psdrive3d1.c

/*! \file
Copyright (c) 2003, The Regents of the University of California, through
Lawrence Berkeley National Laboratory (subject to receipt of any required
approvals from U.S. Dept. of Energy)

All rights reserved.

The source code is distributed under BSD license, see the file License.txt
at the top-level directory.
*/


/*! @file
 * \brief Driver program for PSGSSVX3D example
 *
 * <pre>
 * -- Distributed SuperLU routine (version 9.0) --
 * Lawrence Berkeley National Lab, Georgia Institute of Technology,
 * Oak Ridge National Lab
 * September 10, 2021
 *
 */
#include "superlu_sdefs.h"

/*! \brief
 *
 * <pre>
 * Purpose
 * =======
 *
 * The driver program PSDRIVE3D1.
 *
 * This example illustrates how to use PSGSSVX3D to sovle the systems
 * with the same A but different right-hand side, possibly with
 * different number of right-hand sides.
 * In this case, we factorize A only once in the first call to PSGSSVX3D,
 * and reuse the following data structures in the subsequent call to
 * PSGSSVX3D:
 *        ScalePermstruct  : DiagScale, R, C, perm_r, perm_c
 *        LUstruct         : Glu_persist, Llu
 *        SOLVEstruct      : communication metadata for SpTRSV, SpMV, and
 *                           3D<->2D gather/scatter of {A,B} stored in A3d.
 *
 * The program may be run by typing:
 *    mpiexec -np <p> psdrive3d1 -r <proc rows> -c <proc columns> \
 *                                    -d <proc Z-dimension> <input_file>
 * NOTE: total number of processes p = r * c * d
 *       (d must be a power-of-two, e.g., 1, 2, 4, ...)
 *
 * </pre>
 */

static void matCheck(int n, int m, float* A, int LDA,
       float* B, int LDB)
{
    for(int j=0; j<m;j++)
        for (int i = 0; i < n; ++i) {
	    assert(A[i+ LDA*j] == B[i+ LDB*j]);
	}
    printf("B check passed\n");
    return;
}

static void checkNRFMT(NRformat_loc*A, NRformat_loc*B)
{
    /*
    int_t nnz_loc;
    int_t m_loc;
    int_t fst_row;
    void  *nzval;
    int_t *rowptr;
    int_t *colind;
    */

    assert(A->nnz_loc == B->nnz_loc);
    assert(A->m_loc == B->m_loc);
    assert(A->fst_row == B->fst_row);

#if 0
    double *Aval = (double *)A->nzval, *Bval = (double *)B->nzval;
    Printdouble5("A", A->nnz_loc, Aval);
    Printdouble5("B", B->nnz_loc, Bval);
    fflush(stdout);
#endif

    float * Aval = (float *) A->nzval;
    float * Bval = (float *) B->nzval;
    for (int_t i = 0; i < A->nnz_loc; i++)
    {
        assert( Aval[i] == Bval[i] );
        assert((A->colind)[i] == (B->colind)[i]);
	printf("colind[] correct\n");
    }

    for (int_t i = 0; i < A->m_loc + 1; i++)
    {
        assert((A->rowptr)[i] == (B->rowptr)[i]);
    }

    printf("Matrix check passed\n");

}

int
main (int argc, char *argv[])
{
    superlu_dist_options_t options;
    SuperLUStat_t stat;
    SuperMatrix A;  // Now, A is on all 3D processes
    sScalePermstruct_t ScalePermstruct;
    sLUstruct_t LUstruct;
    sSOLVEstruct_t SOLVEstruct;
    gridinfo3d_t grid;
    float *berr;
    float *b, *xtrue, *b1, *b2;
    int m, n, i, j, m_loc;
    int nprow, npcol, npdep;
    int lookahead, colperm, rowperm, ir;
    int iam, info, ldb, ldx, nrhs;
    char **cpp, c, *suffix;
    FILE *fp, *fopen ();
    extern int cpp_defs ();
    int ii, omp_mpi_level;

    nprow = 1;            /* Default process rows.      */
    npcol = 1;            /* Default process columns.   */
    npdep = 1;            /* replication factor must be power of two */
    nrhs = 1;             /* Number of right-hand side. */
    lookahead = -1;
    colperm = -1;
    rowperm = -1;
    ir = -1;

    /* ------------------------------------------------------------
       INITIALIZE MPI ENVIRONMENT.
       ------------------------------------------------------------ */
    // MPI_Init (&argc, &argv);
    int required = MPI_THREAD_MULTIPLE;
    int provided;
    MPI_Init_thread(&argc, &argv, required, &provided);
    if (provided < required)
    {
        int rank;
        MPI_Comm_rank(MPI_COMM_WORLD, &rank);
        if (!rank) {
	    printf("The MPI library doesn't provide MPI_THREAD_MULTIPLE \n");
	    printf("\tprovided omp_mpi_level: %d\n", provided);
        }
    }

    /* Parse command line argv[]. */
    for (cpp = argv + 1; *cpp; ++cpp)
    {
        if (**cpp == '-')
        {
            c = *(*cpp + 1);
            ++cpp;
            switch (c)
            {
            case 'h':
                printf ("Options:\n");
                printf ("\t-r <int>: process rows    (default %d)\n", nprow);
                printf ("\t-c <int>: process columns (default %d)\n", npcol);
                printf ("\t-d <int>: process Z-dimension (default %d)\n", npdep);
                exit (0);
                break;
            case 'r':
                nprow = atoi (*cpp);
                break;
            case 'c':
                npcol = atoi (*cpp);
                break;
            case 'd':
                npdep = atoi (*cpp);
                break;
            case 'l': lookahead = atoi(*cpp);
                      break;
            case 'p': rowperm = atoi(*cpp);
                      break;
            case 'q': colperm = atoi(*cpp);
                      break;
            case 'i': ir = atoi(*cpp);
                      break;
            }
        }
        else
        {   /* Last arg is considered a filename */
            if (!(fp = fopen (*cpp, "r")))
            {
                ABORT ("File does not exist");
            }
            break;
        }
    }

