olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask Class Reference

Wrapper for a non-blocking block propagation receive request. More...

#include <communicator.hh>

Collaboration diagram for olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask:

Classes
class	ref
	Manual replacement for std::reference_wrapper<RecvTask> More...

Public Member Functions
	RecvTask (MPI_Comm comm, int tag, int rank, const std::vector< std::type_index > &fields, const std::vector< CellID > &cells, ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > &block)
	~RecvTask ()
bool	operator< (const RecvTask &rhs) const
void	receive ()
bool	isDone ()
void	unpack ()
void	wait ()

Detailed Description

template<typename T, typename DESCRIPTOR>
class olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask

Wrapper for a non-blocking block propagation receive request.

Definition at line 415 of file communicator.hh.

Constructor & Destructor Documentation

RecvTask()

template<typename T , typename DESCRIPTOR >

olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask::RecvTask ( MPI_Comm comm, int tag, int rank, const std::vector< std::type_index > & fields, const std::vector< CellID > & cells, ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > & block )

inline

Definition at line 458 of file communicator.hh.

                                                                      :
    _tag(tag),
    _rank(rank),
    _fields(block.getDataRegistry().deviceFieldArrays(fields)),
    _onlyPopulationField(fields.size() == 1 && fields[0] == typeid(descriptors::POPULATION)),
    _cells(cells),
    _target(block),
    _stream(std::make_unique<gpu::cuda::device::Stream>(cudaStreamNonBlocking))
  {
    std::size_t size = 0;
    for (auto& field : fields) {
      size += _target.getCommunicatable(field).size(cells);
    }
    _buffer = gpu::cuda::device::malloc<std::uint8_t>(size);
    _request = std::make_unique<MpiRecvRequest>(
      _buffer.get(), size, _rank, _tag, comm);
  }

References olb::gpu::cuda::device::unique_ptr< T >::get(), olb::ConcreteBlockLattice< T, DESCRIPTOR, PLATFORM >::getCommunicatable(), and olb::Communicatable::size().

Here is the call graph for this function:

~RecvTask()

template<typename T , typename DESCRIPTOR >

olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask::~RecvTask ( )

inline

Definition at line 479 of file communicator.hh.

  {
    wait();
  }

References olb::ConcreteBlockCommunicator< BLOCK >::wait().

Here is the call graph for this function:

Member Function Documentation

isDone()

template<typename T , typename DESCRIPTOR >

bool olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask::isDone ( )

inline

Definition at line 495 of file communicator.hh.

  {
    return _request->isDone();
  }

operator<()

template<typename T , typename DESCRIPTOR >

bool olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask::operator< ( const RecvTask & rhs ) const

inline

Definition at line 484 of file communicator.hh.

  {
    return  _rank  < rhs._rank
        || (_rank == rhs._rank && _tag < rhs._tag);
  }

receive()

template<typename T , typename DESCRIPTOR >

void olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask::receive ( )

inline

Definition at line 490 of file communicator.hh.

  {
    _request->start();
  };

unpack()

template<typename T , typename DESCRIPTOR >

void olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask::unpack ( )

inline

Definition at line 500 of file communicator.hh.

  {
    if (_onlyPopulationField) {
      gpu::cuda::DeviceContext<T,DESCRIPTOR> lattice(_target);
      gpu::cuda::async_scatter_field<descriptors::POPULATION>(_stream->get(), lattice, _cells, _buffer.get());
    } else {
      gpu::cuda::async_scatter_any_fields(_stream->get(), _fields, _cells, _buffer.get());
    }
  }

References olb::gpu::cuda::async_scatter_any_fields(), and olb::gpu::cuda::device::unique_ptr< T >::get().

Here is the call graph for this function:

wait()

template<typename T , typename DESCRIPTOR >

void olb::ConcreteBlockCommunicator< ConcreteBlockLattice< T, DESCRIPTOR, Platform::GPU_CUDA > >::RecvTask::wait ( )

inline

Definition at line 510 of file communicator.hh.

  {
    _stream->synchronize();
  }

---

The documentation for this class was generated from the following file:

• src/core/platform/gpu/cuda/communicator.hh