1
0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-15 02:05:37 +00:00
Grid/lib/communicator/Communicator_mpi3.cc

515 lines
17 KiB
C++
Raw Normal View History

2017-02-07 06:30:02 +00:00
/*************************************************************************************
2016-10-20 17:00:15 +01:00
Grid physics library, www.github.com/paboyle/Grid
Source file: ./lib/communicator/Communicator_mpi.cc
Copyright (C) 2015
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
2017-02-22 18:09:33 +00:00
#include <Grid/GridCore.h>
2018-01-08 11:33:01 +00:00
#include <Grid/communicator/SharedMemory.h>
2017-02-07 06:30:02 +00:00
2016-10-20 17:00:15 +01:00
namespace Grid {
2018-01-08 11:33:01 +00:00
Grid_MPI_Comm CartesianCommunicator::communicator_world;
2018-01-08 11:33:01 +00:00
////////////////////////////////////////////
// First initialise of comms system
////////////////////////////////////////////
void CartesianCommunicator::Init(int *argc, char ***argv)
{
2017-02-07 06:30:02 +00:00
2016-10-20 17:00:15 +01:00
int flag;
2017-02-07 06:30:02 +00:00
int provided;
2016-10-20 17:00:15 +01:00
MPI_Initialized(&flag); // needed to coexist with other libs apparently
if ( !flag ) {
2017-02-07 06:30:02 +00:00
MPI_Init_thread(argc,argv,MPI_THREAD_MULTIPLE,&provided);
//If only 1 comms thread we require any threading mode other than SINGLE, but for multiple comms threads we need MULTIPLE
if( (nCommThreads == 1 && provided == MPI_THREAD_SINGLE) ||
(nCommThreads > 1 && provided != MPI_THREAD_MULTIPLE) )
assert(0);
2016-10-20 17:00:15 +01:00
}
2017-02-07 06:30:02 +00:00
Grid_quiesce_nodes();
2018-02-12 13:27:20 +00:00
// Never clean up as done once.
MPI_Comm_dup (MPI_COMM_WORLD,&communicator_world);
2018-01-08 11:33:01 +00:00
GlobalSharedMemory::Init(communicator_world);
GlobalSharedMemory::SharedMemoryAllocate(
GlobalSharedMemory::MAX_MPI_SHM_BYTES,
GlobalSharedMemory::Hugepages);
2016-10-20 17:00:15 +01:00
}
2018-01-08 11:33:01 +00:00
///////////////////////////////////////////////////////////////////////////
// Use cartesian communicators now even in MPI3
///////////////////////////////////////////////////////////////////////////
void CartesianCommunicator::ShiftedRanks(int dim,int shift,int &source,int &dest)
2016-10-20 17:00:15 +01:00
{
2018-01-08 11:33:01 +00:00
int ierr=MPI_Cart_shift(communicator,dim,shift,&source,&dest);
assert(ierr==0);
}
2016-10-20 17:00:15 +01:00
int CartesianCommunicator::RankFromProcessorCoor(std::vector<int> &coor)
{
int rank;
2018-01-08 11:33:01 +00:00
int ierr=MPI_Cart_rank (communicator, &coor[0], &rank);
assert(ierr==0);
2016-10-20 17:00:15 +01:00
return rank;
2018-01-08 11:33:01 +00:00
}
2016-10-20 17:00:15 +01:00
void CartesianCommunicator::ProcessorCoorFromRank(int rank, std::vector<int> &coor)
{
2018-01-08 11:33:01 +00:00
coor.resize(_ndimension);
int ierr=MPI_Cart_coords (communicator, rank, _ndimension,&coor[0]);
assert(ierr==0);
}
////////////////////////////////////////////////////////////////////////////////////////////////////////
// Initialises from communicator_world
////////////////////////////////////////////////////////////////////////////////////////////////////////
CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
{
MPI_Comm optimal_comm;
////////////////////////////////////////////////////
// Remap using the shared memory optimising routine
// The remap creates a comm which must be freed
////////////////////////////////////////////////////
GlobalSharedMemory::OptimalCommunicator (processors,optimal_comm);
2018-01-08 11:33:01 +00:00
InitFromMPICommunicator(processors,optimal_comm);
SetCommunicator(optimal_comm);
///////////////////////////////////////////////////
2018-02-12 13:27:20 +00:00
// Free the temp communicator
///////////////////////////////////////////////////
2018-02-12 13:27:20 +00:00
MPI_Comm_free(&optimal_comm);
2016-10-20 17:00:15 +01:00
}
2017-06-22 08:14:34 +01:00
//////////////////////////////////
// Try to subdivide communicator
//////////////////////////////////
2018-01-08 11:33:01 +00:00
CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors,const CartesianCommunicator &parent,int &srank)
2017-06-22 08:14:34 +01:00
{
_ndimension = processors.size();
2017-02-07 06:30:02 +00:00
2018-01-08 11:33:01 +00:00
int parent_ndimension = parent._ndimension; assert(_ndimension >= parent._ndimension);
std::vector<int> parent_processor_coor(_ndimension,0);
std::vector<int> parent_processors (_ndimension,1);
2018-01-08 11:33:01 +00:00
// Can make 5d grid from 4d etc...
