Renaming

api/melissa_api.cxx

@@ -66,12 +66,12 @@ int getCommSize()
 }
 
 
-struct ServerRank
+struct ServerRankConnection
 {
   void * data_request_socket;
 
 
-  ServerRank(const char * addr_request)
+  ServerRankConnection(const char * addr_request)
   {
     data_request_socket = zmq_socket (context, ZMQ_REQ);
     assert(data_request_socket);
@@ -84,7 +84,7 @@ struct ServerRank
     D("connect socket %p", data_request_socket);
   }
 
-  ~ServerRank()
+  ~ServerRankConnection()
   {
     D("closing socket %p", data_request_socket);
     zmq_close(data_request_socket);
@@ -174,9 +174,9 @@ struct Server
 Server server;
 
 struct ServerRanks {
-	static map<int, unique_ptr<ServerRank>> ranks;
+	static map<int, unique_ptr<ServerRankConnection>> ranks;
 
-	static ServerRank &get(int server_rank)
+	static ServerRankConnection &get(int server_rank)
 	{
 		auto found = ranks.find(server_rank);
 		if (found == ranks.end())
@@ -185,7 +185,7 @@ struct ServerRanks {
 			// we use unique_ptr's as other wise we would create a ServerRank locally, we than would copy all its values in the ranks map
 			// and then we would destroy it. unfortunately this also closes the zmq connection !
 			auto res = ranks.emplace(server_rank,
-					unique_ptr<ServerRank>(new ServerRank(server.port_names.data() + server_rank * MPI_MAX_PROCESSOR_NAME)));
+					unique_ptr<ServerRankConnection>(new ServerRankConnection(server.port_names.data() + server_rank * MPI_MAX_PROCESSOR_NAME)));
 			return *res.first->second;
 		}
 		else
@@ -194,12 +194,12 @@ struct ServerRanks {
 		}
 	}
 };
-map<int, unique_ptr<ServerRank>> ServerRanks::ranks;
+map<int, unique_ptr<ServerRankConnection>> ServerRanks::ranks;
 
 struct ConnectedServerRank {
   size_t send_count;
   size_t local_vector_offset;
-  ServerRank &server_rank;
+  ServerRankConnection &server_rank;
 };
 
 struct Field {
@@ -422,7 +422,7 @@ void melissa_finalize()  // TODO: when using more serverranks, wait until an end
 	MPI_Barrier(comm);
   //sleep(3);
 	D("End Simulation.");
-  D("server_ranks: %lu", ServerRanks::ranks.size());
+  D("server ranks: %lu", ServerRanks::ranks.size());
 
   phase = PHASE_FINAL;
   // TODO: free all pointers?

common/n_to_m.h

@@ -10,25 +10,25 @@ struct n_to_m {  // TODO: rename datatype into Part
   size_t send_count;
 };
 
-std::vector<n_to_m> calculate_n_to_m(int ranks_server, const std::vector<size_t> &local_vect_sizes_simu)
+std::vector<n_to_m> calculate_n_to_m(const int comm_size_server, const std::vector<size_t> &local_vect_sizes_simu)
 {
-	size_t ranks_simu = local_vect_sizes_simu.size();
+	size_t comm_size_simu = local_vect_sizes_simu.size();
   std::vector <n_to_m> parts;
-  size_t local_vect_sizes_server[ranks_server];
+  size_t local_vect_sizes_server[comm_size_server];
   size_t global_vect_size = 0;
-  for (size_t i = 0; i < ranks_simu; ++i)
+  for (size_t i = 0; i < comm_size_simu; ++i)
   {
     global_vect_size += local_vect_sizes_simu[i];
   }
 
-  for (int i = 0; i < ranks_server; ++i)
+  for (int i = 0; i < comm_size_server; ++i)
   {
     // every server rank gets the same amount
-    local_vect_sizes_server[i] = global_vect_size / ranks_server;
+    local_vect_sizes_server[i] = global_vect_size / comm_size_server;
 
     // let n be the rest of this division
     // the first n server ranks get one more to split the rest fair up...
-    size_t n_rest = global_vect_size - size_t(global_vect_size / ranks_server) * ranks_server;
+    size_t n_rest = global_vect_size - size_t(global_vect_size / comm_size_server) * comm_size_server;
     if (size_t(i) < n_rest)
     {
       local_vect_sizes_server[i]++;

server/server.cxx

@@ -2,6 +2,7 @@
 // TODO 1. refactoring
 //   TODO: use zmq cpp??
 // TODO 2. error prone ness
+		// TODO: check ret values!
 // TODO 3. check with real world sim and DA.
 
