Fix T41193: 2.71 use 100% of CPU at sculpt

Issue wascaused by the famous OpenMP crap in MSVC2013,
so only way is to use openmp threading if number of BVH
nodes is high enough.

Made it 8 for now, which seems to work rather fine on my
laptop and adult dragon from sintel. But maybe it's to be
adjusted a bit more.

5 files changed, 46 insertions, 32 deletions

diff --git a/source/blender/blenkernel/intern/pbvh.c b/source/blender/blenkernel/intern/pbvh.c
index e4e6105fe8c..7bdce4e19d5 100644
--- a/source/blender/blenkernel/intern/pbvh.c
+++ b/source/blender/blenkernel/intern/pbvh.c
@@ -50,6 +50,13 @@
 
 #define STACK_FIXED_DEPTH   100
 
+/* Setting zero so we can catch bugs in OpenMP/PBVH. */
+#ifdef DEBUG
+#  define PBVH_OMP_LIMIT 0
+#else
+#  define PBVH_OMP_LIMIT 8
+#endif
+
 typedef struct PBVHStack {
 	PBVHNode *node;
 	int revisiting;
@@ -965,7 +972,7 @@ static void pbvh_update_normals(PBVH *bvh, PBVHNode **nodes,
 	 *   can only update vertices marked with ME_VERT_PBVH_UPDATE.
 	 */
 
-#pragma omp parallel for private(n) schedule(static)
+#pragma omp parallel for private(n) schedule(static) if (totnode > PBVH_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHNode *node = nodes[n];
 
@@ -1009,7 +1016,7 @@ static void pbvh_update_normals(PBVH *bvh, PBVHNode **nodes,
 		}
 	}
 
-#pragma omp parallel for private(n) schedule(static)
+#pragma omp parallel for private(n) schedule(static) if (totnode > PBVH_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHNode *node = nodes[n];
 
@@ -1046,7 +1053,7 @@ void pbvh_update_BB_redraw(PBVH *bvh, PBVHNode **nodes, int totnode, int flag)
 	int n;
 
 	/* update BB, redraw flag */
-#pragma omp parallel for private(n) schedule(static)
+#pragma omp parallel for private(n) schedule(static) if (totnode > PBVH_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHNode *node = nodes[n];
 
diff --git a/source/blender/editors/sculpt_paint/paint_mask.c b/source/blender/editors/sculpt_paint/paint_mask.c
index b3f81f379f3..8faa4cfaf33 100644
--- a/source/blender/editors/sculpt_paint/paint_mask.c
+++ b/source/blender/editors/sculpt_paint/paint_mask.c
@@ -115,7 +115,7 @@ static int mask_flood_fill_exec(bContext *C, wmOperator *op)
 
 	sculpt_undo_push_begin("Mask flood fill");
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (i = 0; i < totnode; i++) {
 		PBVHVertexIter vi;
 
@@ -236,7 +236,7 @@ int ED_sculpt_mask_box_select(struct bContext *C, ViewContext *vc, const rcti *r
 
 			BKE_pbvh_search_gather(pbvh, BKE_pbvh_node_planes_contain_AABB, clip_planes_final, &nodes, &totnode);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 			for (i = 0; i < totnode; i++) {
 				PBVHVertexIter vi;
 				bool any_masked = false;
@@ -385,7 +385,7 @@ static int paint_mask_gesture_lasso_exec(bContext *C, wmOperator *op)
 				/* gather nodes inside lasso's enclosing rectangle (should greatly help with bigger meshes) */
 				BKE_pbvh_search_gather(pbvh, BKE_pbvh_node_planes_contain_AABB, clip_planes_final, &nodes, &totnode);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 				for (i = 0; i < totnode; i++) {
 					PBVHVertexIter vi;
 					bool any_masked = false;
diff --git a/source/blender/editors/sculpt_paint/sculpt.c b/source/blender/editors/sculpt_paint/sculpt.c
index aa43b7dcde4..261da79e35c 100644
--- a/source/blender/editors/sculpt_paint/sculpt.c
+++ b/source/blender/editors/sculpt_paint/sculpt.c
@@ -396,7 +396,7 @@ static void paint_mesh_restore_co(Sculpt *sd, Object *ob)
 	 * entries might be inserted by sculpt_undo_push_node() into the
 	 * GHash used internally by BM_log_original_vert_co() by a
 	 * different thread. [#33787] */
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP && !ss->bm)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && !ss->bm && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		SculptUndoNode *unode;
 		SculptUndoType type = (brush->sculpt_tool == SCULPT_TOOL_MASK ?
@@ -996,7 +996,7 @@ static void calc_area_normal(Sculpt *sd, Object *ob, float an[3], PBVHNode **nod
 	
