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

449 lines
15 KiB
C++
Raw Normal View History

/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./benchmarks/Benchmark_comms.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 */
2016-07-07 22:31:07 +01:00
#include <Grid/Grid.h>
2015-05-02 23:42:30 +01:00
using namespace std;
using namespace Grid;
struct time_statistics{
double mean;
double err;
double min;
double max;
void statistics(std::vector<double> v){
double sum = std::accumulate(v.begin(), v.end(), 0.0);
mean = sum / v.size();
std::vector<double> diff(v.size());
std::transform(v.begin(), v.end(), diff.begin(), [=](double x) { return x - mean; });
double sq_sum = std::inner_product(diff.begin(), diff.end(), diff.begin(), 0.0);
err = std::sqrt(sq_sum / (v.size()*(v.size() - 1)));
auto result = std::minmax_element(v.begin(), v.end());
min = *result.first;
max = *result.second;
}
};
void header(){
std::cout <<GridLogMessage << " L "<<"\t"<<" Ls "<<"\t"
<<std::setw(11)<<"bytes"<<"MB/s uni (err/min/max)"<<"\t\t"<<"MB/s bidi (err/min/max)"<<std::endl;
};
2015-05-02 23:42:30 +01:00
int main (int argc, char ** argv)
{
Grid_init(&argc,&argv);
Coordinate simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
Coordinate mpi_layout = GridDefaultMpi();
int threads = GridThread::GetThreads();
std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
2015-05-02 23:42:30 +01:00
2017-05-29 21:47:32 +01:00
int Nloop=100;
2015-05-02 23:51:43 +01:00
int nmu=0;
int maxlat=32;
2016-11-01 11:35:43 +00:00
for(int mu=0;mu<Nd;mu++) if (mpi_layout[mu]>1) nmu++;
std::cout << GridLogMessage << "Number of iterations to average: "<< Nloop << std::endl;
std::vector<double> t_time(Nloop);
time_statistics timestat;
2015-05-02 23:51:43 +01:00
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << "= Benchmarking sequential halo exchange in "<<nmu<<" dimensions"<<std::endl;
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
header();
2015-05-02 23:51:43 +01:00
2019-07-20 07:13:28 +01:00
for(int lat=8;lat<=maxlat;lat+=4){
for(int Ls=8;Ls<=8;Ls*=2){
2015-05-02 23:51:43 +01:00
2019-07-20 07:13:28 +01:00
Coordinate latt_size ({lat*mpi_layout[0],
lat*mpi_layout[1],
lat*mpi_layout[2],
lat*mpi_layout[3]});
2015-05-03 09:44:47 +01:00
GridCartesian Grid(latt_size,simd_layout,mpi_layout);
2017-05-29 21:47:32 +01:00
RealD Nrank = Grid._Nprocessors;
RealD Nnode = Grid.NodeCount();
RealD ppn = Nrank/Nnode;
2015-05-03 09:44:47 +01:00
2020-11-12 19:29:58 +00:00
std::vector<std::vector<HalfSpinColourVectorD> > xbuf(8);
std::vector<std::vector<HalfSpinColourVectorD> > rbuf(8);
2015-05-03 09:44:47 +01:00
2017-08-25 09:25:54 +01:00
for(int mu=0;mu<8;mu++){
xbuf[mu].resize(lat*lat*lat*Ls);
rbuf[mu].resize(lat*lat*lat*Ls);
// std::cout << " buffers " << std::hex << (uint64_t)&xbuf[mu][0] <<" " << (uint64_t)&rbuf[mu][0] <<std::endl;
}
2019-07-30 22:51:04 +01:00
uint64_t bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
2015-05-03 09:44:47 +01:00
int ncomm;
2015-05-02 23:51:43 +01:00
for(int i=0;i<Nloop;i++){
double start=usecond();
2015-05-02 23:42:30 +01:00
2015-05-02 23:51:43 +01:00
ncomm=0;
for(int mu=0;mu<4;mu++){
if (mpi_layout[mu]>1 ) {
ncomm++;
int comm_proc=1;
int xmit_to_rank;
int recv_from_rank;
{
2018-01-08 11:27:14 +00:00
std::vector<CommsRequest_t> requests;
2015-05-02 23:51:43 +01:00
Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
Grid.SendToRecvFrom((void *)&xbuf[mu][0],
