Hi,

We currently have an issue with the unpacking of USD files in Houdini and its slowness in unpacking large amount of curves.

Process:
We are exporting our hair grooms from Houdini as .USD (binary) files. We are then bringing those back into Houdini and blasting away 99% of the hairs to use the rest of the curves for simulation. Currently when reading in the .USD back into Houdini in /obj we are getting a massive slowdown when unpacking the large amount of curves (around 5million).

However, if we export our hair grooms as Alembic and then reading those back into Houdini for simulation we are NOT experiencing the same slow unpacking.


We have gone through and optimized the .usd file so it only contains a minimum amount of attributes as well, but it is still very slow compared to Alembic as seen in above screenshots of the performance monitor.

So our questions are:

1. Why is the USD unpacking so much slower than Alembic?
2. Is there anything we can do to export a better optimized .USD file with large amounts of curves?


Please see below screenshots of the performance monitor

.USD




.Alembic




Thanks,
K

Hey all, just bumping this.. I've had similar results whilst trying this.

USD unpack seems to be fairly slow when dealing with lots curves.
Is this a curve specific thing, ie: is there something that can be done on the initial curve USD file that would help with faster unpacking? Or is this a current limitation of the unpack node?

Also, just to check.. is this even the correct approach to use if we want to deal with the curves directly in Houdini?

Thanks for any help!

Anyone able to answer or have any insight on this issue?

I'd suggest submitting an example file to support that reproduces the issue. Offhand, I don't see why there should be a significant performance difference between Alembic and USD for importing curves.