Just out from a discussion with some other users, and knowing this has been a recurrent question for a long time, i thought a thread should be created to help each other to find the best pratice and tips to get the best most out of PDG, in term of speed.

What are in your experience the sensible points which will cause a massive speed drop ?

One of the things I added to every HDA that’s being processed is a timer. This consists of two Python nodes, one at the start of the graph and one at the end, that grab the current time in milliseconds. Since I know when the node started and when it ended, I can calculate how long it took to execute that workItem

Will be happy to hear your feedback about your experience with PDG in a production context


Vincent

Use your scheduler threads (slots) wisely.

Don't overload you GPU/CPU. I usually have a set of schedulers one for 1 core, one for 4 cores, one for all cores etc. Depending on the task, I'll set them to use the appropriate scheduler to maximize the CPU and bandwidth.

- GPU... I have ofen come to case the node refuse to output because of open CL Ram issue. How does it works in PDG? I mean is it not risky to use OpenCL when use multicores but with one GPU ?

- What about "Services"? it's new in 19.5... It's not clear for me how to well use that feature...

- In my experience chaining multiples HDA together is not advised? If the data exchanging between both are not large, have you experimented some deal-breaker issue please ?


Cheers