portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-16 14:57:05 +01:00
Code Releases Activity

Update to use shared memory to contain the stencil comms buffers

Browse Source 
Tested on 2.1.1.1 1.2.1.1 4.1.1.1 1.4.1.1 2.2.1.1 subnode decompositions
This commit is contained in:
azusayamaguchi
2016-10-24 17:30:43 +01:00
parent ea25a4d9ac
commit b6a65059a2
13 changed files with 706 additions and 458 deletions
Download Patch File Download Diff File Expand all files Collapse all files

132
lib/communicator/Communicator_base.cc Normal file
View File

@ -0,0 +1,132 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./lib/communicator/Communicator_none.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 */
#include "Grid.h"
namespace Grid {
///////////////////////////////////////////////////////////////
// Info that is setup once and indept of cartesian layout
///////////////////////////////////////////////////////////////
int CartesianCommunicator::ShmRank;
int CartesianCommunicator::ShmSize;
int CartesianCommunicator::GroupRank;
int CartesianCommunicator::GroupSize;
int CartesianCommunicator::WorldRank;
int CartesianCommunicator::WorldSize;
int CartesianCommunicator::Slave;
void * CartesianCommunicator::ShmCommBuf;
/////////////////////////////////
// Alloc, free shmem region
/////////////////////////////////
void *CartesianCommunicator::ShmBufferMalloc(size_t bytes){
// bytes = (bytes+sizeof(vRealD))&(~(sizeof(vRealD)-1));// align up bytes
void *ptr = (void *)heap_top;
heap_top += bytes;
heap_bytes+= bytes;
assert(heap_bytes < MAX_MPI_SHM_BYTES);
return ptr;
}
void *CartesianCommunicator::ShmBufferFreeAll(void) {
heap_top =(size_t)ShmBufferSelf();
heap_bytes=0;
}
/////////////////////////////////
// Grid information queries
/////////////////////////////////
int CartesianCommunicator::IsBoss(void) { return _processor==0; };
int CartesianCommunicator::BossRank(void) { return 0; };
int CartesianCommunicator::ThisRank(void) { return _processor; };
const std::vector<int> & CartesianCommunicator::ThisProcessorCoor(void) { return _processor_coor; };
const std::vector<int> & CartesianCommunicator::ProcessorGrid(void) { return _processors; };
int CartesianCommunicator::ProcessorCount(void) { return _Nprocessors; };
////////////////////////////////////////////////////////////////////////////////
// very VERY rarely (Log, serial RNG) we need world without a grid
////////////////////////////////////////////////////////////////////////////////
int CartesianCommunicator::RankWorld(void) { return WorldRank; };
int CartesianCommunicator::Ranks (void) { return WorldSize; };
int CartesianCommunicator::Nodes (void) { return GroupSize; };
int CartesianCommunicator::Cores (void) { return ShmSize; };
int CartesianCommunicator::NodeRank (void) { return GroupRank; };
int CartesianCommunicator::CoreRank (void) { return ShmRank; };
void CartesianCommunicator::GlobalSum(ComplexF &c)
{
GlobalSumVector((float *)&c,2);
}
void CartesianCommunicator::GlobalSumVector(ComplexF *c,int N)
{
GlobalSumVector((float *)c,2*N);
}
void CartesianCommunicator::GlobalSum(ComplexD &c)
{
GlobalSumVector((double *)&c,2);
}
void CartesianCommunicator::GlobalSumVector(ComplexD *c,int N)
{
GlobalSumVector((double *)c,2*N);
}
#ifndef GRID_COMMS_MPI3
void CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
void *xmit,
int xmit_to_rank,
void *recv,
int recv_from_rank,
int bytes)
{
SendToRecvFromBegin(list,xmit,xmit_to_rank,recv,recv_from_rank,bytes);
}
void CartesianCommunicator::StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &waitall)
{
SendToRecvFromComplete(waitall);
}
void StencilBarrier(void){};
commVector<uint8_t> CartesianCommunicator::ShmBufStorageVector;
void *CartesianCommunicator::ShmBufferSelf(void) { return ShmCommBuf; }
void *CartesianCommunicator::ShmBuffer(int rank) {
if (rank != ShmRank ) return NULL;
else return ShmCommBuf;
}
void *CartesianCommunicator::ShmBufferTranslate(int rank,void * local_p) {
if (rank != ShmRank ) return NULL;
else return local_p;
}
void CartesianCommunicator::ShmInitGeneric(void){
ShmBufStorageVector.resize(MAX_MPI_SHM_BYTES);
ShmCommBuf=(void *)&ShmBufStorageVector[0];
}
#endif
}

313
lib/communicator/Communicator_base.h
View File

@ -40,169 +40,188 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
#ifdef GRID_COMMS_SHMEM
#include <mpp/shmem.h>
#endif
namespace Grid {
class CartesianCommunicator {
public:
// 65536 ranks per node adequate for now
// 128MB shared memory for comms enought for 48^4 local vol comms
// Give external control (command line override?) of this
static const int MAXLOG2RANKSPERNODE = 16;
static const uint64_t MAX_MPI_SHM_BYTES = 128*1024*1024;
// Communicator should know nothing of the physics grid, only processor grid.
