Saves the geometry to disk with the last control settings.

Use a TOP network to cook and save the cache files to disk in a separate background process. This allows you to keep working on other parts of Houdini. The cached frames will show up as they are completed by the other process.

Stops all background caching processes, if any exist.

When this is Frame Range the node writes a cache file for every frame (see Start/End/Inc below). If you don’t want pre-generate all frames at once, or if you know you only need to regenerate a single changed frame, you can set this to Single Frame. This makes the node write out a cache file only for the current frame (or the frame you set in Override frame).

Use this when you are caching a sequence that is generated using a simulation, that makes each frame dependent on it’s previous frame.

When Evaluate as is Single Frame and the checkbox to the left is on, cache this explicit frame number instead of the current frame from the timeline.

Specifies the range of frames to render (start frame, end frame, and increment). All values may be floating point values. The range is inclusive.

These parameters determine the values of the local variables for the output driver.

When Evaluate as is Frame range, this is the number of fractional sub-frames to divide each frame into. Caching sub-frames increases the quality of motion blur rendering for cached animated geometry. If you set Substeps greater than 1 and you've set File Path to Explicit, make sure your explicit path expression can handle fractional frames, by using $T (floating point time in seconds) or $FF (fractional frame number) instead of $F (whole frame number).

Clamps the current frame if it is earlier than this frame. This can be used to load the Start frame’s geometry for frames before the Start/End range, which would otherwise output empty geometry.

Clamps the current frame if it is later than this frame. This can be used to load the End frame’s geometry for frames after the Start/End range, which would otherwise output empty geometry.

Some file formats, in particular .bgeo, .bgeo.sc, and .geo, contain meta-data in the header than can be loaded without loading the entire file. This is useful for dealing with large data sets.

When loading the geometry as Packed Disk Primitive or when delayed loading, you can set the viewport display to a lighter representation of the packed geometry. This setting only applies to rendering in the viewport, at render time, the full geometry will be rendered.

When loading the geometry as Packed Disk Primitive, this toggle controls whether the file name stored with the primitive will have variables expanded. By default, variables will be preserved. For example, a file such as $HIP/geo/tree.bgeo will be portable if the value of the $HIP variable changes. This is only important if you are saving the packed primitive to disk and may move the geometry file.

When storing variables in the packed primitive (i.e. the toggle is turned off), the primitive itself performs variable expansion. Since the primitive can be accessed in applications other than Houdini (mantra, gplay, gconvert, etc.), the expansion of variables is limited in function. For example, using variables local to the .hip file or using expressions which reference other nodes may not work as expected.

If this parameter is set, packed primitives and other shared data will not be loaded immediately but will be loaded only as needed. This can be useful when opening a large scene to reduce load times.

When creating Packed Disk Sequence primitives, this parameter specifies which geometry from the sequence will be used at render time. The index is linear, with values between 0 and the number of geometry files in the disk sequence (not based on the frame number). Index values out of range will automatically wrap to generate animation cycles.

The index values can have non-integer values. If the topology matches between disk files, the geometry will be blended at render time. See also the vm_pack_sequencesubsteps rendering property.

<path_to_image>/frame`padzero(3, $F+1)`.tga

When creating Packed Disk Sequence primitives, this parameter specifies behavior when Sequence Index is outside of the frame range.

This parameter overrides the TOP scheduler for this node.

A space separated list of attributes to delete from the geometry before saving it. If you know there are attributes on the input geometry you don’t need in the rest of the network, you can decrease the cache file size and make cache loading faster by preventing the attributes from being included in the cache. The names will match at any level (vertex, point, primitive, or detail). The node will never delete the P (position) point attribute. You can use patterns, for example * ^v to delete all attributes except P and v.

A space separated list of groups to delete from the geometry before saving it. This can be useful with simulation data, where there can be groups in the output you will never use. You can decrease the cache file size and make cache loading faster by preventing these groups from being included in the cache.

If you know you don’t need full precision for certain attributes in the input, this multi-parm lets you convert the attributes to lower-precision types before saving the geometry. This can make the cache files smaller and faster to load (depending on how many uses of the attribute there are). Set the number or click the plus or minus buttons to set the number of conversions to do.

See the Attribute Cast SOP help for more information.

In a “cast” multiparm instance, the level (vertex, point, primitive, or detail) of the attribute(s) you want to convert to lower precision to save space.

In a “cast” multiparm instance, a space-separated list of attributes at the given level (set by the Class parameter above) you want to convert to lower precision to save space.

In a “cast” multiparm instance, the precision to convert the matching attributes down to, to save space.

Whether the node will automatically turn on Load from disk when you click Save to disk or Save to disk in background.

When File Path is Constructed and this is on, when you click Save to disk or Save to disk in background the node automatically converts the expression in Base name to a literal string. This prevents the cache from breaking if you duplicate or rename the scene file or node.

When writing out cache files, automatically create any intermediate directories in the file path that don’t exist.

Forces all simulation OPs to be reset. This includes DOP Networks, POP SOPs, and other OPs that cache their results.

This is the safest way to render out a simulation, because it starts the simulation from scratch and discards any partial simulations you might have done with different parameters.

A percentage complete value is printed out as files are written. This is in the style expected by Pixar’s Alfred render queue.

When saving more than one frame, save in a background thread. This can make saving faster for large file sizes, but may use more memory since it will retain the output geometry until the save completes.

If saving the geometry to disk fails due to a disk writing error, Houdini will usually error the output node immediately. This is desirable for most cases where a failure to save means an illegal path, which is not recoverable. However, sometimes files fail to save due to network issues. If the number of save retries is non-zero, Houdini will re-attempt to save this number of times. Each time will be accompanied with an output to the console of the failure to save and a five second wait in the hopes that the network will clear up.

Uses the settings in a particular take while rendering. Choose Current to use the current take when rendering.

This is what the node should do (in Load from disk mode) when the node can’t find (or cleanly load) a cache file. Report error sets an error on the node, preventing the rest of the network from cooking. This makes it obvious when a problem happens. No geometry (the default) sets a warning on the node rather than an error, and outputs empty geometry. This can be useful if your network has a way of dealing with missing cache frames, or if they're not important.

The total number of frames to cache in memory. These sorted by filename, not by time, so if multiple frames evaluate to the same file name they will share the same cached geometry.

Attempt to predict the next frame that will be needed and load it in the background. This lets computation be overlapped with file IO.

If you are blending multiple frames, the Cache Frames needs to be large enough for only the net-new frame to be loaded each cook. Otherwise the simple predictive model of the prefetcher will be confused and stop working.

Run this HScript after each frame’s data has finished writing to disk. This is always after the corresponding Post-Frame Script and always before the Post-Render script, but the order otherwise is undefined. When Save in Background is enabled this allows delaying script actions until the file has finished saving.

Contains the expression that computes the current frame number to cache. Do not edit this expression.

Contains the expression that generates the frame number as a padded string, including a dot prefix. Do not edit this expression.

Contains an expression that generates the “version” part of the path, including a v prefix. Do not edit this expression.

Contains an expression that generates the parent directory path for cache files. Do not edit this expression.

Contains an expression that generates the filename for cache files. Do not edit this expression.

Contains an expression that generates the descriptive text displayed next to this node (below the name and badges) in the network editor. This is set to output the cache filename pattern. Do not edit this expression.

Contains an expression that generates the resolved, full path of the cache file for the current frame. Do not edit this expression.