• Cycles: More adaptive sampling refactoring.
• Fixed typo.
• Merge branch 'master' into cycles_adaptive_sampling

__popcnt on Windows is not supported on all CPUs.

• Cycles: Refactored PMJ sampler.
• Merge commit '153e001c743bf0f6bc259966418446441e00e200^' into cycles_adaptive_sampling
• Merge branch 'master' into cycles_adaptive_sampling

Thick line segments could be emulated through spline primitives at the expense of wasted memory.

I noticed the same with the Koro file.

I don't consider renaming "attrib" to "attr" an improvement in code quality, for example.

I think it should be possible to extend this to fix the problems with overlapping geometry too by implementing the algorithms from "Robust Iterative Find-Next-Hit Ray Traversal"
http://www.sci.utah.edu/~wald/Publications/2018/nexthit/nexthit-pgv18.pdf

Please don't do any mindless search/replace of symbol names or files. Please.

Please don't.
Refactors like this are a huge annoyance for people maintaining branches and patches that haven't been merged yet. I don't see how replacing "util" with "utils" is supposed to be an improvement, the only thing it does is create unnecessary work and noise in the git history.

srqt() isn't that much of a performance issue any more as it once was. That said, the newer mapping should give us a better sample distribution.

This was high on my to-do list, great to see that you've started work on this! Let me know if there's anything I can help you with.

Or should I just go ahead and check in this without review, since it's a bug fix for Embree which is turned off by default anyway?

• Cycles: Use lerp for Optix motion blur.

• build_env: Raised minimum Embree version to 3.7.0.

Note that this patch will require that platform maintainers update the Embree library to 3.7.0.

This time diffing against the correct master, hopefully.

• Merge branch 'master' of git.blender.org:blender into cycles_adaptive_sampling
• Merge remote-tracking branch 'origin/master' into cycles_adaptive_sampling
• Merge remote-tracking branch 'upstream/master' into cycles_adaptive_sampling
• Cycles: Fix Adaptive Sampling passes after AOV introduction.

WIP, currently limited to CPU rendering only:
https://github.com/skwerner/blender/blob/cycles_viewport_denoising/intern/cycles/render/buffers.cpp#L425

I have a patch for viewport denoising that supports OIDN - unfortunately it works different than this one. We should coordinate for a unified approach.

The 2.82 ship has sailed, no more new feature commits are supposed to happen at this stage. My plan is to polish this some more, add OptiX support, maybe split it into two diffs (PMJ sampler can be separated) and then get this in properly for 2.83.

@Brecht Van Lommel (brecht) Did you get a chance to take a look at this? Should I commit this for 2.82?

I can't share details about that particular scene.

You'd have to build your own, and I consider this patch nothing more than a proof of concept. In my limited testing, it appeared to break CUDA rendering and I haven't tested it BRDFs other than diffuse and principled surface.

I was made aware that Appleseed has introduced a workaround for the shadow terminator. Their method, just like the fix for the bump map issue, is pushing the shadow terminator towards the light source, that is darkening the surface overall.

On some of the larger scenes that I threw at it on a 20C/40T machine, the benefits are tremendous. Cycles render preparation time went from 40min to 8min. Might be worth benchmarking this on Linux nonetheless.

@Lukas Stockner (lukasstockner97) It looks to me like this is a thing specific to the principled BSDF, which if I'm not mistaken was added after your denoiser? A quick look in the debugger tells me that for most settings, the principled BSDF adds the white specular component to the albedo when it might be beneficial to only use the diffuse albedo.

It doesn't work quite right yet with spot lights with a non-zero radius. That said, I'm not happy with the way spot lights behave in Cycles in general, but that will be another patch.

I was unable to reproduce this on 10.15.1, testing with a GeForce GT750M and an AMD WX7100.
@ken wong (ken), does your Mac have a P3 display by any chance?

To my knowledge, macOS itself does not make any use of function keys. Cmd F2 is free.

I don't think there's any way around this without ignoring backfacing geometric normals.

For what it's worth, a little insight in what other renderers do (in the case without bump maps)
LuxRender: Nothing.
"the black triangles on the edge are unavoidable, their geometric normal points away from the light source so there is no way to fix them aside increasing tessellation and/or using softer shadows (due to a larger light source, etc.)"
https://forums.luxcorerender.org/viewtopic.php?f=5&t=1286

Read the SIGGRAPH paper more closely - it applies to problems introduced by bump/normal mapped surfaces only. It is implemented and working.

Commenting.

Code deduplication.

Come to think of it, it should be possible to apply the same ray clipping to distance sampling as well.