fuse migrate_from_patch into one call

apps/Delaunay/delaunay_rxmesh.cuh

@@ -187,7 +187,7 @@ inline uint32_t count_non_delaunay_edges(TriMesh& mesh)
 
  float lambda = angle_between_three_vertices(p0, p2, p1);
  float gamma = angle_between_three_vertices(p1, p3, p0);
- if (lambda + gamma > M_PI) {
+ if (lambda + gamma - 0.00001 > M_PI) {
  num_non_delaunay++;
  }
  }
@@ -314,6 +314,7 @@ inline void delaunay_rxmesh(rxmesh::RXMeshDynamic& rx, bool with_verify = true)
  }
  CUDA_ERROR(cudaMemcpy(
  &h_flipped, d_flipped, sizeof(int), cudaMemcpyDeviceToHost));
+ // break;
  }
 
  CUDA_ERROR(cudaProfilerStop());

include/rxmesh/cavity_manager.cuh

@@ -393,6 +393,12 @@ struct CavityManager
  */
  __device__ __inline__ bool migrate(cooperative_groups::thread_block& block);
 
+ /**
+ * @brief migrate vertices/edges/faces from neighbor patches to this patch
+ */
+ __device__ __inline__ bool migrate_v2(
+ cooperative_groups::thread_block& block);
+
  /**
  * @brief given a neighbor patch (q), migrate vertices (and edges and faces
  * connected to these vertices) marked in migrate_mask_v to the patch
@@ -404,12 +410,13 @@ struct CavityManager
  const uint32_t q,
  const Bitmask& migrate_mask_v);
 
- /**
- * @brief given a neighbor patch (q), migrate vertices (and edges and faces
- * connected to these vertices) marked in m_s_migrate_mask_v and
- * m_s_ribbonize_v to the patch managed by this cavity manager
- */
- __device__ __inline__ bool migrate_from_patch(
+ __device__ __inline__ bool migrate_from_patch_v2(
+ cooperative_groups::thread_block& block,
+ const uint8_t q_stash_id,
+ const uint32_t q);
+
+
+ __device__ __inline__ bool soft_migrate_from_patch(
  cooperative_groups::thread_block& block,
  const uint8_t q_stash_id,
  const uint32_t q);
@@ -419,11 +426,13 @@ struct CavityManager
  * of q_vertex in this patch. If it does not exist, create such a copy.
  */
  template <typename FuncT>
- __device__ __inline__ LPPair migrate_vertex(const uint32_t q,
- const uint16_t q_num_vertices,
- const uint16_t q_vertex,
- PatchInfo& q_patch_info,
- FuncT should_migrate);
+ __device__ __inline__ LPPair migrate_vertex(
+ const uint32_t q,
+ const uint16_t q_num_vertices,
+ const uint16_t q_vertex,
+ PatchInfo& q_patch_info,
+ FuncT should_migrate,
+ bool add_to_connect_cavity_bdry_v = false);
 
 
  /**
@@ -593,20 +602,23 @@ struct CavityManager
  // active elements bitmask
  Bitmask m_s_active_mask_v, m_s_active_mask_e, m_s_active_mask_f;
 
- // indicate if a the vertex should be migrated
- Bitmask m_s_migrate_mask_v;
-
  Bitmask m_s_src_mask_v, m_s_src_mask_e;
  Bitmask m_s_src_connect_mask_v, m_s_src_connect_mask_e;
 
  // indicate if the mesh element should change ownership
  Bitmask m_s_ownership_change_mask_v, m_s_ownership_change_mask_e,
  m_s_ownership_change_mask_f;
 
+ // indicate if the vertex is on the cavity boundary and is owned
  Bitmask m_s_owned_cavity_bdry_v;
 
- // indicate if the vertex should be ribbonized
- Bitmask m_s_ribbonize_v;
+ // indicate if the vertex is on the cavity boundary and not owned i.e.,
+ // the vertex should be migrated
+ Bitmask m_s_not_owned_cavity_bdry_v;
+
+ // indicate if the vertex is connected to a vertex on the cavity boundary
+ // i.e., need to be ribbonized
+ Bitmask m_s_connect_cavity_bdry_v;
 
  // indicate which patch (in the patch stash) should be locked
  Bitmask m_s_patches_to_lock_mask;