Merge branch 'master' into blender2.8

Conflicts:
	source/blender/editors/animation/anim_draw.c

66 files changed, 1619 insertions, 1550 deletions

diff --git a/CMakeLists.txt b/CMakeLists.txt
index a107dd78fbb..e16f0cddfd7 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -447,6 +447,7 @@ option(WITH_BOOST					"Enable features depending on boost" ON)
 
 # Unit testsing
 option(WITH_GTESTS "Enable GTest unit testing" OFF)
+option(WITH_OPENGL_TESTS "Enable OpenGL related unit testing (Experimental)" OFF)
 
 
 # Documentation
diff --git a/intern/cycles/CMakeLists.txt b/intern/cycles/CMakeLists.txt
index 79c1c3e3e82..31fa15a2042 100644
--- a/intern/cycles/CMakeLists.txt
+++ b/intern/cycles/CMakeLists.txt
@@ -74,6 +74,7 @@ elseif(CMAKE_COMPILER_IS_GNUCC)
 	if(CXX_HAS_AVX2)
 		set(CYCLES_AVX2_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mfma -mlzcnt -mbmi -mbmi2 -mf16c -mfpmath=sse")
 	endif()
+	set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -ffast-math -fno-finite-math-only")
 elseif(CMAKE_CXX_COMPILER_ID MATCHES "Clang")
 	check_cxx_compiler_flag(-msse CXX_HAS_SSE)
 	check_cxx_compiler_flag(-mavx CXX_HAS_AVX)
@@ -89,6 +90,7 @@ elseif(CMAKE_CXX_COMPILER_ID MATCHES "Clang")
 	if(CXX_HAS_AVX2)
 		set(CYCLES_AVX2_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mfma -mlzcnt -mbmi -mbmi2 -mf16c")
 	endif()
+	set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -ffast-math -fno-finite-math-only")
 endif()
 
 if(CXX_HAS_SSE)
diff --git a/intern/cycles/blender/blender_mesh.cpp b/intern/cycles/blender/blender_mesh.cpp
index 85117cfff7b..a1ff81e750a 100644
--- a/intern/cycles/blender/blender_mesh.cpp
+++ b/intern/cycles/blender/blender_mesh.cpp
@@ -27,6 +27,7 @@
 #include "subd_patch.h"
 #include "subd_split.h"
 
+#include "util_algorithm.h"
 #include "util_foreach.h"
 #include "util_logging.h"
 #include "util_math.h"
@@ -525,69 +526,177 @@ static void attr_create_uv_map(Scene *scene,
 }
 
 /* Create vertex pointiness attributes. */
+
+/* Compare vertices by sum of their coordinates. */
+class VertexAverageComparator {
+public:
+	VertexAverageComparator(const array<float3>& verts)
+	        : verts_(verts) {
+	}
+
+	bool operator()(const int& vert_idx_a, const int& vert_idx_b)
+	{
+		const float3 &vert_a = verts_[vert_idx_a];
+		const float3 &vert_b = verts_[vert_idx_b];
+		if(vert_a == vert_b) {
+			/* Special case for doubles, so we ensure ordering. */
+			return vert_idx_a > vert_idx_b;
+		}
+		const float x1 = vert_a.x + vert_a.y + vert_a.z;
+		const float x2 = vert_b.x + vert_b.y + vert_b.z;
+		return x1 < x2;
+	}
+
+protected:
+	const array<float3>& verts_;
+};
+
 static void attr_create_pointiness(Scene *scene,
                                    Mesh *mesh,
                                    BL::Mesh& b_mesh,
                                    bool subdivision)
 {
-	if(mesh->need_attribute(scene, ATTR_STD_POINTINESS)) {
-		const int numverts = b_mesh.vertices.length();
-		AttributeSet& attributes = (subdivision)? mesh->subd_attributes: mesh->attributes;
-		Attribute *attr = attributes.add(ATTR_STD_POINTINESS);
-		float *data = attr->data_float();
-		int *counter = new int[numverts];
-		float *raw_data = new float[numverts];
-		float3 *edge_accum = new float3[numverts];
-
-		/* Calculate pointiness using single ring neighborhood. */
-		memset(counter, 0, sizeof(int) * numverts);
-		memset(raw_data, 0, sizeof(float) * numverts);
-		memset(edge_accum, 0, sizeof(float3) * numverts);
-		BL::Mesh::edges_iterator e;
-		int i = 0;
-		for(b_mesh.edges.begin(e); e != b_mesh.edges.end(); ++e, ++i) {
-			int v0 = b_mesh.edges[i].vertices()[0],
-			    v1 = b_mesh.edges[i].vertices()[1];
-			float3 co0 = get_float3(b_mesh.vertices[v0].co()),
-			       co1 = get_float3(b_mesh.vertices[v1].co());
-			float3 edge = normalize(co1 - co0);
-			edge_accum[v0] += edge;
-			edge_accum[v1] += -edge;
-			++counter[v0];
-			++counter[v1];
-		}
-		i = 0;
-		BL::Mesh::vertices_iterator v;
-		for(b_mesh.vertices.begin(v); v != b_mesh.vertices.end(); ++v, ++i) {
-			if(counter[i] > 0) {
-				float3 normal = get_float3(b_mesh.vertices[i].normal());
-				float angle = safe_acosf(dot(normal, edge_accum[i] / counter[i]));
-				raw_data[i] = angle * M_1_PI_F;
+	if(!mesh->need_attribute(scene, ATTR_STD_POINTINESS)) {
+		return;
+	}
+	const int num_verts = b_mesh.vertices.length();
+	/* STEP 1: Find out duplicated vertices and point duplicates to a single
+	 *         original vertex.
+	 */
+	vector<int> sorted_vert_indeices(num_verts);
+	for(int vert_index = 0; vert_index < num_verts; ++vert_index) {
+		sorted_vert_indeices[vert_index] = vert_index;
+	}
+	VertexAverageComparator compare(mesh->verts);
+	sort(sorted_vert_indeices.begin(), sorted_vert_indeices.end(), compare);
+	/* This array stores index of the original vertex for the given vertex
+	 * index.
+	 */
+	vector<int> vert_orig_index(num_verts);
+	for(int sorted_vert_index = 0;
+	    sorted_vert_index < num_verts;
+	    ++sorted_vert_index)
+	{
+		const int vert_index = sorted_vert_indeices[sorted_vert_index];
+		const float3 &vert_co = mesh->verts[vert_index];
+		bool found = false;
+		for(int other_sorted_vert_index = sorted_vert_index + 1;
+		    other_sorted_vert_index < num_verts;
+		    ++other_sorted_vert_index)
+		{
+			const int other_vert_index =
+			        sorted_vert_indeices[other_sorted_vert_index];
+			const float3 &other_vert_co = mesh->verts[other_vert_index];
+			/* We are too far away now, we wouldn't have duplicate. */
+			if ((other_vert_co.x + other_vert_co.y + other_vert_co.z) -
+			    (vert_co.x + vert_co.y + vert_co.z) > 3 * FLT_EPSILON)
+			{
+				break;
 			}
-			else {
-				raw_data[i] = 0.0f;
+			/* Found duplicate. */
+			if(len_squared(other_vert_co - vert_co) < FLT_EPSILON) {
+				found = true;
+				vert_orig_index[vert_index] = other_vert_index;
+				break;
 			}
 		}
-
-		/* Blur vertices to approximate 2 ring neighborhood. */
-		memset(counter, 0, sizeof(int) * numverts);
-		memcpy(data, raw_data, sizeof(float) * numverts);
-		i = 0;
-		for(b_mesh.edges.begin(e); e != b_mesh.edges.end(); ++e, ++i) {
-			int v0 = b_mesh.edges[i].vertices()[0],
-			    v1 = b_mesh.edges[i].vertices()[1];
-			data[v0] += raw_data[v1];
-			data[v1] += raw_data[v0];
-			++counter[v0];
-			++counter[v1];
+		if(!found) {
+			vert_orig_index[vert_index] = vert_index;
 		}
-		for(i = 0; i < numverts; ++i) {
-			data[i] /= counter[i] + 1;
+	}
+	/* Make sure we always points to the very first orig vertex. */
+	for(int vert_index = 0; vert_index < num_verts; ++vert_index) {
+		int orig_index = vert_orig_index[vert_index];
+		while(orig_index != vert_orig_index[orig_index]) {
+			orig_index = vert_orig_index[orig_index];
 		}
-
-		delete [] counter;
-		delete [] raw_data;
-		delete [] edge_accum;
+		vert_orig_index[vert_index] = orig_index;
+	}
+	sorted_vert_indeices.free_memory();
+	/* STEP 2: Calculate vertex normals taking into account their possible
+	 *         duplicates which gets "welded" together.
+	 */
+	vector<float3> vert_normal(num_verts, make_float3(0.0f, 0.0f, 0.0f));
+	/* First we accumulate all vertex normals in the original index. */
+	for(int vert_index = 0; vert_index < num_verts; ++vert_index) {
+		const float3 normal = get_float3(b_mesh.vertices[vert_index].normal());
+		const int orig_index = vert_orig_index[vert_index];
+		vert_normal[orig_index] += normal;
+	}
+	/* Then we normalize the accumulated result and flush it to all duplicates
+	 * as well.
+	 */
+	for(int vert_index = 0; vert_index < num_verts; ++vert_index) {
+		const int orig_index = vert_orig_index[vert_index];
+		vert_normal[vert_index] = normalize(vert_normal[orig_index]);
+	}
+	/* STEP 3: Calculate pointiness using single ring neighborhood. */
+	vector<int> counter(num_verts, 0);
+	vector<float> raw_data(num_verts, 0.0f);
+	vector<float3> edge_accum(num_verts, make_float3(0.0f, 0.0f, 0.0f));
+	BL::Mesh::edges_iterator e;
+	EdgeMap visited_edges;
+	int edge_index = 0;
+	memset(&counter[0], 0, sizeof(int) * counter.size());
+	for(b_mesh.edges.begin(e); e != b_mesh.edges.end(); ++e, ++edge_index) {
+		const int v0 = vert_orig_index[b_mesh.edges[edge_index].vertices()[0]],
+		          v1 = vert_orig_index[b_mesh.edges[edge_index].vertices()[1]];
+		if(visited_edges.exists(v0, v1)) {
+			continue;
+		}
+		visited_edges.insert(v0, v1);
+		float3 co0 = get_float3(b_mesh.vertices[v0].co()),
+		       co1 = get_float3(b_mesh.vertices[v1].co());
+		float3 edge = normalize(co1 - co0);
+		edge_accum[v0] += edge;
+		edge_accum[v1] += -edge;
+		++counter[v0];
+		++counter[v1];
+	}
+	for(int vert_index = 0; vert_index < num_verts; ++vert_index) {
+		const int orig_index = vert_orig_index[vert_index];
+		if(orig_index != vert_index) {
+			/* Skip duplicates, they'll be overwritten later on. */
+			continue;
+		}
+		if(counter[vert_index] > 0) {
+			const float3 normal = vert_normal[vert_index];
+			const float angle =
+			        safe_acosf(dot(normal,
+			                       edge_accum[vert_index] / counter[vert_index]));
+			raw_data[vert_index] = angle * M_1_PI_F;
+		}
+		else {
+			raw_data[vert_index] = 0.0f;
+		}
+	}
+	/* STEP 3: Blur vertices to approximate 2 ring neighborhood. */
+	AttributeSet& attributes = (subdivision)? mesh->subd_attributes: mesh->attributes;
+	Attribute *attr = attributes.add(ATTR_STD_POINTINESS);
+	float *data = attr->data_float();
+	memcpy(data, &raw_data[0], sizeof(float) * raw_data.size());
+	memset(&counter[0], 0, sizeof(int) * counter.size());
+	edge_index = 0;
+	visited_edges.clear();
+	for(b_mesh.edges.begin(e); e != b_mesh.edges.end(); ++e, ++edge_index) {
+		const int v0 = vert_orig_index[b_mesh.edges[edge_index].vertices()[0]],
+		          v1 = vert_orig_index[b_mesh.edges[edge_index].vertices()[1]];
+		if(visited_edges.exists(v0, v1)) {
+			continue;
+		}
+		visited_edges.insert(v0, v1);
+		data[v0] += raw_data[v1];
+		data[v1] += raw_data[v0];
+		++counter[v0];
+		++counter[v1];
+	}
+	for(int vert_index = 0; vert_index < num_verts; ++vert_index) {
+		data[vert_index] /= counter[vert_index] + 1;
+	}
+	/* STEP 4: Copy attribute to the duplicated vertices. */
+	for(int vert_index = 0; vert_index < num_verts; ++vert_index) {
+		const int orig_index = vert_orig_index[vert_index];
+		data[vert_index] = data[orig_index];
 	}
 }
 
@@ -656,9 +765,6 @@ static void create_mesh(Scene *scene,
 			generated[i++] = get_float3(v->undeformed_co())*size - loc;
 	}
 
-	/* Create needed vertex attributes. */
-	attr_create_pointiness(scene, mesh, b_mesh, subdivision);
-
 	/* create faces */
 	vector<int> nverts(numfaces);
 	vector<int> face_flags(numfaces, FACE_FLAG_NONE);
@@ -718,6 +824,7 @@ static void create_mesh(Scene *scene,
 	/* Create all needed attributes.
 	 * The calculate functions will check whether they're needed or not.
 	 */
+	attr_create_pointiness(scene, mesh, b_mesh, subdivision);
 	attr_create_vertex_color(scene, mesh, b_mesh, nverts, face_flags, subdivision);
 	attr_create_uv_map(scene, mesh, b_mesh, nverts, face_flags, subdivision, subdivide_uvs);
 
@@ -1178,4 +1285,3 @@ void BlenderSync::sync_mesh_motion(BL::Object& b_ob,
 }
 
 CCL_NAMESPACE_END
-
diff --git a/intern/cycles/blender/blender_shader.cpp b/intern/cycles/blender/blender_shader.cpp
index f6ec015e1b8..8baa53fc2ec 100644
--- a/intern/cycles/blender/blender_shader.cpp
+++ b/intern/cycles/blender/blender_shader.cpp
@@ -609,7 +609,8 @@ static ShaderNode *add_node(Scene *scene,
 			bool is_builtin = b_image.packed_file() ||
 			                  b_image.source() == BL::Image::source_GENERATED ||
 			                  b_image.source() == BL::Image::source_MOVIE ||
-			                  b_engine.is_preview();
+			                  (b_engine.is_preview() &&
+			                   b_image.source() != BL::Image::source_SEQUENCE);
 
 			if(is_builtin) {
 				/* for builtin images we're using image datablock name to find an image to
@@ -662,7 +663,8 @@ static ShaderNode *add_node(Scene *scene,
 			bool is_builtin = b_image.packed_file() ||
 			                  b_image.source() == BL::Image::source_GENERATED ||
 			                  b_image.source() == BL::Image::source_MOVIE ||
-			                  b_engine.is_preview();
+			                  (b_engine.is_preview() &&
+			                   b_image.source() != BL::Image::source_SEQUENCE);
 
 			if(is_builtin) {
 				int scene_frame = b_scene.frame_current();
diff --git a/intern/cycles/blender/blender_util.h b/intern/cycles/blender/blender_util.h
index b67834cdea3..4411181dbcc 100644
--- a/intern/cycles/blender/blender_util.h
+++ b/intern/cycles/blender/blender_util.h
@@ -19,6 +19,7 @@
 
 #include "mesh.h"
 
+#include "util_algorithm.h"
 #include "util_map.h"
 #include "util_path.h"
 #include "util_set.h"
@@ -79,6 +80,7 @@ static inline BL::Mesh object_to_mesh(BL::BlendData& data,
 			}
 			else {
 				me.split_faces();
+				me.calc_normals_split();
 			}
 		}
 		if(subdivision_type == Mesh::SUBDIVISION_NONE) {
@@ -786,6 +788,35 @@ struct ParticleSystemKey {
 	}
 };
 
+class EdgeMap {
+public:
+	EdgeMap() {
+	}
+
+	void clear() {
+		edges_.clear();
+	}
+
+	void insert(int v0, int v1) {
+		get_sorted_verts(v0, v1);
+		edges_.insert(std::pair<int, int>(v0, v1));
+	}
+
+	bool exists(int v0, int v1) {
+		get_sorted_verts(v0, v1);
+		return edges_.find(std::pair<int, int>(v0, v1)) != edges_.end();
+	}
+
+protected:
+	void get_sorted_verts(int& v0, int& v1) {
+		if(v0 > v1) {
+			swap(v0, v1);
+		}
+	}
+
+	set< std::pair<int, int> > edges_;
+};
+
 CCL_NAMESPACE_END
 
 #endif /* __BLENDER_UTIL_H__ */
diff --git a/intern/cycles/bvh/bvh.cpp b/intern/cycles/bvh/bvh.cpp
index 874a4246d1d..1fb2f371a0f 100644
--- a/intern/cycles/bvh/bvh.cpp
+++ b/intern/cycles/bvh/bvh.cpp
@@ -81,6 +81,7 @@ void BVH::build(Progress& progress)
 	                   pack.prim_type,
 	                   pack.prim_index,
 	                   pack.prim_object,
+	                   pack.prim_time,
 	                   params,
 	                   progress);
 	BVHNode *root = bvh_build.run();
@@ -256,6 +257,10 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
 	pack.leaf_nodes.resize(leaf_nodes_size);
 	pack.object_node.resize(objects.size());
 
+	if(params.num_motion_curve_steps > 0 || params.num_motion_triangle_steps > 0) {
+		pack.prim_time.resize(prim_index_size);
+	}
+
 	int *pack_prim_index = (pack.prim_index.size())? &pack.prim_index[0]: NULL;
 	int *pack_prim_type = (pack.prim_type.size())? &pack.prim_type[0]: NULL;
 	int *pack_prim_object = (pack.prim_object.size())? &pack.prim_object[0]: NULL;
@@ -264,6 +269,7 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
 	uint *pack_prim_tri_index = (pack.prim_tri_index.size())? &pack.prim_tri_index[0]: NULL;
 	int4 *pack_nodes = (pack.nodes.size())? &pack.nodes[0]: NULL;
 	int4 *pack_leaf_nodes = (pack.leaf_nodes.size())? &pack.leaf_nodes[0]: NULL;
+	float2 *pack_prim_time = (pack.prim_time.size())? &pack.prim_time[0]: NULL;
 
 	/* merge */
 	foreach(Object *ob, objects) {
@@ -309,6 +315,7 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
 			int *bvh_prim_type = &bvh->pack.prim_type[0];
 			uint *bvh_prim_visibility = &bvh->pack.prim_visibility[0];
 			uint *bvh_prim_tri_index = &bvh->pack.prim_tri_index[0];
+			float2 *bvh_prim_time = bvh->pack.prim_time.size()? &bvh->pack.prim_time[0]: NULL;
 
 			for(size_t i = 0; i < bvh_prim_index_size; i++) {
 				if(bvh->pack.prim_type[i] & PRIMITIVE_ALL_CURVE) {
@@ -324,6 +331,9 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
 				pack_prim_type[pack_prim_index_offset] = bvh_prim_type[i];
 				pack_prim_visibility[pack_prim_index_offset] = bvh_prim_visibility[i];
 				pack_prim_object[pack_prim_index_offset] = 0;  // unused for instances
+				if(bvh_prim_time != NULL) {
+					pack_prim_time[pack_prim_index_offset] = bvh_prim_time[i];
+				}
 				pack_prim_index_offset++;
 			}
 		}
diff --git a/intern/cycles/bvh/bvh.h b/intern/cycles/bvh/bvh.h
index 35f4d305883..08f41fc736f 100644
--- a/intern/cycles/bvh/bvh.h
+++ b/intern/cycles/bvh/bvh.h
@@ -68,6 +68,8 @@ struct PackedBVH {
 	array<int> prim_index;
 	/* mapping from BVH primitive index, to the object id of that primitive. */
 	array<int> prim_object;
+	/* Time range of BVH primitive. */
+	array<float2> prim_time;
 
 	/* index of the root node. */
 	int root_index;
diff --git a/intern/cycles/bvh/bvh_build.cpp b/intern/cycles/bvh/bvh_build.cpp
index a2f8b33cb0b..fcbc50f4f6f 100644
--- a/intern/cycles/bvh/bvh_build.cpp
+++ b/intern/cycles/bvh/bvh_build.cpp
@@ -93,12 +93,14 @@ BVHBuild::BVHBuild(const vector<Object*>& objects_,
                    array<int>& prim_type_,
                    array<int>& prim_index_,
                    array<int>& prim_object_,
+                   array<float2>& prim_time_,
                    const BVHParams& params_,
                    Progress& progress_)
  : objects(objects_),
    prim_type(prim_type_),
    prim_index(prim_index_),
    prim_object(prim_object_),
+   prim_time(prim_time_),
    params(params_),
    progress(progress_),
    progress_start_time(0.0),
@@ -465,6 +467,9 @@ BVHNode* BVHBuild::run()
 	}
 	spatial_free_index = 0;
 
+	need_prim_time = params.num_motion_curve_steps > 0 ||
+	                 params.num_motion_triangle_steps > 0;
+
 	/* init progress updates */
 	double build_start_time;
 	build_start_time = progress_start_time = time_dt();
@@ -475,6 +480,12 @@ BVHNode* BVHBuild::run()
 	prim_type.resize(references.size());
 	prim_index.resize(references.size());
 	prim_object.resize(references.size());
+	if(need_prim_time) {
+		prim_time.resize(references.size());
+	}
+	else {
+		prim_time.resize(0);
+	}
 
 	/* build recursively */
 	BVHNode *rootnode;
@@ -849,6 +860,9 @@ BVHNode *BVHBuild::create_object_leaf_nodes(const BVHReference *ref, int start,
 		prim_type[start] = ref->prim_type();
 		prim_index[start] = ref->prim_index();
 		prim_object[start] = ref->prim_object();
+		if(need_prim_time) {
+			prim_time[start] = make_float2(ref->time_from(), ref->time_to());
+		}
 
 		uint visibility = objects[ref->prim_object()]->visibility;
 		BVHNode *leaf_node =  new LeafNode(ref->bounds(), visibility, start, start+1);
@@ -896,6 +910,7 @@ BVHNode* BVHBuild::create_leaf_node(const BVHRange& range,
 	vector<int, LeafStackAllocator> p_type[PRIMITIVE_NUM_TOTAL];
 	vector<int, LeafStackAllocator> p_index[PRIMITIVE_NUM_TOTAL];
 	vector<int, LeafStackAllocator> p_object[PRIMITIVE_NUM_TOTAL];
+	vector<float2, LeafStackAllocator> p_time[PRIMITIVE_NUM_TOTAL];
 	vector<BVHReference, LeafReferenceStackAllocator> p_ref[PRIMITIVE_NUM_TOTAL];
 
 	/* TODO(sergey): In theory we should be able to store references. */
@@ -918,6 +933,8 @@ BVHNode* BVHBuild::create_leaf_node(const BVHRange& range,
 			p_type[type_index].push_back(ref.prim_type());
 			p_index[type_index].push_back(ref.prim_index());
 			p_object[type_index].push_back(ref.prim_object());
+			p_time[type_index].push_back(make_float2(ref.time_from(),
+			                                         ref.time_to()));
 
 			bounds[type_index].grow(ref.bounds());
 			visibility[type_index] |= objects[ref.prim_object()]->visibility;
@@ -947,9 +964,13 @@ BVHNode* BVHBuild::create_leaf_node(const BVHRange& range,
 	vector<int, LeafStackAllocator> local_prim_type,
 	                                local_prim_index,
 	                                local_prim_object;
+	vector<float2, LeafStackAllocator> local_prim_time;
 	local_prim_type.resize(num_new_prims);
 	local_prim_index.resize(num_new_prims);
 	local_prim_object.resize(num_new_prims);
+	if(need_prim_time) {
+		local_prim_time.resize(num_new_prims);
+	}
 	for(int i = 0; i < PRIMITIVE_NUM_TOTAL; ++i) {
 		int num = (int)p_type[i].size();
 		if(num != 0) {
@@ -962,6 +983,9 @@ BVHNode* BVHBuild::create_leaf_node(const BVHRange& range,
 				local_prim_type[index] = p_type[i][j];
 				local_prim_index[index] = p_index[i][j];
 				local_prim_object[index] = p_object[i][j];
+				if(need_prim_time) {
+					local_prim_time[index] = p_time[i][j];
+				}
 				if(params.use_unaligned_nodes && !alignment_found) {
 					alignment_found =
 						unaligned_heuristic.compute_aligned_space(p_ref[i][j],
@@ -1028,11 +1052,17 @@ BVHNode* BVHBuild::create_leaf_node(const BVHRange& range,
 				prim_type.reserve(reserve);
 				prim_index.reserve(reserve);
 				prim_object.reserve(reserve);
+				if(need_prim_time) {
+					prim_time.reserve(reserve);
+				}
 			}
 
 			prim_type.resize(range_end);
 			prim_index.resize(range_end);
 			prim_object.resize(range_end);
+			if(need_prim_time) {
+				prim_time.resize(range_end);
+			}
 		}
 		spatial_spin_lock.unlock();
 
@@ -1041,6 +1071,9 @@ BVHNode* BVHBuild::create_leaf_node(const BVHRange& range,
 			memcpy(&prim_type[start_index], &local_prim_type[0], new_leaf_data_size);
 			memcpy(&prim_index[start_index], &local_prim_index[0], new_leaf_data_size);
 			memcpy(&prim_object[start_index], &local_prim_object[0], new_leaf_data_size);
+			if(need_prim_time) {
+				memcpy(&prim_time[start_index], &local_prim_time[0], sizeof(float2)*num_new_leaf_data);
+			}
 		}
 	}
 	else {
@@ -1053,6 +1086,9 @@ BVHNode* BVHBuild::create_leaf_node(const BVHRange& range,
 			memcpy(&prim_type[start_index], &local_prim_type[0], new_leaf_data_size);
 			memcpy(&prim_index[start_index], &local_prim_index[0], new_leaf_data_size);
 			memcpy(&prim_object[start_index], &local_prim_object[0], new_leaf_data_size);
+			if(need_prim_time) {
+				memcpy(&prim_time[start_index], &local_prim_time[0], sizeof(float2)*num_new_leaf_data);
+			}
 		}
 	}
 
diff --git a/intern/cycles/bvh/bvh_build.h b/intern/cycles/bvh/bvh_build.h
index ee3cde66a2f..430efc3e0f6 100644
--- a/intern/cycles/bvh/bvh_build.h
+++ b/intern/cycles/bvh/bvh_build.h
@@ -48,6 +48,7 @@ public:
 	         array<int>& prim_type,
 	         array<int>& prim_index,
 	         array<int>& prim_object,
+	         array<float2>& prim_time,
 	         const BVHParams& params,
 	         Progress& progress);
 	~BVHBuild();
@@ -112,6 +113,9 @@ protected:
 	array<int>& prim_type;
 	array<int>& prim_index;
 	array<int>& prim_object;
+	array<float2>& prim_time;
+
+	bool need_prim_time;
 