int _Nprocessors; // How many in all
std::vector<int> _processors; // Which dimensions get relayed out over processors lanes.
int _processor; // linear processor rank
std::vector<int> _processor_coor; // linear processor coordinate
unsigned long _ndimension;
int _Nprocessors; // How many in all
std::vector<int> _processors; // Which dimensions get relayed out over processors lanes.
int _processor; // linear processor rank
std::vector<int> _processor_coor; // linear processor coordinate
unsigned long _ndimension;
#ifdef GRID_COMMS_MPI
MPI_Comm communicator;
typedef MPI_Request CommsRequest_t;
#elif GRID_COMMS_MPI3
int shm_mode;
MPI_Comm communicator;
typedef MPI_Request CommsRequest_t;
const int MAXLOG2RANKSPERNODE = 16; // 65536 ranks per node adequate for now
const uint64_t MAX_MPI_SHM_BYTES = 256*1024*1024; // 256MB shared memory for comms enought for 48^4 local vol comms
std::vector<int> WorldDims;
std::vector<int> GroupDims;
std::vector<int> ShmDims;
std::vector<int> GroupCoor;
std::vector<int> ShmCoor;
std::vector<int> WorldCoor;
static std::vector<int> GroupRanks;
static std::vector<int> MyGroup;
static int ShmSetup;
static MPI_Win ShmWindow;
static MPI_Comm ShmComm;
void * ShmCommBuf;
std::vector<void *> ShmCommBufs;
int WorldRank;
int WorldSize;
static int ShmRank;
static int ShmSize;
static int GroupSize;
static int GroupRank;
std::vector<int> LexicographicToWorldRank;
#if defined (GRID_COMMS_MPI) || defined (GRID_COMMS_MPI3)
MPI_Comm communicator;
static MPI_Comm communicator_world;
typedef MPI_Request CommsRequest_t;
#else
typedef int CommsRequest_t;
typedef int CommsRequest_t;
#endif
static void Init(int *argc, char ***argv);
////////////////////////////////////////////////////////////////////
// Helper functionality for SHM Windows common to all other impls
////////////////////////////////////////////////////////////////////
// Longer term; drop this in favour of a master / slave model with
// cartesian communicator on a subset of ranks, slave ranks controlled
// by group leader with data xfer via shared memory
////////////////////////////////////////////////////////////////////
#ifdef GRID_COMMS_MPI3
std::vector<int> WorldDims;
std::vector<int> GroupDims;
std::vector<int> ShmDims;
std::vector<int> GroupCoor;
std::vector<int> ShmCoor;
std::vector<int> WorldCoor;
static std::vector<int> GroupRanks;
static std::vector<int> MyGroup;
static int ShmSetup;
static MPI_Win ShmWindow;
static MPI_Comm ShmComm;
std::vector<int> LexicographicToWorldRank;
static std::vector<void *> ShmCommBufs;
#else
static void ShmInitGeneric(void);
static commVector<uint8_t> ShmBufStorageVector;
#endif
static void * ShmCommBuf;
size_t heap_top;
size_t heap_bytes;
void *ShmBufferSelf(void);
void *ShmBuffer(int rank);
void *ShmBufferTranslate(int rank,void * local_p);
void *ShmBufferMalloc(size_t bytes);
void *ShmBufferFreeAll(void) ;
////////////////////////////////////////////////
// Must call in Grid startup
////////////////////////////////////////////////
static void Init(int *argc, char ***argv);
////////////////////////////////////////////////
// Constructor of any given grid
////////////////////////////////////////////////
CartesianCommunicator(const std::vector<int> &pdimensions_in);
////////////////////////////////////////////////////////////////////////////////////////
// Wraps MPI_Cart routines, or implements equivalent on other impls
////////////////////////////////////////////////////////////////////////////////////////
void ShiftedRanks(int dim,int shift,int & source, int & dest);
int RankFromProcessorCoor(std::vector<int> &coor);
void ProcessorCoorFromRank(int rank,std::vector<int> &coor);
/////////////////////////////////
// Grid information and queries
/////////////////////////////////
static int ShmRank;
static int ShmSize;
static int GroupSize;
static int GroupRank;
static int WorldRank;
static int WorldSize;
static int Slave;
int IsBoss(void) ;
int BossRank(void) ;
int ThisRank(void) ;
const std::vector<int> & ThisProcessorCoor(void) ;
const std::vector<int> & ProcessorGrid(void) ;
int ProcessorCount(void) ;
static int Ranks (void);
static int Nodes (void);
static int Cores (void);
static int NodeRank (void);
static int CoreRank (void);
// Constructor
CartesianCommunicator(const std::vector<int> &pdimensions_in);
////////////////////////////////////////////////////////////////////////////////
// very VERY rarely (Log, serial RNG) we need world without a grid
////////////////////////////////////////////////////////////////////////////////
static int RankWorld(void) ;
static void BroadcastWorld(int root,void* data, int bytes);
////////////////////////////////////////////////////////////
// Reduction
////////////////////////////////////////////////////////////
void GlobalSum(RealF &);
