> Not sure if you want extra people to test it on actual GPU? @Sergey , not really - the changes are quite straightforward and there is no expectation of functionality changes (well, besides…
Adding @Sergey for review. Sergey, could you please take a look?
According to my performance testing on different machines with Intel GPUs, I haven't noticed any significant performance difference, just small variations. So, these changes are purely functional…
I can also confirm that all works as expected and this changes are addressing the reported problem with unnecessary denoiser recreation: Before this Patch:
Time:00:02.08
Just some thoughts on the overall situation with this part of the code: the entire reason we even have this logic about denoiser_params
versus effective_denoiser_params
is that we're trying to…
> Not sure if @Sirgienko wants to have a look here before landing. @Alaska @Sergey , Yes, I would like to take a look - I will do it today and tomorrow, so I kindly ask do not land this…
@Anti-Distinctlyminty There is an easy way to check, if something wrong with arguments or something wrong with an enviromental setup: you could try to run this command `blender -b --python-expr…
@Sergey, if you will be fine with this changes, then we wound need to backport them to the 4.2 LTS stable branch in order to fix it there as wel.