 	/* Build parameters. */
 	BVHParams params;
diff --git a/intern/cycles/bvh/bvh_params.h b/intern/cycles/bvh/bvh_params.h
index 65f9da1c194..7b309504728 100644
--- a/intern/cycles/bvh/bvh_params.h
+++ b/intern/cycles/bvh/bvh_params.h
@@ -104,6 +104,7 @@ public:
 		primitive_mask = PRIMITIVE_ALL;
 
 		num_motion_curve_steps = 0;
+		num_motion_triangle_steps = 0;
 	}
 
 	/* SAH costs */
diff --git a/intern/cycles/kernel/bvh/bvh.h b/intern/cycles/kernel/bvh/bvh.h
index 36798982653..321983c1abc 100644
--- a/intern/cycles/kernel/bvh/bvh.h
+++ b/intern/cycles/kernel/bvh/bvh.h
@@ -357,7 +357,7 @@ ccl_device_inline float3 ray_offset(float3 P, float3 Ng)
 #endif
 }
 
-#if defined(__SHADOW_RECORD_ALL__) || defined (__VOLUME_RECORD_ALL__)
+#if defined(__VOLUME_RECORD_ALL__) || (defined(__SHADOW_RECORD_ALL__) && defined(__KERNEL_CPU__))
 /* ToDo: Move to another file? */
 ccl_device int intersections_compare(const void *a, const void *b)
 {
@@ -373,5 +373,28 @@ ccl_device int intersections_compare(const void *a, const void *b)
 }
 #endif
 
-CCL_NAMESPACE_END
+#if defined(__SHADOW_RECORD_ALL__)
+ccl_device_inline void sort_intersections(Intersection *hits, uint num_hits)
+{
+#ifdef __KERNEL_GPU__
+	/* Use bubble sort which has more friendly memory pattern on GPU. */
+	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 = hits[j];
+				hits[j] = hits[j + 1];
+				hits[j + 1] = tmp;
+				swapped = true;
+			}
+		}
+		--num_hits;
+	} while(swapped);
+#else
+	qsort(hits, num_hits, sizeof(Intersection), intersections_compare);
+#endif
+}
+#endif  /* __SHADOW_RECORD_ALL__ | __VOLUME_RECORD_ALL__ */
 
+CCL_NAMESPACE_END
diff --git a/intern/cycles/kernel/geom/geom_curve.h b/intern/cycles/kernel/geom/geom_curve.h
index 9de335403ce..712b67a1b55 100644
--- a/intern/cycles/kernel/geom/geom_curve.h
+++ b/intern/cycles/kernel/geom/geom_curve.h
@@ -229,6 +229,15 @@ ccl_device_forceinline bool bvh_cardinal_curve_intersect(KernelGlobals *kg, Inte
 	float3 P, float3 dir, uint visibility, int object, int curveAddr, float time,int type, uint *lcg_state, float difl, float extmax)
 #endif
 {
+	const bool is_curve_primitive = (type & PRIMITIVE_CURVE);
+
+	if(!is_curve_primitive && kernel_data.bvh.use_bvh_steps) {
+		const float2 prim_time = kernel_tex_fetch(__prim_time, curveAddr);
+		if(time < prim_time.x || time > prim_time.y) {
+			return false;
+		}
+	}
+
 	int segment = PRIMITIVE_UNPACK_SEGMENT(type);
 	float epsilon = 0.0f;
 	float r_st, r_en;
@@ -257,7 +266,7 @@ ccl_device_forceinline bool bvh_cardinal_curve_intersect(KernelGlobals *kg, Inte
 
 #ifdef __KERNEL_AVX2__
 		avxf P_curve_0_1, P_curve_2_3;
-		if(type & PRIMITIVE_CURVE) {
+		if(is_curve_primitive) {
 			P_curve_0_1 = _mm256_loadu2_m128(&kg->__curve_keys.data[k0].x, &kg->__curve_keys.data[ka].x);
 			P_curve_2_3 = _mm256_loadu2_m128(&kg->__curve_keys.data[kb].x, &kg->__curve_keys.data[k1].x);
 		}
@@ -268,7 +277,7 @@ ccl_device_forceinline bool bvh_cardinal_curve_intersect(KernelGlobals *kg, Inte
 #else  /* __KERNEL_AVX2__ */
 		ssef P_curve[4];
 
-		if(type & PRIMITIVE_CURVE) {
+		if(is_curve_primitive) {
 			P_curve[0] = load4f(&kg->__curve_keys.data[ka].x);
 			P_curve[1] = load4f(&kg->__curve_keys.data[k0].x);
 			P_curve[2] = load4f(&kg->__curve_keys.data[k1].x);
@@ -363,7 +372,7 @@ ccl_device_forceinline bool bvh_cardinal_curve_intersect(KernelGlobals *kg, Inte
 
 		float4 P_curve[4];
 
-		if(type & PRIMITIVE_CURVE) {
+		if(is_curve_primitive) {
 			P_curve[0] = kernel_tex_fetch(__curve_keys, ka);
 			P_curve[1] = kernel_tex_fetch(__curve_keys, k0);
 			P_curve[2] = kernel_tex_fetch(__curve_keys, k1);
@@ -689,6 +698,15 @@ ccl_device_forceinline bool bvh_curve_intersect(KernelGlobals *kg, Intersection
 #  define dot3(x, y) dot(x, y)
 #endif
 
+	const bool is_curve_primitive = (type & PRIMITIVE_CURVE);
+
+	if(!is_curve_primitive && kernel_data.bvh.use_bvh_steps) {
+		const float2 prim_time = kernel_tex_fetch(__prim_time, curveAddr);
+		if(time < prim_time.x || time > prim_time.y) {
+			return false;
+		}
+	}
+
 	int segment = PRIMITIVE_UNPACK_SEGMENT(type);
 	/* curve Intersection check */
 	int flags = kernel_data.curve.curveflags;
@@ -703,7 +721,7 @@ ccl_device_forceinline bool bvh_curve_intersect(KernelGlobals *kg, Intersection
 #ifndef __KERNEL_SSE2__
 	float4 P_curve[2];
 
-	if(type & PRIMITIVE_CURVE) {
+	if(is_curve_primitive) {
 		P_curve[0] = kernel_tex_fetch(__curve_keys, k0);
 		P_curve[1] = kernel_tex_fetch(__curve_keys, k1);
 	}
@@ -738,7 +756,7 @@ ccl_device_forceinline bool bvh_curve_intersect(KernelGlobals *kg, Intersection
 #else
 	ssef P_curve[2];
 	
-	if(type & PRIMITIVE_CURVE) {
+	if(is_curve_primitive) {
 		P_curve[0] = load4f(&kg->__curve_keys.data[k0].x);
 		P_curve[1] = load4f(&kg->__curve_keys.data[k1].x);
 	}
diff --git a/intern/cycles/kernel/kernel_globals.h b/intern/cycles/kernel/kernel_globals.h
index 8e66a3a0340..2b52a2d2f48 100644
--- a/intern/cycles/kernel/kernel_globals.h
+++ b/intern/cycles/kernel/kernel_globals.h
@@ -76,7 +76,10 @@ typedef struct KernelGlobals {
 #ifdef __KERNEL_CUDA__
 
 __constant__ KernelData __data;
-typedef struct KernelGlobals {} KernelGlobals;
+typedef struct KernelGlobals {
+	/* NOTE: Keep the size in sync with SHADOW_STACK_MAX_HITS. */
+	Intersection hits_stack[64];
+} KernelGlobals;
 
 #  ifdef __KERNEL_CUDA_TEX_STORAGE__
 #    define KERNEL_TEX(type, ttype, name) ttype name;
diff --git a/intern/cycles/kernel/kernel_shadow.h b/intern/cycles/kernel/kernel_shadow.h
index 2981f6ac566..06a77a208cb 100644
--- a/intern/cycles/kernel/kernel_shadow.h
+++ b/intern/cycles/kernel/kernel_shadow.h
@@ -16,9 +16,84 @@
 
 CCL_NAMESPACE_BEGIN
 
-#ifdef __SHADOW_RECORD_ALL__
+/* Attenuate throughput accordingly to the given intersection event.
+ * Returns true if the throughput is zero and traversal can be aborted.
+ */
+ccl_device_forceinline bool shadow_handle_transparent_isect(
+        KernelGlobals *kg,
+        ShaderData *shadow_sd,
+        ccl_addr_space PathState *state,
+#    ifdef __VOLUME__
+        struct PathState *volume_state,
+#    endif
+        Intersection *isect,
+        Ray *ray,
+        float3 *throughput)
+{
+#ifdef __VOLUME__
+	/* Attenuation between last surface and next surface. */
+	if(volume_state->volume_stack[0].shader != SHADER_NONE) {
+		Ray segment_ray = *ray;
+		segment_ray.t = isect->t;
+		kernel_volume_shadow(kg,
+		                     shadow_sd,
+		                     volume_state,
+		                     &segment_ray,
+		                     throughput);
+	}
+#endif
+	/* Setup shader data at surface. */
+	shader_setup_from_ray(kg, shadow_sd, isect, ray);
+	/* Attenuation from transparent surface. */
+	if(!(ccl_fetch(shadow_sd, flag) & SD_HAS_ONLY_VOLUME)) {
+		path_state_modify_bounce(state, true);
+		shader_eval_surface(kg,
+		                    shadow_sd,
+		                    NULL,
+		                    state,
+		                    0.0f,
+		                    PATH_RAY_SHADOW,
+		                    SHADER_CONTEXT_SHADOW);
+		path_state_modify_bounce(state, false);
+		*throughput *= shader_bsdf_transparency(kg, shadow_sd);
+	}
+	/* Stop if all light is blocked. */
+	if(is_zero(*throughput)) {
+		return true;
+	}
+#ifdef __VOLUME__
+	/* Exit/enter volume. */
+	kernel_volume_stack_enter_exit(kg, shadow_sd, volume_state->volume_stack);
+#endif
+	return false;
+}
+
+/* Special version which only handles opaque shadows. */
+ccl_device bool shadow_blocked_opaque(KernelGlobals *kg,
+                                      ShaderData *shadow_sd,
+                                      ccl_addr_space PathState *state,
+                                      Ray *ray,
+                                      Intersection *isect,
+                                      float3 *shadow)
+{
+	const bool blocked = scene_intersect(kg,
+	                                     *ray,
+	                                     PATH_RAY_SHADOW_OPAQUE,
+	                                     isect,
+	                                     NULL,
+	                                     0.0f, 0.0f);
+#ifdef __VOLUME__
+	if(!blocked && state->volume_stack[0].shader != SHADER_NONE) {
+		/* Apply attenuation from current volume shader. */
+		kernel_volume_shadow(kg, shadow_sd, state, ray, shadow);
+	}
+#endif
+	return blocked;
+}
 
-/* Shadow function to compute how much light is blocked, CPU variation.
+#ifdef __TRANSPARENT_SHADOWS__
+#  ifdef __SHADOW_RECORD_ALL__
+/* Shadow function to compute how much light is blocked,
  *
  * We trace a single ray. If it hits any opaque surface, or more than a given
  * number of transparent surfaces is hit, then we consider the geometry to be
@@ -36,261 +111,355 @@ CCL_NAMESPACE_BEGIN
  * or there is a performance increase anyway due to avoiding the need to send
  * two rays with transparent shadows.
  *
- * This is CPU only because of qsort, and malloc or high stack space usage to
- * record all these intersections. */
+ * On CPU it'll handle all transparent bounces (by allocating storage for
+ * intersections when they don't fit into the stack storage).
+ *
+ * On GPU it'll only handle SHADOW_STACK_MAX_HITS-1 intersections, so this
+ * is something to be kept an eye on.
+ */
 
-#define STACK_MAX_HITS 64
+#    define SHADOW_STACK_MAX_HITS 64
 
-ccl_device_inline bool shadow_blocked(KernelGlobals *kg, ShaderData *shadow_sd, PathState *state, Ray *ray, float3 *shadow)
+/* Actual logic with traversal loop implementation which is free from device
+ * specific tweaks.
+ *
+ * Note that hits array should be as big as max_hits+1.
+ */
+ccl_device bool shadow_blocked_transparent_all_loop(KernelGlobals *kg,
+                                                    ShaderData *shadow_sd,
+                                                    ccl_addr_space PathState *state,
+                                                    Ray *ray,
+                                                    Intersection *hits,
+                                                    uint max_hits,
+                                                    float3 *shadow)
 {
-	*shadow = make_float3(1.0f, 1.0f, 1.0f);
-
-	if(ray->t == 0.0f)
-		return false;
-	
-	bool blocked;
-
-	if(kernel_data.integrator.transparent_shadows) {
-		/* check transparent bounces here, for volume scatter which can do
-		 * lighting before surface path termination is checked */
-		if(state->transparent_bounce >= kernel_data.integrator.transparent_max_bounce)
-			return true;
-
-		/* intersect to find an opaque surface, or record all transparent surface hits */
-		Intersection hits_stack[STACK_MAX_HITS];
-		Intersection *hits = hits_stack;
-		const int transparent_max_bounce = kernel_data.integrator.transparent_max_bounce;
-		uint max_hits = transparent_max_bounce - state->transparent_bounce - 1;
-
-		/* prefer to use stack but use dynamic allocation if too deep max hits
-		 * we need max_hits + 1 storage space due to the logic in
-		 * scene_intersect_shadow_all which will first store and then check if
-		 * the limit is exceeded */
-		if(max_hits + 1 > STACK_MAX_HITS) {
-			if(kg->transparent_shadow_intersections == NULL) {
-				kg->transparent_shadow_intersections =
-				    (Intersection*)malloc(sizeof(Intersection)*(transparent_max_bounce + 1));
+	/* Intersect to find an opaque surface, or record all transparent
+	 * surface hits.
+	 */
+	uint num_hits;
+	const bool blocked = scene_intersect_shadow_all(kg,
+	                                                ray,
+	                                                hits,
+	                                                max_hits,
+	                                                &num_hits);
+	/* If no opaque surface found but we did find transparent hits,
+	 * shade them.
+	 */
+	if(!blocked && num_hits > 0) {
+		float3 throughput = make_float3(1.0f, 1.0f, 1.0f);
+		float3 Pend = ray->P + ray->D*ray->t;
+		float last_t = 0.0f;
+		int bounce = state->transparent_bounce;
+		Intersection *isect = hits;
+#    ifdef __VOLUME__
+		PathState ps = *state;
+#    endif
+		sort_intersections(hits, num_hits);
+		for(int hit = 0; hit < num_hits; hit++, isect++) {
+			/* Adjust intersection distance for moving ray forward. */
+			float new_t = isect->t;
+			isect->t -= last_t;
+			/* Skip hit if we did not move forward, step by step raytracing
+			 * would have skipped it as well then.
+			 */
+			if(last_t == new_t) {
+				continue;
 			}
-			hits = kg->transparent_shadow_intersections;
-		}
-
-		uint num_hits;
-		blocked = scene_intersect_shadow_all(kg, ray, hits, max_hits, &num_hits);
-
-		/* if no opaque surface found but we did find transparent hits, shade them */
-		if(!blocked && num_hits > 0) {
-			float3 throughput = make_float3(1.0f, 1.0f, 1.0f);
-			float3 Pend = ray->P + ray->D*ray->t;
-			float last_t = 0.0f;
-			int bounce = state->transparent_bounce;
-			Intersection *isect = hits;
+			last_t = new_t;
+			/* Attenuate the throughput. */
+			if(shadow_handle_transparent_isect(kg,
+			                                   shadow_sd,
+			                                   state,
 #ifdef __VOLUME__
-			PathState ps = *state;
+			                                   &ps,
 #endif
-
-			qsort(hits, num_hits, sizeof(Intersection), intersections_compare);
-
-			for(int hit = 0; hit < num_hits; hit++, isect++) {
-				/* adjust intersection distance for moving ray forward */
-				float new_t = isect->t;
-				isect->t -= last_t;
-
-				/* skip hit if we did not move forward, step by step raytracing
-				 * would have skipped it as well then */
-				if(last_t == new_t)
-					continue;
-
-				last_t = new_t;
-
-#ifdef __VOLUME__
-				/* attenuation between last surface and next surface */
-				if(ps.volume_stack[0].shader != SHADER_NONE) {
-					Ray segment_ray = *ray;
-					segment_ray.t = isect->t;
-					kernel_volume_shadow(kg, shadow_sd, &ps, &segment_ray, &throughput);
-				}
-#endif
-
-				/* setup shader data at surface */
-				shader_setup_from_ray(kg, shadow_sd, isect, ray);
-
-				/* attenuation from transparent surface */
-				if(!(shadow_sd->flag & SD_HAS_ONLY_VOLUME)) {
-					path_state_modify_bounce(state, true);
-					shader_eval_surface(kg, shadow_sd, NULL, state, 0.0f, PATH_RAY_SHADOW, SHADER_CONTEXT_SHADOW);
-					path_state_modify_bounce(state, false);
-
-					throughput *= shader_bsdf_transparency(kg, shadow_sd);
-				}
-
-				/* stop if all light is blocked */
-				if(is_zero(throughput)) {
-					return true;
-				}
-
-				/* move ray forward */
-				ray->P = shadow_sd->P;
-				if(ray->t != FLT_MAX) {
-					ray->D = normalize_len(Pend - ray->P, &ray->t);
-				}
-
-#ifdef __VOLUME__
-				/* exit/enter volume */
-				kernel_volume_stack_enter_exit(kg, shadow_sd, ps.volume_stack);
-#endif
-
-				bounce++;
+			                                   isect,
+			                                   ray,
+			                                   &throughput))
+			{
+				return true;
 			}
-
-#ifdef __VOLUME__
-			/* attenuation for last line segment towards light */
-			if(ps.volume_stack[0].shader != SHADER_NONE)
-				kernel_volume_shadow(kg, shadow_sd, &ps, ray, &throughput);
-#endif
-
-			*shadow = throughput;
-
-			return is_zero(throughput);
+			/* Move ray forward. */
+			ray->P = ccl_fetch(shadow_sd, P);
+			if(ray->t != FLT_MAX) {
+				ray->D = normalize_len(Pend - ray->P, &ray->t);
+			}
+			bounce++;
 		}
+#    ifdef __VOLUME__
+		/* Attenuation for last line segment towards light. */
+		if(ps.volume_stack[0].shader != SHADER_NONE) {
+			kernel_volume_shadow(kg, shadow_sd, &ps, ray, &throughput);
+		}
+#    endif
+		*shadow = throughput;
+		return is_zero(throughput);
 	}
-	else {
-		Intersection isect;
-		blocked = scene_intersect(kg, *ray, PATH_RAY_SHADOW_OPAQUE, &isect, NULL, 0.0f, 0.0f);
-	}
-
-#ifdef __VOLUME__
+#    ifdef __VOLUME__
 	if(!blocked && state->volume_stack[0].shader != SHADER_NONE) {
-		/* apply attenuation from current volume shader */
+		/* Apply attenuation from current volume shader/ */
 		kernel_volume_shadow(kg, shadow_sd, state, ray, shadow);
 	}
-#endif
-
+#    endif
 	return blocked;
 }
 
-#undef STACK_MAX_HITS
-
-#else
+/* Here we do all device specific trickery before invoking actual traversal
+ * loop to help readability of the actual logic.
+ */
+ccl_device bool shadow_blocked_transparent_all(KernelGlobals *kg,
+                                               ShaderData *shadow_sd,
+                                               ccl_addr_space PathState *state,
+                                               Ray *ray,
+                                               uint max_hits,
+                                               float3 *shadow)
+{
+#    ifdef __KERNEL_CUDA__
+	Intersection *hits = kg->hits_stack;
+#    else
+	Intersection hits_stack[SHADOW_STACK_MAX_HITS];
+	Intersection *hits = hits_stack;
+#    endif
+#    ifndef __KERNEL_GPU__
+	/* Prefer to use stack but use dynamic allocation if too deep max hits
+	 * we need max_hits + 1 storage space due to the logic in
+	 * scene_intersect_shadow_all which will first store and then check if
+	 * the limit is exceeded.
+	 *
+	 * Ignore this on GPU because of slow/unavailable malloc().
+	 */
+	if(max_hits + 1 > SHADOW_STACK_MAX_HITS) {
+		if(kg->transparent_shadow_intersections == NULL) {
+			const int transparent_max_bounce = kernel_data.integrator.transparent_max_bounce;
+			kg->transparent_shadow_intersections =
+				(Intersection*)malloc(sizeof(Intersection)*(transparent_max_bounce + 1));
+		}
+		hits = kg->transparent_shadow_intersections;
+	}
+#    endif  /* __KERNEL_GPU__ */
+	/* Invoke actual traversal. */
+	return shadow_blocked_transparent_all_loop(kg,
+	                                           shadow_sd,
+	                                           state,
+	                                           ray,
+	                                           hits,
+	                                           max_hits,
+	                                           shadow);
+}
+#  endif  /* __SHADOW_RECORD_ALL__ */
 
-/* Shadow function to compute how much light is blocked, GPU variation.
+#  ifdef __KERNEL_GPU__
+/* Shadow function to compute how much light is blocked,
  *
  * Here we raytrace from one transparent surface to the next step by step.
  * To minimize overhead in cases where we don't need transparent shadows, we
  * first trace a regular shadow ray. We check if the hit primitive was
  * potentially transparent, and only in that case start marching. this gives
- * one extra ray cast for the cases were we do want transparency. */
+ * one extra ray cast for the cases were we do want transparency.
+ */
 
-ccl_device_noinline bool shadow_blocked(KernelGlobals *kg,
-                                        ShaderData *shadow_sd,
-                                        ccl_addr_space PathState *state,
-                                        ccl_addr_space Ray *ray_input,
-                                        float3 *shadow)
+/* This function is only implementing device-independent traversal logic
+ * which requires some precalculation done.
+ */
+ccl_device bool shadow_blocked_transparent_stepped_loop(
+        KernelGlobals *kg,
+        ShaderData *shadow_sd,
+        ccl_addr_space PathState *state,
+        Ray *ray,
+        Intersection *isect,
+        const bool blocked,
+        const bool is_transparent_isect,
+        float3 *shadow)
 {
-	*shadow = make_float3(1.0f, 1.0f, 1.0f);
+	if(blocked && is_transparent_isect) {
+		float3 throughput = make_float3(1.0f, 1.0f, 1.0f);
+		float3 Pend = ray->P + ray->D*ray->t;
+		int bounce = state->transparent_bounce;
+#    ifdef __VOLUME__
+		PathState ps = *state;
+#    endif
+		for(;;) {
+			if(bounce >= kernel_data.integrator.transparent_max_bounce) {
+				return true;
+			}
+			if(!scene_intersect(kg,
+			                    *ray,
+			                    PATH_RAY_SHADOW_TRANSPARENT,
+			                    isect,
+			                    NULL,
+			                    0.0f, 0.0f))
+			{
+				break;
+			}
+			if(!shader_transparent_shadow(kg, isect)) {
+				return true;
+			}
+			/* Attenuate the throughput. */
+			if(shadow_handle_transparent_isect(kg,
+			                                   shadow_sd,
+			                                   state,
+#ifdef __VOLUME__
+			                                   &ps,
+#endif
+			                                   isect,
+			                                   ray,
+			                                   &throughput))
+			{
+				return true;
+			}
+			/* Move ray forward. */
+			ray->P = ray_offset(ccl_fetch(shadow_sd, P), -ccl_fetch(shadow_sd, Ng));
+			if(ray->t != FLT_MAX) {
+				ray->D = normalize_len(Pend - ray->P, &ray->t);
+			}
+			bounce++;
+		}
+#    ifdef __VOLUME__
+		/* Attenuation for last line segment towards light. */
+		if(ps.volume_stack[0].shader != SHADER_NONE) {
+			kernel_volume_shadow(kg, shadow_sd, &ps, ray, &throughput);
+		}
+#    endif
+		*shadow *= throughput;
+		return is_zero(throughput);
+	}
+#    ifdef __VOLUME__
+	if(!blocked && state->volume_stack[0].shader != SHADER_NONE) {
+		/* Apply attenuation from current volume shader. */
+		kernel_volume_shadow(kg, shadow_sd, state, ray, shadow);
+	}
+#    endif
+	return blocked;
+}
 
-	if(ray_input->t == 0.0f)
-		return false;
+ccl_device bool shadow_blocked_transparent_stepped(
+        KernelGlobals *kg,
+        ShaderData *shadow_sd,
+        ccl_addr_space PathState *state,
+        Ray *ray,
+        Intersection *isect,
+        float3 *shadow)
+{
+	const bool blocked = scene_intersect(kg,
+	                                     *ray,
+	                                     PATH_RAY_SHADOW_OPAQUE,
+	                                     isect,
+	                                     NULL,
+	                                     0.0f, 0.0f);
+	const bool is_transparent_isect = blocked
+	        ? shader_transparent_shadow(kg, isect)
+	        : false;
+	return shadow_blocked_transparent_stepped_loop(kg,
+	                                               shadow_sd,
+	                                               state,
+	                                               ray,
+	                                               isect,
+	                                               blocked,
+	                                               is_transparent_isect,
+	                                               shadow);
+}
 
