fix: Fix misuse of CommID

src/coreComponents/mesh/mpiCommunications/CommID.hpp

@@ -44,15 +44,16 @@ class CommID
    */
   ~CommID();
 
-  /// default copy constructor
-  CommID( CommID const & ) = default;
 
   /**
    * Move constructor
    * @param src The source to move data from.
    */
   CommID( CommID && src );
 
+  /// deleted default copy constructor
+  CommID( CommID const & ) = delete;
+
   /// deleted copy assignment operator
   CommID & operator=( CommID const & ) = delete;
 

src/coreComponents/mesh/mpiCommunications/CommunicationTools.cpp

@@ -121,7 +121,7 @@ void CommunicationTools::assignGlobalIndices( ObjectManagerBase & manager,
 
   integer const numNeighbors = LvArray::integerConversion< integer >( neighbors.size() );
 
-  MPI_iCommData commData( getCommID() );
+  MPI_iCommData commData;
   commData.resize( numNeighbors );
 
   array1d< int > receiveBufferSizes( numNeighbors );
@@ -340,11 +340,11 @@ CommunicationTools::assignNewGlobalIndices( ElementRegionManager & elementManage
  * @return The data received from all the @p neighbors. Data at index @p i coming from neighbor at index @p i in the list of @p neighbors.
  */
 template< class DATA_PROVIDER >
-array1d< array1d< globalIndex > > exchange( int commId,
-                                            std::vector< NeighborCommunicator > & neighbors,
+array1d< array1d< globalIndex > > exchange( std::vector< NeighborCommunicator > & neighbors,
                                             DATA_PROVIDER const & data )
 {
-  MPI_iCommData commData( commId );
+  MPI_iCommData commData;
+  int commId = commData.commID();
   integer const numNeighbors = LvArray::integerConversion< integer >( neighbors.size() );
   commData.resize( numNeighbors );
   for( integer i = 0; i < numNeighbors; ++i )
@@ -391,7 +391,7 @@ CommunicationTools::buildNeighborPartitionBoundaryObjects( ObjectManagerBase & m
   {
     return std::cref( globalPartitionBoundaryObjectsIndices );
   };
-  array1d< array1d< globalIndex > > const neighborPartitionBoundaryObjects = exchange( getCommID(), allNeighbors, data );
+  array1d< array1d< globalIndex > > const neighborPartitionBoundaryObjects = exchange( allNeighbors, data );
 
   integer const numNeighbors = LvArray::integerConversion< integer >( allNeighbors.size() );
   for( integer i = 0; i < numNeighbors; ++i )
@@ -566,7 +566,7 @@ void CommunicationTools::findMatchedPartitionBoundaryNodes( NodeManager & nodeMa
   {
     return req.at( allNeighbors[i].neighborRank() );
   };
-  array1d< array1d< globalIndex > > const nodesRequestedByNeighbors = exchange( getCommID(), allNeighbors, data );
+  array1d< array1d< globalIndex > > const nodesRequestedByNeighbors = exchange( allNeighbors, data );
 
   // Then we store the requested nodes for each receiver.
   for( integer i = 0; i < numNeighbors; ++i )
@@ -804,7 +804,7 @@ void CommunicationTools::setupGhosts( MeshLevel & meshLevel,
                                       bool const unorderedComms )
 {
   GEOS_MARK_FUNCTION;
-  MPI_iCommData commData( getCommID() );
+  MPI_iCommData commData;
   commData.resize( neighbors.size() );
 
   NodeManager & nodeManager = meshLevel.getNodeManager();
@@ -1092,7 +1092,7 @@ void CommunicationTools::synchronizeFields( FieldIdentifiers const & fieldsToBeS
                                             std::vector< NeighborCommunicator > & neighbors,
                                             bool onDevice )
 {
-  MPI_iCommData icomm( getCommID() );
+  MPI_iCommData icomm;
   icomm.resize( neighbors.size() );
   synchronizePackSendRecvSizes( fieldsToBeSync, mesh, neighbors, icomm, onDevice );
   synchronizePackSendRecv( fieldsToBeSync, mesh, neighbors, icomm, onDevice );

src/coreComponents/mesh/mpiCommunications/MPI_iCommData.cpp

@@ -14,13 +14,14 @@
  */
 
 #include "MPI_iCommData.hpp"
+#include "CommunicationTools.hpp"
 
 namespace geos
 {
 
-MPI_iCommData::MPI_iCommData( int const inputCommID ):
+MPI_iCommData::MPI_iCommData():
   m_size( 0 ),
-  m_commID( inputCommID ),      // CommunicationTools::getInstance().getCommID() ),
+  m_commID( CommunicationTools::getInstance().getCommID() ),
   m_mpiSendBufferRequest(),
   m_mpiRecvBufferRequest(),
   m_mpiSendBufferStatus(),

src/coreComponents/mesh/mpiCommunications/MPI_iCommData.hpp

@@ -36,7 +36,7 @@ class MPI_iCommData
    * @param inputCommID The CommID integer that indicates what communication
    *   pipeline to use for a set of neighbor communications.
    */
-  MPI_iCommData( int const inputCommID );
+  MPI_iCommData();
 
   /// Default destructor
   ~MPI_iCommData();
@@ -96,7 +96,7 @@ class MPI_iCommData
   int m_size;
 
   /// The integer ID for the set of communication pipelines
-  int m_commID;
+  CommID m_commID;
 
   /// A collection of field names keyed on object keys to pack/unpack from
   /// communication pipeline.

