Subsurf: Avoid global lock for loops and orig index layers

This is a bit annoying to have per-DM locking, but it's way better (as in, up to
4 times better) for playback speed when having lots of subsurf objects,

1 files changed, 15 insertions, 11 deletions

diff --git a/source/blender/blenkernel/intern/subsurf_ccg.c b/source/blender/blenkernel/intern/subsurf_ccg.c
index f8025f8df84..1386c23e234 100644
--- a/source/blender/blenkernel/intern/subsurf_ccg.c
+++ b/source/blender/blenkernel/intern/subsurf_ccg.c
@@ -90,9 +90,6 @@
 /* assumes MLoop's are layed out 4 for each poly, in order */
 #define USE_LOOP_LAYOUT_FAST
 
-static ThreadRWMutex loops_cache_rwlock = BLI_RWLOCK_INITIALIZER;
-static ThreadRWMutex origindex_cache_rwlock = BLI_RWLOCK_INITIALIZER;
-
 static CCGDerivedMesh *getCCGDerivedMesh(CCGSubSurf *ss,
                                          int drawInteriorEdges,
                                          int useSubsurfUv,
@@ -1492,7 +1489,7 @@ static void ccgDM_copyFinalLoopArray(DerivedMesh *dm, MLoop *mloop)
 	/* DMFlagMat *faceFlags = ccgdm->faceFlags; */ /* UNUSED */
 
 	if (!ccgdm->ehash) {
-		BLI_rw_mutex_lock(&loops_cache_rwlock, THREAD_LOCK_WRITE);
+		BLI_rw_mutex_lock(&ccgdm->loops_cache_rwlock, THREAD_LOCK_WRITE);
 		if (!ccgdm->ehash) {
 			MEdge *medge;
 
@@ -1503,10 +1500,10 @@ static void ccgDM_copyFinalLoopArray(DerivedMesh *dm, MLoop *mloop)
 				BLI_edgehash_insert(ccgdm->ehash, medge[i].v1, medge[i].v2, SET_INT_IN_POINTER(i));
 			}
 		}
-		BLI_rw_mutex_unlock(&loops_cache_rwlock);
+		BLI_rw_mutex_unlock(&ccgdm->loops_cache_rwlock);
 	}
 
-	BLI_rw_mutex_lock(&loops_cache_rwlock, THREAD_LOCK_READ);
+	BLI_rw_mutex_lock(&ccgdm->loops_cache_rwlock, THREAD_LOCK_READ);
 	totface = ccgSubSurf_getNumFaces(ss);
 	mv = mloop;
 	for (index = 0; index < totface; index++) {
@@ -1549,7 +1546,7 @@ static void ccgDM_copyFinalLoopArray(DerivedMesh *dm, MLoop *mloop)
 			}
 		}
 	}
-	BLI_rw_mutex_unlock(&loops_cache_rwlock);
+	BLI_rw_mutex_unlock(&ccgdm->loops_cache_rwlock);
 }
 
 static void ccgDM_copyFinalPolyArray(DerivedMesh *dm, MPoly *mpoly)
@@ -4050,6 +4047,10 @@ static void ccgDM_release(DerivedMesh *dm)
 			MEM_freeN(ccgdm->edgeMap);
 			MEM_freeN(ccgdm->faceMap);
 		}
+
+		BLI_rw_mutex_end(&ccgdm->loops_cache_rwlock);
+		BLI_rw_mutex_end(&ccgdm->origindex_cache_rwlock);
+
 		MEM_freeN(ccgdm);
 	}
 }
@@ -4064,14 +4065,14 @@ static void *ccgDM_get_vert_data_layer(DerivedMesh *dm, int type)
 		int a, index, totnone, totorig;
 
 		/* Avoid re-creation if the layer exists already */
-		BLI_rw_mutex_lock(&origindex_cache_rwlock, THREAD_LOCK_READ);
+		BLI_rw_mutex_lock(&ccgdm->origindex_cache_rwlock, THREAD_LOCK_READ);
 		origindex = DM_get_vert_data_layer(dm, CD_ORIGINDEX);
-		BLI_rw_mutex_unlock(&origindex_cache_rwlock);
+		BLI_rw_mutex_unlock(&ccgdm->origindex_cache_rwlock);
 		if (origindex) {
 			return origindex;
 		}
 
-		BLI_rw_mutex_lock(&origindex_cache_rwlock, THREAD_LOCK_WRITE);
+		BLI_rw_mutex_lock(&ccgdm->origindex_cache_rwlock, THREAD_LOCK_WRITE);
 		DM_add_vert_layer(dm, CD_ORIGINDEX, CD_CALLOC, NULL);
 		origindex = DM_get_vert_data_layer(dm, CD_ORIGINDEX);
 
@@ -4086,7 +4087,7 @@ static void *ccgDM_get_vert_data_layer(DerivedMesh *dm, int type)
 			CCGVert *v = ccgdm->vertMap[index].vert;
 			origindex[a] = ccgDM_getVertMapIndex(ccgdm->ss, v);
 		}
-		BLI_rw_mutex_unlock(&origindex_cache_rwlock);
+		BLI_rw_mutex_unlock(&ccgdm->origindex_cache_rwlock);
 
 		return origindex;
 	}
@@ -5041,6 +5042,9 @@ static CCGDerivedMesh *getCCGDerivedMesh(CCGSubSurf *ss,
 	ccgdm->dm.numLoopData = ccgdm->dm.numPolyData * 4;
 	ccgdm->dm.numTessFaceData = 0;
 
+	BLI_rw_mutex_init(&ccgdm->loops_cache_rwlock);
+	BLI_rw_mutex_init(&ccgdm->origindex_cache_rwlock);
+
 	return ccgdm;
 }