int pad = _ndimension-parent_ndimension;
for(int d=0;d<parent_ndimension;d++){
parent_processor_coor[pad+d]=parent._processor_coor[d];
parent_processors [pad+d]=parent._processors[d];
2016-10-20 17:00:15 +01:00
}
2018-01-08 11:33:01 +00:00
//////////////////////////////////////////////////////////////////////////////////////////////////////
// split the communicator
//////////////////////////////////////////////////////////////////////////////////////////////////////
// int Nparent = parent._processors ;
// std::cout << " splitting from communicator "<<parent.communicator <<std::endl;
int Nparent;
MPI_Comm_size(parent.communicator,&Nparent);
// std::cout << " Parent size "<<Nparent <<std::endl;
int childsize=1;
for(int d=0;d<processors.size();d++) {
childsize *= processors[d];
2016-10-20 17:00:15 +01:00
}
2018-01-08 11:33:01 +00:00
int Nchild = Nparent/childsize;
assert (childsize * Nchild == Nparent);
// std::cout << " child size "<<childsize <<std::endl;
std::vector<int> ccoor(_ndimension); // coor within subcommunicator
std::vector<int> scoor(_ndimension); // coor of split within parent
std::vector<int> ssize(_ndimension); // coor of split within parent
2016-10-20 17:00:15 +01:00
2017-02-07 06:30:02 +00:00
for(int d=0;d<_ndimension;d++){
2018-01-08 11:33:01 +00:00
ccoor[d] = parent_processor_coor[d] % processors[d];
scoor[d] = parent_processor_coor[d] / processors[d];
ssize[d] = parent_processors[d] / processors[d];
2017-02-07 06:30:02 +00:00
}
2018-01-08 11:33:01 +00:00
// rank within subcomm ; srank is rank of subcomm within blocks of subcomms
int crank;
// Mpi uses the reverse Lexico convention to us; so reversed routines called
Lexicographic::IndexFromCoorReversed(ccoor,crank,processors); // processors is the split grid dimensions
Lexicographic::IndexFromCoorReversed(scoor,srank,ssize); // ssize is the number of split grids
MPI_Comm comm_split;
if ( Nchild > 1 ) {
if(0){
std::cout << GridLogMessage<<"Child communicator of "<< std::hex << parent.communicator << std::dec<<std::endl;
std::cout << GridLogMessage<<" parent grid["<< parent._ndimension<<"] ";
for(int d=0;d<parent._ndimension;d++) std::cout << parent._processors[d] << " ";
std::cout<<std::endl;
std::cout << GridLogMessage<<" child grid["<< _ndimension <<"] ";
for(int d=0;d<processors.size();d++) std::cout << processors[d] << " ";
std::cout<<std::endl;
std::cout << GridLogMessage<<" old rank "<< parent._processor<<" coor ["<< parent._ndimension <<"] ";
for(int d=0;d<parent._ndimension;d++) std::cout << parent._processor_coor[d] << " ";
std::cout<<std::endl;
std::cout << GridLogMessage<<" new split "<< srank<<" scoor ["<< _ndimension <<"] ";
for(int d=0;d<processors.size();d++) std::cout << scoor[d] << " ";
std::cout<<std::endl;
std::cout << GridLogMessage<<" new rank "<< crank<<" coor ["<< _ndimension <<"] ";
for(int d=0;d<processors.size();d++) std::cout << ccoor[d] << " ";
std::cout<<std::endl;
//////////////////////////////////////////////////////////////////////////////////////////////////////
// Declare victory
//////////////////////////////////////////////////////////////////////////////////////////////////////
std::cout << GridLogMessage<<"Divided communicator "<< parent._Nprocessors<<" into "
<< Nchild <<" communicators with " << childsize << " ranks"<<std::endl;
std::cout << " Split communicator " <<comm_split <<std::endl;
}
////////////////////////////////////////////////////////////////
// Split the communicator
////////////////////////////////////////////////////////////////