+#  endif  /* __KERNEL_GPU__ */
+#endif /* __TRANSPARENT_SHADOWS__ */
+
+ccl_device_inline bool shadow_blocked(KernelGlobals *kg,
+                                      ShaderData *shadow_sd,
+                                      ccl_addr_space PathState *state,
+                                      ccl_addr_space Ray *ray_input,
+                                      float3 *shadow)
+{
+	/* Special trickery for split kernel: some data is coming from the
+	 * global memory.
+	 */
 #ifdef __SPLIT_KERNEL__
 	Ray private_ray = *ray_input;
 	Ray *ray = &private_ray;
-#else
-	Ray *ray = ray_input;
-#endif
-
-#ifdef __SPLIT_KERNEL__
 	Intersection *isect = &kg->isect_shadow[SD_THREAD];
-#else
+#else  /* __SPLIT_KERNEL__ */
+	Ray *ray = ray_input;
 	Intersection isect_object;
 	Intersection *isect = &isect_object;
-#endif
-
-	bool blocked = scene_intersect(kg, *ray, PATH_RAY_SHADOW_OPAQUE, isect, NULL, 0.0f, 0.0f);
-
+#endif  /* __SPLIT_KERNEL__ */
+	/* Some common early checks. */
+	*shadow = make_float3(1.0f, 1.0f, 1.0f);
+	if(ray->t == 0.0f) {
+		return false;
+	}
+	/* Do actual shadow shading. */
+	/* First of all, we check if integrator requires transparent shadows.
+	 * if not, we use simplest and fastest ever way to calculate occlusion.
+	 */
 #ifdef __TRANSPARENT_SHADOWS__
-	if(blocked && kernel_data.integrator.transparent_shadows) {
-		if(shader_transparent_shadow(kg, isect)) {
-			float3 throughput = make_float3(1.0f, 1.0f, 1.0f);
-			float3 Pend = ray->P + ray->D*ray->t;
-			int bounce = state->transparent_bounce;
-#ifdef __VOLUME__
-			PathState ps = *state;
+	if(!kernel_data.integrator.transparent_shadows)
 #endif
-
-			for(;;) {
-				if(bounce >= kernel_data.integrator.transparent_max_bounce)
-					return true;
-
-				if(!scene_intersect(kg, *ray, PATH_RAY_SHADOW_TRANSPARENT, isect, NULL, 0.0f, 0.0f))
-				{
-#ifdef __VOLUME__
-					/* attenuation for last line segment towards light */
-					if(ps.volume_stack[0].shader != SHADER_NONE)
-						kernel_volume_shadow(kg, shadow_sd, &ps, ray, &throughput);
-#endif
-
-					*shadow *= throughput;
-
-					return false;
-				}
-
-				if(!shader_transparent_shadow(kg, isect)) {
-					return true;
-				}
-
-#ifdef __VOLUME__
-				/* attenuation between last surface and next surface */
-				if(ps.volume_stack[0].shader != SHADER_NONE) {
-					Ray segment_ray = *ray;
-					segment_ray.t = isect->t;
-					kernel_volume_shadow(kg, shadow_sd, &ps, &segment_ray, &throughput);
-				}
-#endif
-
-				/* setup shader data at surface */
-				shader_setup_from_ray(kg, shadow_sd, isect, ray);
-
-				/* attenuation from transparent surface */
-				if(!(ccl_fetch(shadow_sd, flag) & SD_HAS_ONLY_VOLUME)) {
-					path_state_modify_bounce(state, true);
-					shader_eval_surface(kg, shadow_sd, NULL, state, 0.0f, PATH_RAY_SHADOW, SHADER_CONTEXT_SHADOW);
-					path_state_modify_bounce(state, false);
-
-					throughput *= shader_bsdf_transparency(kg, shadow_sd);
-				}
-
-				/* stop if all light is blocked */
-				if(is_zero(throughput)) {
-					return true;
-				}
-
-				/* move ray forward */
-				ray->P = ray_offset(ccl_fetch(shadow_sd, P), -ccl_fetch(shadow_sd, Ng));
-				if(ray->t != FLT_MAX) {
-					ray->D = normalize_len(Pend - ray->P, &ray->t);
-				}
-
-#ifdef __VOLUME__
-				/* exit/enter volume */
-				kernel_volume_stack_enter_exit(kg, shadow_sd, ps.volume_stack);
-#endif
-
-				bounce++;
-			}
-		}
+	{
+		return shadow_blocked_opaque(kg,
+		                             shadow_sd,
+		                             state,
+		                             ray,
+		                             isect,
+		                             shadow);
 	}
-#ifdef __VOLUME__
-	else if(!blocked && state->volume_stack[0].shader != SHADER_NONE) {
-		/* apply attenuation from current volume shader */
-		kernel_volume_shadow(kg, shadow_sd, state, ray, shadow);
+#ifdef __TRANSPARENT_SHADOWS__
+#  ifdef __SHADOW_RECORD_ALL__
+	/* For the transparent shadows we try to use record-all logic on the
+	 * devices which supports this.
+	 */
+	const int transparent_max_bounce = kernel_data.integrator.transparent_max_bounce;
+	/* Check transparent bounces here, for volume scatter which can do
+	 * lighting before surface path termination is checked.
+	 */
+	if(state->transparent_bounce >= transparent_max_bounce) {
+		return true;
 	}
-#endif
-#endif
-
-	return blocked;
+	const uint max_hits = transparent_max_bounce - state->transparent_bounce - 1;
+#    ifdef __KERNEL_GPU__
+	/* On GPU we do trickey with tracing opaque ray first, this avoids speed
+	 * regressions in some files.
+	 *
+	 * TODO(sergey): Check why using record-all behavior causes slowdown in such
+	 * cases. Could that be caused by a higher spill pressure?
+	 */
+	const bool blocked = scene_intersect(kg,
+	                                     *ray,
+	                                     PATH_RAY_SHADOW_OPAQUE,
+	                                     isect,
+	                                     NULL,
+	                                     0.0f, 0.0f);
+	const bool is_transparent_isect = blocked
+	        ? shader_transparent_shadow(kg, isect)
+	        : false;
+	if(!blocked || !is_transparent_isect ||
+	   max_hits + 1 >= SHADOW_STACK_MAX_HITS)
+	{
+		return shadow_blocked_transparent_stepped_loop(kg,
+		                                               shadow_sd,
+		                                               state,
+		                                               ray,
+		                                               isect,
+		                                               blocked,
+		                                               is_transparent_isect,
+		                                               shadow);
+	}
+#    endif  /* __KERNEL_GPU__ */
+	return shadow_blocked_transparent_all(kg,
+	                                      shadow_sd,
+	                                      state,
+	                                      ray,
+	                                      max_hits,
+	                                      shadow);
+#  else  /* __SHADOW_RECORD_ALL__ */
+	/* Fallback to a slowest version which works on all devices. */
+	return shadow_blocked_transparent_stepped(kg,
+	                                          shadow_sd,
+	                                          state,
+	                                          ray,
+	                                          isect,
+	                                          shadow);
+#  endif  /* __SHADOW_RECORD_ALL__ */
+#endif  /* __TRANSPARENT_SHADOWS__ */
 }
 
-#endif
+#undef SHADOW_STACK_MAX_HITS
 
 CCL_NAMESPACE_END
-
diff --git a/intern/cycles/kernel/kernel_textures.h b/intern/cycles/kernel/kernel_textures.h
index 8d5bb75a428..cb1a3f40dee 100644
--- a/intern/cycles/kernel/kernel_textures.h
+++ b/intern/cycles/kernel/kernel_textures.h
@@ -32,6 +32,7 @@ KERNEL_TEX(uint, texture_uint, __prim_visibility)
 KERNEL_TEX(uint, texture_uint, __prim_index)
 KERNEL_TEX(uint, texture_uint, __prim_object)
 KERNEL_TEX(uint, texture_uint, __object_node)
+KERNEL_TEX(float2, texture_float2, __prim_time)
 
 /* objects */
 KERNEL_TEX(float4, texture_float4, __objects)
@@ -177,7 +178,6 @@ KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_085)
 KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_086)
 KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_087)
 KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_088)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_089)
 
 #  else
 /* bindless textures */
diff --git a/intern/cycles/kernel/kernel_types.h b/intern/cycles/kernel/kernel_types.h
index 8c271c75e44..8250eaf6073 100644
--- a/intern/cycles/kernel/kernel_types.h
+++ b/intern/cycles/kernel/kernel_types.h
@@ -84,6 +84,7 @@ CCL_NAMESPACE_BEGIN
 #  define __VOLUME_SCATTER__
 #  define __SUBSURFACE__
 #  define __CMJ__
+#  define __SHADOW_RECORD_ALL__
 #endif  /* __KERNEL_CUDA__ */
 
 #ifdef __KERNEL_OPENCL__
@@ -1201,7 +1202,8 @@ typedef struct KernelBVH {
 	int have_curves;
 	int have_instancing;
 	int use_qbvh;
-	int pad1, pad2;
+	int use_bvh_steps;
+	int pad1;
 } KernelBVH;
 static_assert_align(KernelBVH, 16);
 
diff --git a/intern/cycles/kernel/kernels/cuda/kernel.cu b/intern/cycles/kernel/kernels/cuda/kernel.cu
index eb2b6ea5414..090ab2c50c2 100644
--- a/intern/cycles/kernel/kernels/cuda/kernel.cu
+++ b/intern/cycles/kernel/kernels/cuda/kernel.cu
@@ -130,8 +130,10 @@ kernel_cuda_path_trace(float *buffer, uint *rng_state, int sample, int sx, int s
 	int x = sx + blockDim.x*blockIdx.x + threadIdx.x;
 	int y = sy + blockDim.y*blockIdx.y + threadIdx.y;
 
-	if(x < sx + sw && y < sy + sh)
-		kernel_path_trace(NULL, buffer, rng_state, sample, x, y, offset, stride);
+	if(x < sx + sw && y < sy + sh) {
+		KernelGlobals kg;
+		kernel_path_trace(&kg, buffer, rng_state, sample, x, y, offset, stride);
+	}
 }
 
 #ifdef __BRANCHED_PATH__
@@ -142,8 +144,10 @@ kernel_cuda_branched_path_trace(float *buffer, uint *rng_state, int sample, int
 	int x = sx + blockDim.x*blockIdx.x + threadIdx.x;
 	int y = sy + blockDim.y*blockIdx.y + threadIdx.y;
 
-	if(x < sx + sw && y < sy + sh)
-		kernel_branched_path_trace(NULL, buffer, rng_state, sample, x, y, offset, stride);
+	if(x < sx + sw && y < sy + sh) {
+		KernelGlobals kg;
+		kernel_branched_path_trace(&kg, buffer, rng_state, sample, x, y, offset, stride);
+	}
 }
 #endif
 
@@ -154,8 +158,9 @@ kernel_cuda_convert_to_byte(uchar4 *rgba, float *buffer, float sample_scale, int
 	int x = sx + blockDim.x*blockIdx.x + threadIdx.x;
 	int y = sy + blockDim.y*blockIdx.y + threadIdx.y;
 
-	if(x < sx + sw && y < sy + sh)
+	if(x < sx + sw && y < sy + sh) {
 		kernel_film_convert_to_byte(NULL, rgba, buffer, sample_scale, x, y, offset, stride);
+	}
 }
 
 extern "C" __global__ void
@@ -165,8 +170,9 @@ kernel_cuda_convert_to_half_float(uchar4 *rgba, float *buffer, float sample_scal
 	int x = sx + blockDim.x*blockIdx.x + threadIdx.x;
 	int y = sy + blockDim.y*blockIdx.y + threadIdx.y;
 
-	if(x < sx + sw && y < sy + sh)
+	if(x < sx + sw && y < sy + sh) {
 		kernel_film_convert_to_half_float(NULL, rgba, buffer, sample_scale, x, y, offset, stride);
+	}
 }
 
 extern "C" __global__ void
@@ -183,7 +189,8 @@ kernel_cuda_shader(uint4 *input,
 	int x = sx + blockDim.x*blockIdx.x + threadIdx.x;
 
 	if(x < sx + sw) {
-		kernel_shader_evaluate(NULL,
+		KernelGlobals kg;
+		kernel_shader_evaluate(&kg,
 		                       input,
 		                       output,
 		                       output_luma,
@@ -200,8 +207,10 @@ kernel_cuda_bake(uint4 *input, float4 *output, int type, int filter, int sx, int
 {
 	int x = sx + blockDim.x*blockIdx.x + threadIdx.x;
 
-	if(x < sx + sw)
-		kernel_bake_evaluate(NULL, input, output, (ShaderEvalType)type, filter, x, offset, sample);
+	if(x < sx + sw) {
+		KernelGlobals kg;
+		kernel_bake_evaluate(&kg, input, output, (ShaderEvalType)type, filter, x, offset, sample);
+	}
 }
 #endif
 
diff --git a/intern/cycles/kernel/svm/svm_image.h b/intern/cycles/kernel/svm/svm_image.h
index 2afdf61b476..0d6efb47223 100644
--- a/intern/cycles/kernel/svm/svm_image.h
+++ b/intern/cycles/kernel/svm/svm_image.h
@@ -144,7 +144,6 @@ ccl_device float4 svm_image_texture(KernelGlobals *kg, int id, float x, float y,
 		case 86: r = kernel_tex_image_interp(__tex_image_byte4_086, x, y); break;
 		case 87: r = kernel_tex_image_interp(__tex_image_byte4_087, x, y); break;
 		case 88: r = kernel_tex_image_interp(__tex_image_byte4_088, x, y); break;
-		case 89: r = kernel_tex_image_interp(__tex_image_byte4_089, x, y); break;
 		default:
 			kernel_assert(0);
 			return make_float4(0.0f, 0.0f, 0.0f, 0.0f);
diff --git a/intern/cycles/render/bake.h b/intern/cycles/render/bake.h
index 25f5eb3c897..aed9c5a8e75 100644
--- a/intern/cycles/render/bake.h
+++ b/intern/cycles/render/bake.h
@@ -73,7 +73,7 @@ public:
 
 	bool need_update;
 
-	int total_pixel_samples;
+	size_t total_pixel_samples;
 
 private:
 	BakeData *m_bake_data;
diff --git a/intern/cycles/render/light.cpp b/intern/cycles/render/light.cpp
index 2245c861d5a..6a4557506c3 100644
--- a/intern/cycles/render/light.cpp
+++ b/intern/cycles/render/light.cpp
@@ -486,10 +486,18 @@ static void background_cdf(int start,
                            float2 *cond_cdf)
 {
 	/* Conditional CDFs (rows, U direction). */
+	/* NOTE: It is possible to have some NaN pixels on background
+	 * which will ruin CDF causing wrong shading. We replace such
+	 * pixels with black.
+	 */
 	for(int i = start; i < end; i++) {
 		float sin_theta = sinf(M_PI_F * (i + 0.5f) / res);
 		float3 env_color = (*pixels)[i * res];
 		float ave_luminance = average(env_color);
+		/* TODO(sergey): Consider adding average_safe(). */
+		if(!isfinite(ave_luminance)) {
+			ave_luminance = 0.0f;
+		}
 
 		cond_cdf[i * cdf_count].x = ave_luminance * sin_theta;
 		cond_cdf[i * cdf_count].y = 0.0f;
@@ -497,6 +505,9 @@ static void background_cdf(int start,
 		for(int j = 1; j < res; j++) {
 			env_color = (*pixels)[i * res + j];
 			ave_luminance = average(env_color);
+			if(!isfinite(ave_luminance)) {
+				ave_luminance = 0.0f;
+			}
 
 			cond_cdf[i * cdf_count + j].x = ave_luminance * sin_theta;
 			cond_cdf[i * cdf_count + j].y = cond_cdf[i * cdf_count + j - 1].y + cond_cdf[i * cdf_count + j - 1].x / res;
diff --git a/intern/cycles/render/mesh.cpp b/intern/cycles/render/mesh.cpp
index c42b32919d4..b7660297f3e 100644
--- a/intern/cycles/render/mesh.cpp
+++ b/intern/cycles/render/mesh.cpp
@@ -1873,9 +1873,14 @@ void MeshManager::device_update_bvh(Device *device, DeviceScene *dscene, Scene *
 		dscene->prim_object.reference((uint*)&pack.prim_object[0], pack.prim_object.size());
 		device->tex_alloc("__prim_object", dscene->prim_object);
 	}
+	if(pack.prim_time.size()) {
+		dscene->prim_time.reference((float2*)&pack.prim_time[0], pack.prim_time.size());
+		device->tex_alloc("__prim_time", dscene->prim_time);
+	}
 
 	dscene->data.bvh.root = pack.root_index;
 	dscene->data.bvh.use_qbvh = scene->params.use_qbvh;
+	dscene->data.bvh.use_bvh_steps = (scene->params.num_bvh_time_steps != 0);
 }
 
 void MeshManager::device_update_flags(Device * /*device*/,
@@ -2152,6 +2157,7 @@ void MeshManager::device_free(Device *device, DeviceScene *dscene)
 	device->tex_free(dscene->prim_visibility);
 	device->tex_free(dscene->prim_index);
 	device->tex_free(dscene->prim_object);
+	device->tex_free(dscene->prim_time);
 	device->tex_free(dscene->tri_shader);
 	device->tex_free(dscene->tri_vnormal);
 	device->tex_free(dscene->tri_vindex);
@@ -2173,6 +2179,7 @@ void MeshManager::device_free(Device *device, DeviceScene *dscene)
 	dscene->prim_visibility.clear();
 	dscene->prim_index.clear();
 	dscene->prim_object.clear();
+	dscene->prim_time.clear();
 	dscene->tri_shader.clear();
 	dscene->tri_vnormal.clear();
 	dscene->tri_vindex.clear();
diff --git a/intern/cycles/render/scene.h b/intern/cycles/render/scene.h
index 8768682043f..9f398c444f4 100644
--- a/intern/cycles/render/scene.h
+++ b/intern/cycles/render/scene.h
@@ -69,6 +69,7 @@ public:
 	device_vector<uint> prim_visibility;
 	device_vector<uint> prim_index;
 	device_vector<uint> prim_object;
+	device_vector<float2> prim_time;
 
 	/* mesh */
 	device_vector<uint> tri_shader;
diff --git a/make.bat b/make.bat
index 337a5517656..dd82b1ca8a7 100644
--- a/make.bat
+++ b/make.bat
@@ -5,8 +5,8 @@ REM This is for users who like to configure & build Blender with a single comman
 setlocal ENABLEEXTENSIONS
 set BLENDER_DIR=%~dp0
 set BLENDER_DIR_NOSPACES=%BLENDER_DIR: =%
-if not "%BLENDER_DIR%"=="%BLENDER_DIR_NOSPACES%" ( 
-	echo There are spaces detected in the build path "%BLENDER_DIR%", this is currently not supported, exiting.... 
+if not "%BLENDER_DIR%"=="%BLENDER_DIR_NOSPACES%" (
+	echo There are spaces detected in the build path "%BLENDER_DIR%", this is currently not supported, exiting....
 	goto EOF
 )
 set BUILD_DIR=%BLENDER_DIR%..\build_windows
@@ -79,7 +79,7 @@ if NOT "%1" == "" (
 	set NOBUILD=1
 	)	else if "%1" == "showhash" (
 		for /f "delims=" %%i in ('git rev-parse HEAD') do echo Branch_hash=%%i
-		cd release/datafiles/locale 
+		cd release/datafiles/locale
 		for /f "delims=" %%i in ('git rev-parse HEAD') do echo Locale_hash=%%i
 		cd %~dp0
 		cd release/scripts/addons
@@ -132,13 +132,13 @@ if "%BUILD_ARCH%"=="x64" (
 
 
 if "%target%"=="Release" (
-		rem for vc12 check for both cuda 7.5 and 8 
+		rem for vc12 check for both cuda 7.5 and 8
 		if "%CUDA_PATH%"=="" (
 			echo Cuda Not found, aborting!
 			goto EOF
 		)
 		set BUILD_CMAKE_ARGS=%BUILD_CMAKE_ARGS% ^
-		-C"%BLENDER_DIR%\build_files\cmake\config\blender_release.cmake" 
+		-C"%BLENDER_DIR%\build_files\cmake\config\blender_release.cmake"
 )
 
 :DetectMSVC
@@ -157,7 +157,7 @@ if DEFINED MSVC_VC_DIR goto msvc_detect_finally
 if DEFINED MSVC_VC_DIR call "%MSVC_VC_DIR%\vcvarsall.bat"
 if DEFINED MSVC_VC_DIR goto sanity_checks
 
-rem MSVC Build environment 2017 and up. 
+rem MSVC Build environment 2017 and up.
 for /F "usebackq skip=2 tokens=1-2*" %%A IN (`REG QUERY "HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\Microsoft\VisualStudio\SXS\VS7" /v %BUILD_VS_VER%.0 2^>nul`) DO set MSVC_VS_DIR=%%C
 if DEFINED MSVC_VS_DIR goto msvc_detect_finally_2017
 REM Check 32 bits
@@ -202,7 +202,7 @@ if NOT EXIST %BLENDER_DIR%..\lib\nul (
 if "%TARGET%"=="" (
 	echo Error: Convenience target not set
 	echo This is required for building, aborting!
-	echo . 
+	echo .
 	goto HELP
 )
 
@@ -266,15 +266,15 @@ echo.
 echo At any point you can optionally modify your build configuration by editing:
 echo "%BUILD_DIR%\CMakeCache.txt", then run "make" again to build with the changes applied.
 echo.
-echo Blender successfully built, run from: "%BUILD_DIR%\bin\%BUILD_TYPE%"
+echo Blender successfully built, run from: "%BUILD_DIR%\bin\%BUILD_TYPE%\blender.exe"
 echo.
 goto EOF
 :HELP
 		echo.
 		echo Convenience targets
-		echo - release ^(identical to the offical blender.org builds^)
+		echo - release ^(identical to the official blender.org builds^)
 		echo - full ^(same as release minus the cuda kernels^)
-		echo - lite 
+		echo - lite
 		echo - headless
 		echo - cycles
 		echo - bpy
@@ -289,11 +289,10 @@ goto EOF
 		echo - with_tests ^(enable building unit tests^)
 		echo - debug ^(Build an unoptimized debuggable build^)
 		echo - packagename [newname] ^(override default cpack package name^)
-		echo - x86 ^(override host autodetect and build 32 bit code^)
-		echo - x64 ^(override host autodetect and build 64 bit code^)
+		echo - x86 ^(override host auto-detect and build 32 bit code^)
+		echo - x64 ^(override host auto-detect and build 64 bit code^)
 		echo - 2013 ^(build with visual studio 2013^)
 		echo - 2015 ^(build with visual studio 2015^) [EXPERIMENTAL]
 		echo.
 
 :EOF
-
diff --git a/release/datafiles/blender_icons.svg b/release/datafiles/blender_icons.svg
index e9c114ba1bd..d88788fa904 100644
--- a/release/datafiles/blender_icons.svg
+++ b/release/datafiles/blender_icons.svg
@@ -14,7 +14,7 @@
    height="640"
    id="svg2"
    sodipodi:version="0.32"
-   inkscape:version="0.91 r"
+   inkscape:version="0.91 r13725"
    version="1.0"
    sodipodi:docname="blender_icons.svg"
    inkscape:output_extension="org.inkscape.output.svg.inkscape"
@@ -31338,6 +31338,26 @@
          d="m 125.5,433.5 23,0 0,41 -33,0 0,-31 10,-10 z"
          style="display:inline;fill:url(#linearGradient13110);fill-opacity:1;fill-rule:evenodd;stroke:none;stroke-width:1;marker:none" />
     </clipPath>
+    <linearGradient
+       inkscape:collect="always"
+       xlink:href="#linearGradient1610-6"
+       id="linearGradient18199"
+       gradientUnits="userSpaceOnUse"
+       x1="189.76083"
+       y1="248.13905"
+       x2="116.05637"
+       y2="183.6826" />
+    <radialGradient
+       inkscape:collect="always"
+       xlink:href="#linearGradient22562"
+       id="radialGradient23167-6"
+       gradientUnits="userSpaceOnUse"
+       gradientTransform="matrix(0.99220964,-0.12457927,0.11585516,0.92272644,-34.13325,22.766225)"
+       cx="-0.78262758"
+       cy="294.63174"
+       fx="-0.78262758"
+       fy="294.63174"
+       r="6.6750002" />
   </defs>
   <sodipodi:namedview
      id="base"
@@ -31349,16 +31369,16 @@
      objecttolerance="10000"
      inkscape:pageopacity="0.0"
      inkscape:pageshadow="2"
-     inkscape:zoom="14.413868"
-     inkscape:cx="480.24726"
-     inkscape:cy="269.95478"
+     inkscape:zoom="19.997864"
+     inkscape:cx="462.52244"
+     inkscape:cy="435.14241"
      inkscape:document-units="px"
-     inkscape:current-layer="layer1"
+     inkscape:current-layer="g23149-4"
      showgrid="true"
-     inkscape:window-width="1680"
-     inkscape:window-height="1020"
-     inkscape:window-x="0"
-     inkscape:window-y="30"
+     inkscape:window-width="1920"
+     inkscape:window-height="1025"
+     inkscape:window-x="-8"
+     inkscape:window-y="-8"
      inkscape:snap-nodes="true"
      inkscape:snap-bbox="true"
      showguides="true"
@@ -92660,6 +92680,56 @@
            style="opacity:0.51999996;fill:url(#radialGradient21448-8-143);fill-opacity:1;fill-rule:evenodd;stroke:none" />
       </g>
     </g>
+    <g
+       transform="translate(335.99871,21.048284)"
+       style="display:inline;enable-background:new"
+       id="ICON_ROTATE-7">
+      <rect
+         y="178"
+         x="110"
+         height="16"
+         width="16"
+         id="rect37989-8"
+         style="display:inline;overflow:visible;visibility:visible;opacity:0;fill:#b3b3b3;fill-opacity:1;fill-rule:evenodd;stroke:none;stroke-width:2.79999995;marker:none;enable-background:accumulate" />
+      <path
+         style="display:inline;overflow:visible;visibility:visible;fill:none;stroke:#000000;stroke-width:2.4000001;stroke-linecap:square;stroke-linejoin:round;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1;marker:none;enable-background:new"
+         d="m 114.5,192.5 -3,0 0,-3 m 13,0 0,3 -3,0 m -0.25,-13 3.25,0 0,3 m -13,0 0,-3 3,0"
+         id="path37498-7"
+         sodipodi:nodetypes="cccccccccccc"
+         inkscape:connector-curvature="0" />
+      <path
+         sodipodi:nodetypes="cccccccccccc"
+         id="rect38140-7"
+         d="m 114.5,192.5 -3,0 0,-3 m 13,0 0,3 -3,0 m -0.25,-13 3.25,0 0,3 m -13,0 0,-3 3,0"
+         style="display:inline;overflow:visible;visibility:visible;fill:none;stroke:url(#linearGradient18199);stroke-width:1;stroke-linecap:square;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dashoffset:0;stroke-opacity:1;marker:none;enable-background:new"
+         inkscape:connector-curvature="0" />
+      <g
+         transform="matrix(0.59971056,0,0,0.59971056,116.78278,9.7425599)"
+         style="display:inline;enable-background:new"
+         id="g23145-9">
+        <g
+           id="g23149-4">
+          <path
+             id="path39832-9"
+             style="display:inline;overflow:visible;visibility:visible;fill:none;stroke:#1a1a1a;stroke-width:4.66725159;stroke-linecap:square;stroke-linejoin:round;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1;marker:none;enable-background:accumulate"
+             d="m -4.3682386,287.81345 1.5,0 c 0.999089,0 2.07885534,1.30514 2.50490386,2.78207 1.06592652,3.69512 2.80867074,9.82446 5.88525404,9.96406 2.6782554,0 1.6181317,-5.11535 3.1736046,-5.26275 l 0.25,0"
+             sodipodi:nodetypes="cssccc"
+             inkscape:connector-curvature="0" />
+          <path
+             sodipodi:nodetypes="ccscc"
+             d="m 9.3647983,295.22328 -0.4793018,0 c -2.2335161,0 0.1796731,4.94901 -3.4398065,5.09984 -4.44796752,0.18536 -5.37272213,-12.59185 -8.0767581,-12.56237 l -2,0"
+             style="display:inline;overflow:visible;visibility:visible;fill:none;stroke:#a8df84;stroke-width:2.93474906;stroke-linecap:square;stroke-linejoin:round;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1;marker:none;enable-background:accumulate"
+             id="path39834-2"
+             inkscape:connector-curvature="0" />
+        </g>
+        <path
+           id="path39836-9"
+           style="display:inline;overflow:visible;visibility:visible;opacity:0.35;fill:none;stroke:url(#radialGradient23167-6);stroke-width:3.53503864;stroke-linecap:round;stroke-linejoin:round;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1;marker:none;enable-background:accumulate"
+           d="M 5.6770841,300.48165 C 0.7393262,300.21066 0.54777814,287.99792 -2.9522219,287.99792"
+           sodipodi:nodetypes="cc"
+           inkscape:connector-curvature="0" />
+      </g>
+    </g>
   </g>
   <g
      inkscape:groupmode="layer"
diff --git a/release/datafiles/blender_icons16/icon16_normalize_fcurves.dat b/release/datafiles/blender_icons16/icon16_normalize_fcurves.dat
new file mode 100644
index 00000000000..fe118a23d79
--- /dev/null
+++ b/release/datafiles/blender_icons16/icon16_normalize_fcurves.dat
diff --git a/release/datafiles/blender_icons32/icon32_normalize_fcurves.dat b/release/datafiles/blender_icons32/icon32_normalize_fcurves.dat
new file mode 100644
index 00000000000..5b1f546a563
--- /dev/null
+++ b/release/datafiles/blender_icons32/icon32_normalize_fcurves.dat
diff --git a/release/scripts/freestyle/modules/parameter_editor.py b/release/scripts/freestyle/modules/parameter_editor.py
index 93305cb7c5a..b093920a4cb 100644
--- a/release/scripts/freestyle/modules/parameter_editor.py
+++ b/release/scripts/freestyle/modules/parameter_editor.py
@@ -1170,6 +1170,7 @@ class Seed:
 