void GlobalSumVector(RealF *,int N);
void GlobalSum(RealD &);
void GlobalSumVector(RealD *,int N);
void GlobalSum(uint32_t &);
void GlobalSum(uint64_t &);
void GlobalSum(ComplexF &c);
void GlobalSumVector(ComplexF *c,int N);
void GlobalSum(ComplexD &c);
void GlobalSumVector(ComplexD *c,int N);
template<class obj> void GlobalSum(obj &o){
typedef typename obj::scalar_type scalar_type;
int words = sizeof(obj)/sizeof(scalar_type);
scalar_type * ptr = (scalar_type *)& o;
GlobalSumVector(ptr,words);
}
////////////////////////////////////////////////////////////
// Face exchange, buffer swap in translational invariant way
////////////////////////////////////////////////////////////
void SendToRecvFrom(void *xmit,
int xmit_to_rank,
void *recv,
int recv_from_rank,
int bytes);
void SendRecvPacket(void *xmit,
void *recv,
int xmit_to_rank,
int recv_from_rank,
int bytes);
void SendToRecvFromBegin(std::vector<CommsRequest_t> &list,
void *xmit,
int xmit_to_rank,
void *recv,
int recv_from_rank,
int bytes);
void SendToRecvFromComplete(std::vector<CommsRequest_t> &waitall);
// Wraps MPI_Cart routines
void ShiftedRanks(int dim,int shift,int & source, int & dest);
int RankFromProcessorCoor(std::vector<int> &coor);
void ProcessorCoorFromRank(int rank,std::vector<int> &coor);
void StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
void *xmit,
int xmit_to_rank,
void *recv,
int recv_from_rank,
int bytes);
void StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &waitall);
void StencilBarrier(void);
// Helper function for SHM Windows in MPI3
void *ShmBufferSelf(void);
void *ShmBuffer(int rank);
/////////////////////////////////
// Grid information queries
/////////////////////////////////
int IsBoss(void) { return _processor==0; };
int BossRank(void) { return 0; };
int ThisRank(void) { return _processor; };
const std::vector<int> & ThisProcessorCoor(void) { return _processor_coor; };
const std::vector<int> & ProcessorGrid(void) { return _processors; };
int ProcessorCount(void) { return _Nprocessors; };
////////////////////////////////////////////////////////////
// Reduction
////////////////////////////////////////////////////////////
void GlobalSum(RealF &);
void GlobalSumVector(RealF *,int N);
void GlobalSum(RealD &);
void GlobalSumVector(RealD *,int N);
void GlobalSum(uint32_t &);
void GlobalSum(uint64_t &);
void GlobalSum(ComplexF &c)
{
GlobalSumVector((float *)&c,2);
}
void GlobalSumVector(ComplexF *c,int N)
{
GlobalSumVector((float *)c,2*N);
}
void GlobalSum(ComplexD &c)
{
GlobalSumVector((double *)&c,2);
}
void GlobalSumVector(ComplexD *c,int N)
{
GlobalSumVector((double *)c,2*N);
}
template<class obj> void GlobalSum(obj &o){
typedef typename obj::scalar_type scalar_type;
int words = sizeof(obj)/sizeof(scalar_type);
scalar_type * ptr = (scalar_type *)& o;
GlobalSumVector(ptr,words);
}
////////////////////////////////////////////////////////////
// Face exchange, buffer swap in translational invariant way
////////////////////////////////////////////////////////////
void SendToRecvFrom(void *xmit,
int xmit_to_rank,
void *recv,
int recv_from_rank,
int bytes);
void SendRecvPacket(void *xmit,
void *recv,
int xmit_to_rank,
int recv_from_rank,
int bytes);
void SendToRecvFromBegin(std::vector<CommsRequest_t> &list,
void *xmit,
int xmit_to_rank,
void *recv,
int recv_from_rank,
int bytes);
void SendToRecvFromComplete(std::vector<CommsRequest_t> &waitall);
void StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
void *xmit,
int xmit_to_rank,
void *recv,
int recv_from_rank,
int bytes);
void StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &waitall)
{
SendToRecvFromComplete(waitall);
}
////////////////////////////////////////////////////////////
// Barrier
////////////////////////////////////////////////////////////
void Barrier(void);
////////////////////////////////////////////////////////////
// Broadcast a buffer and composite larger
////////////////////////////////////////////////////////////
void Broadcast(int root,void* data, int bytes);
template<class obj> void Broadcast(int root,obj &data)
////////////////////////////////////////////////////////////
// Barrier
////////////////////////////////////////////////////////////
void Barrier(void);
////////////////////////////////////////////////////////////
// Broadcast a buffer and composite larger
////////////////////////////////////////////////////////////
void Broadcast(int root,void* data, int bytes);
template<class obj> void Broadcast(int root,obj &data)
{
Broadcast(root,(void *)&data,sizeof(data));
};
static void BroadcastWorld(int root,void* data, int bytes);
};
}

41
lib/communicator/Communicator_mpi.cc
View File

@ -30,19 +30,28 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
namespace Grid {
// Should error check all MPI calls.
