Merge pull request elalish#106 from pca006132/optimizations

Optimizations

collider/src/collider.cpp

@@ -266,7 +266,7 @@ Collider::Collider(const VecDH<Box>& leafBB,
  */
 template <typename T>
 SparseIndices Collider::Collisions(const VecDH<T>& querriesIn) const {
-  int maxOverlaps = 1 << 20;
+  int maxOverlaps = querriesIn.size() * 4;
   SparseIndices querryTri(maxOverlaps);
   int nOverlaps = 0;
   while (1) {
@@ -282,7 +282,11 @@ SparseIndices Collider::Collisions(const VecDH<T>& querriesIn) const {
       break;
     else {  // if not enough memory was allocated, guess how much will be needed
       int lastQuery = querryTri.Get(0).H().back();
-      maxOverlaps *= 2 * static_cast<float>(querriesIn.size()) / lastQuery;
+      float ratio = static_cast<float>(querriesIn.size()) / lastQuery;
+      if (ratio > 1000) // do not trust the ratio if it is too large
+        maxOverlaps *= 2;
+      else
+        maxOverlaps *= 2 * ratio;
       querryTri.Resize(maxOverlaps);
     }
   }

manifold/src/face_op.cpp

@@ -42,6 +42,9 @@ void Manifold::Impl::Face2Tri(const VecDH<int>& faceEdge,
   const VecH<Halfedge>& halfedge = halfedge_.H();
   const VecH<glm::vec3>& faceNormal = faceNormal_.H();
   meshRelation_.triBary.resize(0);
+  triVerts.reserve(faceEdge.size());
+  triNormal.reserve(faceEdge.size());
+  meshRelation_.triBary.H().reserve(faceEdge.size()*3);
 
   for (int face = 0; face < faceEdgeH.size() - 1; ++face) {
     const int firstEdge = faceEdgeH[face];
@@ -50,12 +53,17 @@ void Manifold::Impl::Face2Tri(const VecDH<int>& faceEdge,
     ALWAYS_ASSERT(numEdge >= 3, topologyErr, "face has less than three edges.");
     const glm::vec3 normal = faceNormal[face];
 
-    std::map<int, int> vertBary;
-    for (int j = firstEdge; j < lastEdge; ++j)
-      vertBary[halfedge[j].startVert] = halfedgeBary.H()[j];
-    const int startTri = triVerts.size();
+    auto linearSearch = [](const int* mapping, int value) {
+      int i = 0;
+      while (mapping[i] != value)
+        ++i;
+      return i;
+    };
 
     if (numEdge == 3) {  // Single triangle
+      int mapping[3] = {halfedge[firstEdge].startVert,
+                        halfedge[firstEdge + 1].startVert,
+                        halfedge[firstEdge + 2].startVert};
       glm::ivec3 tri(halfedge[firstEdge].startVert,
                      halfedge[firstEdge + 1].startVert,
                      halfedge[firstEdge + 2].startVert);
@@ -71,7 +79,16 @@ void Manifold::Impl::Face2Tri(const VecDH<int>& faceEdge,
 
       triVerts.push_back(tri);
       triNormal.push_back(normal);
+      meshRelation_.triBary.H().push_back(faceRef.H()[face]);
+      for (int k : {0, 1, 2}) {
+        int index = linearSearch(mapping, tri[k]);
+        meshRelation_.triBary.H().back().vertBary[k] = halfedgeBary.H()[firstEdge + index];
+      }
     } else if (numEdge == 4) {  // Pair of triangles
+      int mapping[4] = {halfedge[firstEdge].startVert,
+                        halfedge[firstEdge + 1].startVert,
+                        halfedge[firstEdge + 2].startVert,
+                        halfedge[firstEdge + 3].startVert};
       const glm::mat3x2 projection = GetAxisAlignedProjection(normal);
       auto triCCW = [&projection, &vertPos, this](const glm::ivec3 tri) {
         return CCW(projection * vertPos[tri[0]], projection * vertPos[tri[1]],
@@ -111,13 +128,22 @@ void Manifold::Impl::Face2Tri(const VecDH<int>& faceEdge,
         }
       }
 
-      triVerts.push_back(tri0);
-      triNormal.push_back(normal);
-      triVerts.push_back(tri1);
-      triNormal.push_back(normal);
+      for (auto tri : { tri0, tri1 }) {
+        triVerts.push_back(tri);
+        triNormal.push_back(normal);
+        meshRelation_.triBary.H().push_back(faceRef.H()[face]);
+        for (int k : {0, 1, 2}) {
+          int index = linearSearch(mapping, tri[k]);
+          meshRelation_.triBary.H().back().vertBary[k] = halfedgeBary.H()[firstEdge + index];
+        }
+      }
     } else {  // General triangulation
       const glm::mat3x2 projection = GetAxisAlignedProjection(normal);
 
