minor fixes

include/rxmesh/cavity_manager2.cuh

@@ -338,7 +338,7 @@ struct CavityManager2
     /**
      * @brief try to add an edge in the cavity graph that connects the two nodes
      * that represents the cavities c0 and c1. If we can not add the edge
-     * (because if the space constraints), we deactivate the cavity with more
+     * (because of the space constraints), we deactivate the cavity with more
      * overlaps
      */
     __device__ __forceinline__ void add_edge_to_cavity_graph(const uint16_t c0,

include/rxmesh/cavity_manager_impl2.cuh

@@ -1191,10 +1191,16 @@ CavityManager2<blockThreads, cop>::add_edge_to_cavity_graph(const uint16_t c0,
 
     // decide which cavity to deactivate
     // we choose the one with more overlaps
-    // printf("\n deactiaving !!!!!!!!!!! ");
-    if (c0_id > c1_id) {
+    if (c0_id > c1_id && c0_id != MAX_OVERLAP_CAVITIES) {
         clear(c0, c0_id);
-    } else {
+    } else if (c0_id > c1_id && c1_id != MAX_OVERLAP_CAVITIES) {
+        clear(c1, c1_id);
+    }
+
+    // if we can prioritize based on overlapping size
+    if (c0_id != MAX_OVERLAP_CAVITIES) {
+        clear(c0, c0_id);
+    } else if (c1_id != MAX_OVERLAP_CAVITIES) {
         clear(c1, c1_id);
     }
 }

include/rxmesh/kernels/rxmesh_queries.cuh

@@ -416,6 +416,9 @@ __device__ __forceinline__ void orient_edges_around_vertices(
         // TODO if the vertex is not owned by this patch, then there is no
         // reason to orient its edges because no serious computation is done on
         // it
+        if (patch_info.is_deleted(LocalVertexT(v))) {
+            continue;
+        }
 
         int start = int(s_output_offset[v]);
         int end   = int(s_output_offset[v + 1]);
@@ -601,6 +604,10 @@ __device__ __forceinline__ void v_v(cooperative_groups::thread_block& block,
     const uint32_t* s_ev_duplicate32 =
         reinterpret_cast<const uint32_t*>(s_ev_duplicate);
     for (uint32_t v = threadIdx.x; v < num_vertices; v += blockThreads) {
+        if (detail::is_deleted(v, active_mask_v)) {
+            continue;
+        }
+
         uint32_t start = s_output_offset[v];
         uint32_t end   = s_output_offset[v + 1];
 
@@ -616,6 +623,10 @@ __device__ __forceinline__ void v_v(cooperative_groups::thread_block& block,
             // uint16_t v1 = s_ev_duplicate[2 * edge + 1];
 
             assert(v0 != INVALID16 && v1 != INVALID16);
+
+            assert(!detail::is_deleted(v0, active_mask_v));
+            assert(!detail::is_deleted(v1, active_mask_v));
+
             assert(v0 == v || v1 == v);
             // s_output_value[e] = (v0 == v) ? v1 : v0;
             s_output_value[e] = (v0 == v) * v1 + (v1 == v) * v0;