I'm guessing this gives a measure performance regression. It would be good to understand how big it is. Maybe we'll have to accept it.
@brecht I couldn't measure any performance differences: …
The noise is still continuous even after this change I think, though not differentiable. So maybe clarify the comment as well.
Since it's cutting a chunk of size 100000.0 out and pasting them…
Was the
isinf
there to fix this same issue?
Maybe, or maybe not. It was added by Brecht in 8cf374d401 (diff-c)
…
So it's highly likely the issue we're seeing in OSL is something precision related in the noise function that Cycles own noise implementation handles better.
Not really, both start to have…
I can confirm. The issue was introduced in 4.1 between
ca9c6372c94c
(Sep 28) and09caf5969010
(Nov 28)**System Information** Operating system: Windows-10-10.0.22621-SP0…
I intended it for main
branch, but having it in 4.1 doesn't hurt.
Also, since the Musgrave node doesn't exist anymore from 4.1 onward this issue could be seen as resolved.
@lichtwerk d98005da5d fixes the NAN problem, but only partially. Basically before d98005da5d you'd get NANs for both very low and very high Lacunarity values. d98005da5d fixes the problem for very…