xmit_to_rank,
(void *)&rbuf[mu][0],
recv_from_rank,
bytes);
2015-05-02 23:51:43 +01:00
}
comm_proc = mpi_layout[mu]-1;
{
2018-01-08 11:27:14 +00:00
std::vector<CommsRequest_t> requests;
2015-05-02 23:51:43 +01:00
Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
Grid.SendToRecvFrom((void *)&xbuf[mu+4][0],
xmit_to_rank,
(void *)&rbuf[mu+4][0],
recv_from_rank,
bytes);
2015-05-02 23:51:43 +01:00
}
}
}
Grid.Barrier();
2017-05-29 21:47:32 +01:00
double stop=usecond();
t_time[i] = stop-start; // microseconds
2015-05-02 23:51:43 +01:00
}
timestat.statistics(t_time);
2015-05-29 14:11:34 +01:00
2017-05-29 21:47:32 +01:00
double dbytes = bytes*ppn;
double xbytes = dbytes*2.0*ncomm;
2015-05-02 23:51:43 +01:00
double rbytes = xbytes;
double bidibytes = xbytes+rbytes;
std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
<<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
<<std::right<< xbytes/timestat.mean<<" "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
<<xbytes/timestat.max <<" "<< xbytes/timestat.min
<< "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< " " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
<< bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
2015-05-02 23:51:43 +01:00
}
}
2016-11-09 12:11:03 +00:00
2016-11-01 11:35:43 +00:00
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << "= Benchmarking concurrent STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
header();
2016-11-01 11:35:43 +00:00
2019-07-20 07:13:28 +01:00
for(int lat=8;lat<=maxlat;lat+=4){
for(int Ls=8;Ls<=8;Ls*=2){
2016-11-01 11:35:43 +00:00
Coordinate latt_size ({lat*mpi_layout[0],
2019-07-20 07:13:28 +01:00
lat*mpi_layout[1],
lat*mpi_layout[2],
lat*mpi_layout[3]});
2016-11-01 11:35:43 +00:00
GridCartesian Grid(latt_size,simd_layout,mpi_layout);
2017-05-29 21:47:32 +01:00
RealD Nrank = Grid._Nprocessors;
RealD Nnode = Grid.NodeCount();
RealD ppn = Nrank/Nnode;
2016-11-01 11:35:43 +00:00
std::vector<HalfSpinColourVectorD *> xbuf(8);
std::vector<HalfSpinColourVectorD *> rbuf(8);
Grid.ShmBufferFreeAll();
2019-07-30 22:51:04 +01:00
uint64_t bytes = lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
2016-11-01 11:35:43 +00:00
for(int d=0;d<8;d++){
2019-07-30 22:51:04 +01:00
xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
2016-11-01 11:35:43 +00:00
}
int ncomm;
2017-05-29 21:47:32 +01:00
double dbytes;
2016-11-01 11:35:43 +00:00
for(int i=0;i<Nloop;i++){
2017-05-29 21:47:32 +01:00
double start=usecond();
dbytes=0;
ncomm=0;
2016-11-01 11:35:43 +00:00
2018-01-08 11:27:14 +00:00
std::vector<CommsRequest_t> requests;
2016-11-01 11:35:43 +00:00
for(int mu=0;mu<4;mu++){
2017-05-29 21:47:32 +01:00
2016-11-01 11:35:43 +00:00
if (mpi_layout[mu]>1 ) {
ncomm++;
int comm_proc=1;
int xmit_to_rank;
int recv_from_rank;
Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
2017-05-29 21:47:32 +01:00
dbytes+=
Grid.StencilSendToRecvFromBegin(requests,
(void *)&xbuf[mu][0],
xmit_to_rank,1,
2017-05-29 21:47:32 +01:00
(void *)&rbuf[mu][0],
recv_from_rank,1,
2022-11-15 21:37:15 +00:00
bytes,bytes,mu);
2016-11-01 11:35:43 +00:00
comm_proc = mpi_layout[mu]-1;
Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
2017-05-29 21:47:32 +01:00
dbytes+=
Grid.StencilSendToRecvFromBegin(requests,
(void *)&xbuf[mu+4][0],
xmit_to_rank,1,
2017-05-29 21:47:32 +01:00
(void *)&rbuf[mu+4][0],
recv_from_rank,1,
2022-11-15 21:37:15 +00:00
bytes,bytes,mu+4);
