MultTest.cpp

Go to the documentation of this file.

/****************************************************************/
/* Parallel Combinatorial BLAS Library (for Graph Computations) */
/* version 1.6 -------------------------------------------------*/
/* date: 6/15/2017 ---------------------------------------------*/
/* authors: Ariful Azad, Aydin Buluc  --------------------------*/
/****************************************************************/
/*
 Copyright (c) 2010-2017, The Regents of the University of California
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal
 in the Software without restriction, including without limitation the rights
 to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 copies of the Software, and to permit persons to whom the Software is
 furnished to do so, subject to the following conditions:
 
 The above copyright notice and this permission notice shall be included in
 all copies or substantial portions of the Software.
 
 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 THE SOFTWARE.
 */

#include <mpi.h>
#include <sys/time.h> 
#include <iostream>
#include <functional>
#include <algorithm>
#include <vector>
#include <sstream>
#include "CombBLAS/CombBLAS.h"

using namespace std;
using namespace combblas;

#ifdef TIMING
double cblas_alltoalltime;
double cblas_allgathertime;
#endif

#ifdef _OPENMP
int cblas_splits = omp_get_max_threads(); 
#else
int cblas_splits = 1;
#endif


// Simple helper class for declarations: Just the numerical type is templated 
// The index type and the sequential matrix type stays the same for the whole code
// In this case, they are "int" and "SpDCCols"
template <class NT>
class PSpMat 
{ 
public: 
    typedef SpDCCols < int64_t, NT > DCCols;
    typedef SpParMat < int64_t, NT, DCCols > MPI_DCCols;
};

int main(int argc, char* argv[])
{
    int nprocs, myrank;
    MPI_Init(&argc, &argv);
    MPI_Comm_size(MPI_COMM_WORLD,&nprocs);
    MPI_Comm_rank(MPI_COMM_WORLD,&myrank);

    if(argc < 6)
    {
        if(myrank == 0)
        {
            cout << "Usage: ./MultTest <MatrixA> <MatrixB> <MatrixC> <vecX> <vecY>" << endl;
            cout << "<MatrixA>,<MatrixB>,<MatrixC> are absolute addresses, and files should be in triples format" << endl;
        }
        MPI_Finalize(); 
        return -1;
    }               
    {
        string Aname(argv[1]);      
        string Bname(argv[2]);
        string Cname(argv[3]);
        string V1name(argv[4]);
        string V2name(argv[5]);

        ifstream vecinpx(V1name.c_str());
        ifstream vecinpy(V2name.c_str());

        MPI_Barrier(MPI_COMM_WORLD);    
        typedef PlusTimesSRing<double, double> PTDOUBLEDOUBLE;  
        typedef SelectMaxSRing<bool, int64_t> SR;   

        shared_ptr<CommGrid> fullWorld;
        fullWorld.reset( new CommGrid(MPI_COMM_WORLD, 0, 0) );

            // construct objects
            PSpMat<double>::MPI_DCCols A(fullWorld);
            PSpMat<double>::MPI_DCCols B(fullWorld);
            PSpMat<double>::MPI_DCCols C(fullWorld);
            PSpMat<double>::MPI_DCCols CControl(fullWorld);
            FullyDistVec<int64_t, double> ycontrol(fullWorld);
            FullyDistVec<int64_t, double> x(fullWorld);
            FullyDistSpVec<int64_t, double> spycontrol(fullWorld);
            FullyDistSpVec<int64_t, double> spx(fullWorld);
    
        A.ParallelReadMM(Aname, true, maximum<double>());
#ifndef NOGEMM
        B.ParallelReadMM(Bname, true, maximum<double>());

        CControl.ParallelReadMM(Cname, true, maximum<double>());
#endif
        x.ReadDistribute(vecinpx, 0);
        spx.ReadDistribute(vecinpx, 0);
        ycontrol.ReadDistribute(vecinpy,0);
        spycontrol.ReadDistribute(vecinpy,0);

        FullyDistVec<int64_t, double> y = SpMV<PTDOUBLEDOUBLE>(A, x);
        if (ycontrol == y)
        {
            SpParHelper::Print("Dense SpMV (fully dist) working correctly\n");  
        }
        else
        {
            SpParHelper::Print("ERROR in Dense SpMV, go fix it!\n");    
            y.ParallelWrite("ycontrol_dense.txt",true);
        }

        //FullyDistSpVec<int64_t, double> spy = SpMV<PTDOUBLEDOUBLE>(A, spx);
        
        FullyDistSpVec<int64_t, double> spy(spx.getcommgrid(), A.getnrow());
        SpMV<PTDOUBLEDOUBLE>(A, spx, spy, false);
        
        if (spycontrol == spy)
        {
            SpParHelper::Print("Sparse SpMV (fully dist) working correctly\n"); 
        }
        else
        {
            SpParHelper::Print("ERROR in Sparse SpMV, go fix it!\n");   
            spy.ParallelWrite("ycontrol_sparse.txt",true);
        }
        