///////////////////////////////////////////////////////////////////////////////////////////////////
// Info that is setup once and indept of cartesian layout
///////////////////////////////////////////////////////////////////////////////////////////////////
MPI_Comm CartesianCommunicator::communicator_world;
// Should error check all MPI calls.
void CartesianCommunicator::Init(int *argc, char ***argv) {
int flag;
MPI_Initialized(&flag); // needed to coexist with other libs apparently
if ( !flag ) {
MPI_Init(argc,argv);
}
}
int Rank(void) {
int pe;
MPI_Comm_rank(MPI_COMM_WORLD,&pe);
return pe;
MPI_Comm_dup (MPI_COMM_WORLD,&communicator_world);
MPI_Comm_rank(communicator_world,&WorldRank);
MPI_Comm_size(communicator_world,&WorldSize);
ShmRank=0;
ShmSize=1;
GroupRank=WorldRank;
GroupSize=WorldSize;
Slave =0;
ShmInitGeneric();
}
CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
@ -54,7 +63,7 @@ CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
_processors = processors;
_processor_coor.resize(_ndimension);
MPI_Cart_create(MPI_COMM_WORLD, _ndimension,&_processors[0],&periodic[0],1,&communicator);
MPI_Cart_create(communicator_world, _ndimension,&_processors[0],&periodic[0],1,&communicator);
MPI_Comm_rank(communicator,&_processor);
MPI_Cart_coords(communicator,_processor,_ndimension,&_processor_coor[0]);
@ -67,15 +76,6 @@ CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
assert(Size==_Nprocessors);
}
void *CartesianCommunicator::ShmBufferSelf(void)
{
return NULL;
}
void *CartesianCommunicator::ShmBuffer(int rank)
{
return NULL;
}
void CartesianCommunicator::GlobalSum(uint32_t &u){
int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT32_T,MPI_SUM,communicator);
assert(ierr==0);
@ -194,14 +194,17 @@ void CartesianCommunicator::Broadcast(int root,void* data, int bytes)
communicator);
assert(ierr==0);
}
///////////////////////////////////////////////////////
// Should only be used prior to Grid Init finished.
// Check for this?
///////////////////////////////////////////////////////
void CartesianCommunicator::BroadcastWorld(int root,void* data, int bytes)
{
int ierr= MPI_Bcast(data,
bytes,
MPI_BYTE,
root,
MPI_COMM_WORLD);
communicator_world);
assert(ierr==0);
}

388
lib/communicator/Communicator_mpi3.cc
View File

@ -1,4 +1,3 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
@ -33,26 +32,197 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
namespace Grid {
///////////////////////////////////////////////////////////////////////////////////////////////////
// Info that is setup once and indept of cartesian layout
///////////////////////////////////////////////////////////////////////////////////////////////////
int CartesianCommunicator::ShmSetup = 0;
// Global used by Init and nowhere else. How to hide?
int Rank(void) {
int pe;
MPI_Comm_rank(MPI_COMM_WORLD,&pe);
return pe;
MPI_Comm CartesianCommunicator::communicator_world;
MPI_Comm CartesianCommunicator::ShmComm;
MPI_Win CartesianCommunicator::ShmWindow;
std::vector<int> CartesianCommunicator::GroupRanks;
std::vector<int> CartesianCommunicator::MyGroup;
std::vector<void *> CartesianCommunicator::ShmCommBufs;
void *CartesianCommunicator::ShmBufferSelf(void)
{
return ShmCommBufs[ShmRank];
}
// Should error check all MPI calls.
void *CartesianCommunicator::ShmBuffer(int rank)
{
int gpeer = GroupRanks[rank];
if (gpeer == MPI_UNDEFINED){
return NULL;
} else {
return ShmCommBufs[gpeer];
}
}
void *CartesianCommunicator::ShmBufferTranslate(int rank,void * local_p)
{
int gpeer = GroupRanks[rank];
if (gpeer == MPI_UNDEFINED){
return NULL;
} else {
uint64_t offset = (uint64_t)local_p - (uint64_t)ShmCommBufs[ShmRank];
uint64_t remote = (uint64_t)ShmCommBufs[gpeer]+offset;
return (void *) remote;
}
}
void CartesianCommunicator::Init(int *argc, char ***argv) {
int flag;
MPI_Initialized(&flag); // needed to coexist with other libs apparently
if ( !flag ) {
MPI_Init(argc,argv);
}
}
////////////////////////////////////////////////////////////////////////////////////////////////////////////
// Want to implement some magic ... Group sub-cubes into those on same node
//
////////////////////////////////////////////////////////////////////////////////////////////////////////////
MPI_Comm_dup (MPI_COMM_WORLD,&communicator_world);
MPI_Comm_rank(communicator_world,&WorldRank);
MPI_Comm_size(communicator_world,&WorldSize);
/////////////////////////////////////////////////////////////////////
// Split into groups that can share memory
/////////////////////////////////////////////////////////////////////
MPI_Comm_split_type(communicator_world, MPI_COMM_TYPE_SHARED, 0, MPI_INFO_NULL,&ShmComm);
MPI_Comm_rank(ShmComm ,&ShmRank);
MPI_Comm_size(ShmComm ,&ShmSize);
GroupSize = WorldSize/ShmSize;
/////////////////////////////////////////////////////////////////////
// find world ranks in our SHM group (i.e. which ranks are on our node)
