main.cpp File Reference

#include "header.h"
#include "SparseMatrix.h"
#include <sys/time.h>
#include <math.h>
#include <queue>
#include <unistd.h>
#include "../scheduling/Clock.h"
#include "../msg/DiagonalMsg.h"
#include "../msg/SparseMsg.h"
#include "../mpi/PostMaster.h"
#include "../shell/Shell.h"

Include dependency graph for main.cpp:

Go to the source code of this file.

Functions
void	checkChildren (Id parent, const string &info)
unsigned int	getNumCores ()
Id	init (int argc, char **argv, bool &doUnitTests, bool &doRegressionTests, unsigned int &benchmark)
int	main (int argc, char **argv)
void	mooseBenchmarks (unsigned int option)
void	mpiTests ()
void	nonMpiTests (Shell *s)
void	processTests (Shell *s)
void	speedTestMultiNodeIntFireNetwork (unsigned int size, unsigned int runsteps)

Variables
bool	quitFlag = 0

Function Documentation

void checkChildren	(	Id	parent,
		const string &	info
	)

Definition at line 136 of file main.cpp.

References Neutral::children(), Id::element(), Id::eref(), and Element::getName().

{
    vector< Id > ret;
    Neutral::children( parent.eref(), ret );
    cout << info << " checkChildren of " <<
         parent.element()->getName() << ": " <<
         ret.size() << " children\n";
    for ( vector< Id >::iterator i = ret.begin(); i != ret.end(); ++i )
    {
        cout << i->element()->getName() << endl;
    }
}

Here is the call graph for this function:

unsigned int getNumCores

(

)

Definition at line 93 of file main.cpp.

Referenced by init().

{
    unsigned int numCPU = 0;
#ifdef WIN_32
    SYSTEM_INFO sysinfo;
    GetSystemInfo( &sysinfo );

    numCPU = sysinfo.dwNumberOfProcessors;
#endif

#ifdef LINUX
    numCPU = sysconf( _SC_NPROCESSORS_ONLN );
#endif

#ifdef MACOSX
    int mib[4];
    size_t len = sizeof(numCPU);

    /* set the mib for hw.ncpu */
    mib[0] = CTL_HW;
    mib[1] = HW_AVAILCPU;  // alternatively, try HW_NCPU;

    /* get the number of CPUs from the system */
    sysctl(mib, 2, &numCPU, &len, NULL, 0);

    if( numCPU < 1 )
    {
        mib[1] = HW_NCPU;
        sysctl( mib, 2, &numCPU, &len, NULL, 0 );
    }
#endif
    if ( numCPU < 1 )
    {
        cout << "No CPU information available. Assuming single core." << endl;
        numCPU = 1;
    }
    return numCPU;
}

Here is the caller graph for this function:

Id init	(	int	argc,
		char **	argv,
		bool &	doUnitTests,
		bool &	doRegressionTests,
		unsigned int &	benchmark
	)

Here we allow the user to override the automatic identification of processor configuration

Sets up the Elements that represent each class of Msg.

Definition at line 150 of file main.cpp.

Referenced by benchmarkMsg(), Ksolve::initCinfo(), Gsolve::initCinfo(), moose::CompartmentBase::initCinfo(), NonlinearSystem::initialize(), main(), SteadyState::settle(), testAssortedMsg(), testCopyMsgOps(), testShellAddMsg(), and testSparseMatrix2().

{
    unsigned int numCores = getNumCores();
    int numNodes = 1;
    int myNode = 0;
    bool isInfinite = 0;
    int opt;
    benchmark = 0; // Default, means don't do any benchmarks.
    Cinfo::rebuildOpIndex();
#ifdef USE_MPI
    /*
    // OpenMPI does not use argc or argv.
    // unsigned int temp_argc = 1;
    int provided;
    MPI_Init_thread( &argc, &argv, MPI_THREAD_SERIALIZED, &provided );
    */
    MPI_Init( &argc, &argv );

    MPI_Comm_size( MPI_COMM_WORLD, &numNodes );
    MPI_Comm_rank( MPI_COMM_WORLD, &myNode );
    /*
       if ( provided < MPI_THREAD_SERIALIZED && myNode == 0 ) {
       cout << "Warning: This MPI implementation does not like multithreading: " << provided << "\n";
       }
       */
    // myNode = MPI::COMM_WORLD.Get_rank();
#endif