 	zero_v3(an);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -1602,7 +1602,7 @@ static void smooth(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode,
 	for (iteration = 0; iteration <= count; ++iteration) {
 		float strength = (iteration != count) ? 1.0f : last;
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 		for (n = 0; n < totnode; n++) {
 			switch (type) {
 				case PBVH_GRIDS:
@@ -1638,7 +1638,7 @@ static void do_mask_brush_draw(Sculpt *sd, Object *ob, PBVHNode **nodes, int tot
 	int n;
 
 	/* threaded loop over nodes */
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -1691,7 +1691,7 @@ static void do_draw_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode)
 	mul_v3_fl(offset, bstrength);
 
 	/* threaded loop over nodes */
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -1746,7 +1746,7 @@ static void do_crease_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnod
 	if (brush->sculpt_tool == SCULPT_TOOL_BLOB) flippedbstrength *= -1.0f;
 
 	/* threaded loop over nodes */
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -1789,7 +1789,7 @@ static void do_pinch_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode
 	float bstrength = ss->cache->bstrength;
 	int n;
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -1836,7 +1836,7 @@ static void do_grab_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode)
 		add_v3_v3(grab_delta, ss->cache->sculpt_normal_symm);
 	}
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -1884,7 +1884,7 @@ static void do_nudge_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode
 	cross_v3_v3v3(tmp, ss->cache->sculpt_normal_symm, grab_delta);
 	cross_v3_v3v3(cono, tmp, ss->cache->sculpt_normal_symm);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -1932,7 +1932,7 @@ static void do_snake_hook_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int to
 		add_v3_v3(grab_delta, ss->cache->sculpt_normal_symm);
 	}
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -1972,7 +1972,7 @@ static void do_thumb_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode
 	cross_v3_v3v3(tmp, ss->cache->sculpt_normal_symm, grab_delta);
 	cross_v3_v3v3(cono, tmp, ss->cache->sculpt_normal_symm);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2015,7 +2015,7 @@ static void do_rotate_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnod
 	static const int flip[8] = { 1, -1, -1, 1, -1, 1, 1, -1 };
 	float angle = ss->cache->vertex_rotation * flip[ss->cache->mirror_symmetry_pass];
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2068,7 +2068,7 @@ static void do_layer_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode
 
 	mul_v3_v3v3(offset, ss->cache->scale, ss->cache->sculpt_normal_symm);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2131,7 +2131,7 @@ static void do_inflate_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totno
 	float bstrength = ss->cache->bstrength;
 	int n;
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2176,7 +2176,7 @@ static void calc_flatten_center(Sculpt *sd, Object *ob, PBVHNode **nodes, int to
 
 	zero_v3(fc);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2283,7 +2283,7 @@ static void calc_area_normal_and_flatten_center(Sculpt *sd, Object *ob,
 	/* for flatten center */
 	zero_v3(fc);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2535,7 +2535,7 @@ static void do_flatten_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totno
 	mul_v3_fl(temp, displace);
 	add_v3_v3(fc, temp);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2607,7 +2607,7 @@ static void do_clay_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode)
 
