diff --git a/check.py b/check.py
index cd2e2f7b21e551c6e3562ca862636fef93c57e44..11072fc430b763c80c0af955c519355d7355191f 100644
--- a/check.py
+++ b/check.py
@@ -7,7 +7,7 @@ from sklearn.neighbors import NearestNeighbors
 
 ndims = 5
 k     = 500 
-p     = 2 
+p     = 10 
 
 with open("bb/top_nodes.csv","r") as f:
     l = f.readlines() 
diff --git a/src/tree/tree.c b/src/tree/tree.c
index 6183f7e1b1c684e1c8e3b6f05dfedc148a2cb8d3..061007441c1058e43bc1710b6b7a0300897d8d20 100644
--- a/src/tree/tree.c
+++ b/src/tree/tree.c
@@ -1381,6 +1381,7 @@ void mpi_ngbh_search(global_context_t* ctx, datapoint_info_t* dp_info, top_kdtre
     float_t* __snd_points = (float_t*)malloc(tot_points_snd * (1 + ctx -> dims) * sizeof(float_t)); 
 
 
+
     /* copy data to send in contiguous memory */
     for(int i = 0; i < ctx -> world_size; ++i)
     {
@@ -1404,25 +1405,43 @@ void mpi_ngbh_search(global_context_t* ctx, datapoint_info_t* dp_info, top_kdtre
 
     int work_batch_stride = 1 + ctx -> dims;
 
-    heap_node* __heap_batches_to_snd = (heap_node*)malloc(k * tot_points_rcv * sizeof(heap_node));
-    heap_node* __heap_batches_to_rcv = (heap_node*)malloc(k * tot_points_snd * sizeof(heap_node));
+    heap_node* __heap_batches_to_snd = (heap_node*)malloc((uint64_t)k * (uint64_t)tot_points_rcv * sizeof(heap_node));
+    heap_node* __heap_batches_to_rcv = (heap_node*)malloc((uint64_t)k * (uint64_t)tot_points_snd * sizeof(heap_node));
 
-    rcv_displ[0] = 0;
-    snd_displ[0] = 0;
     /*
-    rcv_count[0] = point_to_rcv_count[0] * k * sizeof(heap_node);
-    snd_count[0] = point_to_snd_count[0] * k * sizeof(heap_node); 
+    for(int i = 0; i < ctx -> world_size; ++i)
+    {
+        if(i == ctx -> mpi_rank)
+        {
+            DB_PRINT("[RANK %d]\t", ctx -> mpi_rank);
+            DB_PRINT("tot point rcv %d tot point send %d \t",tot_points_rcv, tot_points_snd);
+            DB_PRINT("bytes rcv %lu snd %lu", (uint64_t)k * (uint64_t)tot_points_rcv * sizeof(heap_node), (uint64_t)k * (uint64_t)tot_points_snd * sizeof(heap_node));
+            DB_PRINT("\n");
+        }
+    }
     */
+    
+    if( __heap_batches_to_rcv == NULL)
+    {
+        DB_PRINT("Rank %d failed to allocate rcv_heaps %luB required\n",ctx -> mpi_rank, (uint64_t)k * (uint64_t)tot_points_rcv * sizeof(heap_node));
+    }
+
+    if( __heap_batches_to_snd == NULL)
+    {
+        DB_PRINT("Rank %d failed to allocate snd_heaps %luB required\n",ctx -> mpi_rank, (uint64_t)k * (uint64_t)tot_points_snd * sizeof(heap_node));
+    }
+
+    MPI_Barrier(ctx -> mpi_communicator);
+
+
+    rcv_displ[0] = 0;
+    snd_displ[0] = 0;
     rcv_count[0] = point_to_rcv_count[0];
     snd_count[0] = point_to_snd_count[0]; 
 
 
     for(int i = 1; i < ctx -> world_size; ++i)
     {
-        /*
-        rcv_count[i] = point_to_rcv_count[i] * k * sizeof(heap_node); 
-        snd_count[i] = point_to_snd_count[i] * k * sizeof(heap_node); 
-        */
 
         rcv_count[i] = point_to_rcv_count[i]; 
         snd_count[i] = point_to_snd_count[i]; 
@@ -1434,14 +1453,13 @@ void mpi_ngbh_search(global_context_t* ctx, datapoint_info_t* dp_info, top_kdtre
     heap_node** heap_batches_per_node = (heap_node**)malloc(ctx -> world_size * sizeof(heap_node*));
     for(int p = 0; p < ctx -> world_size; ++p) 
     {
-        //heap_batches_per_node[p] = NULL;
-        //heap_batches_per_node[p] = __heap_batches_to_snd + rcv_displ[p] / sizeof(heap_node);
         heap_batches_per_node[p] = __heap_batches_to_snd + rcv_displ[p] * k;
     }
 
     /* compute everything */
     MPI_DB_PRINT("[MASTER] Working on recieved points\n");
 
+
     for(int p = 0; p < ctx -> world_size; ++p)
     {
         if(point_to_rcv_count[p] > 0)
@@ -1467,7 +1485,9 @@ void mpi_ngbh_search(global_context_t* ctx, datapoint_info_t* dp_info, top_kdtre
      * counts are inverted since I have to recieve as many batches as points I
      * Have originally sended
      */
+
     MPI_DB_PRINT("[MASTER] Sending out results\n");
+    /*
     for(int i = 0; i < ctx -> world_size; ++i)
     {
         if(i == ctx -> mpi_rank)
@@ -1479,6 +1499,7 @@ void mpi_ngbh_search(global_context_t* ctx, datapoint_info_t* dp_info, top_kdtre
             DB_PRINT("\n");
         }
     }
+    */
 
 
     MPI_Datatype MPI_my_heap;
@@ -1821,16 +1842,23 @@ void simulate_master_read_and_scatter(int dims, size_t n, global_context_t *ctx)
 		// std_g0163178_Me14_091_0000
         
         /* 10^6 points ca.*/
-		//data = read_data_file(ctx,"../norm_data/std_LR_091_0001",MY_TRUE);
+		data = read_data_file(ctx,"../norm_data/std_LR_091_0001",MY_TRUE);
 
         /* 10^7 ~ 8M points */
-		//data = read_data_file(ctx,"../norm_data/std_g0163178_Me14_091_0001",MY_TRUE);
-		data = read_data_file(ctx,"../norm_data/std_g5503149_091_0001",MY_TRUE);
+        
+        //data = read_data_file(ctx,"../norm_data/std_g0144846_Me14_091_0001",MY_TRUE);
+
+        //88M BREAKS
+		//data = read_data_file(ctx,"../norm_data/std_g5503149_091_0000",MY_TRUE);
+
+        //
+        //34 M
+		//data = read_data_file(ctx,"../norm_data/std_g1212639_091_0001",MY_TRUE);
         ctx->dims = 5;
 
 		// ctx -> n_points = 48*5*2000;
 		ctx->n_points = ctx->n_points / ctx->dims;
-		ctx->n_points = ctx->n_points / 2;
+		//ctx->n_points = ctx->n_points / 2;
         //ctx -> n_points = ctx -> world_size * 1000;
 		mpi_printf(ctx, "Read %lu points in %u dims\n", ctx->n_points, ctx->dims);
 	}