2016-11-01 11:35:43 +00:00
}
}
Grid.StencilSendToRecvFromComplete(requests,0);
2016-11-01 11:35:43 +00:00
Grid.Barrier();
2017-05-29 21:47:32 +01:00
double stop=usecond();
t_time[i] = stop-start; // microseconds
2016-11-01 11:35:43 +00:00
}
timestat.statistics(t_time);
2017-05-29 21:47:32 +01:00
dbytes=dbytes*ppn;
double xbytes = dbytes*0.5;
2018-01-25 23:46:47 +00:00
// double rbytes = dbytes*0.5;
2017-05-29 21:47:32 +01:00
double bidibytes = dbytes;
2016-11-01 11:35:43 +00:00
std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
<<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
<<std::right<< xbytes/timestat.mean<<" "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
<<xbytes/timestat.max <<" "<< xbytes/timestat.min
<< "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< " " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
<< bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
2016-11-01 11:35:43 +00:00
}
}
2016-11-09 12:11:03 +00:00
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << "= Benchmarking sequential STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
header();
2016-11-09 12:11:03 +00:00
2019-07-20 07:13:28 +01:00
for(int lat=8;lat<=maxlat;lat+=4){
for(int Ls=8;Ls<=8;Ls*=2){
2016-11-09 12:11:03 +00:00
Coordinate latt_size ({lat*mpi_layout[0],
2019-07-20 07:13:28 +01:00
lat*mpi_layout[1],
lat*mpi_layout[2],
lat*mpi_layout[3]});
2016-11-09 12:11:03 +00:00
GridCartesian Grid(latt_size,simd_layout,mpi_layout);
2017-05-29 21:47:32 +01:00
RealD Nrank = Grid._Nprocessors;
RealD Nnode = Grid.NodeCount();
RealD ppn = Nrank/Nnode;
2016-11-09 12:11:03 +00:00
std::vector<HalfSpinColourVectorD *> xbuf(8);
std::vector<HalfSpinColourVectorD *> rbuf(8);
Grid.ShmBufferFreeAll();
2019-07-30 22:51:04 +01:00
uint64_t bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
2016-11-09 12:11:03 +00:00
for(int d=0;d<8;d++){
2019-07-30 22:51:04 +01:00
xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
2016-11-09 12:11:03 +00:00
}
int ncomm;
2017-05-29 21:47:32 +01:00
double dbytes;
2016-11-09 12:11:03 +00:00
for(int i=0;i<Nloop;i++){
2017-05-29 21:47:32 +01:00
double start=usecond();
2016-11-09 12:11:03 +00:00
2018-01-08 11:27:14 +00:00
std::vector<CommsRequest_t> requests;
2017-05-29 21:47:32 +01:00
dbytes=0;
2016-11-09 12:11:03 +00:00
ncomm=0;
for(int mu=0;mu<4;mu++){
if (mpi_layout[mu]>1 ) {
ncomm++;
int comm_proc=1;
int xmit_to_rank;
int recv_from_rank;
Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
2017-05-29 21:47:32 +01:00
dbytes+=
Grid.StencilSendToRecvFromBegin(requests,
(void *)&xbuf[mu][0],
xmit_to_rank,1,
2017-05-29 21:47:32 +01:00
(void *)&rbuf[mu][0],
recv_from_rank,1,
2022-11-15 21:37:15 +00:00
bytes,bytes,mu);
Grid.StencilSendToRecvFromComplete(requests,mu);
2017-02-07 06:07:39 +00:00
requests.resize(0);
2016-11-09 12:11:03 +00:00
comm_proc = mpi_layout[mu]-1;
Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
2017-05-29 21:47:32 +01:00
dbytes+=
Grid.StencilSendToRecvFromBegin(requests,
(void *)&xbuf[mu+4][0],
xmit_to_rank,1,
2017-05-29 21:47:32 +01:00
(void *)&rbuf[mu+4][0],
recv_from_rank,1,
2022-11-15 21:37:15 +00:00
bytes,bytes,mu+4);
Grid.StencilSendToRecvFromComplete(requests,mu+4);
2016-11-09 12:11:03 +00:00
requests.resize(0);
}
}
2017-05-29 21:47:32 +01:00
Grid.Barrier();
double stop=usecond();
t_time[i] = stop-start; // microseconds
2016-11-09 12:11:03 +00:00
}
timestat.statistics(t_time);
