Search - User list
Full Version: PDG geometry ROP slower than standard File Cache
Root » PDG/TOPs » PDG geometry ROP slower than standard File Cache
I'm using a regular TOP's geometry rop to point to a SOP and then render it to disk locally. However it's extremely slow compared to just using a regular file cache node in SOP's to do the same save to disk operation. Anyone else notice this or have any insight as to why this might be happening?

When using TOPs, the jobs for cooking the ROP Geometry work items run out of process. This is what allows it to be done in the background, without blocking the UI. If your SOP network is lightweight there's going to be a lot of overhead from spawning a process and waiting for it report back its results (a few seconds).

The ROP Geometry has settings for batching multiple frames into a single job. For example, if you're writing out 240 frames it can run those in a single external process, rather than one process per frame.
You can also set the number of threads the external Houdini process uses when cooking your SOP network. On the Schedulers -> Local Scheduler tab, enable the Houdini Max Threads parameter and enter a custom value. By default, when that parm is disabled, PDG attempts to set the value automatically based on your other scheduler settings to avoid over-subscribing the machine, e.g. by running 8 instances of Mantra each using the maximum thread count.
Ah, thank you! Increasing the Frames per Batch had a huge impact on the speed. That's good to know!
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Powered by DjangoBB