 _seed = Seed()
 
+
 def get_dashed_pattern(linestyle):
     """Extracts the dashed pattern from the various UI options """
     pattern = []
@@ -1185,6 +1186,15 @@ def get_dashed_pattern(linestyle):
     return pattern
 
 
+def get_grouped_objects(group):
+    for ob in group.objects:
+        if ob.dupli_type == 'GROUP' and ob.dupli_group is not None:
+            for dupli in get_grouped_objects(ob.dupli_group):
+                yield dupli
+        else:
+            yield ob
+
+
 integration_types = {
     'MEAN': IntegrationType.MEAN,
     'MIN': IntegrationType.MIN,
@@ -1267,7 +1277,7 @@ def process(layer_name, lineset_name):
     # prepare selection criteria by group of objects
     if lineset.select_by_group:
         if lineset.group is not None:
-            names = {getQualifiedObjectName(ob): True for ob in lineset.group.objects}
+            names = {getQualifiedObjectName(ob): True for ob in get_grouped_objects(lineset.group)}
             upred = ObjectNamesUP1D(names, lineset.group_negation == 'EXCLUSIVE')
             selection_criteria.append(upred)
     # prepare selection criteria by image border
diff --git a/release/scripts/startup/bl_ui/properties_data_modifier.py b/release/scripts/startup/bl_ui/properties_data_modifier.py
index d66fb08bcd6..b24b574f37b 100644
--- a/release/scripts/startup/bl_ui/properties_data_modifier.py
+++ b/release/scripts/startup/bl_ui/properties_data_modifier.py
@@ -1320,7 +1320,9 @@ class DATA_PT_modifiers(ModifierButtonsPanel, Panel):
         row.prop(md, "thickness_vertex_group", text="Factor")
 
         col.prop(md, "use_crease", text="Crease Edges")
-        col.prop(md, "crease_weight", text="Crease Weight")
+        row = col.row()
+        row.active = md.use_crease
+        row.prop(md, "crease_weight", text="Crease Weight")
 
         col = split.column()
 
diff --git a/release/scripts/startup/bl_ui/properties_object.py b/release/scripts/startup/bl_ui/properties_object.py
index 3ff7a248c60..4a596981983 100644
--- a/release/scripts/startup/bl_ui/properties_object.py
+++ b/release/scripts/startup/bl_ui/properties_object.py
@@ -152,6 +152,33 @@ class OBJECT_PT_relations(ObjectButtonsPanel, Panel):
         sub.active = (parent is not None)
 
 
+class OBJECT_PT_relations_extras(ObjectButtonsPanel, Panel):
+    bl_label = "Relations Extras"
+    bl_options = {'DEFAULT_CLOSED'}
+
+    def draw(self, context):
+        layout = self.layout
+
+        ob = context.object
+
+        split = layout.split()
+
+        if context.scene.render.engine != 'BLENDER_GAME':
+            col = split.column()
+            col.label(text="Tracking Axes:")
+            col.prop(ob, "track_axis", text="Axis")
+            col.prop(ob, "up_axis", text="Up Axis")
+
+        col = split.column()
+        col.prop(ob, "use_slow_parent")
+        row = col.row()
+        row.active = ((ob.parent is not None) and (ob.use_slow_parent))
+        row.prop(ob, "slow_parent_offset", text="Offset")
+
+        layout.prop(ob, "use_extra_recalc_object")
+        layout.prop(ob, "use_extra_recalc_data")
+
+
 class GROUP_MT_specials(Menu):
     bl_label = "Group Specials"
 
@@ -296,33 +323,6 @@ class OBJECT_PT_duplication(ObjectButtonsPanel, Panel):
             layout.prop(ob, "dupli_group", text="Group")
 
 
-class OBJECT_PT_relations_extras(ObjectButtonsPanel, Panel):
-    bl_label = "Relations Extras"
-    bl_options = {'DEFAULT_CLOSED'}
-
-    def draw(self, context):
-        layout = self.layout
-
-        ob = context.object
-
-        split = layout.split()
-
-        if context.scene.render.engine != 'BLENDER_GAME':
-            col = split.column()
-            col.label(text="Tracking Axes:")
-            col.prop(ob, "track_axis", text="Axis")
-            col.prop(ob, "up_axis", text="Up Axis")
-
-        col = split.column()
-        col.prop(ob, "use_slow_parent")
-        row = col.row()
-        row.active = ((ob.parent is not None) and (ob.use_slow_parent))
-        row.prop(ob, "slow_parent_offset", text="Offset")
-
-        layout.prop(ob, "use_extra_recalc_object")
-        layout.prop(ob, "use_extra_recalc_data")
-
-
 from bl_ui.properties_animviz import (
         MotionPathButtonsPanel,
         OnionSkinButtonsPanel,
diff --git a/release/scripts/startup/bl_ui/space_graph.py b/release/scripts/startup/bl_ui/space_graph.py
index 446df9e6e79..cec3b26f35a 100644
--- a/release/scripts/startup/bl_ui/space_graph.py
+++ b/release/scripts/startup/bl_ui/space_graph.py
@@ -42,10 +42,11 @@ class GRAPH_HT_header(Header):
 
         dopesheet_filter(layout, context)
 
-        layout.prop(st, "use_normalization", text="Normalize")
-        row = layout.row()
-        row.active = st.use_normalization
-        row.prop(st, "use_auto_normalization", text="Auto")
+        row = layout.row(align=True)
+        row.prop(st, "use_normalization", icon='NORMALIZE_FCURVES', text="Normalize", toggle=True)
+        sub = row.row(align=True)
+        sub.active = st.use_normalization
+        sub.prop(st, "use_auto_normalization", icon='FILE_REFRESH', text="", toggle=True)
 
         row = layout.row(align=True)
 
diff --git a/release/scripts/startup/bl_ui/space_sequencer.py b/release/scripts/startup/bl_ui/space_sequencer.py
index 6bb516cf929..8ab5b4724b8 100644
--- a/release/scripts/startup/bl_ui/space_sequencer.py
+++ b/release/scripts/startup/bl_ui/space_sequencer.py
@@ -652,17 +652,39 @@ class SEQUENCER_PT_effect(SequencerButtonsPanel, Panel):
             col.prop(strip, "rotation_start", text="Rotation")
 
         elif strip.type == 'MULTICAM':
-            layout.prop(strip, "multicam_source")
-
-            row = layout.row(align=True)
-            sub = row.row(align=True)
-            sub.scale_x = 2.0
-
-            sub.operator("screen.animation_play", text="", icon='PAUSE' if context.screen.is_animation_playing else 'PLAY')
+            col = layout.column(align=True)
+            strip_channel = strip.channel
+
+            col.prop(strip, "multicam_source", text="Source Channel")
+
+            # The multicam strip needs at least 2 strips to be useful
+            if strip_channel > 2:
+                BT_ROW = 4
+
+                col.label("Cut To:")
+                row = col.row()
+
+                for i in range(1, strip_channel):
+                    if (i % BT_ROW) == 1:
+                        row = col.row(align=True)
+
+                    # Workaround - .active has to have a separate UI block to work
+                    if i == strip.multicam_source:
+                        sub = row.row(align=True)
+                        sub.active = False
+                        sub.operator("sequencer.cut_multicam", text="%d" % i).camera = i
+                    else:
+                        sub_1 = row.row(align=True)
+                        sub_1.active = True
+                        sub_1.operator("sequencer.cut_multicam", text="%d" % i).camera = i
+
+                if strip.channel > BT_ROW and (strip_channel - 1) % BT_ROW:
+                    for i in range(strip.channel, strip_channel + ((BT_ROW + 1 - strip_channel) % BT_ROW)):
+                        row.label("")
+            else:
+                col.separator()
+                col.label(text="Two or more channels are needed below this strip.", icon="INFO")
 
-            row.label("Cut To")
-            for i in range(1, strip.channel):
-                row.operator("sequencer.cut_multicam", text="%d" % i).camera = i
 
         elif strip.type == 'TEXT':
             col = layout.column()
diff --git a/release/scripts/startup/bl_ui/space_time.py b/release/scripts/startup/bl_ui/space_time.py
index 508e62e4f56..1f62d8d6968 100644
--- a/release/scripts/startup/bl_ui/space_time.py
+++ b/release/scripts/startup/bl_ui/space_time.py
@@ -49,7 +49,10 @@ class TIME_HT_header(Header):
             row.prop(scene, "frame_preview_start", text="Start")
             row.prop(scene, "frame_preview_end", text="End")
 
-        layout.prop(scene, "frame_current", text="")
+        if scene.show_subframe:
+            layout.prop(scene, "frame_float", text="")
+        else:
+            layout.prop(scene, "frame_current", text="")
 
         layout.separator()
 
@@ -135,6 +138,7 @@ class TIME_MT_view(Menu):
 
         layout.prop(st, "show_frame_indicator")
         layout.prop(scene, "show_keys_from_selected_only")
+        layout.prop(scene, "show_subframe")
 
         layout.separator()
 
diff --git a/release/scripts/startup/bl_ui/space_view3d.py b/release/scripts/startup/bl_ui/space_view3d.py
index 9ed59d830e3..cba7502dd96 100644
--- a/release/scripts/startup/bl_ui/space_view3d.py
+++ b/release/scripts/startup/bl_ui/space_view3d.py
@@ -1722,6 +1722,7 @@ class VIEW3D_MT_brush_paint_modes(Menu):
         layout.prop(brush, "use_paint_weight", text="Weight Paint")
         layout.prop(brush, "use_paint_image", text="Texture Paint")
 
+
 # ********** Vertex paint menu **********
 
 
@@ -1791,6 +1792,7 @@ class VIEW3D_MT_vertex_group(Menu):
             layout.operator("object.vertex_group_remove", text="Remove Active Group").all = False
             layout.operator("object.vertex_group_remove", text="Remove All Groups").all = True
 
+
 # ********** Weight paint menu **********
 
 
@@ -1829,6 +1831,7 @@ class VIEW3D_MT_paint_weight(Menu):
 
         layout.operator("paint.weight_set")
 
+
 # ********** Sculpt menu **********
 
 
@@ -1982,6 +1985,7 @@ class VIEW3D_MT_particle_specials(Menu):
 class VIEW3D_MT_particle_showhide(ShowHideMenu, Menu):
     _operator_name = "particle"
 
+
 # ********** Pose Menu **********
 
 
@@ -2255,6 +2259,7 @@ class VIEW3D_MT_bone_options_disable(Menu, BoneOptions):
     bl_label = "Disable Bone Options"
     type = 'DISABLE'
 
+
 # ********** Edit Menus, suffix from ob.type **********
 
 
@@ -2422,6 +2427,7 @@ class VIEW3D_MT_edit_mesh_vertices(Menu):
         with_bullet = bpy.app.build_options.bullet
 
         layout.operator("mesh.merge")
+        layout.operator("mesh.remove_doubles")
         layout.operator("mesh.rip_move")
         layout.operator("mesh.rip_move_fill")
         layout.operator("mesh.rip_edge_move")
@@ -2444,7 +2450,6 @@ class VIEW3D_MT_edit_mesh_vertices(Menu):
         if with_bullet:
             layout.operator("mesh.convex_hull")
         layout.operator("mesh.vertices_smooth")
-        layout.operator("mesh.remove_doubles")
 
         layout.operator("mesh.blend_from_shape")
 
@@ -2601,6 +2606,7 @@ class VIEW3D_MT_edit_mesh_clean(Menu):
         layout.operator("mesh.face_make_planar")
         layout.operator("mesh.vert_connect_nonplanar")
         layout.operator("mesh.vert_connect_concave")
+        layout.operator("mesh.remove_doubles")
         layout.operator("mesh.fill_holes")
 
 
diff --git a/source/blender/alembic/intern/abc_archive.cc b/source/blender/alembic/intern/abc_archive.cc
index 0985a06d732..5f8fc1a3739 100644
--- a/source/blender/alembic/intern/abc_archive.cc
+++ b/source/blender/alembic/intern/abc_archive.cc
@@ -113,25 +113,25 @@ static OArchive create_archive(std::ostream *ostream,
                                Alembic::Abc::MetaData &md,
                                bool ogawa)
 {
-    md.set(Alembic::Abc::kApplicationNameKey, "Blender");
+	md.set(Alembic::Abc::kApplicationNameKey, "Blender");
 	md.set(Alembic::Abc::kUserDescriptionKey, scene_name);
 
-    time_t raw_time;
-    time(&raw_time);
-    char buffer[128];
+	time_t raw_time;
+	time(&raw_time);
+	char buffer[128];
 
 #if defined _WIN32 || defined _WIN64
-    ctime_s(buffer, 128, &raw_time);
+	ctime_s(buffer, 128, &raw_time);
 #else
-    ctime_r(&raw_time, buffer);
+	ctime_r(&raw_time, buffer);
 #endif
 
-    const std::size_t buffer_len = strlen(buffer);
-    if (buffer_len > 0 && buffer[buffer_len - 1] == '\n') {
-        buffer[buffer_len - 1] = '\0';
-    }
+	const std::size_t buffer_len = strlen(buffer);
+	if (buffer_len > 0 && buffer[buffer_len - 1] == '\n') {
+		buffer[buffer_len - 1] = '\0';
+	}
 
-    md.set(Alembic::Abc::kDateWrittenKey, buffer);
+	md.set(Alembic::Abc::kDateWrittenKey, buffer);
 
 	ErrorHandler::Policy policy = ErrorHandler::kThrowPolicy;
 
diff --git a/source/blender/alembic/intern/abc_curves.cc b/source/blender/alembic/intern/abc_curves.cc
index 282777f3af0..0542255d84b 100644
--- a/source/blender/alembic/intern/abc_curves.cc
+++ b/source/blender/alembic/intern/abc_curves.cc
@@ -361,7 +361,7 @@ void read_curve_sample(Curve *cu, const ICurvesSchema &schema, const float time)
  * object directly and create a new DerivedMesh from that. Also we might need to
  * create new or delete existing NURBS in the curve.
  */
-DerivedMesh *AbcCurveReader::read_derivedmesh(DerivedMesh */*dm*/, const float time, int /*read_flag*/, const char **/*err_str*/)
+DerivedMesh *AbcCurveReader::read_derivedmesh(DerivedMesh * /*dm*/, const float time, int /*read_flag*/, const char ** /*err_str*/)
 {
 	ISampleSelector sample_sel(time);
 	const ICurvesSchema::Sample sample = m_curves_schema.getValue(sample_sel);
diff --git a/source/blender/alembic/intern/abc_exporter.cc b/source/blender/alembic/intern/abc_exporter.cc
index 28d54672d20..45c31279b50 100644
--- a/source/blender/alembic/intern/abc_exporter.cc
+++ b/source/blender/alembic/intern/abc_exporter.cc
@@ -47,7 +47,7 @@ extern "C" {
 
 #ifdef WIN32
 /* needed for MSCV because of snprintf from BLI_string */
-#	include "BLI_winstuff.h"
+#  include "BLI_winstuff.h"
 #endif
 
 #include "BKE_anim.h"
diff --git a/source/blender/alembic/intern/abc_points.cc b/source/blender/alembic/intern/abc_points.cc
index 4c78f3e83c7..fc84759b1d9 100644
--- a/source/blender/alembic/intern/abc_points.cc
+++ b/source/blender/alembic/intern/abc_points.cc
@@ -200,7 +200,7 @@ void read_points_sample(const IPointsSchema &schema,
 	read_mverts(config.mvert, positions, vnormals);
 }
 
-DerivedMesh *AbcPointsReader::read_derivedmesh(DerivedMesh *dm, const float time, int /*read_flag*/, const char **/*err_str*/)
+DerivedMesh *AbcPointsReader::read_derivedmesh(DerivedMesh *dm, const float time, int /*read_flag*/, const char ** /*err_str*/)
 {
 	ISampleSelector sample_sel(time);
 	const IPointsSchema::Sample sample = m_schema.getValue(sample_sel);
diff --git a/source/blender/alembic/intern/abc_transform.cc b/source/blender/alembic/intern/abc_transform.cc
index e2fc7674c4e..2c6ef09326c 100644
--- a/source/blender/alembic/intern/abc_transform.cc
+++ b/source/blender/alembic/intern/abc_transform.cc
@@ -122,7 +122,7 @@ Imath::Box3d AbcTransformWriter::bounds()
 	return Imath::transform(bounds, m_matrix);
 }
 
-bool AbcTransformWriter::hasAnimation(Object */*ob*/) const
+bool AbcTransformWriter::hasAnimation(Object * /*ob*/) const
 {
 	/* TODO(kevin): implement this. */
 	return true;
@@ -146,6 +146,6 @@ bool AbcEmptyReader::valid() const
 
 void AbcEmptyReader::readObjectData(Main *bmain, float /*time*/)
 {
-	m_object = BKE_object_add_only_object(bmain, OB_EMPTY, m_object_name.c_str());
+	m_object = BKE_object_add_only_object(bmain, OB_EMPTY, m_data_name.c_str());
 	m_object->data = NULL;
 }
diff --git a/source/blender/blenkernel/BKE_bvhutils.h b/source/blender/blenkernel/BKE_bvhutils.h
index bf45a27e51c..cb72f0859d5 100644
--- a/source/blender/blenkernel/BKE_bvhutils.h
+++ b/source/blender/blenkernel/BKE_bvhutils.h
@@ -54,7 +54,6 @@ typedef struct BVHTreeFromEditMesh {
 	/* default callbacks to bvh nearest and raycast */
 	BVHTree_NearestPointCallback nearest_callback;
 	BVHTree_RayCastCallback raycast_callback;
-	BVHTree_NearestToRayCallback nearest_to_ray_callback;
 
 	struct BMEditMesh *em;
 
@@ -75,7 +74,6 @@ typedef struct BVHTreeFromMesh {
 	/* default callbacks to bvh nearest and raycast */
 	BVHTree_NearestPointCallback nearest_callback;
 	BVHTree_RayCastCallback raycast_callback;
-	BVHTree_NearestToRayCallback nearest_to_ray_callback;
 
 	/* Vertex array, so that callbacks have instante access to data */
 	const struct MVert *vert;
@@ -104,7 +102,7 @@ typedef struct BVHTreeFromMesh {
  * The tree is build in mesh space coordinates, this means special care must be made on queries
  * so that the coordinates and rays are first translated on the mesh local coordinates.
  * Reason for this is that bvh_from_mesh_* can use a cache in some cases and so it becomes possible to reuse a BVHTree.
- * 
+ *
  * free_bvhtree_from_mesh should be called when the tree is no longer needed.
  */
 BVHTree *bvhtree_from_editmesh_verts(
@@ -118,7 +116,7 @@ BVHTree *bvhtree_from_editmesh_verts_ex(
 BVHTree *bvhtree_from_mesh_verts(
         struct BVHTreeFromMesh *data, struct DerivedMesh *mesh, float epsilon, int tree_type, int axis);
 BVHTree *bvhtree_from_mesh_verts_ex(
-        struct BVHTreeFromMesh *data, struct MVert *vert, const int numVerts,
+        struct BVHTreeFromMesh *data, const struct MVert *vert, const int numVerts,
         const bool vert_allocated, const BLI_bitmap *mask, int verts_num_active,
         float epsilon, int tree_type, int axis);
 
@@ -135,8 +133,8 @@ BVHTree *bvhtree_from_mesh_edges(
         float epsilon, int tree_type, int axis);
 BVHTree *bvhtree_from_mesh_edges_ex(
         struct BVHTreeFromMesh *data,
-        struct MVert *vert, const bool vert_allocated,
-        struct MEdge *edge, const int edges_num, const bool edge_allocated,
+        const struct MVert *vert, const bool vert_allocated,
+        const struct MEdge *edge, const int edges_num, const bool edge_allocated,
         const BLI_bitmap *edges_mask, int edges_num_active,
         float epsilon, int tree_type, int axis);
 
@@ -145,8 +143,8 @@ BVHTree *bvhtree_from_mesh_faces(
         int tree_type, int axis);
 BVHTree *bvhtree_from_mesh_faces_ex(
         struct BVHTreeFromMesh *data,
-        struct MVert *vert, const bool vert_allocated,
-        struct MFace *face, const int numFaces, const bool face_allocated,
+        const struct MVert *vert, const bool vert_allocated,
+        const struct MFace *face, const int numFaces, const bool face_allocated,
         const BLI_bitmap *mask, int numFaces_active,
         float epsilon, int tree_type, int axis);
 
diff --git a/source/blender/blenkernel/BKE_object.h b/source/blender/blenkernel/BKE_object.h
index c1655d3e9c6..b3b67d0924a 100644
--- a/source/blender/blenkernel/BKE_object.h
+++ b/source/blender/blenkernel/BKE_object.h
@@ -140,8 +140,6 @@ void BKE_boundbox_init_from_minmax(struct BoundBox *bb, const float min[3], cons
 void BKE_boundbox_calc_center_aabb(const struct BoundBox *bb, float r_cent[3]);
 void BKE_boundbox_calc_size_aabb(const struct BoundBox *bb, float r_size[3]);
 void BKE_boundbox_minmax(const struct BoundBox *bb, float obmat[4][4], float r_min[3], float r_max[3]);
-struct BoundBox *BKE_boundbox_ensure_minimum_dimensions(
-        struct BoundBox *bb, struct BoundBox *bb_temp, const float epsilon);
 
 struct BoundBox *BKE_object_boundbox_get(struct Object *ob);
 void BKE_object_dimensions_get(struct Object *ob, float vec[3]);
diff --git a/source/blender/blenkernel/intern/appdir.c b/source/blender/blenkernel/intern/appdir.c
index b1dcc40279f..9236a1b5853 100644
--- a/source/blender/blenkernel/intern/appdir.c
+++ b/source/blender/blenkernel/intern/appdir.c
@@ -121,7 +121,7 @@ static bool test_path(char *targetpath, const char *path_base, const char *path_
 	if (path_sep) BLI_join_dirfile(tmppath, sizeof(tmppath), path_base, path_sep);
 	else BLI_strncpy(tmppath, path_base, sizeof(tmppath));
 
-	/* rare cases folder_name is omitted (when looking for ~/.blender/2.xx dir only) */
+	/* rare cases folder_name is omitted (when looking for ~/.config/blender/2.xx dir only) */
 	if (folder_name)
 		BLI_make_file_string("/", targetpath, tmppath, folder_name);
 	else
diff --git a/source/blender/blenkernel/intern/bvhutils.c b/source/blender/blenkernel/intern/bvhutils.c
index d0e0c82e3be..c0e4ef37a93 100644
--- a/source/blender/blenkernel/intern/bvhutils.c
+++ b/source/blender/blenkernel/intern/bvhutils.c
@@ -376,45 +376,6 @@ static void mesh_edges_spherecast(void *userdata, int index, const BVHTreeRay *r
 	}
 }
 
-#define V3_MUL_ELEM(a, b) \
-	(a)[0] * (b)[0], \
-	(a)[1] * (b)[1], \
-	(a)[2] * (b)[2]
-
-/* Callback to bvh tree nearest edge to ray.
- * The tree must have been built using bvhtree_from_mesh_edges.
- * userdata must be a BVHMeshCallbackUserdata built from the same mesh as the tree. */
-static void mesh_edges_nearest_to_ray(
-        void *userdata, const float ray_co[3], const float ray_dir[3],
-        const float scale[3], int index, BVHTreeNearest *nearest)
-{
-	struct BVHTreeFromMesh *data = userdata;
-	const MVert *vert = data->vert;
-	const MEdge *e = &data->edge[index];
-
-	const float t0[3]        = {V3_MUL_ELEM(vert[e->v1].co, scale)};
-	const float t1[3]        = {V3_MUL_ELEM(vert[e->v2].co, scale)};
-	const float origin_sc[3] = {V3_MUL_ELEM(ray_co, scale)};
-	const float dir_sc[3]    = {V3_MUL_ELEM(ray_dir, scale)};
-
-	float depth, point[3];
-	const float dist_sq = dist_squared_ray_to_seg_v3(origin_sc, dir_sc, t0, t1, point, &depth);
-
-	if (dist_sq < nearest->dist_sq) {
-		nearest->dist_sq = dist_sq;
-		nearest->index = index;
-
-		point[0] /= scale[0];
-		point[1] /= scale[1];
-		point[2] /= scale[2];
-
-		copy_v3_v3(nearest->co, point);
-		sub_v3_v3v3(nearest->no, t0, t1);
-	}
-}
-
-#undef V3_MUL_ELEM
-
 /** \} */
 
