MPI.cpp File Reference

Implementations of the MPI wrapper functions declared in MPI.hpp: retry-aware Bcast/ Alltoall/Alltoallv/ Allreduce/Allgather/ Barrier variants, lazy waits, singleton definitions, CUDA-aware probe. More...

#include <ctime>
#include <cstdio>
#include <cstdlib>
#include <iostream>
#include <chrono>
#include <thread>
#include "MPI.hpp"
#include "Profiling.hpp"

Include dependency graph for MPI.cpp:

Go to the source code of this file.

Namespaces
namespace	DNDS
	the host side operators are provided as implemented
namespace	DNDS::Debug
namespace	DNDS::MPI

Macros
#define	__start_timer   PerformanceTimer::Instance().StartTimer(PerformanceTimer::Comm)
#define	__stop_timer   PerformanceTimer::Instance().StopTimer(PerformanceTimer::Comm)

Functions
bool	DNDS::Debug::IsDebugged ()
	Whether the current process is running under a debugger. Implemented via /proc/self/status TracerPid on Linux.
void	DNDS::Debug::MPIDebugHold (const MPIInfo &mpi)
	If isDebugging is set, block every rank in a busy-wait loop so the user can attach a debugger and inspect state. Exit by setting isDebugging = false in the debugger.
void	DNDS::assert_false_info_mpi (const char *expr, const char *file, int line, const std::string &info, const DNDS::MPIInfo &mpi)
	MPI-aware assertion-failure reporter.
std::string	DNDS::getTimeStamp (const MPIInfo &mpi)
	Format a human-readable timestamp using the calling rank as context.
MPI_int	DNDS::MPI::Bcast (void *buf, MPI_int num, MPI_Datatype type, MPI_int source_rank, MPI_Comm comm)
	dumb wrapper
MPI_int	DNDS::MPI::Alltoall (void *send, MPI_int sendNum, MPI_Datatype typeSend, void *recv, MPI_int recvNum, MPI_Datatype typeRecv, MPI_Comm comm)
	Wrapper over MPI_Alltoall (fixed per-peer count).
MPI_int	DNDS::MPI::Alltoallv (void *send, MPI_int *sendSizes, MPI_int *sendStarts, MPI_Datatype sendType, void *recv, MPI_int *recvSizes, MPI_int *recvStarts, MPI_Datatype recvType, MPI_Comm comm)
	Wrapper over MPI_Alltoallv (variable per-peer counts + displacements).
MPI_int	DNDS::MPI::Allreduce (const void *sendbuf, void *recvbuf, MPI_int count, MPI_Datatype datatype, MPI_Op op, MPI_Comm comm)
	Wrapper over MPI_Allreduce.
MPI_int	DNDS::MPI::Scan (const void *sendbuf, void *recvbuf, MPI_int count, MPI_Datatype datatype, MPI_Op op, MPI_Comm comm)
	Wrapper over MPI_Scan (inclusive prefix reduction).
MPI_int	DNDS::MPI::Allgather (const void *sendbuf, MPI_int sendcount, MPI_Datatype sendtype, void *recvbuf, MPI_int recvcount, MPI_Datatype recvtype, MPI_Comm comm)
	Wrapper over MPI_Allgather.
MPI_int	DNDS::MPI::Barrier (MPI_Comm comm)
	Wrapper over MPI_Barrier.
MPI_int	DNDS::MPI::BarrierLazy (MPI_Comm comm, uint64_t checkNanoSecs)
	Polling barrier that sleeps checkNanoSecs ns between MPI_Test calls. Reduces CPU spin when many ranks wait unevenly.
MPI_int	DNDS::MPI::WaitallLazy (MPI_int count, MPI_Request *reqs, MPI_Status *statuses, uint64_t checkNanoSecs=10000000)
	Like WaitallAuto but sleeps checkNanoSecs ns between polls.
MPI_int	DNDS::MPI::WaitallAuto (MPI_int count, MPI_Request *reqs, MPI_Status *statuses)
	Wait on an array of requests, choosing between MPI_Waitall and the lazy-poll variant based on CommStrategy settings.
bool	DNDS::MPI::isCudaAware ()
	Runtime probe: is the current MPI implementation configured with CUDA-aware support? Affects whether arrays are transferred on-device or via the host round-trip.

Variables
bool	DNDS::Debug::isDebugging = false
	Flag consulted by MPIDebugHold and assert_false_info_mpi.
std::mutex	DNDS::HDF_mutex
	Global mutex serialising host-side HDF5 calls.

Detailed Description

Implementations of the MPI wrapper functions declared in MPI.hpp: retry-aware Bcast/ Alltoall/Alltoallv/ Allreduce/Allgather/ Barrier variants, lazy waits, singleton definitions, CUDA-aware probe.

Definition in file MPI.cpp.

Macro Definition Documentation

__start_timer

#define __start_timer   PerformanceTimer::Instance().StartTimer(PerformanceTimer::Comm)

Definition at line 147 of file MPI.cpp.

__stop_timer

#define __stop_timer   PerformanceTimer::Instance().StopTimer(PerformanceTimer::Comm)

Definition at line 148 of file MPI.cpp.