    while ( ( opt = getopt( argc, argv, "hiqurn:b:B:" ) ) != -1 )
    {
        switch ( opt )
        {
        case 'i' : // infinite loop, used for multinode debugging, to give gdb something to attach to.
            isInfinite = 1;
            break;
        case 'n': // Multiple nodes
            numNodes = (unsigned int)atoi( optarg );
            break;
        case 'b': // Benchmark:
        {
            string s(optarg);
            if ( s == "ee" )
                benchmark = 1;
            else if ( s == "gsl" )
                benchmark = 2;
            else if ( s == "gssa" )
                benchmark = 3;
            else if ( s[0] == 'i' )
                benchmark = 4;
            else if ( s[0] == 'h' )
                benchmark = 5;
            else if ( s[0] == 'm' )
                benchmark = 6;
            else
                cout << "Unknown benchmark, " << optarg << ", skipping\n";
        }
        break;
        case 'B': // Benchmark plus dump data: handle later.
            break;
        case 'u': // Do unit tests, pass back.
            doUnitTests = 1;
            break;
        case 'r': // Do regression tests: pass back
            doRegressionTests = 1;
            break;
        case 'q': // quit immediately after completion.
            quitFlag = 1;
            break;
        case 'h': // help
        default:
            cout << "Usage: moose -help -infiniteLoop -unit_tests -regression_tests -quit -n numNodes -benchmark [ee gsl gssa intFire hhNet msg_<msgType>_<size>]\n";

            exit( 1 );
        }
    }
    if ( myNode == 0 )
    {
#ifndef QUIET_MODE
        cout << "on node " << myNode << ", numNodes = "
             << numNodes << ", numCores = " << numCores << endl;
#endif
    }

    Id shellId;
    Element* shelle =
        new GlobalDataElement( shellId, Shell::initCinfo(), "root", 1 );

    Id clockId = Id::nextId();
    assert( clockId.value() == 1 );
    Id classMasterId = Id::nextId();
    Id postMasterId = Id::nextId();

    Shell* s = reinterpret_cast< Shell* >( shellId.eref().data() );
    s->setShellElement( shelle );
    s->setHardware( numCores, numNodes, myNode );
    s->loadBalance();

    unsigned int numMsg = Msg::initMsgManagers();

    new GlobalDataElement( clockId, Clock::initCinfo(), "clock", 1 );
    new GlobalDataElement( classMasterId, Neutral::initCinfo(), "classes", 1);
    new GlobalDataElement( postMasterId, PostMaster::initCinfo(), "postmaster", 1 );

    assert ( shellId == Id() );
    assert( clockId == Id( 1 ) );
    assert( classMasterId == Id( 2 ) );
    assert( postMasterId == Id( 3 ) );



    // s->connectMasterMsg();

    Shell::adopt( shellId, clockId, numMsg++ );
    Shell::adopt( shellId, classMasterId, numMsg++ );
    Shell::adopt( shellId, postMasterId, numMsg++ );

    assert( numMsg == 10 ); // Must be the same on all nodes.

    Cinfo::makeCinfoElements( classMasterId );


    // This will be initialized within the Process loop, and better there
    // as it flags attempts to call the Reduce operations before ProcessLoop
    // Qinfo::clearReduceQ( numCores ); // Initialize the ReduceQ entry.


    // SetGet::setShell();
    // Msg* m = new OneToOneMsg( shelle, shelle );
    // assert ( m != 0 );

    while ( isInfinite ) // busy loop for debugging under gdb and MPI.
        ;

    return shellId;
}

Here is the caller graph for this function:

int main	(	int	argc,
		char **	argv
	)

Definition at line 360 of file main.cpp.

References Id::clearAllElements(), Msg::clearAllMsgs(), PostMaster::clearPending(), ObjId::data(), Element::data(), Shell::doQuit(), doRegressionTests, Shell::doSetClock(), doUnitTests, Shell::doUseClock(), Id::element(), init(), Shell::keepLooping(), Shell::launchParser(), mooseBenchmarks(), mpiTests(), Shell::myNode(), nonMpiTests(), Shell::numNodes(), processTests(), and quitFlag.

{
    bool doUnitTests = 0;
    bool doRegressionTests = 0;
    unsigned int benchmark = 0;
    // This reorders the OpFunc to Fid mapping to ensure it is node and
    // compiler independent.
    Id shellId = init( argc, argv, doUnitTests, doRegressionTests, benchmark );
    // Note that the main loop remains the parser loop, though it may
    // spawn a lot of other stuff.
    Element* shelle = shellId.element();
    Shell* s = reinterpret_cast< Shell* >( shelle->data( 0 ) );
    if ( doUnitTests )
        nonMpiTests( s ); // These tests do not need the process loop.

    if ( Shell::myNode() == 0 )
    {
        if ( Shell::numNodes() > 1 )
        {
            // Use the last clock for the postmaster, so that it is called
            // after everything else has been processed and all messages
            // are ready to send out.
            s->doUseClock( "/postmaster", "process", 9 );
            s->doSetClock( 9, 1.0 ); // Use a sensible default.
        }
#ifdef DO_UNIT_TESTS
        if ( doUnitTests )
        {
            mpiTests();
            processTests( s );
        }
        // if ( doRegressionTests ) regressionTests();
#endif
        // These are outside unit tests because they happen in optimized
        // mode, using a command-line argument. As soon as they are done
        // the system quits, in order to estimate timing.
        if ( benchmark != 0 )
        {
            mooseBenchmarks( benchmark );
            s->doQuit();
        }
        else
        {
            // Here we set off a little event loop to poll user input.
            // It deals with the doQuit call too.
            if(! quitFlag)
                Shell::launchParser();
        }
    }
    else
    {
        PostMaster* p = reinterpret_cast< PostMaster* >( ObjId( 3 ).data());
        while ( Shell::keepLooping() )
        {
            p->clearPending();
        }
    }
    Msg::clearAllMsgs();
    Id::clearAllElements();
#ifdef USE_MPI
    MPI_Finalize();
#endif
    return 0;
}

Here is the call graph for this function:

void mooseBenchmarks

(

unsigned int

option

)

Referenced by main().

Here is the caller graph for this function:

void mpiTests

(

)

These are tests that are MPI safe. They should also run properly on single nodes.

Definition at line 350 of file main.cpp.

Referenced by main().

{
#ifdef DO_UNIT_TESTS
    MOOSE_TEST( "testMpiMsg", testMpiMsg());
    MOOSE_TEST( "testMpiShell", testMpiShell());
    MOOSE_TEST( "testMpiBuiltins", testMpiBuiltins());
    MOOSE_TEST( "testMpiScheduling", testMpiScheduling());
#endif
}

Here is the caller graph for this function:

void nonMpiTests

(

Shell *

s

)

These tests are meant to run on individual nodes, and should not invoke MPI calls. They should not be run when MPI is running. These tests do not use the threaded/MPI event loop and are the most basic of the set.

Definition at line 297 of file main.cpp.

Referenced by main().

{
#ifdef DO_UNIT_TESTS
    if ( Shell::myNode() == 0 )
    {
        unsigned int numNodes = s->numNodes();
        unsigned int numCores = s->numCores();
        if ( numCores > 0 )
            s->setHardware( 1, 1, 0 );
        MOOSE_TEST("testAsync", testAsync());
        MOOSE_TEST("testMsg", testMsg());
        MOOSE_TEST("testShell", testShell());
        MOOSE_TEST("testScheduling", testScheduling());
        MOOSE_TEST("testBuiltints", testBuiltins());
        //MOOSE_TEST("testKinetics", testKinetics());
        MOOSE_TEST("testKsolve", testKsolve());
        //MOOSE_TEST("testKsolveProcess", testKsolveProcess());
        MOOSE_TEST("testBiophysics", testBiophysics());
        MOOSE_TEST("testDiffusion", testDiffusion());
        MOOSE_TEST("testHsolve", testHSolve());
        //MOOSE_TEST("testGeom", testGeom());
        MOOSE_TEST("testMesh", testMesh());
        MOOSE_TEST("testSynapse", testSynapse());
        MOOSE_TEST( "testSigneur", testSigNeur());
#ifdef USE_SMOLDYN
        //MOOSE_TEST(testSmoldyn());
#endif
        s->setHardware( numCores, numNodes, 0 );
    }
#endif
}

Here is the caller graph for this function:

void processTests

(

Shell *

s

)

These tests involve the threaded/MPI process loop and are the next level of tests.

Definition at line 333 of file main.cpp.

Referenced by main().

{
#ifdef DO_UNIT_TESTS
    MOOSE_TEST( "testSchedulingProcess", testSchedulingProcess());
    MOOSE_TEST( "testBuiltinsProcess", testBuiltinsProcess());
    // MOOSE_TEST( "testKineticsProcess", testKineticsProcess());
    MOOSE_TEST( "testBiophysicsProcess", testBiophysicsProcess());
    // MOOSE_TEST( "testKineticSolversProcess", testKineticSolversProcess());
    // MOOSE_TEST( "testSimManager", testSimManager());
    MOOSE_TEST( "testSigNeurProcess", testSigNeurProcess());
#endif
}

Here is the caller graph for this function:

void speedTestMultiNodeIntFireNetwork	(	unsigned int	size,
		unsigned int	runsteps
	)

Variable Documentation

bool quitFlag = 0

Definition at line 132 of file main.cpp.

Referenced by init(), and main().