Performance bottleneck with transformations each frame

[paul_george45MHL]

Hi,

I’m using a vrTimer to update transformations (setTranslation, setRotation, etc.) on multiple nodes every frame (using vrTimer).

Profiling shows a massive CPU hit from these individual API calls.

Is there a way to batch these transformation updates in the v2 API to reduce Python-to-C++ overhead ?

Any tips ?

[michael_nikelsky]

This is not really possible to answer without knowing how your script is actually setting the parameters and how many nodes we are talking about. Especially setting the world space variants require calculating the whole upstream graph transforms and eventually decompose the matrix for each node every time so one optimization would surely be to set local transforms that don´t need additional upstream graph evaluations.