 	/* add_v3_v3v3(p, ss->cache->location, an); */
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2708,7 +2708,7 @@ static void do_clay_strips_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int t
 	mul_m4_m4m4(tmat, mat, scale);
 	invert_m4_m4(mat, tmat);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2772,7 +2772,7 @@ static void do_fill_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode)
 	mul_v3_fl(temp, displace);
 	add_v3_v3(fc, temp);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2836,7 +2836,7 @@ static void do_scrape_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnod
 	mul_v3_fl(temp, displace);
 	add_v3_v3(fc, temp);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -2890,7 +2890,7 @@ static void do_gravity(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode, fl
 	mul_v3_fl(offset, bstrength);
 
 	/* threaded loop over nodes */
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 	for (n = 0; n < totnode; n++) {
 		PBVHVertexIter vd;
 		SculptBrushTest test;
@@ -3059,7 +3059,7 @@ static void do_brush_action(Sculpt *sd, Object *ob, Brush *brush)
 	if (totnode) {
 		float location[3];
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 		for (n = 0; n < totnode; n++) {
 			sculpt_undo_push_node(ob, nodes[n],
 			                      brush->sculpt_tool == SCULPT_TOOL_MASK ?
@@ -3192,7 +3192,7 @@ static void sculpt_combine_proxies(Sculpt *sd, Object *ob)
 		const bool use_orco = ELEM(brush->sculpt_tool, SCULPT_TOOL_GRAB,
 		                           SCULPT_TOOL_ROTATE, SCULPT_TOOL_THUMB);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 		for (n = 0; n < totnode; n++) {
 			PBVHVertexIter vd;
 			PBVHProxyNode *proxies;
@@ -3285,7 +3285,7 @@ static void sculpt_flush_stroke_deform(Sculpt *sd, Object *ob)
 
 		BKE_pbvh_search_gather(ss->pbvh, NULL, NULL, &nodes, &totnode);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 		for (n = 0; n < totnode; n++) {
 			PBVHVertexIter vd;
 
diff --git a/source/blender/editors/sculpt_paint/sculpt_intern.h b/source/blender/editors/sculpt_paint/sculpt_intern.h
index cd79f525d82..248fc78715e 100644
--- a/source/blender/editors/sculpt_paint/sculpt_intern.h
+++ b/source/blender/editors/sculpt_paint/sculpt_intern.h
@@ -130,4 +130,11 @@ void sculpt_vertcos_to_key(Object *ob, KeyBlock *kb, float (*vertCos)[3]);
 
 void sculpt_update_object_bounding_box(struct Object *ob);
 
+/* Setting zero so we can catch bugs in OpenMP/sculpt. */
+#ifdef DEBUG
+#  define SCULPT_OMP_LIMIT 0
+#else
+#  define SCULPT_OMP_LIMIT 8
+#endif
+
 #endif
diff --git a/source/blender/editors/sculpt_paint/sculpt_undo.c b/source/blender/editors/sculpt_paint/sculpt_undo.c
index 614c1f3ef1d..91f80a4fc40 100644
--- a/source/blender/editors/sculpt_paint/sculpt_undo.c
+++ b/source/blender/editors/sculpt_paint/sculpt_undo.c
@@ -291,7 +291,7 @@ static void sculpt_undo_bmesh_restore_generic(bContext *C,
 
 		BKE_pbvh_search_gather(ss->pbvh, NULL, NULL, &nodes, &totnode);
 
-#pragma omp parallel for schedule(guided) if (sd->flags & SCULPT_USE_OPENMP)
+#pragma omp parallel for schedule(guided) if ((sd->flags & SCULPT_USE_OPENMP) && totnode > SCULPT_OMP_LIMIT)
 		for (i = 0; i < totnode; i++) {
 			BKE_pbvh_node_mark_redraw(nodes[i]);
 		}