Cycles: random walk subsurface scattering.

It is basically brute force volume scattering within the mesh, but part
of the SSS code for faster performance. The main difference with actual
volume scattering is that we assume the boundaries are diffuse and that
all lighting is coming through this boundary from outside the volume.

This gives much more accurate results for thin features and low density.
Some challenges remain however:

* Significantly more noisy than BSSRDF. Adding Dwivedi sampling may help
  here, but it's unclear still how much it helps in real world cases.
* Due to this being a volumetric method, geometry like eyes or mouth can
  darken the skin on the outside. We may be able to reduce this effect,
  or users can compensate for it by reducing the scattering radius in
  such areas.
* Sharp corners are quite bright. This matches actual volume rendering
  and results in some other renderers, but maybe not so much real world
  objects.

Differential Revision: https://developer.blender.org/D3054

2 files changed, 62 insertions, 32 deletions

diff --git a/intern/cycles/kernel/geom/geom_motion_triangle_intersect.h b/intern/cycles/kernel/geom/geom_motion_triangle_intersect.h
index 4a19f2ba031..542843edc84 100644
--- a/intern/cycles/kernel/geom/geom_motion_triangle_intersect.h
+++ b/intern/cycles/kernel/geom/geom_motion_triangle_intersect.h
@@ -248,23 +248,30 @@ ccl_device_inline void motion_triangle_intersect_local(
 	motion_triangle_vertices(kg, local_object, prim, time, verts);
 	/* Ray-triangle intersection, unoptimized. */
 	float t, u, v;
-	if(ray_triangle_intersect(P,
-	                          dir,
-	                          tmax,
+	if(!ray_triangle_intersect(P,
+	                           dir,
+	                           tmax,
 #if defined(__KERNEL_SSE2__) && defined(__KERNEL_SSE__)
-	                          (ssef*)verts,
+	                           (ssef*)verts,
 #else
-	                          verts[0], verts[1], verts[2],
+	                           verts[0], verts[1], verts[2],
 #endif
-	                          &u, &v, &t))
+	                           &u, &v, &t))
 	{
+		return;
+	}
+
+	int hit;
+	if(lcg_state) {
+		/* Record up to max_hits intersections. */
 		for(int i = min(max_hits, local_isect->num_hits) - 1; i >= 0; --i) {
 			if(local_isect->hits[i].t == t) {
 				return;
 			}
 		}
+
 		local_isect->num_hits++;
-		int hit;
+
 		if(local_isect->num_hits <= max_hits) {
 			hit = local_isect->num_hits - 1;
 		}
@@ -277,18 +284,29 @@ ccl_device_inline void motion_triangle_intersect_local(
 			if(hit >= max_hits)
 				return;
 		}
-		/* Record intersection. */
-		Intersection *isect = &local_isect->hits[hit];
-		isect->t = t;
-		isect->u = u;
-		isect->v = v;
-		isect->prim = prim_addr;
-		isect->object = object;
-		isect->type = PRIMITIVE_MOTION_TRIANGLE;
-		/* Record geometric normal. */
-		local_isect->Ng[hit] = normalize(cross(verts[1] - verts[0],
-		                                    verts[2] - verts[0]));
 	}
+	else {
+		/* Record closest intersection only. */
+		if(local_isect->num_hits && t > local_isect->hits[0].t) {
+			return;
+		}
+
+		hit = 0;
+		local_isect->num_hits = 1;
+	}
+
+	/* Record intersection. */
+	Intersection *isect = &local_isect->hits[hit];
+	isect->t = t;
+	isect->u = u;
+	isect->v = v;
+	isect->prim = prim_addr;
+	isect->object = object;
+	isect->type = PRIMITIVE_MOTION_TRIANGLE;
+
+	/* Record geometric normal. */
+	local_isect->Ng[hit] = normalize(cross(verts[1] - verts[0],
+	                                       verts[2] - verts[0]));
 }
 #endif  /* __BVH_LOCAL__ */
 
diff --git a/intern/cycles/kernel/geom/geom_triangle_intersect.h b/intern/cycles/kernel/geom/geom_triangle_intersect.h
index 7daa378f81e..a3b23115ae4 100644
--- a/intern/cycles/kernel/geom/geom_triangle_intersect.h
+++ b/intern/cycles/kernel/geom/geom_triangle_intersect.h
@@ -118,28 +118,40 @@ ccl_device_inline void triangle_intersect_local(
 		return;
 	}
 
-	for(int i = min(max_hits, local_isect->num_hits) - 1; i >= 0; --i) {
-		if(local_isect->hits[i].t == t) {
-			return;
+	int hit;
+	if(lcg_state) {
+		/* Record up to max_hits intersections. */
+		for(int i = min(max_hits, local_isect->num_hits) - 1; i >= 0; --i) {
+			if(local_isect->hits[i].t == t) {
+				return;
+			}
 		}
-	}
 
-	local_isect->num_hits++;
-	int hit;
+		local_isect->num_hits++;
+
+		if(local_isect->num_hits <= max_hits) {
+			hit = local_isect->num_hits - 1;
+		}
+		else {
+			/* reservoir sampling: if we are at the maximum number of
+			 * hits, randomly replace element or skip it */
+			hit = lcg_step_uint(lcg_state) % local_isect->num_hits;
 
-	if(local_isect->num_hits <= max_hits) {
-		hit = local_isect->num_hits - 1;
+			if(hit >= max_hits)
+				return;
+		}
 	}
 	else {
-		/* reservoir sampling: if we are at the maximum number of
-		 * hits, randomly replace element or skip it */
-		hit = lcg_step_uint(lcg_state) % local_isect->num_hits;
-
-		if(hit >= max_hits)
+		/* Record closest intersection only. */
+		if(local_isect->num_hits && t > local_isect->hits[0].t) {
 			return;
+		}
+
+		hit = 0;
+		local_isect->num_hits = 1;
 	}
 
-	/* record intersection */
+	/* Record intersection. */
 	Intersection *isect = &local_isect->hits[hit];
 	isect->prim = prim_addr;
 	isect->object = object;