 /*
@@ -459,7 +420,7 @@ static BVHTree *bvhtree_from_editmesh_verts_create_tree(
 
 static BVHTree *bvhtree_from_mesh_verts_create_tree(
         float epsilon, int tree_type, int axis,
-        MVert *vert, const int verts_num,
+        const MVert *vert, const int verts_num,
         const BLI_bitmap *verts_mask, int verts_num_active)
 {
 	BLI_assert(vert != NULL);
@@ -488,31 +449,23 @@ static BVHTree *bvhtree_from_mesh_verts_create_tree(
 
 static void bvhtree_from_mesh_verts_setup_data(
         BVHTreeFromMesh *data, BVHTree *tree, const bool is_cached, float epsilon,
-        MVert *vert, const bool vert_allocated)
+        const MVert *vert, const bool vert_allocated)
 {
 	memset(data, 0, sizeof(*data));
 
-	if (tree) {
-		data->tree = tree;
-		data->cached = is_cached;
+	data->tree = tree;
+	data->cached = is_cached;
 
-		/* a NULL nearest callback works fine
-		 * remember the min distance to point is the same as the min distance to BV of point */
-		data->nearest_callback = NULL;
-		data->raycast_callback = mesh_verts_spherecast;
-		data->nearest_to_ray_callback = NULL;
+	/* a NULL nearest callback works fine
+	 * remember the min distance to point is the same as the min distance to BV of point */
+	data->nearest_callback = NULL;
+	data->raycast_callback = mesh_verts_spherecast;
 
-		data->vert = vert;
-		data->vert_allocated = vert_allocated;
-		//data->face = DM_get_tessface_array(dm, &data->face_allocated);  /* XXX WHY???? */
+	data->vert = vert;
+	data->vert_allocated = vert_allocated;
+	//data->face = DM_get_tessface_array(dm, &data->face_allocated);  /* XXX WHY???? */
 
-		data->sphere_radius = epsilon;
-	}
-	else {
-		if (vert_allocated) {
-			MEM_freeN(vert);
-		}
-	}
+	data->sphere_radius = epsilon;
 }
 
 /* Builds a bvh tree where nodes are the vertices of the given em */
@@ -531,7 +484,6 @@ BVHTree *bvhtree_from_editmesh_verts_ex(
 		data->em = em;
 		data->nearest_callback = NULL;
 		data->raycast_callback = editmesh_verts_spherecast;
-		data->nearest_to_ray_callback = NULL;
 	}
 
 	return tree;
@@ -588,11 +540,18 @@ BVHTree *bvhtree_from_mesh_verts(
 		/* printf("BVHTree is already build, using cached tree\n"); */
 	}
 
-	/* Setup BVHTreeFromMesh */
-	bvhtree_from_mesh_verts_setup_data(
-	        data, tree, true, epsilon, vert, vert_allocated);
-
-	return data->tree;
+	if (tree) {
+		/* Setup BVHTreeFromMesh */
+		bvhtree_from_mesh_verts_setup_data(
+		        data, tree, true, epsilon, vert, vert_allocated);
+	}
+	else {
+		if (vert_allocated) {
+			MEM_freeN(vert);
+		}
+		memset(data, 0, sizeof(*data));
+	}
+	return tree;
 }
 
 /**
@@ -602,7 +561,7 @@ BVHTree *bvhtree_from_mesh_verts(
  * \param verts_num_active if >= 0, number of active verts to add to BVH tree (else will be computed from mask).
  */
 BVHTree *bvhtree_from_mesh_verts_ex(
-        BVHTreeFromMesh *data, MVert *vert, const int verts_num, const bool vert_allocated,
+        BVHTreeFromMesh *data, const MVert *vert, const int verts_num, const bool vert_allocated,
         const BLI_bitmap *verts_mask, int verts_num_active,
         float epsilon, int tree_type, int axis)
 {
@@ -613,7 +572,7 @@ BVHTree *bvhtree_from_mesh_verts_ex(
 	bvhtree_from_mesh_verts_setup_data(
 	        data, tree, false, epsilon, vert, vert_allocated);
 
-	return data->tree;
+	return tree;
 }
 
 /** \} */
@@ -661,7 +620,7 @@ static BVHTree *bvhtree_from_editmesh_edges_create_tree(
 }
 
 static BVHTree *bvhtree_from_mesh_edges_create_tree(
-        MVert *vert, MEdge *edge, const int edge_num,
+        const MVert *vert, const MEdge *edge, const int edge_num,
         const BLI_bitmap *edges_mask, int edges_num_active,
         float epsilon, int tree_type, int axis)
 {
@@ -694,34 +653,26 @@ static BVHTree *bvhtree_from_mesh_edges_create_tree(
 }
 
 static void bvhtree_from_mesh_edges_setup_data(
-        BVHTreeFromMesh *data, BVHTree *tree, const bool is_cached, float epsilon,
-        MVert *vert, const bool vert_allocated, MEdge *edge, const bool edge_allocated)
+        BVHTreeFromMesh *data, BVHTree *tree,
+        const bool is_cached, float epsilon,
+        const MVert *vert, const bool vert_allocated,
+        const MEdge *edge, const bool edge_allocated)
 {
 	memset(data, 0, sizeof(*data));
+
 	data->tree = tree;
 
-	if (data->tree) {
-		data->cached = is_cached;
+	data->cached = is_cached;
 
-		data->nearest_callback = mesh_edges_nearest_point;
-		data->raycast_callback = mesh_edges_spherecast;
-		data->nearest_to_ray_callback = mesh_edges_nearest_to_ray;
+	data->nearest_callback = mesh_edges_nearest_point;
+	data->raycast_callback = mesh_edges_spherecast;
 
-		data->vert = vert;
-		data->vert_allocated = vert_allocated;
-		data->edge = edge;
-		data->edge_allocated = edge_allocated;
+	data->vert = vert;
+	data->vert_allocated = vert_allocated;
+	data->edge = edge;
+	data->edge_allocated = edge_allocated;
 
-		data->sphere_radius = epsilon;
-	}
-	else {
-		if (vert_allocated) {
-			MEM_freeN(vert);
-		}
-		if (edge_allocated) {
-			MEM_freeN(edge);
-		}
-	}
+	data->sphere_radius = epsilon;
 }
 
 /* Builds a bvh tree where nodes are the edges of the given em */
@@ -742,8 +693,6 @@ BVHTree *bvhtree_from_editmesh_edges_ex(
 		data->em = em;
 		data->nearest_callback = NULL;  /* TODO */
 		data->raycast_callback = NULL;  /* TODO */
-		/* TODO: not urgent however since users currently define own callbacks */
-		data->nearest_to_ray_callback = NULL;
 	}
 
 	return tree;
@@ -795,11 +744,21 @@ BVHTree *bvhtree_from_mesh_edges(
 		/* printf("BVHTree is already build, using cached tree\n"); */
 	}
 
-	/* Setup BVHTreeFromMesh */
-	bvhtree_from_mesh_edges_setup_data(
-	        data, tree, true, epsilon, vert, vert_allocated, edge, edge_allocated);
-
-	return data->tree;
+	if (tree) {
+		/* Setup BVHTreeFromMesh */
+		bvhtree_from_mesh_edges_setup_data(
+		        data, tree, true, epsilon, vert, vert_allocated, edge, edge_allocated);
+	}
+	else {
+		if (vert_allocated) {
+			MEM_freeN(vert);
+		}
+		if (edge_allocated) {
+			MEM_freeN(edge);
+		}
+		memset(data, 0, sizeof(*data));
+	}
+	return tree;
 }
 
 /**
@@ -810,8 +769,8 @@ BVHTree *bvhtree_from_mesh_edges(
  */
 BVHTree *bvhtree_from_mesh_edges_ex(
         BVHTreeFromMesh *data,
-        MVert *vert, const bool vert_allocated,
-        MEdge *edge, const int edges_num, const bool edge_allocated,
+        const MVert *vert, const bool vert_allocated,
+        const MEdge *edge, const int edges_num, const bool edge_allocated,
         const BLI_bitmap *edges_mask, int edges_num_active,
         float epsilon, int tree_type, int axis)
 {
@@ -823,7 +782,7 @@ BVHTree *bvhtree_from_mesh_edges_ex(
 	bvhtree_from_mesh_edges_setup_data(
 	        data, tree, false, epsilon, vert, vert_allocated, edge, edge_allocated);
 
-	return data->tree;
+	return tree;
 }
 
 /** \} */
@@ -836,7 +795,7 @@ BVHTree *bvhtree_from_mesh_edges_ex(
 
 static BVHTree *bvhtree_from_mesh_faces_create_tree(
         float epsilon, int tree_type, int axis,
-        MVert *vert, MFace *face, const int faces_num,
+        const MVert *vert, const MFace *face, const int faces_num,
         const BLI_bitmap *faces_mask, int faces_num_active)
 {
 	BVHTree *tree = NULL;
@@ -880,34 +839,23 @@ static BVHTree *bvhtree_from_mesh_faces_create_tree(
 
 static void bvhtree_from_mesh_faces_setup_data(
         BVHTreeFromMesh *data, BVHTree *tree, const bool is_cached, float epsilon,
-        MVert *vert, const bool vert_allocated,
-        MFace *face, const bool face_allocated)
+        const MVert *vert, const bool vert_allocated,
+        const MFace *face, const bool face_allocated)
 {
 	memset(data, 0, sizeof(*data));
 
-	if (tree) {
-		data->tree = tree;
-		data->cached = is_cached;
+	data->tree = tree;
+	data->cached = is_cached;
 
-		data->nearest_callback = mesh_faces_nearest_point;
-		data->raycast_callback = mesh_faces_spherecast;
-		data->nearest_to_ray_callback = NULL;
+	data->nearest_callback = mesh_faces_nearest_point;
+	data->raycast_callback = mesh_faces_spherecast;
 
-		data->vert = vert;
-		data->vert_allocated = vert_allocated;
-		data->face = face;
-		data->face_allocated = face_allocated;
+	data->vert = vert;
+	data->vert_allocated = vert_allocated;
+	data->face = face;
+	data->face_allocated = face_allocated;
 
-		data->sphere_radius = epsilon;
-	}
-	else {
-		if (vert_allocated) {
-			MEM_freeN(vert);
-		}
-		if (face_allocated) {
-			MEM_freeN(face);
-		}
-	}
+	data->sphere_radius = epsilon;
 }
 
 /* Builds a bvh tree where nodes are the tesselated faces of the given dm */
@@ -950,10 +898,21 @@ BVHTree *bvhtree_from_mesh_faces(
 		/* printf("BVHTree is already build, using cached tree\n"); */
 	}
 
-	/* Setup BVHTreeFromMesh */
-	bvhtree_from_mesh_faces_setup_data(data, tree, true, epsilon, vert, vert_allocated, face, face_allocated);
-
-	return data->tree;
+	if (tree) {
+		/* Setup BVHTreeFromMesh */
+		bvhtree_from_mesh_faces_setup_data(
+		        data, tree, true, epsilon, vert, vert_allocated, face, face_allocated);
+	}
+	else {
+		if (vert_allocated) {
+			MEM_freeN(vert);
+		}
+		if (face_allocated) {
+			MEM_freeN(face);
+		}
+		memset(data, 0, sizeof(*data));
+	}
+	return tree;
 }
 
 /**
@@ -964,8 +923,8 @@ BVHTree *bvhtree_from_mesh_faces(
  * \param numFaces_active if >= 0, number of active faces to add to BVH tree (else will be computed from mask).
  */
 BVHTree *bvhtree_from_mesh_faces_ex(
-        BVHTreeFromMesh *data, MVert *vert, const bool vert_allocated,
-        MFace *face, const int numFaces, const bool face_allocated,
+        BVHTreeFromMesh *data, const MVert *vert, const bool vert_allocated,
+        const MFace *face, const int numFaces, const bool face_allocated,
         const BLI_bitmap *faces_mask, int faces_num_active,
         float epsilon, int tree_type, int axis)
 {
@@ -975,9 +934,10 @@ BVHTree *bvhtree_from_mesh_faces_ex(
 	        faces_mask, faces_num_active);
 
 	/* Setup BVHTreeFromMesh */
-	bvhtree_from_mesh_faces_setup_data(data, tree, false, epsilon, vert, vert_allocated, face, face_allocated);
+	bvhtree_from_mesh_faces_setup_data(
+	        data, tree, false, epsilon, vert, vert_allocated, face, face_allocated);
 
-	return data->tree;
+	return tree;
 }
 
 /** \} */
@@ -1088,34 +1048,20 @@ static void bvhtree_from_mesh_looptri_setup_data(
 {
 	memset(data, 0, sizeof(*data));
 
-	if (tree) {
-		data->tree = tree;
-		data->cached = is_cached;
+	data->tree = tree;
+	data->cached = is_cached;
 
-		data->nearest_callback = mesh_looptri_nearest_point;
-		data->raycast_callback = mesh_looptri_spherecast;
-		data->nearest_to_ray_callback = NULL;
+	data->nearest_callback = mesh_looptri_nearest_point;
+	data->raycast_callback = mesh_looptri_spherecast;
 
-		data->vert = vert;
-		data->vert_allocated = vert_allocated;
-		data->loop = mloop;
-		data->loop_allocated = loop_allocated;
-		data->looptri = looptri;
-		data->looptri_allocated = looptri_allocated;
+	data->vert = vert;
+	data->vert_allocated = vert_allocated;
+	data->loop = mloop;
+	data->loop_allocated = loop_allocated;
+	data->looptri = looptri;
+	data->looptri_allocated = looptri_allocated;
 
-		data->sphere_radius = epsilon;
-	}
-	else {
-		if (vert_allocated) {
-			MEM_freeN((void *)vert);
-		}
-		if (loop_allocated) {
-			MEM_freeN((void *)mloop);
-		}
-		if (looptri_allocated) {
-			MEM_freeN((void *)looptri);
-		}
-	}
+	data->sphere_radius = epsilon;
 }
 
 /**
@@ -1160,7 +1106,6 @@ BVHTree *bvhtree_from_editmesh_looptri_ex(
 		data->tree = tree;
 		data->nearest_callback = editmesh_looptri_nearest_point;
 		data->raycast_callback = editmesh_looptri_spherecast;
-		data->nearest_to_ray_callback = NULL;
 		data->sphere_radius = 0.0f;
 		data->em = em;
 		data->cached = bvhCache != NULL;
@@ -1242,14 +1187,28 @@ BVHTree *bvhtree_from_mesh_looptri(
 		/* printf("BVHTree is already build, using cached tree\n"); */
 	}
 
-	/* Setup BVHTreeFromMesh */
-	bvhtree_from_mesh_looptri_setup_data(
-	        data, tree, true, epsilon,
-	        mvert, vert_allocated,
-	        mloop, loop_allocated,
-	        looptri, looptri_allocated);
+	if (tree) {
+		/* Setup BVHTreeFromMesh */
+		bvhtree_from_mesh_looptri_setup_data(
+		        data, tree, true, epsilon,
+		        mvert, vert_allocated,
+		        mloop, loop_allocated,
+		        looptri, looptri_allocated);
+	}
+	else {
+		if (vert_allocated) {
+			MEM_freeN(mvert);
+		}
+		if (loop_allocated) {
+			MEM_freeN(mloop);
+		}
+		if (looptri_allocated) {
+			MEM_freeN((void *)looptri);
+		}
+		memset(data, 0, sizeof(*data));
+	}
 
-	return data->tree;
+	return tree;
 }
 
 BVHTree *bvhtree_from_mesh_looptri_ex(
@@ -1272,7 +1231,7 @@ BVHTree *bvhtree_from_mesh_looptri_ex(
 	        mloop, loop_allocated,
 	        looptri, looptri_allocated);
 
-	return data->tree;
+	return tree;
 }
 
 /** \} */
@@ -1292,29 +1251,27 @@ void free_bvhtree_from_editmesh(struct BVHTreeFromEditMesh *data)
 /* Frees data allocated by a call to bvhtree_from_mesh_*. */
 void free_bvhtree_from_mesh(struct BVHTreeFromMesh *data)
 {
-	if (data->tree) {
-		if (!data->cached) {
-			BLI_bvhtree_free(data->tree);
-		}
-
-		if (data->vert_allocated) {
-			MEM_freeN((void *)data->vert);
-		}
-		if (data->edge_allocated) {
-			MEM_freeN((void *)data->edge);
-		}
-		if (data->face_allocated) {
-			MEM_freeN((void *)data->face);
-		}
-		if (data->loop_allocated) {
-			MEM_freeN((void *)data->loop);
-		}
-		if (data->looptri_allocated) {
-			MEM_freeN((void *)data->looptri);
-		}
+	if (data->tree && !data->cached) {
+		BLI_bvhtree_free(data->tree);
+	}
 
-		memset(data, 0, sizeof(*data));
+	if (data->vert_allocated) {
+		MEM_freeN((void *)data->vert);
+	}
+	if (data->edge_allocated) {
+		MEM_freeN((void *)data->edge);
+	}
+	if (data->face_allocated) {
+		MEM_freeN((void *)data->face);
 	}
+	if (data->loop_allocated) {
+		MEM_freeN((void *)data->loop);
+	}
+	if (data->looptri_allocated) {
+		MEM_freeN((void *)data->looptri);
+	}
+
+	memset(data, 0, sizeof(*data));
 }
 
 
diff --git a/source/blender/blenkernel/intern/cdderivedmesh.c b/source/blender/blenkernel/intern/cdderivedmesh.c
index 483fa977aff..7042b46330b 100644
--- a/source/blender/blenkernel/intern/cdderivedmesh.c
+++ b/source/blender/blenkernel/intern/cdderivedmesh.c
@@ -2428,8 +2428,12 @@ static DerivedMesh *cddm_copy_ex(DerivedMesh *source,
 	dm->cd_flag = source->cd_flag;
 	dm->dirty = source->dirty;
 
-	/* Tessellation data is never copied, so tag it here. */
-	dm->dirty |= DM_DIRTY_TESS_CDLAYERS;
+	/* Tessellation data is never copied, so tag it here.
+	 * Only tag dirty layers if we really ignored tessellation faces.
+	 */
+	if (!copy_tessface_data) {
+		dm->dirty |= DM_DIRTY_TESS_CDLAYERS;
+	}
 
 	CustomData_copy_data(&source->vertData, &dm->vertData, 0, 0, numVerts);
 	CustomData_copy_data(&source->edgeData, &dm->edgeData, 0, 0, numEdges);
diff --git a/source/blender/blenkernel/intern/image.c b/source/blender/blenkernel/intern/image.c
index a2d94ccc478..318f6480aaf 100644
--- a/source/blender/blenkernel/intern/image.c
+++ b/source/blender/blenkernel/intern/image.c
@@ -3159,7 +3159,7 @@ static ImBuf *load_sequence_single(Image *ima, ImageUser *iuser, int frame, cons
 	struct ImBuf *ibuf;
 	char name[FILE_MAX];
 	int flag;
-	ImageUser iuser_t;
+	ImageUser iuser_t = {0};
 
 	/* XXX temp stuff? */
 	if (ima->lastframe != frame)
@@ -3167,8 +3167,12 @@ static ImBuf *load_sequence_single(Image *ima, ImageUser *iuser, int frame, cons
 
 	ima->lastframe = frame;
 
-	if (iuser)
+	if (iuser) {
 		iuser_t = *iuser;
+	}
+	else {
+		/* TODO(sergey): Do we need to initialize something here? */
+	}
 
 	iuser_t.view = view_id;
 	BKE_image_user_file_path(&iuser_t, ima, name);
diff --git a/source/blender/blenkernel/intern/mesh.c b/source/blender/blenkernel/intern/mesh.c
index 134173580ae..d93e66c6171 100644
--- a/source/blender/blenkernel/intern/mesh.c
+++ b/source/blender/blenkernel/intern/mesh.c
@@ -67,6 +67,11 @@
 
 #include "DEG_depsgraph.h"
 
+/* Define for cases when you want extra validation of mesh
+ * after certain modifications.
+ */
+// #undef VALIDATE_MESH
+
 enum {
 	MESHCMP_DVERT_WEIGHTMISMATCH = 1,
 	MESHCMP_DVERT_GROUPMISMATCH,
@@ -2094,104 +2099,296 @@ void BKE_mesh_calc_normals_split(Mesh *mesh)
 	}
 }
 