2017-05-29 21:47:32 +01:00
dbytes=dbytes*ppn;
double xbytes = dbytes*0.5;
2018-01-25 23:46:47 +00:00
// double rbytes = dbytes*0.5;
2017-05-29 21:47:32 +01:00
double bidibytes = dbytes;
2016-11-09 12:11:03 +00:00
std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
<<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
<<std::right<< xbytes/timestat.mean<<" "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
<<xbytes/timestat.max <<" "<< xbytes/timestat.min
<< "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< " " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
<< bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
2016-11-09 12:11:03 +00:00
}
}
2017-06-28 23:27:02 +01:00
#ifdef GRID_OMP
2017-06-28 23:27:02 +01:00
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << "= Benchmarking threaded STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
header();
2019-07-20 07:13:28 +01:00
for(int lat=8;lat<=maxlat;lat+=4){
2017-06-28 23:27:02 +01:00
for(int Ls=8;Ls<=8;Ls*=2){
Coordinate latt_size ({lat*mpi_layout[0],
lat*mpi_layout[1],
lat*mpi_layout[2],
lat*mpi_layout[3]});
2017-06-28 23:27:02 +01:00
GridCartesian Grid(latt_size,simd_layout,mpi_layout);
RealD Nrank = Grid._Nprocessors;
RealD Nnode = Grid.NodeCount();
RealD ppn = Nrank/Nnode;
std::vector<HalfSpinColourVectorD *> xbuf(8);
std::vector<HalfSpinColourVectorD *> rbuf(8);
Grid.ShmBufferFreeAll();
uint64_t bytes = lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
2017-06-28 23:27:02 +01:00
for(int d=0;d<8;d++){
2019-07-30 22:51:04 +01:00
xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
2017-06-28 23:27:02 +01:00
}
int ncomm;
double dbytes;
for(int i=0;i<Nloop;i++){
double start=usecond();
2018-01-08 11:27:14 +00:00
std::vector<CommsRequest_t> requests;
2017-06-28 23:27:02 +01:00
dbytes=0;
ncomm=0;
thread_for(dir,8,{
2017-06-28 23:27:02 +01:00
double tbytes;
int mu =dir % 4;
if (mpi_layout[mu]>1 ) {
ncomm++;
int xmit_to_rank;
int recv_from_rank;
if ( dir == mu ) {
int comm_proc=1;
Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
} else {
int comm_proc = mpi_layout[mu]-1;
Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
}
2018-04-27 11:44:46 +01:00
int tid = omp_get_thread_num();
tbytes= Grid.StencilSendToRecvFrom((void *)&xbuf[dir][0], xmit_to_rank,1,
2022-11-15 22:00:49 +00:00
(void *)&rbuf[dir][0], recv_from_rank,1, bytes,tid);
2017-06-28 23:27:02 +01:00
thread_critical { dbytes+=tbytes; }
2017-06-28 23:27:02 +01:00
}
});
2017-06-28 23:27:02 +01:00
Grid.Barrier();
double stop=usecond();
t_time[i] = stop-start; // microseconds
}
timestat.statistics(t_time);
dbytes=dbytes*ppn;
double xbytes = dbytes*0.5;
2018-01-25 23:46:47 +00:00
// double rbytes = dbytes*0.5;
2017-06-28 23:27:02 +01:00
double bidibytes = dbytes;
std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
<<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
<<std::right<< xbytes/timestat.mean<<" "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
<<xbytes/timestat.max <<" "<< xbytes/timestat.min
<< "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< " " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
<< bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
}
}
#endif
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << "= All done; Bye Bye"<<std::endl;
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
2016-11-09 12:11:03 +00:00
2015-05-02 23:42:30 +01:00
Grid_finalize();
}