On this page |
Overview ¶
TOPs is designed to work with compute farms that may have a variety different filesystems. For example, a TOPs user could be on a Windows machine, but also be using a Linux-based farm. The problem is how to map file paths from one filesystem to another. TOPs provides the PDG Path Map to address this.
Most TOP nodes that do work let you specify input and/or output file paths. In TOPs, each scheduler node can specify a working directory. This is because different render farm software may use different shared network filesystems. In the case you are using a farm scheduler, you should make sure that all files you output are reachable by the farm machines relative to this PDG_DIR
directory.
How to ¶
-
Set the base working directory on the scheduler node. This directory is available to jobs as the
PDG_DIR
environment variable.-
Use separate working directories for each HIP file. This is to avoid having two HIP files writing to the same
PDG_DIR
. Many of the default generated filenames used in parameter defaults are only unique within the HIP file. -
For render farm schedulers, make sure that the directory is inside the network filesystem (like NFS mount or SMB share) and is shared with the render farm client machines.
-
-
When you use
PDG_DIR
orPDG_TEMP
in parameter filenames, use the form__PDG_DIR__
instead of${PDG_DIR}
. If you use${PDG_DIR}
, Houdini will try and fail to expand the variable itself before the dependency graph gets it. Houdini will ignore__PDG_DIR__
syntax, but the PDG scheduler knows to expand that token to the absolute path on the executing machine. -
Put intermediate files under
__PDG_TEMP__
and final output files under__PDG_DIR__
.-
Categorize output files using subdirectories. For example,
__PDG_TEMP__/geo
for intermediate geometry files and__PDG_DIR__/geo
for final geometry output.
-