+      std::map<int, int> vertBary;
+      for (int j = firstEdge; j < lastEdge; ++j)
+        vertBary[halfedge[j].startVert] = halfedgeBary.H()[j];
+
       Polygons polys;
       try {
         polys = Face2Polygons(face, projection, faceEdgeH);
@@ -134,18 +160,15 @@ void Manifold::Impl::Face2Tri(const VecDH<int>& faceEdge,
       for (auto tri : newTris) {
         triVerts.push_back(tri);
         triNormal.push_back(normal);
-      }
-    }
-
-    for (int tri = startTri; tri < triVerts.size(); ++tri) {
-      meshRelation_.triBary.H().push_back(faceRef.H()[face]);
-      for (int k : {0, 1, 2}) {
-        meshRelation_.triBary.H().back().vertBary[k] =
-            vertBary[triVerts[tri][k]];
+        meshRelation_.triBary.H().push_back(faceRef.H()[face]);
+        for (int k : {0, 1, 2}) {
+          meshRelation_.triBary.H().back().vertBary[k] =
+              vertBary[tri[k]];
+        }
       }
     }
   }
-  faceNormal_ = triNormalOut;
+  faceNormal_ = std::move(triNormalOut);
   CreateAndFixHalfedges(triVertsOut);
 }
 

manifold/src/impl.cpp

@@ -507,17 +507,19 @@ void Manifold::Impl::CreateHalfedges(const VecDH<glm::ivec3>& triVerts) {
  */
 void Manifold::Impl::CreateAndFixHalfedges(const VecDH<glm::ivec3>& triVerts) {
   const int numTri = triVerts.size();
+  // drop the old value first to avoid copy
+  halfedge_.resize(0);
   halfedge_.resize(3 * numTri);
   VecDH<TmpEdge> edge(3 * numTri);
-  thrust::for_each_n(zip(countAt(0), triVerts.begin()), numTri,
-                     Tri2Halfedges({halfedge_.ptrH(), edge.ptrH()}));
+  thrust::for_each_n(zip(countAt(0), triVerts.beginD()), numTri,
+                     Tri2Halfedges({halfedge_.ptrD(), edge.ptrD()}));
   // Stable sort is required here so that halfedges from the same face are
   // paired together (the triangles were created in face order). In some
   // degenerate situations the triangulator can add the same internal edge in
   // two different faces, causing this edge to not be 2-manifold. We detect this
   // and fix it by swapping one of the identical edges, so it is important that
   // we have the edges paired according to their face.
-  std::stable_sort(edge.begin(), edge.end());
+  thrust::stable_sort(edge.beginD(), edge.endD());
   thrust::for_each_n(thrust::host, countAt(0), halfedge_.size() / 2,
                      LinkHalfedges({halfedge_.ptrH(), edge.cptrH()}));
   thrust::for_each(thrust::host, countAt(1), countAt(halfedge_.size() / 2),

manifold/src/smoothing.cpp

@@ -470,7 +470,7 @@ Manifold::Impl::MeshRelationD Manifold::Impl::Subdivide(int n) {
   MeshRelationD relation;
   relation.barycentric.resize(numTri * VertsPerTri(n + 1));
   relation.triBary.resize(n * n * numTri);
-  MeshRelationD oldMeshRelation = meshRelation_;
+  MeshRelationD oldMeshRelation = std::move(meshRelation_);
   meshRelation_.barycentric.resize(relation.barycentric.size());
   meshRelation_.triBary.resize(relation.triBary.size());
 

manifold/src/sort.cpp

@@ -292,8 +292,8 @@ void Manifold::Impl::GatherFaces(const VecDH<int>& faceNew2Old) {
 
   if (faceNormal_.size() == NumTri()) Permute(faceNormal_, faceNew2Old);
 
-  VecDH<Halfedge> oldHalfedge(halfedge_);
-  VecDH<glm::vec4> oldHalfedgeTangent(halfedgeTangent_);
+  VecDH<Halfedge> oldHalfedge(std::move(halfedge_));
+  VecDH<glm::vec4> oldHalfedgeTangent(std::move(halfedgeTangent_));
   VecDH<int> faceOld2New(oldHalfedge.size() / 3);
   thrust::scatter(countAt(0), countAt(numTri), faceNew2Old.beginD(),
                   faceOld2New.beginD());