int ierr= MPI_Comm_split(parent.communicator,srank,crank,&comm_split);
assert(ierr==0);
} else {
srank = 0;
2018-02-12 13:27:20 +00:00
int ierr = MPI_Comm_dup (parent.communicator,&comm_split);
assert(ierr==0);
2017-02-07 06:30:02 +00:00
}
2018-01-08 11:33:01 +00:00
//////////////////////////////////////////////////////////////////////////////////////////////////////
// Set up from the new split communicator
//////////////////////////////////////////////////////////////////////////////////////////////////////
InitFromMPICommunicator(processors,comm_split);
//////////////////////////////////////////////////////////////////////////////////////////////////////
// Take the right SHM buffers
//////////////////////////////////////////////////////////////////////////////////////////////////////
SetCommunicator(comm_split);
///////////////////////////////////////////////
2018-02-12 13:27:20 +00:00
// Free the temp communicator
///////////////////////////////////////////////
2018-02-12 13:27:20 +00:00
MPI_Comm_free(&comm_split);
2018-01-08 11:33:01 +00:00
if(0){
std::cout << " ndim " <<_ndimension<<" " << parent._ndimension << std::endl;
for(int d=0;d<processors.size();d++){
std::cout << d<< " " << _processor_coor[d] <<" " << ccoor[d]<<std::endl;
}
}
for(int d=0;d<processors.size();d++){
assert(_processor_coor[d] == ccoor[d] );
2017-02-07 06:30:02 +00:00
}
2018-01-08 11:33:01 +00:00
}
void CartesianCommunicator::InitFromMPICommunicator(const std::vector<int> &processors, MPI_Comm communicator_base)
{
2018-02-12 13:27:20 +00:00
////////////////////////////////////////////////////
// Creates communicator, and the communicator_halo
////////////////////////////////////////////////////
2018-01-08 11:33:01 +00:00
_ndimension = processors.size();
_processor_coor.resize(_ndimension);
/////////////////////////////////
// Count the requested nodes
/////////////////////////////////
2016-10-20 17:00:15 +01:00
_Nprocessors=1;
_processors = processors;
for(int i=0;i<_ndimension;i++){
_Nprocessors*=_processors[i];
}
2018-01-08 11:33:01 +00:00
std::vector<int> periodic(_ndimension,1);
MPI_Cart_create(communicator_base, _ndimension,&_processors[0],&periodic[0],0,&communicator);
MPI_Comm_rank(communicator,&_processor);
MPI_Cart_coords(communicator,_processor,_ndimension,&_processor_coor[0]);
2017-02-07 06:30:02 +00:00
2018-01-08 11:33:01 +00:00
if ( 0 && (communicator_base != communicator_world) ) {
std::cout << "InitFromMPICommunicator Cartesian communicator created with a non-world communicator"<<std::endl;
std::cout << " new communicator rank "<<_processor<< " coor ["<<_ndimension<<"] ";
for(int d=0;d<_processors.size();d++){
std::cout << _processor_coor[d]<<" ";
}
std::cout << std::endl;
}
2017-02-07 06:30:02 +00:00
2018-01-08 11:33:01 +00:00
int Size;
MPI_Comm_size(communicator,&Size);
2017-02-07 06:30:02 +00:00
2018-01-08 11:33:01 +00:00
communicator_halo.resize (2*_ndimension);
for(int i=0;i<_ndimension*2;i++){
MPI_Comm_dup(communicator,&communicator_halo[i]);
2017-02-07 06:30:02 +00:00
}
2018-01-08 11:33:01 +00:00
assert(Size==_Nprocessors);
}
CartesianCommunicator::~CartesianCommunicator()
{
int MPI_is_finalised;
MPI_Finalized(&MPI_is_finalised);
if (communicator && !MPI_is_finalised) {
MPI_Comm_free(&communicator);
2017-11-19 01:39:04 +00:00
for(int i=0;i<communicator_halo.size();i++){
MPI_Comm_free(&communicator_halo[i]);
}
}
}
2016-10-20 17:00:15 +01:00
void CartesianCommunicator::GlobalSum(uint32_t &u){
int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT32_T,MPI_SUM,communicator);