            // Test SpMSpV-bucket for general CSC matrices
            SpParMat < int64_t, double, SpCCols<int64_t,double> >  ACsc (A);
            PreAllocatedSPA<double> SPA(ACsc.seq(), cblas_splits*4);
            FullyDistSpVec<int64_t, double> spy_csc(spx.getcommgrid(), ACsc.getnrow());
            SpMV<PTDOUBLEDOUBLE>(ACsc, spx, spy_csc, false, SPA);
        
            if (spy == spy_csc)
            {
                    SpParHelper::Print("SpMSpV-bucket works correctly for general CSC matrices\n");
            }
            else
            {
                    SpParHelper::Print("SpMSpV-bucket does not work correctly for general CSC matrices, go fix it!\n");
            }

        
#ifndef NOGEMM
        C = Mult_AnXBn_Synch<PTDOUBLEDOUBLE, double, PSpMat<double>::DCCols >(A,B);
        if (CControl == C)
        {
            SpParHelper::Print("Synchronous Multiplication working correctly\n");   
            // C.SaveGathered("CControl.txt");
        }
        else
        {
            SpParHelper::Print("ERROR in Synchronous Multiplication, go fix it!\n");    
        }

        C = Mult_AnXBn_DoubleBuff<PTDOUBLEDOUBLE, double, PSpMat<double>::DCCols >(A,B);
        if (CControl == C)
        {
            SpParHelper::Print("Double buffered multiplication working correctly\n");   
        }
        else
        {
            SpParHelper::Print("ERROR in double buffered multiplication, go fix it!\n");    
        }
#endif
        OptBuf<int32_t, int64_t> optbuf;
        PSpMat<bool>::MPI_DCCols ABool(A);

        spx.Apply(bind1st (multiplies<double>(), 100));
        FullyDistSpVec<int64_t, int64_t> spxint64 (spx);
        //FullyDistSpVec<int64_t, int64_t> spyint64 = SpMV<SR>(ABool, spxint64, false);
        FullyDistSpVec<int64_t, int64_t> spyint64(spxint64.getcommgrid(), ABool.getnrow());
        SpMV<SR>(ABool, spxint64, spyint64, false);

        
        ABool.OptimizeForGraph500(optbuf);
        //FullyDistSpVec<int64_t, int64_t> spyint64buf = SpMV<SR>(ABool, spxint64, false, optbuf);
        FullyDistSpVec<int64_t, int64_t> spyint64buf(spxint64.getcommgrid(), ABool.getnrow());
        SpMV<SR>(ABool, spxint64, spyint64buf, false, optbuf);
        
        
        if (spyint64 == spyint64buf)
        {
            SpParHelper::Print("Graph500 Optimizations are correct\n"); 
        }
        else
        {
            SpParHelper::Print("ERROR in graph500 optimizations, go fix it!\n");    
            spyint64.ParallelWrite("Original_SpMSV.txt",true);
            spyint64buf.ParallelWrite("Buffered_SpMSV.txt",true);
        }
        
       
        ABool.ActivateThreading(cblas_splits);
        //FullyDistSpVec<int64_t, int64_t> spyint64_threaded = SpMV<SR>(ABool, spxint64, false);
        FullyDistSpVec<int64_t, int64_t> spyint64_threaded(spxint64.getcommgrid(), ABool.getnrow());
        SpMV<SR>(ABool, spxint64, spyint64_threaded, false);

        if (spyint64 == spyint64_threaded)
        {
            SpParHelper::Print("Multithreaded Sparse SpMV works\n");    
        }
        else
        {
            SpParHelper::Print("ERROR in multithreaded sparse SpMV, go fix it!\n"); 
        }
        

            // Test SpMSpV-bucket for Boolean CSC matrices
            SpParMat < int64_t, bool, SpCCols<int64_t,bool> >  ABoolCsc (A);
            PreAllocatedSPA<int64_t> SPA1(ABoolCsc.seq(), cblas_splits*4);
            FullyDistSpVec<int64_t, int64_t> spyint64_csc_threaded(spxint64.getcommgrid(), ABoolCsc.getnrow());
            SpMV<SR>(ABoolCsc, spxint64, spyint64_csc_threaded, false, SPA1);
        
            if (spyint64 == spyint64_csc_threaded)
            {
                    SpParHelper::Print("SpMSpV-bucket works correctly for Boolean CSC matrices\n");
            }
            else
            {
                SpParHelper::Print("ERROR in SpMSpV-bucket with Boolean CSC matrices, go fix it!\n");
            }
        
        
        vecinpx.clear();
        vecinpx.close();
        vecinpy.clear();
        vecinpy.close();
    }
    MPI_Finalize();
    return 0;
}