Optimizing a simulation

See caching in the section on importing and exporting simulation data.

If you have a very large simulation that needs too much CPU or memory, but has discrete parts with one-way relationships (for example a group of RBD objects which push around smoke, but are influenced by the smoke), you can break up the large simulation into separate simulations.

For example, you can run the RBD objects as one simulation, and cache the results to sim files. Then in a separate simulation you can import the RBD object sim files and set up one-way affector relationships with the smoke.

You can also break up a large simulation by time. You can run the simulation up to a certain point, and save the entire simulation state to a sim file (using a File dynamics node at the end of the network). Then, you can use the same network to run the second half, but in the Initial state parameter of the DOP Network container node, load the saved midpoint sim file.

The DOP Network container node has several parameters on the Simulation tab to help optimize simulations:

Cache simulation

This lets you turn off caching completely. Turn this option off is saving memory is more important than speeding up interactivity in the viewer.

Cache memory

this is the maximum amount of memory Houdini will use as a simulation cache, in megabytes. The default is 500 MB.

Cache substep data

This includes sub-step solver information in the cache. This is useful if you want to do sub-frame motion blur.

Allow caching to disk

When the in-memory simulation cache is full, Houdini will use disk space. The amount of disk space Houdini will use when this option is on cannot be limited.

Provide data hints

This option checks that the data attachments you make in the network make sense. When the network is finished, or if you don’t want or need the warnings, you can turn this off for a small performance boost (though it might be noticable in a very large simulation with tens of thousands of objects).

Enable automatic resimulation

This option resets the simulation to the current settings whenever you return to frame 1. If setting up the first frame takes a noticeable amount of time, you may want to turn this off. When this option is off, you must click the Resimulate icon to recook the simulation with the current settings.

If a simulations timestep is less than the frame speed (for example, 3/$FPS), the Interpolate display data option on the DOP Network container node will smooth the transitions between timesteps as the animation plays.

Optimizing a simulation

Dynamics

Learning dynamics ¶

Colliding objects ¶

Simulation types ¶

Next steps ¶