-/* Spli faces based on the edge angle.
- * Matches behavior of face splitting in render engines.
+/* Split faces helper functions. */
+
+enum {
+	/* Vertex is adjacent to some loop which normal is different,
+	 * hence split of this vertex is required.
+	 */
+	SPLIT_VERT_NEED_SPLIT = (1 << 0),
+	/* Original vertex was already re-used by split logic. */
+	SPLIT_VERT_REUSED     = (1 << 1),
+};
+enum {
+	/* Edge is adjacent to any of vertex tagged for split.
+	 */
+	SPLIT_EDGE_NEED_SPLIT = (1 << 0),
+	/* Original edge was already re-used by split logic. */
+	SPLIT_EDGE_REUSED     = (1 << 1),
+};
+
+/* Tag vertices which normals are not equal to any adjacent loop
+ * and hence split on that vertex is required.
+ *
+ * Returns truth if any of vertex needs to be split.
  */
-void BKE_mesh_split_faces(Mesh *mesh)
+static bool split_faces_tag_verts(const Mesh *mesh, uchar *vert_flags)
+{
+	const int num_polys = mesh->totpoly;
+	const MVert *mvert = mesh->mvert;
+	const MLoop *mloop = mesh->mloop;
+	const MPoly *mpoly = mesh->mpoly;
+	float (*lnors)[3] = CustomData_get_layer(&mesh->ldata, CD_NORMAL);
+	bool has_split_verts = false;
+	for (int poly = 0; poly < num_polys; poly++) {
+		const MPoly *mp = &mpoly[poly];
+		for (int loop = 0; loop < mp->totloop; loop++) {
+			const MLoop *ml = &mloop[mp->loopstart + loop];
+			const MVert *mv = &mvert[ml->v];
+			float vn[3];
+			normal_short_to_float_v3(vn, mv->no);
+			if (len_squared_v3v3(vn, lnors[mp->loopstart + loop]) > FLT_EPSILON) {
+				vert_flags[ml->v] |= SPLIT_VERT_NEED_SPLIT;
+				has_split_verts = true;
+			}
+		}
+	}
+	return has_split_verts;
+}
+
+/* Count number of new vertices to be added.
+ *
+ * Note that one of the loop where split is required will re-use
+ * it's vertex in order to avoid creation of loose vertices.
+ */
+static int split_faces_count_new_verts(const Mesh *mesh, uchar *vert_flags)
+{
+	const int num_polys = mesh->totpoly;
+	const MLoop *mloop = mesh->mloop;
+	const MPoly *mpoly = mesh->mpoly;
+	int num_new_verts = 0;
+	for (int poly = 0; poly < num_polys; poly++) {
+		const MPoly *mp = &mpoly[poly];
+		for (int loop = 0; loop < mp->totloop; loop++) {
+			const MLoop *ml = &mloop[mp->loopstart + loop];
+			if (vert_flags[ml->v] & SPLIT_VERT_NEED_SPLIT) {
+				if (vert_flags[ml->v] & SPLIT_VERT_REUSED) {
+					++num_new_verts;
+				}
+				else {
+					vert_flags[ml->v] |= SPLIT_VERT_REUSED;
+				}
+			}
+		}
+	}
+	return num_new_verts;
+}
+
+/* Tag edges which are adjacent to at least one vertex tagged for split. */
+static void split_faces_tag_edges(Mesh *mesh,
+                                  const uchar *vert_flags,
+                                  uchar *edge_flags)
+{
+	const int num_polys = mesh->totpoly;
+	const MLoop *mloop = mesh->mloop;
+	const MPoly *mpoly = mesh->mpoly;
+	for (int poly = 0; poly < num_polys; poly++) {
+		const MPoly *mp = &mpoly[poly];
+		int loop_prev = mp->totloop - 1;
+		for (int loop = 0; loop < mp->totloop; loop++) {
+			const int poly_loop_prev = mp->loopstart + loop_prev;
+			const MLoop *ml = &mloop[mp->loopstart + loop];
+			const MLoop *ml_prev = &mloop[poly_loop_prev];
+			const int mv_flag = vert_flags[ml->v];
+			const int mv_prev_flag = vert_flags[ml_prev->v];
+			bool need_split = false;
+			if (mv_flag & SPLIT_VERT_NEED_SPLIT) {
+				if (mv_prev_flag & SPLIT_VERT_NEED_SPLIT) {
+					/* Create new edge between twp split vertices. */
+					need_split = true;
+				}
+				else {
+					/* Create new edge from existing vertex to a split one. */
+					need_split = true;
+				}
+			}
+			else if (mv_prev_flag & SPLIT_VERT_NEED_SPLIT) {
+				/* Create new edge from split vertex to existing one. */
+				need_split = true;
+			}
+			if (need_split) {
+				edge_flags[ml_prev->e] |= SPLIT_EDGE_NEED_SPLIT;
+			}
+			loop_prev = loop;
+		}
+	}
+}
+
+/* Count number of new edges to be added.
+ *
+ * Note that one of the loop where split is required will re-use
+ * it's edge in order to avoid creation of loose edges.
+ */
+static int split_faces_count_new_edges(const Mesh *mesh, uchar *edge_flags)
 {
-	const int num_verts = mesh->totvert;
-	const int num_edges = mesh->totedge;
+	const int num_polys = mesh->totpoly;
+	const MLoop *mloop = mesh->mloop;
+	const MPoly *mpoly = mesh->mpoly;
+	int num_new_edges = 0;
+	for (int poly = 0; poly < num_polys; poly++) {
+		const MPoly *mp = &mpoly[poly];
+		for (int loop = 0; loop < mp->totloop; loop++) {
+			const MLoop *ml = &mloop[mp->loopstart + loop];
+			if (edge_flags[ml->e] & SPLIT_EDGE_NEED_SPLIT) {
+				if (edge_flags[ml->e] & SPLIT_EDGE_REUSED) {
+					++num_new_edges;
+				}
+				else {
+					edge_flags[ml->e] |= SPLIT_EDGE_REUSED;
+				}
+			}
+		}
+	}
+	return num_new_edges;
+}
+
+/* Perform actual split of vertices.
+ *
+ * NOTE: Will leave edges in inconsistent state.
+ */
+static void split_faces_split_verts(Mesh *mesh,
+                                    const int num_new_verts,
+                                    uchar *vert_flags)
+{
+	const int num_verts = mesh->totvert - num_new_verts;
 	const int num_polys = mesh->totpoly;
 	MVert *mvert = mesh->mvert;
+	MLoop *mloop = mesh->mloop;
+	MPoly *mpoly = mesh->mpoly;
+	const float (*lnors)[3] = CustomData_get_layer(&mesh->ldata, CD_NORMAL);
+	int num_added_verts = 0;
+	/* Clear reused flag, we need it again. */
+	for (int i = 0; i < num_verts; ++i) {
+		vert_flags[i] &= ~SPLIT_VERT_REUSED;
+	}
+	for (int poly = 0; poly < num_polys; poly++) {
+		MPoly *mp = &mpoly[poly];
+		/* First we split all vertices to get proper flag whether they are
+		 * split or not for all of them before handling edges.
+		 */
+		for (int loop = 0; loop < mp->totloop; loop++) {
+			int poly_loop = mp->loopstart + loop;
+			MLoop *ml = &mloop[poly_loop];
+			if (vert_flags[ml->v] & SPLIT_VERT_NEED_SPLIT) {
+				if ((vert_flags[ml->v] & SPLIT_VERT_REUSED) == 0) {
+					/* Ignore first split on vertex, re-use it instead. */
+					vert_flags[ml->v] |= SPLIT_VERT_REUSED;
+					continue;
+				}
+				/* Create new vertex. */
+				int new_vert = num_verts + num_added_verts;
+				CustomData_copy_data(&mesh->vdata, &mesh->vdata,
+				                     ml->v, new_vert, 1);
+				normal_float_to_short_v3(mvert[new_vert].no,
+				                         lnors[poly_loop]);
+				ml->v = new_vert;
+				num_added_verts++;
+			}
+		}
+	}
+}
+
+/* Perform actual split of edges.
+ *
+ * NOTE: Will correct all edges.
+ */
+static void split_faces_split_edges(Mesh *mesh,
+                                    const int num_new_edges,
+                                    uchar *edge_flags)
+{
+	const int num_edges = mesh->totedge - num_new_edges;
+	const int num_polys = mesh->totpoly;
 	MEdge *medge = mesh->medge;
 	MLoop *mloop = mesh->mloop;
 	MPoly *mpoly = mesh->mpoly;
-	float (*lnors)[3];
-	int poly, num_new_verts = 0;
+	int num_added_edges = 0;
+	/* Clear reused flag, we need it again. */
+	for (int i = 0; i < num_edges; ++i) {
+		edge_flags[i] &= ~SPLIT_EDGE_REUSED;
+	}
+	for (int poly = 0; poly < num_polys; poly++) {
+		MPoly *mp = &mpoly[poly];
+		for (int loop = 0, loop_prev = mp->totloop - 1; loop < mp->totloop; loop++) {
+			const int poly_loop_prev = mp->loopstart + loop_prev;
+			const MLoop *ml = &mloop[mp->loopstart + loop];
+			MLoop *ml_prev = &mloop[poly_loop_prev];
+			MEdge *me_prev = &medge[ml_prev->e];
+			if (edge_flags[ml_prev->e] & SPLIT_EDGE_NEED_SPLIT) {
+				if ((edge_flags[ml_prev->e] & SPLIT_EDGE_REUSED) == 0) {
+					edge_flags[ml_prev->e] |= SPLIT_EDGE_REUSED;
+					me_prev->v1 = ml_prev->v;
+					me_prev->v2 = ml->v;
+				}
+				else {
+					const int index = num_edges + num_added_edges;
+					CustomData_copy_data(&mesh->edata, &mesh->edata,
+					                     ml_prev->e, index, 1);
+					MEdge *me_new = &medge[index];
+					me_new->v1 = ml_prev->v;
+					me_new->v2 = ml->v;
+					ml_prev->e = index;
+					num_added_edges++;
+				}
+			}
+			loop_prev = loop;
+		}
+	}
+}
+
+/* Split faces based on the edge angle.
+ * Matches behavior of face splitting in render engines.
+ */
+void BKE_mesh_split_faces(Mesh *mesh)
+{
+	const int num_verts = mesh->totvert;
+	const int num_edges = mesh->totedge;
+	const int num_polys = mesh->totpoly;
 	if ((mesh->flag & ME_AUTOSMOOTH) == 0) {
 		return;
 	}
+	if (num_polys == 0) {
+		return;
+	}
 	BKE_mesh_tessface_clear(mesh);
 	/* Compute loop normals if needed. */
 	if (!CustomData_has_layer(&mesh->ldata, CD_NORMAL)) {
 		BKE_mesh_calc_normals_split(mesh);
 	}
-	lnors = CustomData_get_layer(&mesh->ldata, CD_NORMAL);
-	/* Count number of vertices to be split. */
-	for (poly = 0; poly < num_polys; poly++) {
-		MPoly *mp = &mpoly[poly];
-		int loop;
-		for (loop = 0; loop < mp->totloop; loop++) {
-			MLoop *ml = &mloop[mp->loopstart + loop];
-			MVert *mv = &mvert[ml->v];
-			float vn[3];
-			normal_short_to_float_v3(vn, mv->no);
-			if (!equals_v3v3(vn, lnors[mp->loopstart + loop])) {
-				num_new_verts++;
-			}
-		}
-	}
-	if (num_new_verts == 0) {
-		/* No new vertices are to be added, can do early exit. */
+	/* Runtime flags. */
+	uchar *vert_flags = MEM_callocN(sizeof(*vert_flags) * num_verts,
+	                                "split faces vert flags");
+	/* Tag vertces and check whether anything is tagged. */
+	if (!split_faces_tag_verts(mesh, vert_flags)) {
+		/* No new vertices to be split added, can do early exit. */
+		MEM_freeN(vert_flags);
 		return;
 	}
+	/* Flush vertex flags to edges. */
+	uchar *edge_flags = MEM_callocN(sizeof(*edge_flags) * num_edges,
+	                                "split faces edge flags");
+	split_faces_tag_edges(mesh, vert_flags, edge_flags);
+	/* Count amount of new geometry. */
+	int num_new_verts = split_faces_count_new_verts(mesh, vert_flags);
+	int num_new_edges =  split_faces_count_new_edges(mesh, edge_flags);
 	/* Reallocate all vert and edge related data. */
 	mesh->totvert += num_new_verts;
-	mesh->totedge += 2 * num_new_verts;
+	mesh->totedge += num_new_edges;
 	CustomData_realloc(&mesh->vdata, mesh->totvert);
 	CustomData_realloc(&mesh->edata, mesh->totedge);
 	/* Update pointers to a newly allocated memory. */
 	BKE_mesh_update_customdata_pointers(mesh, false);
-	mvert = mesh->mvert;
-	medge = mesh->medge;
-	/* Perform actual vertex split. */
-	num_new_verts = 0;
-	for (poly = 0; poly < num_polys; poly++) {
-		MPoly *mp = &mpoly[poly];
-		int loop;
-		for (loop = 0; loop < mp->totloop; loop++) {
-			int poly_loop = mp->loopstart + loop;
-			MLoop *ml = &mloop[poly_loop];
-			MVert *mv = &mvert[ml->v];
-			float vn[3];
-			normal_short_to_float_v3(vn, mv->no);
-			if (!equals_v3v3(vn, lnors[mp->loopstart + loop])) {
-				int poly_loop_prev = mp->loopstart + (loop + mp->totloop - 1) % mp->totloop;
-				MLoop *ml_prev = &mloop[poly_loop_prev];
-				int new_edge_prev, new_edge;
-				/* Cretae new vertex. */
-				int new_vert = num_verts + num_new_verts;
-				CustomData_copy_data(&mesh->vdata, &mesh->vdata,
-				                     ml->v, new_vert, 1);
-				normal_float_to_short_v3(mvert[new_vert].no,
-				                         lnors[poly_loop]);
-				/* Create new edges. */
-				new_edge_prev = num_edges + 2 * num_new_verts;
-				new_edge = num_edges + 2 * num_new_verts + 1;
-				CustomData_copy_data(&mesh->edata, &mesh->edata,
-				                     ml_prev->e, new_edge_prev, 1);
-				CustomData_copy_data(&mesh->edata, &mesh->edata,
-				                     ml->e, new_edge, 1);
-				if (medge[new_edge_prev].v1 == ml->v) {
-					medge[new_edge_prev].v1 = new_vert;
-				}
-				else {
-					medge[new_edge_prev].v2 = new_vert;
-				}
-				if (medge[new_edge].v1 == ml->v) {
-					medge[new_edge].v1 = new_vert;
-				}
-				else {
-					medge[new_edge].v2 = new_vert;
-				}
-
-				ml->v = new_vert;
-				ml_prev->e = new_edge_prev;
-				ml->e = new_edge;
-				num_new_verts++;
-			}
-		}
-	}
+	/* Perform actual split of vertices and adjacent edges. */
+	split_faces_split_verts(mesh, num_new_verts, vert_flags);
+	split_faces_split_edges(mesh, num_new_edges, edge_flags);
+	/* CD_NORMAL is expected to be temporary only, and it's invalid at
+	 * this point anyway.
+	 */
+	CustomData_free_layers(&mesh->ldata, CD_NORMAL, mesh->totloop);
+	MEM_freeN(vert_flags);
+	MEM_freeN(edge_flags);
+#ifdef VALIDATE_MESH
+	BKE_mesh_validate(mesh, true, true);
+#endif
 }
 
 /* settings: 1 - preview, 2 - render */
diff --git a/source/blender/blenkernel/intern/object.c b/source/blender/blenkernel/intern/object.c
index 7b3869b3a94..ad0ce3fef9b 100644
--- a/source/blender/blenkernel/intern/object.c
+++ b/source/blender/blenkernel/intern/object.c
@@ -2241,66 +2241,6 @@ void BKE_boundbox_minmax(const BoundBox *bb, float obmat[4][4], float r_min[3],
 	}
 }
 
-/**
- * Returns a BBox which each dimensions are at least epsilon.
- * \note In case a given dimension needs to be enlarged, its final value will be in [epsilon, 3 * epsilon] range.
- *
- * \param bb the input bbox to check.
- * \param bb_temp the temp bbox to modify (\a bb content is never changed).
- * \param epsilon the minimum dimension to ensure.
- * \return either bb (if nothing needed to be changed) or bb_temp.
- */
-BoundBox *BKE_boundbox_ensure_minimum_dimensions(BoundBox *bb, BoundBox *bb_temp, const float epsilon)
-{
-	if (fabsf(bb->vec[0][0] - bb->vec[4][0]) < epsilon) {
-		/* Flat along X axis... */
-		*bb_temp = *bb;
-		bb = bb_temp;
-		bb->vec[0][0] -= epsilon;
-		bb->vec[1][0] -= epsilon;
-		bb->vec[2][0] -= epsilon;
-		bb->vec[3][0] -= epsilon;
-		bb->vec[4][0] += epsilon;
-		bb->vec[5][0] += epsilon;
-		bb->vec[6][0] += epsilon;
-		bb->vec[7][0] += epsilon;
-	}
-
-	if (fabsf(bb->vec[0][1] - bb->vec[3][1]) < epsilon) {
-		/* Flat along Y axis... */
-		if (bb != bb_temp) {
-			*bb_temp = *bb;
-			bb = bb_temp;
-		}
-		bb->vec[0][1] -= epsilon;
-		bb->vec[1][1] -= epsilon;
-		bb->vec[4][1] -= epsilon;
-		bb->vec[5][1] -= epsilon;
-		bb->vec[2][1] += epsilon;
-		bb->vec[3][1] += epsilon;
-		bb->vec[6][1] += epsilon;
-		bb->vec[7][1] += epsilon;
-	}
-
-	if (fabsf(bb->vec[0][2] - bb->vec[1][2]) < epsilon) {
-		/* Flat along Z axis... */
-		if (bb != bb_temp) {
-			*bb_temp = *bb;
-			bb = bb_temp;
-		}
-		bb->vec[0][2] -= epsilon;
-		bb->vec[3][2] -= epsilon;
-		bb->vec[4][2] -= epsilon;
-		bb->vec[7][2] -= epsilon;
-		bb->vec[1][2] += epsilon;
-		bb->vec[2][2] += epsilon;
-		bb->vec[5][2] += epsilon;
-		bb->vec[6][2] += epsilon;
-	}
-
-	return bb;
-}
-
 BoundBox *BKE_object_boundbox_get(Object *ob)
 {
 	BoundBox *bb = NULL;
diff --git a/source/blender/blenlib/BLI_kdopbvh.h b/source/blender/blenlib/BLI_kdopbvh.h
index 91d39801645..ba565fca522 100644
--- a/source/blender/blenlib/BLI_kdopbvh.h
+++ b/source/blender/blenlib/BLI_kdopbvh.h
@@ -95,10 +95,6 @@ typedef void (*BVHTree_NearestPointCallback)(void *userdata, int index, const fl
 /* callback must update hit in case it finds a nearest successful hit */
 typedef void (*BVHTree_RayCastCallback)(void *userdata, int index, const BVHTreeRay *ray, BVHTreeRayHit *hit);
 
-/* callback must update nearest in case it finds a nearest result */
-typedef void (*BVHTree_NearestToRayCallback)(void *userdata, const float ray_co[3], const float ray_dir[3],
-                                             const float scale[3], int index, BVHTreeNearest *nearest);
-
 /* callback to check if 2 nodes overlap (use thread if intersection results need to be stored) */
 typedef bool (*BVHTree_OverlapCallback)(void *userdata, int index_a, int index_b, int thread);
 
@@ -143,18 +139,6 @@ int BLI_bvhtree_find_nearest(
         BVHTree *tree, const float co[3], BVHTreeNearest *nearest,
         BVHTree_NearestPointCallback callback, void *userdata);
 
-int BLI_bvhtree_find_nearest_to_ray_angle(
-        BVHTree *tree, const float co[3], const float dir[3],
-        const bool ray_is_normalized, const float scale[3],
-        BVHTreeNearest *nearest,
-        BVHTree_NearestToRayCallback callback, void *userdata);
-
-int BLI_bvhtree_find_nearest_to_ray(
-        BVHTree *tree, const float co[3], const float dir[3],
-        const bool ray_is_normalized, const float scale[3],
-        BVHTreeNearest *nearest,
-        BVHTree_NearestToRayCallback callback, void *userdata);
-
 int BLI_bvhtree_ray_cast_ex(
         BVHTree *tree, const float co[3], const float dir[3], float radius, BVHTreeRayHit *hit,
         BVHTree_RayCastCallback callback, void *userdata,
diff --git a/source/blender/blenlib/BLI_math_geom.h b/source/blender/blenlib/BLI_math_geom.h
index 4a85e859c16..f1d9c9571f2 100644
--- a/source/blender/blenlib/BLI_math_geom.h
+++ b/source/blender/blenlib/BLI_math_geom.h
@@ -298,23 +298,6 @@ bool isect_ray_aabb_v3_simple(
         const float bb_min[3], const float bb_max[3],
         float *tmin, float *tmax);
 
-struct NearestRayToAABB_Precalc {
-	float ray_origin[3];
-	float ray_direction[3];
-	float ray_inv_dir[3];
-	float cdot_axis[3];
-	float idiag_sq[3];
-	bool sign[3];
-};
-
-void dist_squared_ray_to_aabb_v3_precalc(
-        struct NearestRayToAABB_Precalc *data,
-        const float ray_origin[3], const float ray_direction[3]);
-float dist_squared_ray_to_aabb_v3(
-        const struct NearestRayToAABB_Precalc *data,
-        const float bb_min[3], const float bb_max[3],
-        bool r_axis_closest[3]);
-
 /* other */
 bool isect_sweeping_sphere_tri_v3(const float p1[3], const float p2[3], const float radius,
                                   const float v0[3], const float v1[3], const float v2[3], float *r_lambda, float ipoint[3]);
diff --git a/source/blender/blenlib/intern/BLI_kdopbvh.c b/source/blender/blenlib/intern/BLI_kdopbvh.c
index b14007a88cb..19d9711922e 100644
--- a/source/blender/blenlib/intern/BLI_kdopbvh.c
+++ b/source/blender/blenlib/intern/BLI_kdopbvh.c
@@ -159,29 +159,6 @@ typedef struct BVHRayCastData {
 	BVHTreeRayHit hit;
 } BVHRayCastData;
 
-typedef struct BVHNearestRayData {
-	BVHTree *tree;
-	BVHTree_NearestToRayCallback callback;
-	void    *userdata;
-
-	struct {
-		bool sign[3];
-		float origin[3];
-		float direction[3];
-
-		float direction_scaled_square[3];
-		float inv_dir[3];
-
-		float cdot_axis[3];
-	} ray;
-
-	bool pick_smallest[3];
-
-	BVHTreeNearest nearest;
-
-	float scale[3];
-} BVHNearestRayData;
-
 /** \} */
 
 
@@ -1900,453 +1877,6 @@ void BLI_bvhtree_ray_cast_all(
 
 /* -------------------------------------------------------------------- */
 
-/** \name BLI_bvhtree_find_nearest_to_ray functions
- *
- * \{ */
-
-static void dist_squared_ray_to_aabb_scaled_v3_precalc(
-        BVHNearestRayData *data,
-        const float ray_origin[3], const float ray_direction[3],
-        const bool ray_is_normalized, const float scale[3])
-{
-	if (scale) {
-		copy_v3_v3(data->scale, scale);
-	}
-	else {
-		copy_v3_fl(data->scale, 1.0f);
-	}
-	/* un-normalize ray */
-	if (ray_is_normalized && scale &&
-	    (data->scale[0] != 1.0f || data->scale[1] != 1.0f || data->scale[2] != 1.0f))
-	{
-		data->ray.direction[0] = ray_direction[0] * data->scale[0];
-		data->ray.direction[1] = ray_direction[1] * data->scale[1];
-		data->ray.direction[2] = ray_direction[2] * data->scale[2];
-
-		mul_v3_v3fl(data->ray.direction, ray_direction, 1 / len_v3(data->ray.direction));
-	}
-	else {
-		copy_v3_v3(data->ray.direction, ray_direction);
-	}
-
-	float dir_sq[3];
-
-	for (int i = 0; i < 3; i++) {
-		data->ray.origin[i] = ray_origin[i];
-		data->ray.inv_dir[i] = (data->ray.direction[i] != 0.0f) ?
-		                       (1.0f / data->ray.direction[i]) : FLT_MAX;
-		/* It has to be in function of `ray.inv_dir`,
-		 * since the division of 1 by 0.0f, can be -inf or +inf */
-		data->ray.sign[i] = (data->ray.inv_dir[i] < 0.0f);
-
-		data->ray.direction_scaled_square[i] = data->ray.direction[i] * data->scale[i];
-
-		dir_sq[i] = SQUARE(data->ray.direction_scaled_square[i]);
-
-		data->ray.direction_scaled_square[i] *= data->scale[i];
-	}
-
-	/* `diag_sq` Length square of each face diagonal */
-	float diag_sq[3] = {
-		dir_sq[1] + dir_sq[2],
-		dir_sq[0] + dir_sq[2],
-		dir_sq[0] + dir_sq[1],
-	};
-
-	data->ray.cdot_axis[0] = (diag_sq[0] != 0.0f) ? data->ray.direction[0] / diag_sq[0] : FLT_MAX;
-	data->ray.cdot_axis[1] = (diag_sq[1] != 0.0f) ? data->ray.direction[1] / diag_sq[1] : FLT_MAX;
-	data->ray.cdot_axis[2] = (diag_sq[2] != 0.0f) ? data->ray.direction[2] / diag_sq[2] : FLT_MAX;
-}
-
-/**
- * Returns the squared distance from a ray to a bound-box `AABB`.
- * It is based on `fast_ray_nearest_hit` solution to obtain
- * the coordinates of the nearest edge of Bound Box to the ray
- */
-MINLINE float dist_squared_ray_to_aabb_scaled_v3__impl(
-        const BVHNearestRayData *data,
-        const float bv[6], float *r_depth_sq, bool r_axis_closest[3])
-{
-
-	/* `tmin` is a vector that has the smaller distances to each of the
-	 * infinite planes of the `AABB` faces (hit in nearest face X plane,
-	 * nearest face Y plane and nearest face Z plane) */
-	float local_bvmin[3], local_bvmax[3];
-
-	if (data->ray.sign[0]) {
-		local_bvmin[0] = bv[1];
-		local_bvmax[0] = bv[0];
-	}
-	else {
-		local_bvmin[0] = bv[0];
-		local_bvmax[0] = bv[1];
-	}
-
-	if (data->ray.sign[1]) {
-		local_bvmin[1] = bv[3];
-		local_bvmax[1] = bv[2];
-	}
-	else {
-		local_bvmin[1] = bv[2];
-		local_bvmax[1] = bv[3];
-	}
-
-	if (data->ray.sign[2]) {
-		local_bvmin[2] = bv[5];
-		local_bvmax[2] = bv[4];
-	}
-	else {
-		local_bvmin[2] = bv[4];
-		local_bvmax[2] = bv[5];
-	}
-
-	sub_v3_v3(local_bvmin, data->ray.origin);
-	sub_v3_v3(local_bvmax, data->ray.origin);
-
-	const float tmin[3] = {
-		local_bvmin[0] * data->ray.inv_dir[0],
-		local_bvmin[1] * data->ray.inv_dir[1],
-		local_bvmin[2] * data->ray.inv_dir[2],
-	};
-
-	/* `tmax` is a vector that has the longer distances to each of the
-	 * infinite planes of the `AABB` faces (hit in farthest face X plane,
-	 * farthest face Y plane and farthest face Z plane) */
-	const float tmax[3] = {
-		local_bvmax[0] * data->ray.inv_dir[0],
-		local_bvmax[1] * data->ray.inv_dir[1],
-		local_bvmax[2] * data->ray.inv_dir[2],
-	};
-	/* `v1` and `v3` is be the coordinates of the nearest `AABB` edge to the ray*/
-	float v1[3], v2[3];
-	/* `rtmin` is the highest value of the smaller distances. == max_axis_v3(tmin)
-	 * `rtmax` is the lowest value of longer distances. == min_axis_v3(tmax)*/
-	float rtmin, rtmax, mul;
-	/* `main_axis` is the axis equivalent to edge close to the ray */
-	int main_axis;
-
-	r_axis_closest[0] = false;
-	r_axis_closest[1] = false;
-	r_axis_closest[2] = false;
-
-	/* *** min_axis_v3(tmax) *** */
-	if ((tmax[0] <= tmax[1]) && (tmax[0] <= tmax[2])) {
-		// printf("# Hit in X %s\n", data->sign[0] ? "min", "max");
-		rtmax = tmax[0];
-		v1[0] = v2[0] = local_bvmax[0];
-		mul = local_bvmax[0] * data->ray.direction_scaled_square[0];
-		main_axis = 3;
-		r_axis_closest[0] = data->ray.sign[0];
-	}
-	else if ((tmax[1] <= tmax[0]) && (tmax[1] <= tmax[2])) {
-		// printf("# Hit in Y %s\n", data->sign[1] ? "min", "max");
-		rtmax = tmax[1];
-		v1[1] = v2[1] = local_bvmax[1];
-		mul = local_bvmax[1] * data->ray.direction_scaled_square[1];
-		main_axis = 2;
-		r_axis_closest[1] = data->ray.sign[1];
-	}
-	else {
-		// printf("# Hit in Z %s\n", data->sign[2] ? "min", "max");
-		rtmax = tmax[2];
-		v1[2] = v2[2] = local_bvmax[2];
-		mul = local_bvmax[2] * data->ray.direction_scaled_square[2];
-		main_axis = 1;
-		r_axis_closest[2] = data->ray.sign[2];
-	}
-
-	/* *** max_axis_v3(tmin) *** */
-	if ((tmin[0] >= tmin[1]) && (tmin[0] >= tmin[2])) {
-		// printf("# To X %s\n", data->sign[0] ? "max", "min");
-		rtmin = tmin[0];
-		v1[0] = v2[0] = local_bvmin[0];
-		mul += local_bvmin[0] * data->ray.direction_scaled_square[0];
-		main_axis -= 3;
-		r_axis_closest[0] = !data->ray.sign[0];
-	}
-	else if ((tmin[1] >= tmin[0]) && (tmin[1] >= tmin[2])) {
-		// printf("# To Y %s\n", data->sign[1] ? "max", "min");
-		rtmin = tmin[1];
-		v1[1] = v2[1] = local_bvmin[1];
-		mul += local_bvmin[1] * data->ray.direction_scaled_square[1];
-		main_axis -= 1;
-		r_axis_closest[1] = !data->ray.sign[1];
-	}
-	else {
-		// printf("# To Z %s\n", data->sign[2] ? "max", "min");
-		rtmin = tmin[2];
-		v1[2] = v2[2] = local_bvmin[2];
-		mul += local_bvmin[2] * data->ray.direction_scaled_square[2];
-		main_axis -= 2;
-		r_axis_closest[2] = !data->ray.sign[2];
-	}
-	/* *** end min/max axis *** */
-
-	if (main_axis < 0)
-		main_axis += 3;
-
-	/* if rtmin < rtmax, ray intersect `AABB` */
-	if (rtmin <= rtmax) {
-#ifdef IGNORE_BEHIND_RAY
-		/* `if rtmax < depth_min`, the whole `AABB` is behind us */
-		if (rtmax < min_depth) {
-			return fallback;
-		}
-#endif
-		const float proj = rtmin * data->ray.direction[main_axis];
-
-		if (data->ray.sign[main_axis])
-			r_axis_closest[main_axis] = (proj - local_bvmax[main_axis]) < (local_bvmin[main_axis] - proj);
-		else
-			r_axis_closest[main_axis] = (proj - local_bvmin[main_axis]) < (local_bvmax[main_axis] - proj);
-
-		//if (r_depth_sq)
-		//	*r_depth_sq = SQUARE(rtmin);
-
-		return 0.0f;
-	}
-#ifdef IGNORE_BEHIND_RAY
-	/* `if rtmin < depth_min`, the whole `AABB` is behing us */
-	else if (rtmin < min_depth) {
-		return fallback;
-	}
-#endif
-
-	if (data->ray.sign[main_axis]) {
-		v1[main_axis] = local_bvmax[main_axis];
-		v2[main_axis] = local_bvmin[main_axis];
-	}
-	else {
-		v1[main_axis] = local_bvmin[main_axis];
-		v2[main_axis] = local_bvmax[main_axis];
-	}
-	{
-		/* `proj` equals to nearest point on the ray closest to the edge `v1 v2` of the `AABB`. */
-		const float proj = mul * data->ray.cdot_axis[main_axis];
-		float depth_sq, r_point[3];
-		if (v1[main_axis] > proj) { /* the nearest point to the ray is the point v1 */
-			r_axis_closest[main_axis] = true;
-			/* `depth` is equivalent the distance of the the projection of v1 on the ray */
-			depth_sq = mul + data->ray.direction_scaled_square[main_axis] * v1[main_axis];
-
-			copy_v3_v3(r_point, v1);
-		}
-		else if (v2[main_axis] < proj) { /* the nearest point of the ray is the point v2 */
-			r_axis_closest[main_axis] = false;
-
-			depth_sq = mul + data->ray.direction_scaled_square[main_axis] * v2[main_axis];
-
-			copy_v3_v3(r_point, v2);
-		}
-		else {  /* the nearest point of the ray is on the edge of the `AABB`. */
-			r_axis_closest[main_axis] = (proj - v1[main_axis]) < (v2[main_axis] - proj);
-
-			depth_sq = mul + data->ray.direction_scaled_square[main_axis] * proj;
-#if 0
-			r_point[0] = main_axis == 0 ? proj : v2[0];
-			r_point[1] = main_axis == 1 ? proj : v2[1];
-			r_point[2] = main_axis == 2 ? proj : v2[2];
-#else
-			v2[main_axis] = proj;
-			copy_v3_v3(r_point, v2);
-#endif
-		}
-		depth_sq *= depth_sq;
-
-		if (r_depth_sq)
-			*r_depth_sq = depth_sq;
-
-		/* TODO: scale can be optional */
-		r_point[0] *= data->scale[0];
-		r_point[1] *= data->scale[1];
-		r_point[2] *= data->scale[2];
-
-		return len_squared_v3(r_point) - depth_sq;
-	}
-}
-
-/**
- * <pre>
- *  + r_point
- *  |
- *  | dist
- *  |
- *  +----depth----+orig <-- dir
- *
- * tangent = dist/depth
- * </pre>
- */
-static float calc_tangent_sq(BVHNearestRayData *data, BVHNode *node)
-{
-	float depth_sq;
-	const float dist_sq = dist_squared_ray_to_aabb_scaled_v3__impl(
-	        data, node->bv, &depth_sq, data->pick_smallest);
-
-	return (dist_sq != 0.0f) ? (dist_sq / depth_sq) : 0.0f;
-}
-
-static float calc_dist_sq_to_ray(BVHNearestRayData *data, BVHNode *node)
-{
-	return dist_squared_ray_to_aabb_scaled_v3__impl(
-	        data, node->bv, NULL,
-	        data->pick_smallest);
-}
-
-static void dfs_find_lowest_tangent_dfs(BVHNearestRayData *data, BVHNode *node)
-{
-	if (node->totnode == 0) {
-		if (data->callback) {
-			data->callback(data->userdata, data->ray.origin, data->ray.direction,
-			               data->scale, node->index, &data->nearest);
-		}
-		else {
-			data->nearest.index = node->index;
-			data->nearest.dist_sq = calc_tangent_sq(data, node);
-			/* TODO: return a value to the data->nearest.co
-			 * not urgent however since users currently define own callbacks */
-		}
-	}
-	else {
-		int i;
-		/* First pick the closest node to dive on */
-		if (data->pick_smallest[node->main_axis]) {
-			for (i = 0; i != node->totnode; i++) {
-				if (calc_tangent_sq(data, node->children[i]) < data->nearest.dist_sq) {
-					dfs_find_lowest_tangent_dfs(data, node->children[i]);
-				}
-			}
-		}
-		else {
-			for (i = node->totnode - 1; i >= 0; i--) {
-				if (calc_tangent_sq(data, node->children[i]) < data->nearest.dist_sq) {
-					dfs_find_lowest_tangent_dfs(data, node->children[i]);
-				}
-			}
-		}
-	}
-}
-
-static void dfs_find_nearest_to_ray_dfs(BVHNearestRayData *data, BVHNode *node)
-{
-	if (node->totnode == 0) {
-		if (data->callback) {
-			data->callback(data->userdata, data->ray.origin, data->ray.direction,
-			               data->scale, node->index, &data->nearest);
-		}
-		else {
-			data->nearest.index = node->index;
-			data->nearest.dist_sq = calc_dist_sq_to_ray(data, node);
-			/* TODO: return a value to the data->nearest.co
-			 * not urgent however since users currently define own callbacks */
-		}
-	}
-	else {
-		int i;
-		/* First pick the closest node to dive on */
-		if (data->pick_smallest[node->main_axis]) {
-			for (i = 0; i != node->totnode; i++) {
-				if (calc_dist_sq_to_ray(data, node->children[i]) < data->nearest.dist_sq) {
-					dfs_find_nearest_to_ray_dfs(data, node->children[i]);
-				}
-			}
-		}
-		else {
-			for (i = node->totnode - 1; i >= 0; i--) {
-				if (calc_dist_sq_to_ray(data, node->children[i]) < data->nearest.dist_sq) {
-					dfs_find_nearest_to_ray_dfs(data, node->children[i]);
-				}
-			}
-		}
-	}
-}
-
-/**
- * Returns the point whose tangent defined by the angle between the point and ray is the lowest
- * nearest.dist_sq returns the angle's tangent
- */
-int BLI_bvhtree_find_nearest_to_ray_angle(
-        BVHTree *tree, const float co[3], const float dir[3],
-        const bool ray_is_normalized, const float scale[3],
-        BVHTreeNearest *nearest,
-        BVHTree_NearestToRayCallback callback, void *userdata)
-{
-	BVHNearestRayData data;
-	BVHNode *root = tree->nodes[tree->totleaf];
-
-	data.tree = tree;
-
-	data.callback = callback;
-	data.userdata = userdata;
-
-	dist_squared_ray_to_aabb_scaled_v3_precalc(&data, co, dir, ray_is_normalized, scale);
-
-	if (nearest) {
-		memcpy(&data.nearest, nearest, sizeof(*nearest));
-	}
-	else {
-		data.nearest.index = -1;
-		data.nearest.dist_sq = FLT_MAX;
-	}
-
-	/* dfs search */
-	if (root) {
-		if (calc_tangent_sq(&data, root) < data.nearest.dist_sq)
-			dfs_find_lowest_tangent_dfs(&data, root);
-	}
-
-	/* copy back results */
-	if (nearest) {
-		memcpy(nearest, &data.nearest, sizeof(*nearest));
-	}
-
-	return data.nearest.index;
-}
-
-/* return the nearest point to ray */
-int BLI_bvhtree_find_nearest_to_ray(
-        BVHTree *tree, const float co[3], const float dir[3],
-        const bool ray_is_normalized, const float scale[3],
-        BVHTreeNearest *nearest,
-        BVHTree_NearestToRayCallback callback, void *userdata)
-{
-	BVHNearestRayData data;
-	BVHNode *root = tree->nodes[tree->totleaf];
-
-	data.tree = tree;
-
-	data.callback = callback;
-	data.userdata = userdata;
-
-	dist_squared_ray_to_aabb_scaled_v3_precalc(&data, co, dir, ray_is_normalized, scale);
-
-	if (nearest) {
-		memcpy(&data.nearest, nearest, sizeof(*nearest));
-	}
-	else {
-		data.nearest.index = -1;
-		data.nearest.dist_sq = FLT_MAX;
-	}
-
-	/* dfs search */
-	if (root) {
-		if (calc_dist_sq_to_ray(&data, root) < data.nearest.dist_sq) {
-			dfs_find_nearest_to_ray_dfs(&data, root);
-		}
-	}
-
-	/* copy back results */
-	if (nearest) {
-		memcpy(nearest, &data.nearest, sizeof(*nearest));
-	}
-
-	return data.nearest.index;
-}
-
-/** \} */
-
-
-/* -------------------------------------------------------------------- */
-
 /** \name BLI_bvhtree_range_query
  *
  * Allocs and fills an array with the indexs of node that are on the given spherical range (center, radius).
diff --git a/source/blender/blenlib/intern/math_geom.c b/source/blender/blenlib/intern/math_geom.c
index aeb6a550cd9..3cf26ccf904 100644
--- a/source/blender/blenlib/intern/math_geom.c
+++ b/source/blender/blenlib/intern/math_geom.c
@@ -2337,224 +2337,6 @@ bool isect_ray_aabb_v3_simple(
 	}
 }
 