src/coreComponents/physicsSolvers/solidMechanics/SolidMechanicsLagrangianFEM.cpp

@@ -59,7 +59,6 @@
   m_maxForce( 0.0 ),
   m_maxNumResolves( 10 ),
   m_strainTheory( 0 ),
-  m_iComm( CommunicationTools::getInstance().getCommID() ),
   m_isFixedStressPoromechanicsUpdate( false )
 {
 
@@ -571,6 +570,7 @@
     solidMechanics::arrayView2dLayoutIncrDisplacement const & uhat = nodes.getField< solidMechanics::incrementalDisplacement >();
     solidMechanics::arrayView2dLayoutAcceleration const & acc = nodes.getField< solidMechanics::acceleration >();
 
+    MPI_iCommData m_iComm;
     FieldIdentifiers fieldsToBeSync;
     fieldsToBeSync.addFields( FieldLocation::Node,
                               { solidMechanics::velocity::key(),

src/coreComponents/physicsSolvers/solidMechanics/SolidMechanicsLagrangianFEM.hpp

@@ -294,7 +294,7 @@ class SolidMechanicsLagrangianFEM : public SolverBase
   real64 m_maxForce = 0.0;
   integer m_maxNumResolves;
   integer m_strainTheory;
-  MPI_iCommData m_iComm;
+//  MPI_iCommData m_iComm;
   bool m_isFixedStressPoromechanicsUpdate;
 
   /// Rigid body modes

src/coreComponents/physicsSolvers/solidMechanics/SolidMechanicsMPM.cpp

@@ -58,7 +58,7 @@
   SolverBase( name, parent ),
   m_solverProfiling( 0 ),
   m_timeIntegrationOption( TimeIntegrationOption::ExplicitDynamic ),
-  m_iComm( CommunicationTools::getInstance().getCommID() ),
+//  m_iComm( CommunicationTools::getInstance().getCommID() ),
   m_prescribedBcTable( 0 ),
   m_boundaryConditionTypes(),
   m_bcTable(),
@@ -940,6 +940,7 @@
   MeshLevel & mesh = grid.getBaseDiscretization();
   NodeManager & nodeManager = mesh.getNodeManager();
 
+  MPI_iCommData m_iComm;
   m_iComm.resize( domain.getNeighbors().size() );
 
 
@@ -1258,6 +1259,9 @@
   fieldsToBeSynced.addFields( FieldLocation::Node, fieldNames );
   std::vector< NeighborCommunicator > & neighbors = domain.getNeighbors();
 
+  MPI_iCommData m_iComm;
+  m_iComm.resize( neighbors.size() );
+
   // (2) Swap send and receive indices so we can sum from ghost to master
   for( size_t n=0; n<neighbors.size(); n++ )
   {

src/coreComponents/physicsSolvers/solidMechanics/SolidMechanicsMPM.hpp

@@ -377,7 +377,6 @@ class SolidMechanicsMPM : public SolverBase
   std::vector< std::string > m_profilingLabels;
 
   TimeIntegrationOption m_timeIntegrationOption;
-  MPI_iCommData m_iComm;
 
   int m_prescribedBcTable;
   array1d< int > m_boundaryConditionTypes; // TODO: Surely there's a way to have just one variable here

src/coreComponents/physicsSolvers/surfaceGeneration/EmbeddedSurfacesParallelSynchronization.cpp

@@ -383,7 +383,7 @@
   //************************************************************************************************
   // We need to send over the new embedded surfaces and related objects for those whose parents are ghosts on neighbors.
 
-  MPI_iCommData commData( CommunicationTools::getInstance().getCommID() );
+  MPI_iCommData commData;
   commData.resize( neighbors.size());
   for( unsigned int neighborIndex=0; neighborIndex<neighbors.size(); ++neighborIndex )
   {
@@ -458,7 +458,7 @@
   //************************************************************************************************
   // We need to send over the new embedded surfaces and related objects for those whose parents are ghosts on neighbors.
 
-  MPI_iCommData commData( CommunicationTools::getInstance().getCommID() );
+  MPI_iCommData commData;
   commData.resize( neighbors.size());
   for( unsigned int neighborIndex=0; neighborIndex<neighbors.size(); ++neighborIndex )
   {
@@ -528,8 +528,7 @@
                                    std::vector< NeighborCommunicator > & neighbors,
                                    string const fractureRegionName )
 {
-  MPI_iCommData commDataJunk( CommunicationTools::getInstance().getCommID() );
-  MPI_iCommData commData( CommunicationTools::getInstance().getCommID() );
+  MPI_iCommData commData;
   commData.resize( neighbors.size());
   for( unsigned int neighborIndex=0; neighborIndex<neighbors.size(); ++neighborIndex )
   {

src/coreComponents/physicsSolvers/surfaceGeneration/ParallelTopologyChange.cpp

@@ -813,7 +813,7 @@
   //   b) the modified objects to the owning ranks.
 
   // pack the buffers, and send the size of the buffers
-  MPI_iCommData commData( CommunicationTools::getInstance().getCommID() );
+  MPI_iCommData commData;
   commData.resize( neighbors.size() );
   for( unsigned int neighborIndex=0; neighborIndex<neighbors.size(); ++neighborIndex )
   {
@@ -905,7 +905,7 @@
 
 
 
-  MPI_iCommData commData2( CommunicationTools::getInstance().getCommID() );
+  MPI_iCommData commData2;
   commData2.resize( neighbors.size());
   for( unsigned int neighborIndex=0; neighborIndex<neighbors.size(); ++neighborIndex )
   {