Added NO_CHANGE option to debug

example/VCluster/0_simple/main.cpp

@@ -4,40 +4,76 @@
 #include "VCluster.hpp"
 
 /*
- * ### WIKI 1 ###
+ *
+ * \page VCluster_0_simple Using Vcluster to communicate across processors
+ *
+ *
+ * [TOC]
  *
  * ## Simple example
  * 
  * This example show several basic functionalities of VCluster
  * 
- * ### WIKI END ###
  * 
  */
-
-
 int main(int argc, char* argv[])
 {
-	//
-	// ### WIKI 2 ###
-	//
-	// Initialize the library and several objects 
-	//
+	/*
+	 *
+	 * \page VCluster_0_simple Using Vcluster to communicate across processors
+	 *
+	 *
+	 * ## Initialization
+	 *
+	 * Before using any functionality the library must be initialized
+	 *
+	 * \snippet Vcluster/0_simple/main.cpp initialization
+	 *
+	 */
+
+	//! \cond [initialization] \endcond
+
 	openfpm_init(&argc,&argv);
 	
-	//
-	// ### WIKI 3 ###
-	//
-	// Get the vcluster object and the number of processor
-	//
+	//! \cond [initialization] \endcond
+
+	/*
+	 *
+	 * \page VCluster_0_simple Using Vcluster to communicate across processors
+	 *
+	 * ### Initialization of Vcluster
+	 *
+	 * Because in general our program is parallel we have more than one processors. With
+	 * the function getProcessingUnits we can querry how many processors are involved in
+	 * out computation
+	 *
+	 * \snippet Vcluster/0_simple/main.cpp initialization
+	 *
+	 */
 	Vcluster & v_cl = create_vcluster();
 	long int N_prc = v_cl.getProcessingUnits();
 
-	//
-	// ### WIKI 3 ###
-	//
-	// We find the maximum of the processors rank, that should be the Number of
-	// processor minus one, only processor 0 print on terminal
-	//
+	/*
+	 *
+	 * \page VCluster_0_simple Using Vcluster to communicate across processors
+	 *
+	 *
+	 * ### min max, sum
+	 *
+	 * with the function getProcessUnitID we can the the id of the processor executing
+	 * the function. This function is equivalent to the MPI rank function.
+	 * Vcluster provide several high and low level functionalities. One is max that
+	 * return the maximum value across processors. There is also the function min
+	 * and sum that return respectively the sum and the minimum across processors.
+	 * All these operations are asynchronous, in order to get the result the function
+	 * execute must be used. In our example the processor 0 print we also print the value
+	 * but can be easily verified that also the other processors has the same value.
+	 *
+	 * \snippet Vcluster/0_simple/main.cpp max calc
+	 *
+	 */
+
+	//! \cond [max calc] \endcond
 
 	long int id = v_cl.getProcessUnitID();
 
@@ -46,12 +82,21 @@ int main(int argc, char* argv[])
 	if (v_cl.getProcessUnitID() == 0)
 		std::cout << "Maximum processor rank: " << id << "\n";
 
-	//
-	// ### WIKI 4 ###
-	//
-	// We sum all the processor ranks the maximum, the result should be that should
-	// be $\frac{(n-1)n}{2}$, only processor 0 print on terminal
-	//
+	//! \cond [max calc] \endcond
+
+	/*
+	 *
+	 * \page VCluster_0_simple Using Vcluster to communicate across processors
+	 *
+	 * We sum all the processor ranks the result should be that should
+	 * be \$\frac{(n-1)n}{2}\$, only processor 0 print on terminal
+	 *
+	 *
+	 * \snippet Vcluster/0_simple/main.cpp sum calc
+	 *
+	 */
+
+	//! \cond [sum calc] \endcond
 
 	size_t id2 = v_cl.getProcessUnitID();
 
@@ -60,11 +105,19 @@ int main(int argc, char* argv[])
 	if (v_cl.getProcessUnitID() == 0)
 		std::cout << "Sum of all processors rank: " << id2 << "\n";
 
-	//
-	// ### WIKI 5 ###
-	//
-	// we can collect information from all processors using the function gather
-	//
+	//! \cond [sum calc] \endcond
+
+	/*
+	 *
+	 * \page VCluster_0_simple Using Vcluster to communicate across processors
+	 *
+	 * We sum all the processor ranks the result should be that should
+	 * be \$\frac{(n-1)n}{2}\$, only processor 0 print on terminal
+	 *
+	 *
+	 * \snippet Vcluster/0_simple/main.cpp max calc
+	 *
+	 */
 
 	long int id3 = v_cl.getProcessUnitID();
 	openfpm::vector<long int> v;

src/Decomposition/CartDecomposition.hpp

@@ -1251,7 +1251,11 @@ public:
 
 	/*! \brief Get the domain Cells
 	 *
-	 * \param shift Shifting point
+	 * It performa a linearization of the domain cells using the extension provided in gs
+	 *
+	 *
+	 * \param shift Cell padding
+	 * \param cell_shift where the domain cell start
 	 * \param gs grid extension
 	 *
 	 */

src/Vector/vector_dist.hpp

@@ -390,7 +390,7 @@ public:
 		Ghost<dim,size_t> g_ext(0);
 		cell_list.Initialize(cd_sm,pbox,pad);
 
-		updateCellList(cell_list);
+		updateCellListSym(cell_list);
 
 		return cell_list;
 	}

src/Vector/vector_dist_comm.hpp

@@ -15,6 +15,7 @@
 
 #define NO_POSITION 1
 #define WITH_POSITION 2
+#define NO_CHANGE_ELEMENTS 4
 
 #define BIND_DEC_TO_GHOST 1
 
@@ -57,7 +58,7 @@ class vector_dist_comm
 	//! particles that must be communicated to the other processors
 	openfpm::vector<openfpm::vector<aggregate<size_t,size_t>>> g_opart;
 
-	// processor rank list of g_opart
+	//! processor rank list of g_opart
 	openfpm::vector<size_t> prc_g_opart;
 
 	//! Sending buffer for the ghost particles position
@@ -76,6 +77,8 @@ class vector_dist_comm
 	//! It store the size of the elements added for each processor that communicate with us (local processor)
 	//! from the last ghost get
 	openfpm::vector<size_t> recv_sz_get;
+	//! Conversion to byte of recv_sz_get
+	openfpm::vector<size_t> recv_sz_get_byte;
 
 	//! The same as recv_sz_get but for put
 	openfpm::vector<size_t> recv_sz_put;
@@ -87,18 +90,21 @@ class vector_dist_comm
 	//! replicated ghost particles that are local
 	size_t lg_m;
 
-	//! process the particle with properties
+	//! process the particle without properties
 	struct proc_without_prp
 	{
+		//! process the particle
 		template<typename T1, typename T2> inline static void proc(size_t lbl, size_t cnt, size_t id, T1 & v_prp, T2 & m_prp)
 		{
 			m_prp.get(lbl).set(cnt, v_prp.get(id));
 		}
 	};
 
+	//! process the particle with properties
 	template<typename prp_object, int ... prp>
 	struct proc_with_prp
 	{
+		//! process the particle
 		template<typename T1, typename T2> inline static void proc(size_t lbl, size_t cnt, size_t id, T1 & v_prp, T2 & m_prp)
 		{
 			// source object type
@@ -151,7 +157,7 @@ class vector_dist_comm
 	/*! \brief Return a valid particle starting from end and tracing back
 	 *
 	 * \param end actual opart particle pointer
-	 * \param actual end particle point
+	 * \param end_id actual end particle point
 	 *
 	 * \return a valid particle
 	 *
@@ -386,7 +392,6 @@ class vector_dist_comm
 	 *
 	 * \param v_pos vector of particle positions
 	 * \param g_pos_send Send buffer to fill
-	 * \param prAlloc_pos Memory object for the send buffer
 	 *
 	 */
 	void fill_send_ghost_pos_buf(openfpm::vector<Point<dim, St>> & v_pos,openfpm::vector<send_pos_vector> & g_pos_send)
@@ -422,7 +427,6 @@ class vector_dist_comm
 	 *
 	 * \param v_prp vector of particle properties
 	 * \param g_send_prp Send buffer to fill
-	 * \param prAlloc_prp Memory object for the send buffer
 	 * \param g_m ghost marker
 	 *
 	 */
@@ -468,7 +472,6 @@ class vector_dist_comm
 	 *
 	 * \param v_prp vector of particle properties
 	 * \param g_send_prp Send buffer to fill
-	 * \param prAlloc_prp Memory object for the send buffer
 	 *
 	 */
 	template<typename send_vector, typename prp_object, int ... prp> void fill_send_ghost_prp_buf(openfpm::vector<prop> & v_prp, openfpm::vector<send_vector> & g_send_prp)
@@ -501,9 +504,9 @@ class vector_dist_comm
 	 *
 	 * \param v_pos vector of particle positions
 	 * \param v_prp vector of particles properties
-	 * \param prc_r List of processor rank involved in the send
 	 * \param prc_sz_r For each processor in the list the size of the message to send
-	 * \param pb send buffer
+	 * \param m_pos sending buffer for position
+	 * \param m_prp sending buffer for properties
 	 *
 	 */
 	void fill_send_map_buf(openfpm::vector<Point<dim, St>> & v_pos, openfpm::vector<prop> & v_prp, openfpm::vector<size_t> & prc_sz_r, openfpm::vector<openfpm::vector<Point<dim,St>>> & m_pos, openfpm::vector<openfpm::vector<prop>> & m_prp)
@@ -542,8 +545,8 @@ class vector_dist_comm
 	 * \param v_pos vector of particle positions
 	 * \param v_prp vector of particle properties
 	 * \param prc_r List of processor rank involved in the send
-	 * \param prc_sz_r For each processor in the list the size of the message to send
-	 * \param pb send buffer
+	 * \param m_pos sending buffer for position
+	 * \param m_prp sending buffer for properties
 	 *
 	 */
 	template<typename prp_object,int ... prp> void fill_send_map_buf_list(openfpm::vector<Point<dim, St>> & v_pos, openfpm::vector<prop> & v_prp, openfpm::vector<size_t> & prc_sz_r, openfpm::vector<openfpm::vector<Point<dim,St>>> & m_pos, openfpm::vector<openfpm::vector<prp_object>> & m_prp)
@@ -871,11 +874,15 @@ public:
         {
                 if (opt & SKIP_LABELLING)
                 {
-                                op_ssend_gg_recv_merge opm(g_m);
-                                v_cl.SSendRecvP_op<op_ssend_gg_recv_merge,send_vector,decltype(v_prp),prp...>(g_send_prp,v_prp,prc_g_opart,opm,prc_recv_get,recv_sz_get);
+                	size_t opt = NONE;
+                	if (opt & NO_CHANGE_ELEMENTS)
+                		opt = RECEIVE_KNOWN;
+
+                	op_ssend_gg_recv_merge opm(g_m);
+                    v_cl.SSendRecvP_op<op_ssend_gg_recv_merge,send_vector,decltype(v_prp),prp...>(g_send_prp,v_prp,prc_g_opart,opm,prc_recv_get,recv_sz_get);
                 }
                 else
-                        v_cl.SSendRecvP<send_vector,decltype(v_prp),prp...>(g_send_prp,v_prp,prc_g_opart,prc_recv_get,recv_sz_get);
+                        v_cl.SSendRecvP<send_vector,decltype(v_prp),prp...>(g_send_prp,v_prp,prc_g_opart,prc_recv_get,recv_sz_get,recv_sz_get_byte);
         }
 
 		if (!(opt & NO_POSITION))

src/Vector/vector_dist_unit_test.hpp

@@ -1501,7 +1501,10 @@ BOOST_AUTO_TEST_CASE( vector_dist_ghost_with_ghost_buffering )
 			++it;
 		}
 
-		vd.ghost_get<0>(SKIP_LABELLING);
+		if (i % 2 == 0)
+			vd.ghost_get<0>(SKIP_LABELLING);
+		else
+			vd.ghost_get<0>(SKIP_LABELLING | NO_CHANGE_ELEMENTS );
 
 		auto it2 = vd.getGhostIterator();
 		bool ret = true;