-void dist_squared_ray_to_aabb_v3_precalc(
-        struct NearestRayToAABB_Precalc *data,
-        const float ray_origin[3], const float ray_direction[3])
-{
-	float dir_sq[3];
-
-	for (int i = 0; i < 3; i++) {
-		data->ray_origin[i] = ray_origin[i];
-		data->ray_direction[i] = ray_direction[i];
-		data->ray_inv_dir[i] = (data->ray_direction[i] != 0.0f) ? (1.0f / data->ray_direction[i]) : FLT_MAX;
-		/* It has to be a function of `ray_inv_dir`,
-		 * since the division of 1 by 0.0f, can be -inf or +inf */
-		data->sign[i] = (data->ray_inv_dir[i] < 0.0f);
-
-		dir_sq[i] = SQUARE(data->ray_direction[i]);
-	}
-
-	/* `diag_sq` Length square of each face diagonal */
-	float diag_sq[3] = {
-		dir_sq[1] + dir_sq[2],
-		dir_sq[0] + dir_sq[2],
-		dir_sq[0] + dir_sq[1],
-	};
-	data->idiag_sq[0] = (diag_sq[0] > FLT_EPSILON) ? (1.0f / diag_sq[0]) : FLT_MAX;
-	data->idiag_sq[1] = (diag_sq[1] > FLT_EPSILON) ? (1.0f / diag_sq[1]) : FLT_MAX;
-	data->idiag_sq[2] = (diag_sq[2] > FLT_EPSILON) ? (1.0f / diag_sq[2]) : FLT_MAX;
-
-	data->cdot_axis[0] = data->ray_direction[0] * data->idiag_sq[0];
-	data->cdot_axis[1] = data->ray_direction[1] * data->idiag_sq[1];
-	data->cdot_axis[2] = data->ray_direction[2] * data->idiag_sq[2];
-}
-
-/**
- * Returns the squared distance from a ray to a bound-box `AABB`.
- * It is based on `fast_ray_nearest_hit` solution to obtain
- * the coordinates of the nearest edge of Bound Box to the ray
- */
-float dist_squared_ray_to_aabb_v3(
-        const struct NearestRayToAABB_Precalc *data,
-        const float bb_min[3], const float bb_max[3],
-        bool r_axis_closest[3])
-{
-	/* `tmin` is a vector that has the smaller distances to each of the
-	 * infinite planes of the `AABB` faces (hit in nearest face X plane,
-	 * nearest face Y plane and nearest face Z plane) */
-	float local_bvmin[3], local_bvmax[3];
-
-	if (data->sign[0] == 0) {
-		local_bvmin[0] = bb_min[0] - data->ray_origin[0];
-		local_bvmax[0] = bb_max[0] - data->ray_origin[0];
-	}
-	else {
-		local_bvmin[0] = bb_max[0] - data->ray_origin[0];
-		local_bvmax[0] = bb_min[0] - data->ray_origin[0];
-	}
-
-	if (data->sign[1] == 0) {
-		local_bvmin[1] = bb_min[1] - data->ray_origin[1];
-		local_bvmax[1] = bb_max[1] - data->ray_origin[1];
-	}
-	else {
-		local_bvmin[1] = bb_max[1] - data->ray_origin[1];
-		local_bvmax[1] = bb_min[1] - data->ray_origin[1];
-	}
-
-	if (data->sign[2] == 0) {
-		local_bvmin[2] = bb_min[2] - data->ray_origin[2];
-		local_bvmax[2] = bb_max[2] - data->ray_origin[2];
-	}
-	else {
-		local_bvmin[2] = bb_max[2] - data->ray_origin[2];
-		local_bvmax[2] = bb_min[2] - data->ray_origin[2];
-	}
-
-	const float tmin[3] = {
-		local_bvmin[0] * data->ray_inv_dir[0],
-		local_bvmin[1] * data->ray_inv_dir[1],
-		local_bvmin[2] * data->ray_inv_dir[2],
-	};
-
-	/* `tmax` is a vector that has the longer distances to each of the
-	 * infinite planes of the `AABB` faces (hit in farthest face X plane,
-	 * farthest face Y plane and farthest face Z plane) */
-	const float tmax[3] = {
-		local_bvmax[0] * data->ray_inv_dir[0],
-		local_bvmax[1] * data->ray_inv_dir[1],
-		local_bvmax[2] * data->ray_inv_dir[2],
-	};
-	/* `v1` and `v3` is be the coordinates of the nearest `AABB` edge to the ray*/
-	float v1[3], v2[3];
-	/* `rtmin` is the highest value of the smaller distances. == max_axis_v3(tmin)
-	 * `rtmax` is the lowest value of longer distances. == min_axis_v3(tmax)*/
-	float rtmin, rtmax, mul, rdist;
-	/* `main_axis` is the axis equivalent to edge close to the ray */
-	int main_axis;
-
-	r_axis_closest[0] = false;
-	r_axis_closest[1] = false;
-	r_axis_closest[2] = false;
-
-	/* *** min_axis_v3(tmax) *** */
-	if ((tmax[0] <= tmax[1]) && (tmax[0] <= tmax[2])) {
-		// printf("# Hit in X %s\n", data->sign[0] ? "min", "max");
-		rtmax = tmax[0];
-		v1[0] = v2[0] = local_bvmax[0];
-		mul = local_bvmax[0] * data->ray_direction[0];
-		main_axis = 3;
-		r_axis_closest[0] = data->sign[0];
-	}
-	else if ((tmax[1] <= tmax[0]) && (tmax[1] <= tmax[2])) {
-		// printf("# Hit in Y %s\n", data->sign[1] ? "min", "max");
-		rtmax = tmax[1];
-		v1[1] = v2[1] = local_bvmax[1];
-		mul = local_bvmax[1] * data->ray_direction[1];
-		main_axis = 2;
-		r_axis_closest[1] = data->sign[1];
-	}
-	else {
-		// printf("# Hit in Z %s\n", data->sign[2] ? "min", "max");
-		rtmax = tmax[2];
-		v1[2] = v2[2] = local_bvmax[2];
-		mul = local_bvmax[2] * data->ray_direction[2];
-		main_axis = 1;
-		r_axis_closest[2] = data->sign[2];
-	}
-
-	/* *** max_axis_v3(tmin) *** */
-	if ((tmin[0] >= tmin[1]) && (tmin[0] >= tmin[2])) {
-		// printf("# To X %s\n", data->sign[0] ? "max", "min");
-		rtmin = tmin[0];
-		v1[0] = v2[0] = local_bvmin[0];
-		mul += local_bvmin[0] * data->ray_direction[0];
-		main_axis -= 3;
-		r_axis_closest[0] = !data->sign[0];
-	}
-	else if ((tmin[1] >= tmin[0]) && (tmin[1] >= tmin[2])) {
-		// printf("# To Y %s\n", data->sign[1] ? "max", "min");
-		rtmin = tmin[1];
-		v1[1] = v2[1] = local_bvmin[1];
-		mul += local_bvmin[1] * data->ray_direction[1];
-		main_axis -= 1;
-		r_axis_closest[1] = !data->sign[1];
-	}
-	else {
-		// printf("# To Z %s\n", data->sign[2] ? "max", "min");
-		rtmin = tmin[2];
-		v1[2] = v2[2] = local_bvmin[2];
-		mul += local_bvmin[2] * data->ray_direction[2];
-		main_axis -= 2;
-		r_axis_closest[2] = !data->sign[2];
-	}
-	/* *** end min/max axis *** */
-
-
-	/* `if rtmax < 0`, the whole `AABB` is behing us */
-	if ((rtmax < 0.0f) && (rtmin < 0.0f)) {
-		return FLT_MAX;
-	}
-
-	if (main_axis < 0) {
-		main_axis += 3;
-	}
-
-	if (data->sign[main_axis] == 0) {
-		v1[main_axis] = local_bvmin[main_axis];
-		v2[main_axis] = local_bvmax[main_axis];
-	}
-	else {
-		v1[main_axis] = local_bvmax[main_axis];
-		v2[main_axis] = local_bvmin[main_axis];
-	}
-
-	/* if rtmin < rtmax, ray intersect `AABB` */
-	if (rtmin <= rtmax) {
-		const float proj = rtmin * data->ray_direction[main_axis];
-		rdist = 0.0f;
-		r_axis_closest[main_axis] = (proj - v1[main_axis]) < (v2[main_axis] - proj);
-	}
-	else {
-		/* `proj` equals to nearest point on the ray closest to the edge `v1 v2` of the `AABB`. */
-		const float proj = mul * data->cdot_axis[main_axis];
-		float depth;
-		if (v1[main_axis] > proj) {  /* the nearest point to the ray is the point v1 */
-			/* `depth` is equivalent the distance from the origin to the point v1,
-			 * Here's a faster way to calculate the dot product of v1 and ray
-			 * (depth = dot_v3v3(v1, data->ray.direction))*/
-			depth = mul + data->ray_direction[main_axis] * v1[main_axis];
-			rdist = len_squared_v3(v1) - SQUARE(depth);
-			r_axis_closest[main_axis] = true;
-		}
-		else if (v2[main_axis] < proj) {  /* the nearest point of the ray is the point v2 */
-			depth = mul + data->ray_direction[main_axis] * v2[main_axis];
-			rdist = len_squared_v3(v2) - SQUARE(depth);
-			r_axis_closest[main_axis] = false;
-		}
-		else {  /* the nearest point of the ray is on the edge of the `AABB`. */
-			float v[2];
-			mul *= data->idiag_sq[main_axis];
-			if (main_axis == 0) {
-				v[0] = (mul * data->ray_direction[1]) - v1[1];
-				v[1] = (mul * data->ray_direction[2]) - v1[2];
-			}
-			else if (main_axis == 1) {
-				v[0] = (mul * data->ray_direction[0]) - v1[0];
-				v[1] = (mul * data->ray_direction[2]) - v1[2];
-			}
-			else {
-				v[0] = (mul * data->ray_direction[0]) - v1[0];
-				v[1] = (mul * data->ray_direction[1]) - v1[1];
-			}
-			rdist = len_squared_v2(v);
-			r_axis_closest[main_axis] = (proj - v1[main_axis]) < (v2[main_axis] - proj);
-		}
-	}
-
-	return rdist;
-}
-
 /* find closest point to p on line through (l1, l2) and return lambda,
  * where (0 <= lambda <= 1) when cp is in the line segment (l1, l2)
  */
diff --git a/source/blender/blenlib/intern/polyfill2d.c b/source/blender/blenlib/intern/polyfill2d.c
index 8d9881e4539..2969b0eccf4 100644
--- a/source/blender/blenlib/intern/polyfill2d.c
+++ b/source/blender/blenlib/intern/polyfill2d.c
@@ -21,8 +21,15 @@
 /** \file blender/blenlib/intern/polyfill2d.c
  *  \ingroup bli
  *
- * A simple implementation of the ear cutting algorithm
- * to triangulate simple polygons without holes.
+ * An ear clipping algorithm to triangulate single boundary polygons.
+ *
+ * Details:
+ *
+ * - The algorithm guarantees all triangles are assigned (number of coords - 2)
+ *   and that triangles will have non-overlapping indices (even for degenerate geometry).
+ * - Self-intersections are considered degenerate (resulting triangles will overlap).
+ * - While multiple polygons aren't supported, holes can still be defined using *key-holes*
+ *   (where the polygon doubles back on its self with *exactly* matching coordinates).
  *
  * \note
  *
@@ -74,6 +81,12 @@ typedef signed char eSign;
 
 #ifdef USE_KDTREE
 /**
+ * Spatial optimization for point-in-triangle intersection checks.
+ * The simple version of this algorithm is ``O(n^2)`` complexity
+ * (every point needing to check the triangle defined by every other point),
+ * Using a binary-tree reduces the complexity to ``O(n log n)``
+ * plus some overhead of creating the tree.
+ *
  * This is a single purpose KDTree based on BLI_kdtree with some modifications
  * to better suit polyfill2d.
  *
diff --git a/source/blender/blenloader/intern/readfile.c b/source/blender/blenloader/intern/readfile.c
index 61a93018f71..8a18533f730 100644
--- a/source/blender/blenloader/intern/readfile.c
+++ b/source/blender/blenloader/intern/readfile.c
@@ -10520,6 +10520,9 @@ static void read_libraries(FileData *basefd, ListBase *mainlist)
 					else {
 						mainptr->curlib->filedata = NULL;
 						mainptr->curlib->id.tag |= LIB_TAG_MISSING;
+						/* Set lib version to current main one... Makes assert later happy. */
+						mainptr->versionfile = mainptr->curlib->versionfile = mainl->versionfile;
+						mainptr->subversionfile = mainptr->curlib->subversionfile = mainl->subversionfile;
 					}
 					
 					if (fd == NULL) {
diff --git a/source/blender/blenloader/intern/writefile.c b/source/blender/blenloader/intern/writefile.c
index 57846f6c030..fda5b9501ee 100644
--- a/source/blender/blenloader/intern/writefile.c
+++ b/source/blender/blenloader/intern/writefile.c
@@ -78,7 +78,7 @@
  * - write #TEST (#RenderInfo struct. 128x128 blend file preview is optional).
  * - write #GLOB (#FileGlobal struct) (some global vars).
  * - write #DNA1 (#SDNA struct)
- * - write #USER (#UserDef struct) if filename is ``~/X.XX/config/startup.blend``.
+ * - write #USER (#UserDef struct) if filename is ``~/.config/blender/X.XX/config/startup.blend``.
  */
 
 
diff --git a/source/blender/editors/animation/anim_draw.c b/source/blender/editors/animation/anim_draw.c
index 2b830657923..1d71ba8dc34 100644
--- a/source/blender/editors/animation/anim_draw.c
+++ b/source/blender/editors/animation/anim_draw.c
@@ -46,6 +46,8 @@
 #include "BLI_dlrbTree.h"
 
 #include "BKE_context.h"
+#include "BKE_curve.h"
+#include "BKE_fcurve.h"
 #include "BKE_global.h"
 #include "BKE_nla.h"
 #include "BKE_mask.h"
@@ -121,7 +123,8 @@ void ANIM_draw_cfra(const bContext *C, View2D *v2d, short flag)
 {
 	Scene *scene = CTX_data_scene(C);
 
-	const float x = (float)(scene->r.cfra * scene->r.framelen);
+	const float time = scene->r.cfra + scene->r.subframe;
+	const float x = (float)(time * scene->r.framelen);
 
 	glLineWidth((flag & DRAWCFRA_WIDE) ? 3.0 : 2.0);
 
@@ -330,7 +333,8 @@ static float normalization_factor_get(Scene *scene, FCurve *fcu, short flag, flo
 
 	fcu->prev_norm_factor = 1.0f;
 	if (fcu->bezt) {
-		BezTriple *bezt;
+		const bool use_preview_only = PRVRANGEON;
+		const BezTriple *bezt;
 		int i;
 		float max_coord = -FLT_MAX;
 		float min_coord = FLT_MAX;
@@ -340,28 +344,77 @@ static float normalization_factor_get(Scene *scene, FCurve *fcu, short flag, flo
 			return 1.0f;
 		}
 
-		if (PRVRANGEON) {
-			for (i = 0, bezt = fcu->bezt; i < fcu->totvert; i++, bezt++) {
-				if (IN_RANGE_INCL(bezt->vec[1][0], scene->r.psfra, scene->r.pefra)) {
-					max_coord = max_ff(max_coord, bezt->vec[0][1]);
-					max_coord = max_ff(max_coord, bezt->vec[1][1]);
-					max_coord = max_ff(max_coord, bezt->vec[2][1]);
-
-					min_coord = min_ff(min_coord, bezt->vec[0][1]);
-					min_coord = min_ff(min_coord, bezt->vec[1][1]);
-					min_coord = min_ff(min_coord, bezt->vec[2][1]);
-				}
+		for (i = 0, bezt = fcu->bezt; i < fcu->totvert; i++, bezt++) {
+			if (use_preview_only && !IN_RANGE_INCL(bezt->vec[1][0],
+			                                       scene->r.psfra,
+			                                       scene->r.pefra))
+			{
+				continue;
 			}
-		}
-		else {
-			for (i = 0, bezt = fcu->bezt; i < fcu->totvert; i++, bezt++) {
-				max_coord = max_ff(max_coord, bezt->vec[0][1]);
+
+			if (i == 0) {
+				/* We ignore extrapolation flags and handle here, and use the
+				 * control point position only. so we normalize "interesting"
+				 * part of the curve.
+				 *
+				 * Here we handle left extrapolation.
+				 */
 				max_coord = max_ff(max_coord, bezt->vec[1][1]);
-				max_coord = max_ff(max_coord, bezt->vec[2][1]);
 
-				min_coord = min_ff(min_coord, bezt->vec[0][1]);
 				min_coord = min_ff(min_coord, bezt->vec[1][1]);
-				min_coord = min_ff(min_coord, bezt->vec[2][1]);
+			}
+			else {
+				const BezTriple *prev_bezt = bezt - 1;
+				if (prev_bezt->ipo == BEZT_IPO_CONST) {
+					/* Constant interpolation: previous CV value is used up
+					 * to the current keyframe.
+					 */
+					max_coord = max_ff(max_coord, bezt->vec[1][1]);
+					min_coord = min_ff(min_coord, bezt->vec[1][1]);
+				}
+				else if (prev_bezt->ipo == BEZT_IPO_LIN) {
+					/* Linear interpolation: min/max using both previous and
+					 * and current CV.
+					 */
+					max_coord = max_ff(max_coord, bezt->vec[1][1]);
+					min_coord = min_ff(min_coord, bezt->vec[1][1]);
+					max_coord = max_ff(max_coord, prev_bezt->vec[1][1]);
+					min_coord = min_ff(min_coord, prev_bezt->vec[1][1]);
+				}
+				else if (prev_bezt->ipo == BEZT_IPO_BEZ) {
+					const int resol = fcu->driver
+					        ? 32
+					        : min_ii((int)(5.0f * len_v2v2(bezt->vec[1], prev_bezt->vec[1])), 32);
+					if (resol < 2) {
+						max_coord = max_ff(max_coord, prev_bezt->vec[1][1]);
+						min_coord = min_ff(min_coord, prev_bezt->vec[1][1]);
+					}
+					else {
+						float data[120];
+						float v1[2], v2[2], v3[2], v4[2];
+
+						v1[0] = prev_bezt->vec[1][0];
+						v1[1] = prev_bezt->vec[1][1];
+						v2[0] = prev_bezt->vec[2][0];
+						v2[1] = prev_bezt->vec[2][1];
+
+						v3[0] = bezt->vec[0][0];
+						v3[1] = bezt->vec[0][1];
+						v4[0] = bezt->vec[1][0];
+						v4[1] = bezt->vec[1][1];
+
+						correct_bezpart(v1, v2, v3, v4);
+
+						BKE_curve_forward_diff_bezier(v1[0], v2[0], v3[0], v4[0], data, resol, sizeof(float) * 3);
+						BKE_curve_forward_diff_bezier(v1[1], v2[1], v3[1], v4[1], data + 1, resol, sizeof(float) * 3);
+
+						for (int j = 0; j <= resol; ++j) {
+							const float *fp = &data[j * 3];
+							max_coord = max_ff(max_coord, fp[1]);
+							min_coord = min_ff(min_coord, fp[1]);
+						}
+					}
+				}
 			}
 		}
 
diff --git a/source/blender/editors/animation/anim_ops.c b/source/blender/editors/animation/anim_ops.c
index c0d6963acbb..bb73cbf03b4 100644
--- a/source/blender/editors/animation/anim_ops.c
+++ b/source/blender/editors/animation/anim_ops.c
@@ -95,7 +95,7 @@ static void change_frame_apply(bContext *C, wmOperator *op)
 {
 	Main *bmain = CTX_data_main(C);
 	Scene *scene = CTX_data_scene(C);
-	int frame = RNA_int_get(op->ptr, "frame");
+	float frame = RNA_float_get(op->ptr, "frame");
 	bool do_snap = RNA_boolean_get(op->ptr, "snap");
 
 	if (do_snap && CTX_wm_space_seq(C)) {
@@ -103,10 +103,15 @@ static void change_frame_apply(bContext *C, wmOperator *op)
 	}
 