/////////////////////////////////////////////////////////////////////
MPI_Group WorldGroup, ShmGroup;
MPI_Comm_group (communicator_world, &WorldGroup);
MPI_Comm_group (ShmComm, &ShmGroup);
std::vector<int> world_ranks(WorldSize);
GroupRanks.resize(WorldSize);
MyGroup.resize(ShmSize);
for(int r=0;r<WorldSize;r++) world_ranks[r]=r;
MPI_Group_translate_ranks (WorldGroup,WorldSize,&world_ranks[0],ShmGroup, &GroupRanks[0]);
///////////////////////////////////////////////////////////////////
// Identify who is in my group and noninate the leader
///////////////////////////////////////////////////////////////////
int g=0;
for(int rank=0;rank<WorldSize;rank++){
if(GroupRanks[rank]!=MPI_UNDEFINED){
assert(g<ShmSize);
MyGroup[g++] = rank;
}
}
std::sort(MyGroup.begin(),MyGroup.end(),std::less<int>());
int myleader = MyGroup[0];
std::vector<int> leaders_1hot(WorldSize,0);
std::vector<int> leaders_group(GroupSize,0);
leaders_1hot [ myleader ] = 1;
///////////////////////////////////////////////////////////////////
// global sum leaders over comm world
///////////////////////////////////////////////////////////////////
int ierr=MPI_Allreduce(MPI_IN_PLACE,&leaders_1hot[0],WorldSize,MPI_INT,MPI_SUM,communicator_world);
assert(ierr==0);
///////////////////////////////////////////////////////////////////
// find the group leaders world rank
///////////////////////////////////////////////////////////////////
int group=0;
for(int l=0;l<WorldSize;l++){
if(leaders_1hot[l]){
leaders_group[group++] = l;
}
}
///////////////////////////////////////////////////////////////////
// Identify the rank of the group in which I (and my leader) live
///////////////////////////////////////////////////////////////////
GroupRank=-1;
for(int g=0;g<GroupSize;g++){
if (myleader == leaders_group[g]){
GroupRank=g;
}
}
assert(GroupRank!=-1);
//////////////////////////////////////////////////////////////////////////////////////////////////////////
// allocate the shared window for our group
//////////////////////////////////////////////////////////////////////////////////////////////////////////
ShmCommBuf = 0;
ierr = MPI_Win_allocate_shared(MAX_MPI_SHM_BYTES,1,MPI_INFO_NULL,ShmComm,&ShmCommBuf,&ShmWindow);
assert(ierr==0);
// KNL hack -- force to numa-domain 1 in flat
#if 0
for(uint64_t page=0;page<MAX_MPI_SHM_BYTES;page+=4096){
void *pages = (void *) ( page + ShmCommBuf );
int status;
int flags=MPOL_MF_MOVE_ALL;
int nodes=1; // numa domain == MCDRAM
unsigned long count=1;
ierr= move_pages(0,count, &pages,&nodes,&status,flags);
if (ierr && (page==0)) perror("numa relocate command failed");
}
#endif
MPI_Win_lock_all (MPI_MODE_NOCHECK, ShmWindow);
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
// Plan: allocate a fixed SHM region. Scratch that is just used via some scheme during stencil comms, with no allocate free.
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
ShmCommBufs.resize(ShmSize);
for(int r=0;r<ShmSize;r++){
MPI_Aint sz;
int dsp_unit;
MPI_Win_shared_query (ShmWindow, r, &sz, &dsp_unit, &ShmCommBufs[r]);
}
//////////////////////////////////////////////////////////////////////////////////////////////////////////
// Verbose for now
//////////////////////////////////////////////////////////////////////////////////////////////////////////
if (WorldRank == 0){
std::cout<<GridLogMessage<< "Grid MPI-3 configuration: detected ";
std::cout<< WorldSize << " Ranks " ;
std::cout<< GroupSize << " Nodes " ;
std::cout<< ShmSize << " with ranks-per-node "<<std::endl;
std::cout<<GridLogMessage <<"Grid MPI-3 configuration: allocated shared memory region of size ";
std::cout<<std::hex << MAX_MPI_SHM_BYTES <<" ShmCommBuf address = "<<ShmCommBuf << std::dec<<std::endl;
for(int g=0;g<GroupSize;g++){
std::cout<<GridLogMessage<<" Node "<<g<<" led by MPI rank "<<leaders_group[g]<<std::endl;
}
std::cout<<GridLogMessage<<" Boss Node Shm Pointers are {";
for(int g=0;g<ShmSize;g++){
std::cout<<std::hex<<ShmCommBufs[g]<<std::dec;
if(g!=ShmSize-1) std::cout<<",";
else std::cout<<"}"<<std::endl;
}
}
for(int g=0;g<GroupSize;g++){
if ( (ShmRank == 0) && (GroupRank==g) ) std::cout<<GridLogMessage<<"["<<g<<"] Node Group "<<g<<" is ranks {";
for(int r=0;r<ShmSize;r++){
if ( (ShmRank == 0) && (GroupRank==g) ) {
std::cout<<MyGroup[r];
if(r<ShmSize-1) std::cout<<",";
else std::cout<<"}"<<std::endl;
}
MPI_Barrier(communicator_world);
}
}
assert(ShmSetup==0); ShmSetup=1;
}
////////////////////////////////////////////////////////////////////////////////////////////////////////////
// Want to implement some magic ... Group sub-cubes into those on same node
////////////////////////////////////////////////////////////////////////////////////////////////////////////
void CartesianCommunicator::ShiftedRanks(int dim,int shift,int &source,int &dest)
{
std::vector<int> coor = _processor_coor;
@ -80,139 +250,13 @@ void CartesianCommunicator::ProcessorCoorFromRank(int rank, std::vector<int> &c
rank = LexicographicToWorldRank[rank];
}
///////////////////////////////////////////////////////////////////////////////////////////////////
// Info that is setup once and indept of cartesian layout
///////////////////////////////////////////////////////////////////////////////////////////////////
int CartesianCommunicator::ShmSetup = 0;
int CartesianCommunicator::ShmRank;
int CartesianCommunicator::ShmSize;
int CartesianCommunicator::GroupRank;
int CartesianCommunicator::GroupSize;
MPI_Comm CartesianCommunicator::ShmComm;
MPI_Win CartesianCommunicator::ShmWindow;
std::vector<int> CartesianCommunicator::GroupRanks;
std::vector<int> CartesianCommunicator::MyGroup;
CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
{
_ndimension = processors.size();
WorldDims = processors;
communicator = MPI_COMM_WORLD;
MPI_Comm_rank(communicator,&WorldRank);
MPI_Comm_size(communicator,&WorldSize);
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
// Plan: allocate a fixed SHM region. Scratch that is just used via some scheme during stencil comms, with no allocate free.
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
// Does every grid need one, or could we share across all grids via a singleton/guard?
int ierr;
if ( !ShmSetup ) {
communicator=communicator_world;
MPI_Comm_split_type(communicator, MPI_COMM_TYPE_SHARED, 0, MPI_INFO_NULL,&ShmComm);
MPI_Comm_rank(ShmComm ,&ShmRank);
MPI_Comm_size(ShmComm ,&ShmSize);
GroupSize = WorldSize/ShmSize;
/////////////////////////////////////////////////////////////////////
// find world ranks in our SHM group (i.e. which ranks are on our node)
/////////////////////////////////////////////////////////////////////
MPI_Group WorldGroup, ShmGroup;
MPI_Comm_group (communicator, &WorldGroup);
MPI_Comm_group (ShmComm, &ShmGroup);
std::vector<int> world_ranks(WorldSize);
GroupRanks.resize(WorldSize);
MyGroup.resize(ShmSize);
for(int r=0;r<WorldSize;r++) world_ranks[r]=r;
MPI_Group_translate_ranks (WorldGroup,WorldSize,&world_ranks[0],ShmGroup, &GroupRanks[0]);
///////////////////////////////////////////////////////////////////
// Identify who is in my group and noninate the leader
///////////////////////////////////////////////////////////////////
int g=0;
for(int rank=0;rank<WorldSize;rank++){
if(GroupRanks[rank]!=MPI_UNDEFINED){
assert(g<ShmSize);
MyGroup[g++] = rank;
}
}
std::sort(MyGroup.begin(),MyGroup.end(),std::greater<int>());
int myleader = MyGroup[0];
std::vector<int> leaders_1hot(WorldSize,0);
std::vector<int> leaders_group(GroupSize,0);
leaders_1hot [ myleader ] = 1;
///////////////////////////////////////////////////////////////////
// global sum leaders over comm world
///////////////////////////////////////////////////////////////////
ierr=MPI_Allreduce(MPI_IN_PLACE,&leaders_1hot[0],WorldSize,MPI_INT,MPI_SUM,communicator);
assert(ierr==0);
///////////////////////////////////////////////////////////////////
// find the group leaders world rank
///////////////////////////////////////////////////////////////////
int group=0;
for(int l=0;l<WorldSize;l++){
if(leaders_1hot[l]){
leaders_group[group++] = l;
}
}
///////////////////////////////////////////////////////////////////
// Identify the rank of the group in which I (and my leader) live
///////////////////////////////////////////////////////////////////
GroupRank=-1;
for(int g=0;g<GroupSize;g++){
if (myleader == leaders_group[g]){
GroupRank=g;
}
}
assert(GroupRank!=-1);
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
// allocate the shared window for our group
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
ShmCommBuf = 0;
ierr = MPI_Win_allocate_shared(MAX_MPI_SHM_BYTES,1,MPI_INFO_NULL,ShmComm,&ShmCommBuf,&ShmWindow);
assert(ierr==0);
for(uint64_t page=0;page<MAX_MPI_SHM_BYTES;page+=4096){
void *pages = (void *) ( page + ShmCommBuf );
int status;
int flags=MPOL_MF_MOVE_ALL;
int nodes=1; // numa domain == MCDRAM
unsigned long count=1;
ierr= move_pages(0,count, &pages,&nodes,&status,flags);
if (ierr && (page==0)) perror("numa relocate command failed");
}
MPI_Win_lock_all (MPI_MODE_NOCHECK, ShmWindow);
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
// Verbose for now
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
std::cout<<GridLogMessage<< "MPI-3 configuration: Ranks per node "<< ShmSize ;
std::cout<< " Nodes "<< GroupSize;
std::cout<< " Ranks "<< WorldSize;
std::cout<< " Shm CommBuf address"<< std::hex <<ShmCommBuf << std::dec<<std::endl;
// Done
ShmSetup=1;
}
ShmCommBufs.resize(ShmSize);
for(int r=0;r<ShmSize;r++){
MPI_Aint sz;
int dsp_unit;
MPI_Win_shared_query (ShmWindow, r, &sz, &dsp_unit, &ShmCommBufs[r]);
}
_ndimension = processors.size();
////////////////////////////////////////////////////////////////
// Assert power of two shm_size.
@ -232,6 +276,8 @@ CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
////////////////////////////////////////////////////////////////
int dim = 0;
std::vector<int> WorldDims = processors;
ShmDims.resize(_ndimension,1);
GroupDims.resize(_ndimension);
@ -346,21 +392,6 @@ void CartesianCommunicator::SendRecvPacket(void *xmit,
}
}
void *CartesianCommunicator::ShmBufferSelf(void)
{
return ShmCommBufs[ShmRank];
}
void *CartesianCommunicator::ShmBuffer(int rank)
{
int gpeer = GroupRanks[rank];
if (gpeer == MPI_UNDEFINED){
return NULL;
} else {
return ShmCommBufs[gpeer];
}
}
// Basic Halo comms primitive
void CartesianCommunicator::SendToRecvFromBegin(std::vector<CommsRequest_t> &list,
void *xmit,
@ -369,6 +400,7 @@ void CartesianCommunicator::SendToRecvFromBegin(std::vector<CommsRequest_t> &lis
int from,
int bytes)
{
#if 1
MPI_Request xrq;
MPI_Request rrq;
@ -387,12 +419,11 @@ void CartesianCommunicator::SendToRecvFromBegin(std::vector<CommsRequest_t> &lis
sequence++;
char *to_ptr = (char *)ShmCommBufs[gdest];
char *from_ptr = (char *)ShmCommBufs[ShmRank];
int small = (bytes<MAX_MPI_SHM_BYTES);
typedef vRealD T;
typedef uint64_t T;
int words = bytes/sizeof(T);
assert(((size_t)bytes &(sizeof(T)-1))==0);
@ -400,13 +431,18 @@ void CartesianCommunicator::SendToRecvFromBegin(std::vector<CommsRequest_t> &lis
if ( small && (gdest !=MPI_UNDEFINED) ) {
char *to_ptr = (char *)ShmCommBufs[gdest];
assert(gme != gdest);
T *ip = (T *)xmit;
T *op = (T *)to_ptr;
PARALLEL_FOR_LOOP
for(int w=0;w<words;w++) {
vstream(op[w],ip[w]);
op[w]=ip[w];
if ( w == 0 ) {
// std::cout << " xmit "<< ShmRank <<" -> "<< gdest<<" " <<std::hex<<op[w]<<std::dec<<std::endl;
}
}
bcopy(&_processor,&to_ptr[bytes],sizeof(_processor));
@ -426,7 +462,10 @@ PARALLEL_FOR_LOOP
T *op = (T *)recv;
PARALLEL_FOR_LOOP
for(int w=0;w<words;w++) {
vstream(op[w],ip[w]);
op[w]=ip[w];
if ( w == 0 ) {
// std::cout << " recv "<< ShmRank <<" <- "<< gfrom<<" " <<std::hex<<op[w]<<std::dec<<std::endl;
}
}
bcopy(&from_ptr[bytes] ,&tag ,sizeof(tag));
bcopy(&from_ptr[bytes+4],&check,sizeof(check));
@ -441,6 +480,19 @@ PARALLEL_FOR_LOOP
MPI_Win_sync (ShmWindow);
MPI_Barrier (ShmComm);
MPI_Win_sync (ShmWindow);
#else
MPI_Request xrq;
MPI_Request rrq;
int rank = _processor;
int ierr;
ierr =MPI_Isend(xmit, bytes, MPI_CHAR,dest,_processor,communicator,&xrq);
ierr|=MPI_Irecv(recv, bytes, MPI_CHAR,from,from,communicator,&rrq);
assert(ierr==0);
list.push_back(xrq);
list.push_back(rrq);
#endif
}
void CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
@ -476,19 +528,29 @@ void CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsRequest_
list.push_back(rrq);
}
StencilBarrier();
}
void CartesianCommunicator::StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &list)
{
SendToRecvFromComplete(list);
}
void CartesianCommunicator::StencilBarrier(void)
{
MPI_Win_sync (ShmWindow);
MPI_Barrier (ShmComm);
MPI_Win_sync (ShmWindow);
}
void CartesianCommunicator::SendToRecvFromComplete(std::vector<CommsRequest_t> &list)
{
int nreq=list.size();
std::vector<MPI_Status> status(nreq);
int ierr = MPI_Waitall(nreq,&list[0],&status[0]);
assert(ierr==0);
}
@ -514,7 +576,7 @@ void CartesianCommunicator::BroadcastWorld(int root,void* data, int bytes)
bytes,
MPI_BYTE,
root,
MPI_COMM_WORLD);
communicator_world);
assert(ierr==0);
}

55
lib/communicator/Communicator_none.cc
View File

@ -28,18 +28,29 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
#include "Grid.h"
namespace Grid {
///////////////////////////////////////////////////////////////////////////////////////////////////
// Info that is setup once and indept of cartesian layout
///////////////////////////////////////////////////////////////////////////////////////////////////
