Cycles: restore Christensen-Burley SSS

There is not enough time before the release to improve Random Walk to handle
all cases this was used for, so restore it for now.

Since there is no more path splitting in cycles-x, this can increase noise in
non-flat areas for the sample number of samples, though fewer rays will be traced
also. This is fundamentally a trade-off we made in the new design and why Random
Walk is a better fit. However the importance resampling we do now does help to
reduce noise.

Differential Revision: https://developer.blender.org/D12800

1 files changed, 24 insertions, 0 deletions

diff --git a/intern/cycles/kernel/bvh/bvh_util.h b/intern/cycles/kernel/bvh/bvh_util.h
index 9f188a93e2c..e16da9755f2 100644
--- a/intern/cycles/kernel/bvh/bvh_util.h
+++ b/intern/cycles/kernel/bvh/bvh_util.h
@@ -113,6 +113,30 @@ ccl_device_inline void sort_intersections(Intersection *hits, uint num_hits)
 }
 #endif /* __SHADOW_RECORD_ALL__ | __VOLUME_RECORD_ALL__ */
 
+/* For subsurface scattering, only sorting a small amount of intersections
+ * so bubble sort is fine for CPU and GPU. */
+ccl_device_inline void sort_intersections_and_normals(Intersection *hits,
+                                                      float3 *Ng,
+                                                      uint num_hits)
+{
+  bool swapped;
+  do {
+    swapped = false;
+    for (int j = 0; j < num_hits - 1; ++j) {
+      if (hits[j].t > hits[j + 1].t) {
+        struct Intersection tmp_hit = hits[j];
+        struct float3 tmp_Ng = Ng[j];
+        hits[j] = hits[j + 1];
+        Ng[j] = Ng[j + 1];
+        hits[j + 1] = tmp_hit;
+        Ng[j + 1] = tmp_Ng;
+        swapped = true;
+      }
+    }
+    --num_hits;
+  } while (swapped);
+}
+
 /* Utility to quickly get flags from an intersection. */
 
 ccl_device_forceinline int intersection_get_shader_flags(const KernelGlobals *ccl_restrict kg,