch("particlesep")*ch("../flipsolver1/limit_sizey")/ch("particlesep")*ch("../flipsolver1/limit_sizez")/ch("particlesep")
This expression creates a scientific notation number based upon the particle separation and the size of the flip container. Values on the right side of the + symbol represent the millions of voxels multiplier. So 06 is times a million, 07 is ten million and 08 is hundreds of millions. If you are 09 or larger (billions), you may want to reduce the voxel count by increasing particle separation. I typically strive for 250 million for final but work lower when developing the effect.
This GIF reports the voxel estimate of 800,000 first. Then shows 8 million then 37 million. Being able to accurately dial in voxel counts helps control performance and memory usage a lot.
Edited by Enivob - Sept. 7, 2017 08:31:53
Using Houdini Indie 20.0
Windows 11 64GB Ryzen 16 core.
nVidia 3050RTX 8BG RAM.
Windows 11 64GB Ryzen 16 core.
nVidia 3050RTX 8BG RAM.
- tamte
- Member
- 8591 posts
- Joined: July 2007
- Offline
- catchyid
- Member
- 39 posts
- Joined: Feb. 2016
- Offline
Thanks Enivob for your help and thanks for the expression
Thanks Tomas,Yes, I am using fluidcompress.
The problem is the sim result is approved and I “should not” resim it, so my goal is to add whitewater only…I can resample the VDBs to lower their resolutions, but I have no clue how to down sample the particles packed geo (actually, I don't understand them much…)
Thanks,
Thanks Tomas,Yes, I am using fluidcompress.
The problem is the sim result is approved and I “should not” resim it, so my goal is to add whitewater only…I can resample the VDBs to lower their resolutions, but I have no clue how to down sample the particles packed geo (actually, I don't understand them much…)
Thanks,
-
- Quick Links