assert(ierr==0);
}
void CartesianCommunicator::GlobalSum(uint64_t &u){
int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT64_T,MPI_SUM,communicator);
assert(ierr==0);
}
2017-06-11 23:14:10 +01:00
void CartesianCommunicator::GlobalXOR(uint32_t &u){
int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT32_T,MPI_BXOR,communicator);
assert(ierr==0);
}
void CartesianCommunicator::GlobalXOR(uint64_t &u){
int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT64_T,MPI_BXOR,communicator);
assert(ierr==0);
}
2016-10-20 17:00:15 +01:00
void CartesianCommunicator::GlobalSum(float &f){
int ierr=MPI_Allreduce(MPI_IN_PLACE,&f,1,MPI_FLOAT,MPI_SUM,communicator);
assert(ierr==0);
}
void CartesianCommunicator::GlobalSumVector(float *f,int N)
{
int ierr=MPI_Allreduce(MPI_IN_PLACE,f,N,MPI_FLOAT,MPI_SUM,communicator);
assert(ierr==0);
}
void CartesianCommunicator::GlobalSum(double &d)
{
int ierr = MPI_Allreduce(MPI_IN_PLACE,&d,1,MPI_DOUBLE,MPI_SUM,communicator);
assert(ierr==0);
}
void CartesianCommunicator::GlobalSumVector(double *d,int N)
{
int ierr = MPI_Allreduce(MPI_IN_PLACE,d,N,MPI_DOUBLE,MPI_SUM,communicator);
assert(ierr==0);
}
// Basic Halo comms primitive
void CartesianCommunicator::SendToRecvFrom(void *xmit,
int dest,
void *recv,
int from,
int bytes)
{
std::vector<CommsRequest_t> reqs(0);
2017-02-07 06:30:02 +00:00
// unsigned long xcrc = crc32(0L, Z_NULL, 0);
// unsigned long rcrc = crc32(0L, Z_NULL, 0);
// xcrc = crc32(xcrc,(unsigned char *)xmit,bytes);
2016-10-20 17:00:15 +01:00
SendToRecvFromBegin(reqs,xmit,dest,recv,from,bytes);
SendToRecvFromComplete(reqs);
2017-02-07 06:30:02 +00:00
// rcrc = crc32(rcrc,(unsigned char *)recv,bytes);
// printf("proc %d SendToRecvFrom %d bytes %lx %lx\n",_processor,bytes,xcrc,rcrc);
2016-10-20 17:00:15 +01:00
}
void CartesianCommunicator::SendRecvPacket(void *xmit,
void *recv,
int sender,
int receiver,
int bytes)
{
MPI_Status stat;
assert(sender != receiver);
int tag = sender;
if ( _processor == sender ) {
MPI_Send(xmit, bytes, MPI_CHAR,receiver,tag,communicator);
}
if ( _processor == receiver ) {
MPI_Recv(recv, bytes, MPI_CHAR,sender,tag,communicator,&stat);
}
}
// Basic Halo comms primitive
void CartesianCommunicator::SendToRecvFromBegin(std::vector<CommsRequest_t> &list,
void *xmit,
int dest,
void *recv,
int from,
int bytes)
{
2017-02-07 06:30:02 +00:00
int myrank = _processor;
2016-10-20 17:00:15 +01:00
int ierr;
2016-10-21 12:07:28 +01:00
if ( CommunicatorPolicy == CommunicatorPolicyConcurrent ) {
2017-02-07 06:30:02 +00:00
MPI_Request xrq;
MPI_Request rrq;
2017-02-07 06:30:02 +00:00
ierr =MPI_Irecv(recv, bytes, MPI_CHAR,from,from,communicator,&rrq);
ierr|=MPI_Isend(xmit, bytes, MPI_CHAR,dest,_processor,communicator,&xrq);
assert(ierr==0);
list.push_back(xrq);
2017-02-07 06:30:02 +00:00
list.push_back(rrq);
} else {
2017-02-07 06:30:02 +00:00
// Give the CPU to MPI immediately; can use threads to overlap optionally
ierr=MPI_Sendrecv(xmit,bytes,MPI_CHAR,dest,myrank,
recv,bytes,MPI_CHAR,from, from,
communicator,MPI_STATUS_IGNORE);
assert(ierr==0);
}
}
2016-10-21 13:35:37 +01:00
2017-08-19 18:18:50 +01:00
double CartesianCommunicator::StencilSendToRecvFrom( void *xmit,
int dest,
void *recv,
int from,
int bytes,int dir)
{
std::vector<CommsRequest_t> list;
2017-08-20 01:10:50 +01:00
double offbytes = StencilSendToRecvFromBegin(list,xmit,dest,recv,from,bytes,dir);
2017-08-19 18:18:50 +01:00
StencilSendToRecvFromComplete(list,dir);
2017-08-20 01:10:50 +01:00
return offbytes;