int CartesianCommunicator::ShmRank;
int CartesianCommunicator::ShmSize;
int CartesianCommunicator::GroupRank;
int CartesianCommunicator::GroupSize;
int CartesianCommunicator::WorldRank;
int CartesianCommunicator::WorldSize;
int CartesianCommunicator::Slave;
void * CartesianCommunicator::ShmCommBuf;
commVector<uint8_t> CartesianCommunicator::ShmBufStorageVector;
void CartesianCommunicator::Init(int *argc, char *** arv)
{
}
int Rank(void ){ return 0; };
void *CartesianCommunicator::ShmBufferSelf(void)
{
return NULL;
}
void *CartesianCommunicator::ShmBuffer(int rank)
{
return NULL;
WorldRank = 0;
WorldSize = 1;
ShmRank=0;
ShmSize=1;
GroupRank=_WorldRank;
GroupSize=_WorldSize;
Slave =0;
ShmInitGeneric();
}
CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
@ -97,30 +108,16 @@ void CartesianCommunicator::SendToRecvFromComplete(std::vector<CommsRequest_t> &
assert(0);
}
void CartesianCommunicator::Barrier(void)
{
}
void CartesianCommunicator::Broadcast(int root,void* data, int bytes)
{
}
void CartesianCommunicator::BroadcastWorld(int root,void* data, int bytes)
{
}
void CartesianCommunicator::Barrier(void){}
void CartesianCommunicator::Broadcast(int root,void* data, int bytes) {}
void CartesianCommunicator::BroadcastWorld(int root,void* data, int bytes) { }
int CartesianCommunicator::RankFromProcessorCoor(std::vector<int> &coor) { return 0;}
void CartesianCommunicator::ProcessorCoorFromRank(int rank, std::vector<int> &coor){ assert(0);}
void CartesianCommunicator::ShiftedRanks(int dim,int shift,int &source,int &dest)
{
source =0;
dest=0;
}
int CartesianCommunicator::RankFromProcessorCoor(std::vector<int> &coor)
{
return 0;
}
void CartesianCommunicator::ProcessorCoorFromRank(int rank, std::vector<int> &coor)
{
}
}

57
lib/communicator/Communicator_shmem.cc
View File

@ -39,25 +39,19 @@ namespace Grid {
BACKTRACEFILE(); \
}\
}
int Rank(void) {
return shmem_my_pe();
}
typedef struct HandShake_t {
uint64_t seq_local;
uint64_t seq_remote;
} HandShake;
static Vector< HandShake > XConnections;
static Vector< HandShake > RConnections;
void *CartesianCommunicator::ShmBufferSelf(void)
{
return NULL;
}
void *CartesianCommunicator::ShmBuffer(int rank)
{
return NULL;
}
///////////////////////////////////////////////////////////////////////////////////////////////////
// Info that is setup once and indept of cartesian layout
///////////////////////////////////////////////////////////////////////////////////////////////////
int CartesianCommunicator::ShmRank;
int CartesianCommunicator::ShmSize;
int CartesianCommunicator::GroupRank;
int CartesianCommunicator::GroupSize;
int CartesianCommunicator::WorldRank;
int CartesianCommunicator::WorldSize;
int CartesianCommunicator::Slave;
void CartesianCommunicator::Init(int *argc, char ***argv) {
shmem_init();
XConnections.resize(shmem_n_pes());
@ -69,7 +63,36 @@ void CartesianCommunicator::Init(int *argc, char ***argv) {
RConnections[pe].seq_remote= 0;
}
shmem_barrier_all();
ShmInitGeneric();
}
// Should error check all MPI calls.
void CartesianCommunicator::Init(int *argc, char ***argv) {
int flag;
MPI_Initialized(&flag); // needed to coexist with other libs apparently
if ( !flag ) {
MPI_Init(argc,argv);
MPI_Comm_dup (MPI_COMM_WORLD,&communicator_world);
MPI_Comm_rank(communicator_world,&_WorldRank);
MPI_Comm_size(communicator_world,&_WorldSize);
_ShmRank=0;
_ShmSize=1;
_GroupRank=_WorldRank;
_GroupSize=_WorldSize;
_Slave =0;
}
}
typedef struct HandShake_t {
uint64_t seq_local;
uint64_t seq_remote;
} HandShake;
static Vector< HandShake > XConnections;
static Vector< HandShake > RConnections;
CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
{
_ndimension = processors.size();