Whatever makes the compiler happy is fine with me.

• Merge 'master' into 'cycles_adaptive_sampling'
• Merge 'master' into 'cycles_adaptive_sampling'
• Merge 'master' into 'cycles_adaptive_sampling'

@Wo!262 (wo262) This scene looks like it needs a lot more samples. The adaptive sampler can only pick up and resolve noise that it finds within the "min samples" it takes - if those initial samples turn out all black, the sampler has no reason to continue sampling this seemingly uniform pixel.

• Cycles: Fixed a typo in sample_is_even().

Pretty pictures for a more complex scene:

Cycles: Moved "Adaptive Sampilng" option out of the advanced section in the UI.

I noticed I hadn't added documentation of the user facing parameters:

There isn't much to see, but we all enjoy pretty pictures:

This is substantial patch and obviously not intended for 2.80. There is still room for improvement, I hope to add more comments/documentation, deduplicate some code and take a closer look at the overall code style.

• Cycles: Added const to some variables.

Both RenderMan and Arnold have sphere lights that emit to the outside only.
RenderMan render:

In D4676#106437, @Sergey Sharybin (sergey) wrote:

If desired, I can also change it to make the sphere a double-sided emitter by switching to uniform sampling when the shading point is inside the sphere (this is what PBRT does).

There is a slight performance hit from the more complex sampling.
Haven't done extensive benchmarking, but a test scene I measured went from 58s -> 59s CPU render time.

Before:

• Added support for debug versions of OpenImageDenoise Windows DLLs.

• Compositor: More fixes for OIDN build on Windows.

• Merge remote-tracking branch 'origin' into openimagedenoise
• Merge branch 'master' of git.blender.org:blender into openimagedenoise
• Fixed padding for OIDN node.
• build_environemnt: Updated OpenImageDenoise to static 0.8.2
• build_environment: OpenImageDenoise fixes for Linux and Windows.
• build_environment: Fixed FindOpenImageDenoise.cmake.
• build_environment: Fixed Windows build with static OpenImageDenoise.

Nevermind, I wrote this up prematurely. Need to double-check things first.

This broke OpenCL/AMD on Windows. See inline comment.

For those who are using the scrambling distance patch, this change is trivial to remove. I haven't tested it myself yet, but I too would expect this to negate some of the performance benefits from the scramble distance patch.

FYI, I'm waiting for Intel to release the next version of OIDN that can build as a static library.

I'm not expecting this patch to land right away, this should probably undergo some wider testing before we can declare it safe.

Applying this to specular shaders introduces new artifacts, so it's only applied to diffuse BSDFs for now.

This patch does not fix T37814.

The fix in "Ray Tracing Gems" applies to a slightly different problem.

Updated tangent space calculation and OSL attributes to use new float2 attributes.

• Cycles: Added Float2 attribute type.
• Cycles: Fixed OSL support of float2 attributes.
• Cycles: Fixed Tangents with float2 attributes.

This can save quite some memory. For example, the Agent 327 03_01_B shrinks its peak Mem from 10931MB to 10226MB, that is without this patch, Cycles is allocating >700MB of zeroes.

What are you doing to get Albedo > 1? I didn't see that happen in my test scenes. Adding clamping to the albedo input will be easy though.

Yes, reverting it allowed make deps to succeed.

This boke make depsfor me. When I test this on a brand new Ubuntu 16.04, OSL fails to find pugixml.

Yes, projects created with cmake -G Xcode create the datatoc and other binaries in a subdirectory named after the current inside bin.