2017-08-19 18:18:50 +01:00
}
2017-02-20 22:47:40 +00:00
double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
void *xmit,
int dest,
void *recv,
int from,
int bytes,int dir)
{
int ncomm =communicator_halo.size();
int commdir=dir%ncomm;
2016-10-21 13:35:37 +01:00
MPI_Request xrq;
MPI_Request rrq;
2016-10-21 13:35:37 +01:00
int ierr;
2018-01-08 11:33:01 +00:00
int gdest = ShmRanks[dest];
int gfrom = ShmRanks[from];
int gme = ShmRanks[_processor];
2017-02-07 06:30:02 +00:00
assert(dest != _processor);
assert(from != _processor);
assert(gme == ShmRank);
2017-02-20 22:47:40 +00:00
double off_node_bytes=0.0;
if ( gfrom ==MPI_UNDEFINED) {
ierr=MPI_Irecv(recv, bytes, MPI_CHAR,from,from,communicator_halo[commdir],&rrq);
assert(ierr==0);
list.push_back(rrq);
2017-02-20 22:47:40 +00:00
off_node_bytes+=bytes;
}
2016-10-21 13:35:37 +01:00
2017-02-07 06:30:02 +00:00
if ( gdest == MPI_UNDEFINED ) {
ierr =MPI_Isend(xmit, bytes, MPI_CHAR,dest,_processor,communicator_halo[commdir],&xrq);
2017-02-07 06:30:02 +00:00
assert(ierr==0);
list.push_back(xrq);
2017-02-20 22:47:40 +00:00
off_node_bytes+=bytes;
2017-02-07 06:30:02 +00:00
}
if ( CommunicatorPolicy == CommunicatorPolicySequential ) {
2017-08-20 01:10:50 +01:00
this->StencilSendToRecvFromComplete(list,dir);
}
2017-02-20 22:47:40 +00:00
return off_node_bytes;
}
void CartesianCommunicator::StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &waitall,int dir)
{
2017-02-07 06:30:02 +00:00
SendToRecvFromComplete(waitall);
}
void CartesianCommunicator::StencilBarrier(void)
{
MPI_Barrier (ShmComm);
2016-10-20 17:00:15 +01:00
}
void CartesianCommunicator::SendToRecvFromComplete(std::vector<CommsRequest_t> &list)
{
int nreq=list.size();
2017-02-07 06:30:02 +00:00
if (nreq==0) return;
2016-10-20 17:00:15 +01:00
std::vector<MPI_Status> status(nreq);
int ierr = MPI_Waitall(nreq,&list[0],&status[0]);
assert(ierr==0);
list.resize(0);
2016-10-20 17:00:15 +01:00
}
void CartesianCommunicator::Barrier(void)
{
int ierr = MPI_Barrier(communicator);
assert(ierr==0);
}
void CartesianCommunicator::Broadcast(int root,void* data, int bytes)
{
int ierr=MPI_Bcast(data,
bytes,
MPI_BYTE,
root,
communicator);
assert(ierr==0);
}
2017-02-07 06:30:02 +00:00
int CartesianCommunicator::RankWorld(void){
int r;
MPI_Comm_rank(communicator_world,&r);
return r;
}
2016-10-20 17:00:15 +01:00
void CartesianCommunicator::BroadcastWorld(int root,void* data, int bytes)
{
int ierr= MPI_Bcast(data,
bytes,
MPI_BYTE,
root,
communicator_world);
2016-10-20 17:00:15 +01:00
assert(ierr==0);
}
2018-01-08 11:33:01 +00:00
void CartesianCommunicator::AllToAll(int dim,void *in,void *out,uint64_t words,uint64_t bytes)
{
std::vector<int> row(_ndimension,1);
assert(dim>=0 && dim<_ndimension);
// Split the communicator
row[dim] = _processors[dim];
int me;
CartesianCommunicator Comm(row,*this,me);
Comm.AllToAll(in,out,words,bytes);
}
void CartesianCommunicator::AllToAll(void *in,void *out,uint64_t words,uint64_t bytes)
{
// MPI is a pain and uses "int" arguments
// 64*64*64*128*16 == 500Million elements of data.
// When 24*4 bytes multiples get 50x 10^9 >>> 2x10^9 Y2K bug.
// (Turns up on 32^3 x 64 Gparity too)
MPI_Datatype object;
int iwords;
int ibytes;
iwords = words;
ibytes = bytes;
assert(words == iwords); // safe to cast to int ?
assert(bytes == ibytes); // safe to cast to int ?
MPI_Type_contiguous(ibytes,MPI_BYTE,&object);
MPI_Type_commit(&object);
MPI_Alltoall(in,iwords,object,out,iwords,object,communicator);
MPI_Type_free(&object);
}
2016-10-20 17:00:15 +01:00
}