 #include <map>
@@ -56,7 +57,7 @@ struct Part
 
 struct EnsembleMember
 {
-	vector<double> state_analysis;  // really tODO use vector! vector.data() will give the same... raw pointer!
+	vector<double> state_analysis;
 	vector<double> state_background;
 	size_t received_state_background = 0;
 
@@ -94,7 +95,6 @@ struct Field {
 		ensemble_members.resize(ensemble_size_);
 	}
 
-	// TODO: naming: server_comm size or ranks_server? same for simu!
 	/// Calculates all the state vector parts that are send between the server and the
 	/// simulations
 	void calculate_parts(int server_comm_size)
@@ -186,7 +186,7 @@ struct ConnectedSimulationRank {
 
 
 	/// returns true if a new was sent.
-	bool try_to_start_task(int simu_rank) {
+	bool try_send_task(int simu_rank) {
 
 		// already working?
 		if (current_task != WANT_WORK)
@@ -299,9 +299,9 @@ struct Simulation  // Model process runner
 		}
 	}
 
-	void try_to_start_task() {// todo: replace fields.begin() by field as there will be only on field soon.
+	void try_send_task() {// low: replace fields.begin() by field as there will be only on field soon.
 		for (auto cs = connected_simulation_ranks.begin(); cs != connected_simulation_ranks.end(); cs++) {
-			cs->second.try_to_start_task(cs->first);
+			cs->second.try_send_task(cs->first);
 		}
 	}
 
@@ -671,6 +671,7 @@ int main(int argc, char * argv[])
 					zmq_msg_size(&data_msg), part.send_count * sizeof(double), field_name, simu_id, simu_rank, simu_state_id, simu_timestamp);
 			D("local server offset %lu, sendcount=%lu", part.local_offset_server, part.send_count);
 
+
 			D("values[0] = %.3f", reinterpret_cast<double*>(zmq_msg_data(&data_msg))[0]);
 			if (simu_timestamp == current_timestamp)
 			{
@@ -695,7 +696,7 @@ int main(int argc, char * argv[])
 			zmq_msg_close(&data_msg);
 
 			// Check if we can answer directly with new data... means starting of a new model task
-			bool got_task = simu.connected_simulation_ranks[simu_rank].try_to_start_task(simu_rank);
+			bool got_task = simu.connected_simulation_ranks[simu_rank].try_send_task(simu_rank);
 
 			// If we could not start a new model task try to schedule a new one. This is initiated by server rank 0
 			if (!got_task && comm_rank == 0)
@@ -704,7 +705,7 @@ int main(int argc, char * argv[])
 					schedule_new_task(simu_id);
 
 					// try to run it directly:
-					simu.connected_simulation_ranks[simu_rank].try_to_start_task(simu_rank);
+					simu.connected_simulation_ranks[simu_rank].try_send_task(simu_rank);
 				}
 			}
 
@@ -740,7 +741,7 @@ int main(int argc, char * argv[])
 					{
 						if (schedule_new_task(simu_it->first)) {
 							// normally we arrive always here if there are not more model task runners than ensemble members.
-							simu_it->second.try_to_start_task();
+							simu_it->second.try_send_task();
 						}
 					}
 				}
@@ -748,8 +749,6 @@ int main(int argc, char * argv[])
 		}
 
 
-		// TODO: check ret values!
-		// TODO: remove compile warnings!
 
 		if (phase == PHASE_SIMULATION && comm_rank != 0)
 		{
@@ -766,7 +765,7 @@ int main(int argc, char * argv[])
 				simu.addTask(new_task.task_id, new_task.task);
 
 				// If so try to start them directly.
-				simu.try_to_start_task();
+				simu.try_send_task();
 			}
 
 		}