 	/* set the new frame number */
-	CFRA = frame;
+	CFRA = (int)frame;
+	if (scene->r.flag & SCER_SHOW_SUBFRAME) {
+		SUBFRA = frame - (int)frame;
+	}
+	else {
+		SUBFRA = 0.0f;
+	}
 	FRAMENUMBER_MIN_CLAMP(CFRA);
-	SUBFRA = 0.0f;
-	
+
 	/* do updates */
 	BKE_sound_seek_scene(bmain, scene);
 	WM_event_add_notifier(C, NC_SCENE | ND_FRAME, scene);
@@ -125,18 +130,18 @@ static int change_frame_exec(bContext *C, wmOperator *op)
 /* ---- */
 
 /* Get frame from mouse coordinates */
-static int frame_from_event(bContext *C, const wmEvent *event)
+static float frame_from_event(bContext *C, const wmEvent *event)
 {
 	ARegion *region = CTX_wm_region(C);
 	Scene *scene = CTX_data_scene(C);
 	float viewx;
-	int frame;
+	float frame;
 
 	/* convert from region coordinates to View2D 'tot' space */
 	viewx = UI_view2d_region_to_view_x(&region->v2d, event->mval[0]);
 	
 	/* round result to nearest int (frames are ints!) */
-	frame = iroundf(viewx);
+	frame = viewx;
 	
 	if (scene->r.flag & SCER_LOCK_FRAME_SELECTION) {
 		CLAMP(frame, PSFRA, PEFRA);
@@ -187,7 +192,7 @@ static int change_frame_invoke(bContext *C, wmOperator *op, const wmEvent *event
 	 * as user could click on a single frame (jump to frame) as well as
 	 * click-dragging over a range (modal scrubbing).
 	 */
-	RNA_int_set(op->ptr, "frame", frame_from_event(C, event));
+	RNA_float_set(op->ptr, "frame", frame_from_event(C, event));
 
 	change_frame_seq_preview_begin(C, event);
 
@@ -215,7 +220,7 @@ static int change_frame_modal(bContext *C, wmOperator *op, const wmEvent *event)
 			break;
 
 		case MOUSEMOVE:
-			RNA_int_set(op->ptr, "frame", frame_from_event(C, event));
+			RNA_float_set(op->ptr, "frame", frame_from_event(C, event));
 			change_frame_apply(C, op);
 			break;
 		
@@ -268,7 +273,7 @@ static void ANIM_OT_change_frame(wmOperatorType *ot)
 	ot->undo_group = "FRAME_CHANGE";
 
 	/* rna */
-	ot->prop = RNA_def_int(ot->srna, "frame", 0, MINAFRAME, MAXFRAME, "Frame", "", MINAFRAME, MAXFRAME);
+	ot->prop = RNA_def_float(ot->srna, "frame", 0, MINAFRAME, MAXFRAME, "Frame", "", MINAFRAME, MAXFRAME);
 	prop = RNA_def_boolean(ot->srna, "snap", false, "Snap", "");
 	RNA_def_property_flag(prop, PROP_SKIP_SAVE);
 }
diff --git a/source/blender/editors/curve/editcurve.c b/source/blender/editors/curve/editcurve.c
index 6b5a5d81d32..276cfb373c2 100644
--- a/source/blender/editors/curve/editcurve.c
+++ b/source/blender/editors/curve/editcurve.c
@@ -1196,6 +1196,11 @@ void ED_curve_editnurb_load(Object *obedit)
 
 		remap_hooks_and_vertex_parents(obedit);
 
+		/* We have to apply shapekeys *before* copying nurbs into newnurb, otherwise the reset to
+		 * refkey/original curve data that has to be done when editing non-refkey shapekey would be useless,
+		 * only affecting editnurb and not ob->data. */
+		calc_shapeKeys(obedit);
+
 		for (nu = editnurb->first; nu; nu = nu->next) {
 			newnu = BKE_nurb_duplicate(nu);
 			BLI_addtail(&newnurb, newnu);
@@ -1207,7 +1212,6 @@ void ED_curve_editnurb_load(Object *obedit)
 
 		cu->nurb = newnurb;
 
-		calc_shapeKeys(obedit);
 		ED_curve_updateAnimPaths(obedit->data);
 
 		BKE_nurbList_free(&oldnurb);
@@ -1228,7 +1232,6 @@ void ED_curve_editnurb_make(Object *obedit)
 		if (actkey) {
 			// XXX strcpy(G.editModeTitleExtra, "(Key) ");
 			undo_editmode_clear();
-			BKE_keyblock_convert_to_curve(actkey, cu, &cu->nurb);
 		}
 
 		if (editnurb) {
@@ -1249,12 +1252,16 @@ void ED_curve_editnurb_make(Object *obedit)
 			nu = nu->next;
 		}
 
-		if (actkey)
-			editnurb->shapenr = obedit->shapenr;
-
 		/* animation could be added in editmode even if there was no animdata in
 		 * object mode hence we always need CVs index be created */
 		init_editNurb_keyIndex(editnurb, &cu->nurb);
+
+		if (actkey) {
+			editnurb->shapenr = obedit->shapenr;
+			/* Apply shapekey to new nurbs of editnurb, not those of original curve (and *after* we generated keyIndex),
+			 * else we do not have valid 'original' data to properly restore curve when leaving editmode. */
+			BKE_keyblock_convert_to_curve(actkey, cu, &editnurb->nurbs);
+		}
 	}
 }
 
diff --git a/source/blender/editors/gpencil/gpencil_edit.c b/source/blender/editors/gpencil/gpencil_edit.c
index 5b011b679a6..e118e490f25 100644
--- a/source/blender/editors/gpencil/gpencil_edit.c
+++ b/source/blender/editors/gpencil/gpencil_edit.c
@@ -953,6 +953,9 @@ static int gp_dissolve_selected_points(bContext *C)
 			/* skip strokes that are invalid for current view */
 			if (ED_gpencil_stroke_can_use(C, gps) == false)
 				continue;
+			/* check if the color is editable */
+			if (ED_gpencil_stroke_color_use(gpl, gps) == false)
+				continue;
 			
 			if (gps->flag & GP_STROKE_SELECT) {
 				bGPDspoint *pt;
@@ -1165,6 +1168,9 @@ static int gp_delete_selected_points(bContext *C)
 			/* skip strokes that are invalid for current view */
 			if (ED_gpencil_stroke_can_use(C, gps) == false)
 				continue;
+			/* check if the color is editable */
+			if (ED_gpencil_stroke_color_use(gpl, gps) == false)
+				continue;
 			
 			
 			if (gps->flag & GP_STROKE_SELECT) {
@@ -1204,7 +1210,7 @@ static int gp_delete_exec(bContext *C, wmOperator *op)
 		case GP_DELETEOP_POINTS:	/* selected points (breaks the stroke into segments) */
 			result = gp_delete_selected_points(C);
 			break;
-
+		
 		case GP_DELETEOP_FRAME:		/* active frame */
 			result = gp_actframe_delete_exec(C, op);
 			break;
diff --git a/source/blender/editors/include/UI_icons.h b/source/blender/editors/include/UI_icons.h
index 8579778ff79..8420591aa3e 100644
--- a/source/blender/editors/include/UI_icons.h
+++ b/source/blender/editors/include/UI_icons.h
@@ -653,9 +653,9 @@ DEF_ICON(IPO_BACK)
 DEF_ICON(IPO_EASE_IN)
 DEF_ICON(IPO_EASE_OUT)
 DEF_ICON(IPO_EASE_IN_OUT)
+DEF_ICON(NORMALIZE_FCURVES)
 #ifndef DEF_ICON_BLANK_SKIP
 	/* available */
-	DEF_ICON(BLANK203)
 	DEF_ICON(BLANK204)
 	DEF_ICON(BLANK205)
 	DEF_ICON(BLANK206)
diff --git a/source/blender/editors/interface/interface_widgets.c b/source/blender/editors/interface/interface_widgets.c
index e3bad758dd7..ab71ff53731 100644
--- a/source/blender/editors/interface/interface_widgets.c
+++ b/source/blender/editors/interface/interface_widgets.c
@@ -1596,7 +1596,7 @@ static void widget_draw_text_icon(uiFontStyle *fstyle, uiWidgetColors *wcol, uiB
 		temp.xmin = temp.xmax - (BLI_rcti_size_y(rect) * 1.08f);
 
 		if (extra_icon_type == UI_BUT_ICONEXTRA_CLEAR) {
-			widget_draw_icon(but, ICON_X, alpha, &temp, false);
+			widget_draw_icon(but, ICON_PANEL_CLOSE, alpha, &temp, false);
 		}
 		else if (extra_icon_type == UI_BUT_ICONEXTRA_EYEDROPPER) {
 			widget_draw_icon(but, ICON_EYEDROPPER, alpha, &temp, false);
diff --git a/source/blender/editors/render/render_opengl.c b/source/blender/editors/render/render_opengl.c
index 9097432a251..55c07140216 100644
--- a/source/blender/editors/render/render_opengl.c
+++ b/source/blender/editors/render/render_opengl.c
@@ -887,13 +887,13 @@ static void write_result_func(TaskPool * __restrict pool,
 	ReportList reports;
 	BKE_reports_init(&reports, oglrender->reports->flag & ~RPT_PRINT);
 	/* Do actual save logic here, depending on the file format. */
+	Scene tmp_scene = *scene;
+	tmp_scene.r.cfra = cfra;
 	if (is_movie) {
 		/* We have to construct temporary scene with proper scene->r.cfra.
 		 * This is because underlying calls do not use r.cfra but use scene
 		 * for that.
 		 */
-		Scene tmp_scene = *scene;
-		tmp_scene.r.cfra = cfra;
 		ok = RE_WriteRenderViewsMovie(&reports,
 		                              rr,
 		                              &tmp_scene,
@@ -917,8 +917,8 @@ static void write_result_func(TaskPool * __restrict pool,
 		                             true,
 		                             NULL);
 
-		BKE_render_result_stamp_info(scene, scene->camera, rr, false);
-		ok = RE_WriteRenderViewsImage(NULL, rr, scene, true, name);
+		BKE_render_result_stamp_info(&tmp_scene, tmp_scene.camera, rr, false);
+		ok = RE_WriteRenderViewsImage(NULL, rr, &tmp_scene, true, name);
 		if (!ok) {
 			BKE_reportf(&reports,
 			            RPT_ERROR,
diff --git a/source/blender/editors/space_view3d/space_view3d.c b/source/blender/editors/space_view3d/space_view3d.c
index a48208b55f7..5605a885584 100644
--- a/source/blender/editors/space_view3d/space_view3d.c
+++ b/source/blender/editors/space_view3d/space_view3d.c
@@ -1258,21 +1258,6 @@ static void space_view3d_listener(bScreen *UNUSED(sc), ScrArea *sa, struct wmNot
 			}
 			break;
 	}
-
-	/* removed since BKE_image_user_frame_calc is now called in view3d_draw_bgpic because screen_ops doesnt call the notifier. */
-#if 0
-	if (wmn->category == NC_SCENE && wmn->data == ND_FRAME) {
-		View3D *v3d = area->spacedata.first;
-		BGpic *bgpic = v3d->bgpicbase.first;
-
-		for (; bgpic; bgpic = bgpic->next) {
-			if (bgpic->ima) {
-				Scene *scene = wmn->reference;
-				BKE_image_user_frame_calc(&bgpic->iuser, scene->r.cfra, 0);
-			}
-		}
-	}
-#endif
 }
 
 const char *view3d_context_dir[] = {
diff --git a/source/blender/editors/transform/transform_snap_object.c b/source/blender/editors/transform/transform_snap_object.c
index 2e95b67d508..03e0b544657 100644
--- a/source/blender/editors/transform/transform_snap_object.c
+++ b/source/blender/editors/transform/transform_snap_object.c
@@ -1051,7 +1051,6 @@ static int dm_looptri_to_poly_index(DerivedMesh *dm, const MLoopTri *lt)
 static bool snapDerivedMesh(
         SnapObjectContext *sctx, SnapData *snapdata,
         Object *ob, DerivedMesh *dm, float obmat[4][4], const unsigned int ob_index,
-        bool do_bb,
         /* read/write args */
         float *ray_depth, float *dist_px,
         /* return args */
@@ -1112,39 +1111,31 @@ static bool snapDerivedMesh(
 	copy_v3_v3(ray_org_local, snapdata->ray_origin);
 	mul_m4_v3(imat, ray_org_local);
 
-	if (do_bb) {
-		BoundBox *bb = BKE_object_boundbox_get(ob);
-
-		if (bb) {
-			BoundBox bb_temp;
-
-			/* We cannot afford a bounding box with some null dimension, which may happen in some cases...
-			 * Threshold is rather high, but seems to be needed to get good behavior, see T46099. */
-			bb = BKE_boundbox_ensure_minimum_dimensions(bb, &bb_temp, 1e-1f);
-
-			/* In vertex and edges you need to get the pixel distance from ray to BoundBox, see T46816. */
-			if (ELEM(snapdata->snap_to, SCE_SNAP_MODE_VERTEX, SCE_SNAP_MODE_EDGE)) {
-				float dist_px_sq = dist_squared_to_projected_aabb_simple(
-					    lpmat, snapdata->win_half, ray_min_dist, snapdata->mval,
-					    ray_org_local, ray_normal_local, bb->vec[0], bb->vec[6]);
-				if (dist_px_sq > SQUARE(*dist_px))
-				{
-					return retval;
-				}
+	/* Test BoundBox */
+	BoundBox *bb = BKE_object_boundbox_get(ob);
+	if (bb) {
+		/* In vertex and edges you need to get the pixel distance from ray to BoundBox, see: T46099, T46816 */
+		if (ELEM(snapdata->snap_to, SCE_SNAP_MODE_VERTEX, SCE_SNAP_MODE_EDGE)) {
+			float dist_px_sq = dist_squared_to_projected_aabb_simple(
+				    lpmat, snapdata->win_half, ray_min_dist, snapdata->mval,
+				    ray_org_local, ray_normal_local, bb->vec[0], bb->vec[6]);
+			if (dist_px_sq > SQUARE(*dist_px))
+			{
+				return retval;
 			}
-			else {
-				/* was BKE_boundbox_ray_hit_check, see: cf6ca226fa58 */
-				if (!isect_ray_aabb_v3_simple(
-					ray_start_local, ray_normal_local, bb->vec[0], bb->vec[6], NULL, NULL))
-				{
-					return retval;
-				}
+		}
+		else {
+			/* was BKE_boundbox_ray_hit_check, see: cf6ca226fa58 */
+			if (!isect_ray_aabb_v3_simple(
+				ray_start_local, ray_normal_local, bb->vec[0], bb->vec[6], NULL, NULL))
+			{
+				return retval;
 			}
-			/* was local_depth, see: T47838 */
-			len_diff = dist_aabb_to_plane(bb->vec[0], bb->vec[6], ray_start_local, ray_normal_local);
-			if (len_diff < 0) len_diff = 0.0f;
-			need_ray_start_correction_init = false;
 		}
+		/* was local_depth, see: T47838 */
+		len_diff = dist_aabb_to_plane(bb->vec[0], bb->vec[6], ray_start_local, ray_normal_local);
+		if (len_diff < 0) len_diff = 0.0f;
+		need_ray_start_correction_init = false;
 	}
 
 	SnapObjectData_Mesh *sod = NULL;
@@ -1295,10 +1286,17 @@ static bool snapDerivedMesh(
 	}
 	/* SCE_SNAP_MODE_VERTEX or SCE_SNAP_MODE_EDGE */
 	else {
+
+		/* Warning: the depth_max is currently being used only in perspective view.
+		 * It is not correct to limit the maximum depth for elements obtained with nearest
+		 * since this limitation depends on the normal and the size of the occlusion face.
+		 * And more... ray_depth is being confused with Z-depth here... (varies only the precision) */
+		const float ray_depth_max_global = *ray_depth + snapdata->depth_range[0];
+
 		Nearest2dUserData neasrest2d = {
 		    .dist_px_sq = SQUARE(*dist_px),
 		    .r_axis_closest = {1.0f, 1.0f, 1.0f},
-		    .depth_range = {snapdata->depth_range[0], *ray_depth + snapdata->depth_range[0]},
+		    .depth_range = {snapdata->depth_range[0], ray_depth_max_global},
 		    .userdata = treedata,
 		    .get_edge_verts = (Nearest2DGetEdgeVertsCallback)get_dm_edge_verts,
 		    .copy_vert_no = (Nearest2DCopyVertNoCallback)copy_dm_vert_no,
@@ -1650,7 +1648,6 @@ static bool snapObject(
 			}
 			retval = snapDerivedMesh(
 			        sctx, snapdata, ob, dm, obmat, ob_index,
-			        true,
 			        ray_depth, dist_px,
 			        r_loc, r_no,
 			        r_index, r_hit_list);
diff --git a/source/blender/makesdna/DNA_scene_types.h b/source/blender/makesdna/DNA_scene_types.h
index 085bde11619..03958b004ed 100644
--- a/source/blender/makesdna/DNA_scene_types.h
+++ b/source/blender/makesdna/DNA_scene_types.h
@@ -1765,6 +1765,7 @@ typedef struct Scene {
 #define SCER_LOCK_FRAME_SELECTION	(1<<1)
 	/* timeline/keyframe jumping - only selected items (on by default) */
 #define SCE_KEYS_NO_SELONLY	(1<<2)
+#define SCER_SHOW_SUBFRAME	(1<<3)
 
 /* mode (int now) */
 #define R_OSA			0x0001
diff --git a/source/blender/makesrna/intern/rna_scene.c b/source/blender/makesrna/intern/rna_scene.c
index ba542a457b2..211063fa03b 100644
--- a/source/blender/makesrna/intern/rna_scene.c
+++ b/source/blender/makesrna/intern/rna_scene.c
@@ -807,6 +807,21 @@ static void rna_Scene_frame_current_set(PointerRNA *ptr, int value)
 	data->r.cfra = value;
 }
 
+static float rna_Scene_frame_float_get(PointerRNA *ptr)
+{
+	Scene *data = (Scene *)ptr->data;
+	return (float)data->r.cfra + data->r.subframe;
+}
+
+static void rna_Scene_frame_float_set(PointerRNA *ptr, float value)
+{
+	Scene *data = (Scene *)ptr->data;
+	/* if negative frames aren't allowed, then we can't use them */
+	FRAMENUMBER_MIN_CLAMP(value);
+	data->r.cfra = (int)value;
+	data->r.subframe = value - data->r.cfra;
+}
+
 static float rna_Scene_frame_current_final_get(PointerRNA *ptr)
 {
 	Scene *scene = (Scene *)ptr->data;
@@ -887,6 +902,12 @@ static void rna_Scene_preview_range_end_frame_set(PointerRNA *ptr, int value)
 	data->r.pefra = value;
 }
 
+static void rna_Scene_show_subframe_update(Main *UNUSED(bmain), Scene *UNUSED(current_scene), PointerRNA *ptr)
+{
+	Scene *scene = (Scene *)ptr->id.data;
+	scene->r.subframe = 0.0f;
+}
+
 static void rna_Scene_frame_update(Main *bmain, Scene *UNUSED(current_scene), PointerRNA *ptr)
 {
 	Scene *scene = (Scene *)ptr->id.data;
@@ -8470,8 +8491,19 @@ void RNA_def_scene(BlenderRNA *brna)
 	prop = RNA_def_property(srna, "frame_subframe", PROP_FLOAT, PROP_TIME);
 	RNA_def_property_float_sdna(prop, NULL, "r.subframe");
 	RNA_def_property_ui_text(prop, "Current Sub-Frame", "");
-	RNA_def_property_clear_flag(prop, PROP_ANIMATABLE | PROP_EDITABLE);
-	
+	RNA_def_property_clear_flag(prop, PROP_ANIMATABLE);
+	RNA_def_property_range(prop, 0.0f, 1.0f);
+	RNA_def_property_ui_range(prop, 0.0f, 1.0f, 0.01, 2);
+	RNA_def_property_update(prop, NC_SCENE | ND_FRAME, "rna_Scene_frame_update");
+
+	prop = RNA_def_property(srna, "frame_float", PROP_FLOAT, PROP_TIME);
+	RNA_def_property_ui_text(prop, "Current Sub-Frame", "");
+	RNA_def_property_clear_flag(prop, PROP_ANIMATABLE);
+	RNA_def_property_range(prop, MINAFRAME, MAXFRAME);
+	RNA_def_property_ui_range(prop, MINAFRAME, MAXFRAME, 0.1, 2);
+	RNA_def_property_float_funcs(prop, "rna_Scene_frame_float_get", "rna_Scene_frame_float_set", NULL);
+	RNA_def_property_update(prop, NC_SCENE | ND_FRAME, "rna_Scene_frame_update");
+
 	prop = RNA_def_property(srna, "frame_start", PROP_INT, PROP_TIME);
 	RNA_def_property_clear_flag(prop, PROP_ANIMATABLE);
 	RNA_def_property_int_sdna(prop, NULL, "r.sfra");
@@ -8536,7 +8568,15 @@ void RNA_def_scene(BlenderRNA *brna)
 	RNA_def_property_int_funcs(prop, NULL, "rna_Scene_preview_range_end_frame_set", NULL);
 	RNA_def_property_ui_text(prop, "Preview Range End Frame", "Alternative end frame for UI playback");
 	RNA_def_property_update(prop, NC_SCENE | ND_FRAME, NULL);
-	
+
+	/* Subframe for moblur debug. */
+	prop = RNA_def_property(srna, "show_subframe", PROP_BOOLEAN, PROP_NONE);
+	RNA_def_property_clear_flag(prop, PROP_ANIMATABLE);
+	RNA_def_property_boolean_sdna(prop, NULL, "r.flag", SCER_SHOW_SUBFRAME);
+	RNA_def_property_ui_text(prop, "Show Subframe",
+	                         "Show current scene subframe and allow set it using interface tools");
+	RNA_def_property_update(prop, NC_SCENE | ND_FRAME, "rna_Scene_show_subframe_update");
+
 	/* Timeline / Time Navigation settings */
 	prop = RNA_def_property(srna, "show_keys_from_selected_only", PROP_BOOLEAN, PROP_NONE);
 	RNA_def_property_boolean_negative_sdna(prop, NULL, "flag", SCE_KEYS_NO_SELONLY);
diff --git a/tests/python/CMakeLists.txt b/tests/python/CMakeLists.txt
index 3e383614dc0..faa31669f81 100644
--- a/tests/python/CMakeLists.txt
+++ b/tests/python/CMakeLists.txt
@@ -95,6 +95,11 @@ add_test(bevel ${TEST_BLENDER_EXE}
 	--python-text run_tests
 )
 
+add_test(split_faces ${TEST_BLENDER_EXE}
+    ${TEST_SRC_DIR}/modeling/split_faces_test.blend
+    --python-text run_tests
+)
+
 # ------------------------------------------------------------------------------
 # IO TESTS
 
@@ -417,7 +422,12 @@ if(WITH_CYCLES)
 					-idiff "${OPENIMAGEIO_IDIFF}"
 				)
 			endif()
-		endmacro()	
+		endmacro()
+		if(WITH_OPENGL_TESTS)
+			add_cycles_render_test(opengl)
+		endif()
+		add_cycles_render_test(image)
+		add_cycles_render_test(mblur)
 		add_cycles_render_test(reports)
 		add_cycles_render_test(render)
 		add_cycles_render_test(shader)
diff --git a/tests/python/cycles_render_tests.py b/tests/python/cycles_render_tests.py
index 64a71da301a..a030cc5e0de 100755
--- a/tests/python/cycles_render_tests.py
+++ b/tests/python/cycles_render_tests.py
@@ -47,20 +47,41 @@ def printMessage(type, status, message):
 
 
 def render_file(filepath):
-    command = (
-        BLENDER,
-        "--background",
-        "-noaudio",
-        "--factory-startup",
-        "--enable-autoexec",
-        filepath,
-        "-E", "CYCLES",
-        # Run with OSL enabled
-        # "--python-expr", "import bpy; bpy.context.scene.cycles.shading_system = True",
-        "-o", TEMP_FILE_MASK,
-        "-F", "PNG",
-        "-f", "1",
+    dirname = os.path.dirname(filepath)
+    basedir = os.path.dirname(dirname)
+    subject = os.path.basename(dirname)
+    if subject == 'opengl':
+        command = (
+            BLENDER,
+            "--window-geometry", "0", "0", "1", "1",
+            "-noaudio",
+            "--factory-startup",
+            "--enable-autoexec",
+            filepath,
+            "-E", "CYCLES",
+            # Run with OSL enabled
+            # "--python-expr", "import bpy; bpy.context.scene.cycles.shading_system = True",
+            "-o", TEMP_FILE_MASK,
+            "-F", "PNG",
+            '--python', os.path.join(basedir,
+                                     "util",
+                                     "render_opengl.py")
         )
+    else:
+        command = (
+            BLENDER,
+            "--background",
+            "-noaudio",
+            "--factory-startup",
+            "--enable-autoexec",
+            filepath,
+            "-E", "CYCLES",
+            # Run with OSL enabled
+            # "--python-expr", "import bpy; bpy.context.scene.cycles.shading_system = True",
+            "-o", TEMP_FILE_MASK,
+            "-F", "PNG",
+            "-f", "1",
+            )
     try:
         output = subprocess.check_output(command)
         if VERBOSE: