Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix MPICheckpointWriter #370

Merged
merged 4 commits into from
Jan 13, 2025
Merged

Fix MPICheckpointWriter #370

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f93bb05
Fix incorrect xmlfile bool parsing in MPICheckpointWriter::readXML
cniethammer Dec 23, 2024
aef0537
Replace tabs in log output with whitespaces for better readability
cniethammer Dec 23, 2024
a1d1b34
Add documentation for readXML of MPICheckpointWriter
cniethammer Dec 23, 2024
deacfc7
Clean up outcommented code
cniethammer Dec 23, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion examples/Generators/mkesfera/config.xml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -109,7 +109,7 @@
<outputprefix>mkesfera</outputprefix>
<appendTimestamp>true</appendTimestamp>
<datarep>native</datarep>
<measureTime>1</measureTime>
<measureTime>true</measureTime>
<mpi_info>
<hint> <key>striping_factor</key> <value>4</value> </hint>
<hint> <key>striping_unit</key> <value>65536</value> </hint>
Expand Down
2 changes: 1 addition & 1 deletion examples/all-options.xml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -417,7 +417,7 @@
<outputprefix>default</outputprefix>
<appendTimestamp>true</appendTimestamp>
<datarep>native</datarep>
<measureTime>1</measureTime>
<measureTime>true</measureTime>
</outputplugin>

<!-- more output plugins -->
Expand Down
113 changes: 27 additions & 86 deletions src/io/MPICheckpointWriter.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,7 +28,6 @@
extern Simulation* global_simulation;

const char MPICheckpointWriter::_magicVersion[] = "MarDyn20150211trunk";
// int32_t
const int MPICheckpointWriter::_endiannesstest = 0x0a0b0c0d;

MPICheckpointWriter::MPICheckpointWriter(unsigned long writeFrequency, std::string outputPrefix, bool incremental, std::string datarep)
Expand All @@ -48,43 +47,34 @@ void MPICheckpointWriter::readXML(XMLfileUnits& xmlconfig)
{
_writeFrequency = 1;
xmlconfig.getNodeValue("writefrequency", _writeFrequency);
Log::global_log->info() << "[MPICheckpointWriter]\twrite frequency: " << _writeFrequency << std::endl;
Log::global_log->info() << "[MPICheckpointWriter] write frequency: " << _writeFrequency << std::endl;

_outputPrefix = "mardyn";
xmlconfig.getNodeValue("outputprefix", _outputPrefix);
Log::global_log->info() << "[MPICheckpointWriter]\toutput prefix: " << _outputPrefix << std::endl;
Log::global_log->info() << "[MPICheckpointWriter] output prefix: " << _outputPrefix << std::endl;

_incremental = false;
int incremental = 1;
xmlconfig.getNodeValue("incremental", incremental);
//_incremental = (incremental != 0);
if(incremental > 0) {
_incremental = true;
Log::global_log->info() << "[MPICheckpointWriter]\tusing incremental numbers in file names" << std::endl;
xmlconfig.getNodeValue("incremental", _incremental);
if(_incremental) {
Log::global_log->info() << "[MPICheckpointWriter] using incremental numbers in file names" << std::endl;
}

_appendTimestamp = false;
int appendTimestamp = 0;
xmlconfig.getNodeValue("appendTimestamp", appendTimestamp);
//_appendTimestamp = (appendTimestamp != 0);
if(appendTimestamp > 0) {
_appendTimestamp = true;
Log::global_log->info() << "[MPICheckpointWriter]\tappend timestamp to file names" << std::endl;
xmlconfig.getNodeValue("appendTimestamp", _appendTimestamp);
if(_appendTimestamp) {
Log::global_log->info() << "[MPICheckpointWriter] appending timestamps to file names" << std::endl;
}

_datarep = ""; // -> NULL
//_datarep = "external32"; // "native", "internal", "external32"
xmlconfig.getNodeValue("datarep", _datarep);
if(!_datarep.empty())
Log::global_log->info() << "[MPICheckpointWriter]\tdata representation: " << _datarep << std::endl;
Log::global_log->info() << "[MPICheckpointWriter] data representation: " << _datarep << std::endl;

_measureTime = false;
int measureTime = 0;
xmlconfig.getNodeValue("measureTime", measureTime);
//_measureTime = (measureTime != 0);
if(measureTime > 0) {
_measureTime = true;
Log::global_log->info() << "[MPICheckpointWriter]\texecution wall time will be measured" << std::endl;
xmlconfig.getNodeValue("measureTime", _measureTime);
if(_measureTime) {
Log::global_log->info() << "[MPICheckpointWriter] execution wall time will be measured" << std::endl;
}

if(xmlconfig.changecurrentnode("mpi_info")) {
Expand All @@ -102,9 +92,9 @@ void MPICheckpointWriter::readXML(XMLfileUnits& xmlconfig)
if(_particlesbuffersize)
{
#ifdef ENABLE_MPI
Log::global_log->info() << "[MPICheckpointWriter]\tparticles buffer size: " << _particlesbuffersize << std::endl;
Log::global_log->info() << "[MPICheckpointWriter] particles buffer size: " << _particlesbuffersize << std::endl;
#else
Log::global_log->info() << "[MPICheckpointWriter]\tparticles buffer size (" << _particlesbuffersize << ") only used in parallel/MPI version" << std::endl;
Log::global_log->info() << "[MPICheckpointWriter] particles buffer size (" << _particlesbuffersize << ") only used in parallel/MPI version" << std::endl;
#endif
}
}
Expand Down Expand Up @@ -143,7 +133,6 @@ void MPICheckpointWriter::endStep(ParticleContainer *particleContainer, DomainDe

if(_incremental)
{ /* align file numbers with preceding '0's in the required range from 0 to _numberOfTimesteps. */

unsigned long numTimesteps = _simulation.getNumTimesteps();
int num_digits = (int) ceil( log( double( numTimesteps / _writeFrequency ) ) / log(10.) );
filenamestream << "-" << aligned_number( simstep / _writeFrequency, num_digits, '0' );
Expand All @@ -164,16 +153,15 @@ void MPICheckpointWriter::endStep(ParticleContainer *particleContainer, DomainDe
filenamestream << ".MPIrestart.dat";

std::string filename = filenamestream.str();
Log::global_log->info() << "[MPICheckpointWriter]\tfilename: " << filename << std::endl;
Log::global_log->info() << "[MPICheckpointWriter] filename: " << filename << std::endl;

unsigned long numParticles_global = domain->getglobalNumMolecules(true, particleContainer, domainDecomp);
unsigned long numParticles = particleContainer->getNumberOfParticles(); // local
unsigned long numbb{1ul};
#ifdef ENABLE_MPI
Log::global_log->info() << "[MPICheckpointWriter]\tnumber of particles: " << numParticles_global
<< "\t(*" << sizeof(ParticleData) << "=" << numParticles_global*sizeof(ParticleData) << " Bytes in memory)"
Log::global_log->info() << "[MPICheckpointWriter] number of particles: " << numParticles_global
<< " (*" << sizeof(ParticleData) << "=" << numParticles_global*sizeof(ParticleData) << " Bytes in memory)"
<< std::endl;
//global_log->set_mpi_output_all()
int num_procs;
MPI_CHECK( MPI_Comm_size(MPI_COMM_WORLD, &num_procs) );
unsigned long gap=7+3+sizeof(unsigned long)+num_procs*(6*sizeof(double)+2*sizeof(unsigned long));
Expand All @@ -182,10 +170,8 @@ void MPICheckpointWriter::endStep(ParticleContainer *particleContainer, DomainDe
double mpistarttime=0; // =0 to prevent Jenkins/gcc complaining about uninitialized mpistarttime [-Werror=uninitialized]
if(_measureTime)
{
//if(ownrank==0) Log::global_log->debug() << "MPICheckpointWriter (" << filename << ")\tstart measuring time" << std::endl;
MPI_CHECK( MPI_Barrier(MPI_COMM_WORLD) );
mpistarttime=MPI_Wtime();
// global_simulation->timers()->start("MPI_CHECKPOINT_WRITER_INPUT"); // should use Timer instead
}
MPI_File mpifh;
// arg 2 type cast due to old MPI (<=V2) implementations (should be const char* now)
Expand Down Expand Up @@ -235,7 +221,7 @@ void MPICheckpointWriter::endStep(ParticleContainer *particleContainer, DomainDe
startidx=0;
}
MPI_CHECK( MPI_Exscan(&numParticles, &startidx, 1, MPI_UNSIGNED_LONG, MPI_SUM, MPI_COMM_WORLD) );
//

mpioffset=64+7+3+sizeof(unsigned long)+ownrank*(6*sizeof(double)+2*sizeof(unsigned long));
double bbmin[3],bbmax[3];
bbmin[0]=domainDecomp->getBoundingBoxMin(0,domain);
Expand All @@ -260,10 +246,10 @@ void MPICheckpointWriter::endStep(ParticleContainer *particleContainer, DomainDe
mpioffset+=sizeof(unsigned long);
MPI_CHECK( MPI_File_write_at(mpifh,mpioffset,&numParticles,1,MPI_UNSIGNED_LONG,&mpistat) );
mpioffset+=sizeof(unsigned long);
Log::global_log->debug() << "[MPICheckpointWriter](" << ownrank << ")\tBB " << ":\t"
Log::global_log->debug() << "[MPICheckpointWriter](" << ownrank << ") BB " << ": "
<< bbmin[0] << ", " << bbmin[1] << ", " << bbmin[2] << " - "
<< bbmax[0] << ", " << bbmax[1] << ", " << bbmax[2]
<< "\tstarting index=" << startidx << " numParticles=" << numParticles << std::endl;
<< " starting index=" << startidx << " numParticles=" << numParticles << std::endl;
//
MPI_Datatype mpidtParticleM, mpidtParticleD;
ParticleData::getMPIType(mpidtParticleM);
Expand All @@ -285,73 +271,35 @@ void MPICheckpointWriter::endStep(ParticleContainer *particleContainer, DomainDe
Log::global_log->debug()
<< "[MPICheckpointWriter]("
<< ownrank
<< ")\twriting molecule data for "
<< ") writing molecule data for "
<< numParticles
<< " particles of size "
<< mpidtParticleDts
<< std::endl;
//unsigned long writecounter=0;
if(_particlesbuffersize>0)
{
ParticleData* particleStructBuffer=new ParticleData[_particlesbuffersize];
unsigned long bufidx=0;
for (auto pos = particleContainer->iterator(ParticleIterator::ONLY_INNER_AND_BOUNDARY); pos.isValid(); ++pos) {
// Log::global_log->debug()
// << "MPICheckpointWriter["
// << ownrank
// << "]\t"
// << pos->getID()
// << "\t"
// << pos->componentid()
// << "\t"
// << pos->r(0)
// << ","
// << pos->r(1)
// << ","
// << pos->r(2)
// << std::endl;
ParticleData::MoleculeToParticleData(particleStructBuffer[bufidx], *pos);
++bufidx;
if(bufidx==_particlesbuffersize)
{
//global_log->debug() << "MPICheckpointWriter[" << ownrank << "]\twriting" << _particlesbuffersize << " particles" << std::endl
if(bufidx==_particlesbuffersize) {
MPI_CHECK( MPI_File_write(mpifh, particleStructBuffer, _particlesbuffersize, mpidtParticleD, &mpistat) );
//++writecounter;
bufidx=0;
}
}
if(bufidx>0)
{
//global_log->debug() << "MPICheckpointWriter[" << ownrank << "]\twriting" << bufidx << " particles" << std::endl
MPI_CHECK( MPI_File_write(mpifh, particleStructBuffer, bufidx, mpidtParticleD, &mpistat) );
//++writecounter;
}
delete[] particleStructBuffer;
}
else
{
ParticleData particleStruct;
for (auto pos = particleContainer->iterator(ParticleIterator::ONLY_INNER_AND_BOUNDARY); pos.isValid(); ++pos) {
// Log::global_log->debug()
// << "MPICheckpointWriter["
// << ownrank
// << "]\t"
// << pos->getID()
// << "\t"
// << pos->componentid()
// << "\t"
// << pos->r(0)
// << ","
// << pos->r(1)
// << ","
// << pos->r(2)
// << std::endl;
ParticleData::MoleculeToParticleData(particleStruct, *pos);
// Log::global_log->debug() << "MPICheckpointWriter[" << ownrank << "]\twriting particle" << std::endl
MPI_CHECK( MPI_File_write(mpifh, &particleStruct, 1, mpidtParticleD, &mpistat) );
//++writecounter;
// saving a struct directly will also save padding zeros...
//mpioffset+=mpidtParticleMsize;
}
}

Expand All @@ -360,33 +308,28 @@ void MPICheckpointWriter::endStep(ParticleContainer *particleContainer, DomainDe
{
MPI_CHECK( MPI_Barrier(MPI_COMM_WORLD) );
double mpimeasuredtime=MPI_Wtime()-mpistarttime;
// global_simulation->timers()->stop("MPI_CHECKPOINT_WRITER_INPUT");
// double mpimeasuredtime=global_simulation->timers()->getTime("MPI_CHECKPOINT_WRITER_INPUT");
if(ownrank==0) {
Log::global_log->info() << "[MPICheckpointWriter]\tmeasured time: " << mpimeasuredtime << " sec (par., "
Log::global_log->info() << "[MPICheckpointWriter] measured time: " << mpimeasuredtime << " sec (par., "
<< num_procs << " proc.; " << numParticles_global << "*" << mpidtParticleDts
<< "=" << numParticles_global * mpidtParticleDts << " Bytes)" << std::endl;
}
}
#else
Log::global_log->info() << "[MPICheckpointWriter]\tnumber of particles: " << numParticles_global
<< "\t(*" << 2*sizeof(unsigned long)+13*sizeof(double) << "=" << numParticles_global*(2*sizeof(unsigned long)+13*sizeof(double)) << " Bytes in memory)"
Log::global_log->info() << "[MPICheckpointWriter] number of particles: " << numParticles_global
<< " (*" << 2*sizeof(unsigned long)+13*sizeof(double) << "=" << numParticles_global*(2*sizeof(unsigned long)+13*sizeof(double)) << " Bytes in memory)"
<< std::endl;
unsigned long gap=7+3+sizeof(unsigned long)+(6*sizeof(double)+2*sizeof(unsigned long));
unsigned int i;
unsigned int offset=0;
if (!_datarep.empty()) {
Log::global_log->warning() << "[MPICheckpointWriter]\tsetting data representation (" << _datarep
Log::global_log->warning() << "[MPICheckpointWriter] setting data representation (" << _datarep
<< ") is not supported (yet) in sequential version" << std::endl;
}
// should use Timer instead
struct timeval tod_start;
if(_measureTime) {
//global_log->debug() << "MPICheckpointWriter (" << filename << ")\tstart measuring time" << std::endl;
gettimeofday( &tod_start, NULL );
// global_simulation->timers()->start("MPI_CHECKPOINT_WRITER_INPUT");
}
//
std::ofstream ostrm(filename.c_str(),std::ios::out|std::ios::binary);
ostrm << _magicVersion;
offset+=strlen(_magicVersion);
Expand Down Expand Up @@ -440,9 +383,7 @@ void MPICheckpointWriter::endStep(ParticleContainer *particleContainer, DomainDe
struct timeval tod_end;
gettimeofday( &tod_end, NULL );
double measuredtime=(double)(tod_end.tv_sec-tod_start.tv_sec)+(double)(tod_end.tv_usec-tod_start.tv_usec)/1.E6;
// global_simulation->timers()->stop("MPI_CHECKPOINT_WRITER_INPUT");
// double measuredtime=global_simulation->timers()->getTime("MPI_CHECKPOINT_WRITER_INPUT");
Log::global_log->info() << "[MPICheckpointWriter]\tmeasured time: " << measuredtime << " sec (seq.)" << std::endl;
Log::global_log->info() << "[MPICheckpointWriter] measured time: " << measuredtime << " sec (seq.)" << std::endl;
}
#endif
}
Expand Down
15 changes: 14 additions & 1 deletion src/io/MPICheckpointWriter.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,8 +38,21 @@ class MPICheckpointWriter : public PluginBase {
MPICheckpointWriter(unsigned long writeFrequency
, std::string outputPrefix, bool incremental=true
, std::string datarep=std::string(""));
//~MPICheckpointWriter() {};

/** @brief Read in XML configuration for MPICheckpointWriter and all its included objects.
*
* The following xml object structure is handled by this method:
* \code{.xml}
<outputplugin name="MPICheckpointWriter">
<writefrequency>INTEGER</writefrequency> <!-- Frequency in which the output is written; Default: 1 -->
<outputprefix>STRING</outputprefix> <!-- Prefix of the output file; Default: "mardyn" -->
<incremental>BOOL</incremental> <!-- Checkpoint files will get individual numbers; Default: false -->
<appendTimestamp>BOOL</appendTimestamp> <!-- Append timestamp to checkpoint files; Default: false -->
<datarep>STRING</datarep> <!-- MPI I/O output representation to use, valid values are "native", "internal", "external32"; Default: "" -->
<mpi_info><!-- see MPI_Info_object class documentation --></mpi_info> <!-- MPI infos to be used for MPI file I/O writing the checkpoint files -->
</outputplugin>
\endcode
*/
void readXML(XMLfileUnits& xmlconfig);

void init(ParticleContainer *particleContainer,
Expand Down
Loading