    /* ------------------------------------------------------------
       INITIALIZE THE SUPERLU PROCESS GRID.
       ------------------------------------------------------------ */
    superlu_gridinit3d (MPI_COMM_WORLD, nprow, npcol, npdep, &grid);

    if(grid.iam==0) {
	MPI_Query_thread(&omp_mpi_level);
	switch (omp_mpi_level) {
	case MPI_THREAD_SINGLE:
	    printf("MPI_Query_thread with MPI_THREAD_SINGLE\n");
	    fflush(stdout);
	    break;
	case MPI_THREAD_FUNNELED:
	    printf("MPI_Query_thread with MPI_THREAD_FUNNELED\n");
	    fflush(stdout);
	    break;
	case MPI_THREAD_SERIALIZED:
	    printf("MPI_Query_thread with MPI_THREAD_SERIALIZED\n");
	    fflush(stdout);
	    break;
	case MPI_THREAD_MULTIPLE:
	    printf("MPI_Query_thread with MPI_THREAD_MULTIPLE\n");
	    fflush(stdout);
	    break;
	}
	fflush(stdout);
    }

    /* Bail out if I do not belong in the grid. */
    iam = grid.iam;
    if (iam == -1)     goto out;
    if (!iam) {
	int v_major, v_minor, v_bugfix;
#ifdef __INTEL_COMPILER
	printf("__INTEL_COMPILER is defined\n");
#endif
	printf("__STDC_VERSION__ %ld\n", __STDC_VERSION__);

	superlu_dist_GetVersionNumber(&v_major, &v_minor, &v_bugfix);
	printf("Library version:\t%d.%d.%d\n", v_major, v_minor, v_bugfix);

	printf("Input matrix file:\t%s\n", *cpp);
	printf("3D process grid: %d X %d X %d\n", nprow, npcol, npdep);
	//printf("2D Process grid: %d X %d\n", (int)grid.nprow, (int)grid.npcol);
	fflush(stdout);
    }

#if ( DEBUGlevel>=1 )
    CHECK_MALLOC (iam, "Enter main()");
#endif

    /* ------------------------------------------------------------
       GET THE MATRIX FROM FILE AND SETUP THE RIGHT HAND SIDE.
       ------------------------------------------------------------ */
    for (ii = 0; ii<strlen(*cpp); ii++) {
	if((*cpp)[ii]=='.'){
	    suffix = &((*cpp)[ii+1]);
	    // printf("%s\n", suffix);
	}
    }

    // *fp0 = *fp;
    screate_matrix_postfix3d(&A, nrhs, &b, &ldb,
                             &xtrue, &ldx, fp, suffix, &(grid));
    //printf("ldx %d, ldb %d\n", ldx, ldb);

#if 0  // following code is only for checking *Gather* routine
    NRformat_loc *Astore, *Astore0;
    float* B2d;
    NRformat_loc Atmp = dGatherNRformat_loc(
                            (NRformat_loc *) A.Store,
                            b, ldb, nrhs, &B2d,
                            &grid);
    Astore = &Atmp;
    SuperMatrix Aref;
    float *bref, *xtrueref;
    if ( grid.zscp.Iam == 0 )  // only in process layer 0
    {
        screate_matrix_postfix(&Aref, nrhs, &bref, &ldb,
                               &xtrueref, &ldx, fp0,
                               suffix, &(grid.grid2d));
        Astore0 = (NRformat_loc *) Aref.Store;

	/*
	if ( (grid.grid2d).iam == 0 ) {
	    printf(" iam %d\n", 0);
	    checkNRFMT(Astore, Astore0);
	} else if ((grid.grid2d).iam == 1 ) {
	    printf(" iam %d\n", 1);
	    checkNRFMT(Astore, Astore0);
	}
	*/

	// bref, xtrueref are created on 2D
        matCheck(Astore->m_loc, nrhs, B2d, Astore->m_loc, bref, ldb);
    }
    // MPI_Finalize(); exit(0);
#endif

    /* Save two copies of the RHS */
    if ( !(b1 = floatMalloc_dist(ldb * nrhs)) )
        ABORT("Malloc fails for b1[]");
    if ( !(b2 = floatMalloc_dist(ldb * nrhs)) )
        ABORT("Malloc fails for b1[]");
    for (j = 0; j < nrhs; ++j) {
        for (i = 0; i < ldb; ++i) {
	    b1[i+j*ldb] = b[i+j*ldb];
	    b2[i+j*ldb] = b[i+j*ldb];
        }
    }

    if (!(berr = floatMalloc_dist (nrhs)))
        ABORT ("Malloc fails for berr[].");

    /* ------------------------------------------------------------
       1. SOLVE THE LINEAR SYSTEM FOR THE FIRST TIME, WITH 1 RHS.
       ------------------------------------------------------------*/
    /* Set the default input options:
       options.Fact              = DOFACT;
       options.Equil             = YES;
       options.ParSymbFact       = NO;
       options.ColPerm           = METIS_AT_PLUS_A;
       options.RowPerm           = LargeDiag_MC64;
       options.ReplaceTinyPivot  = NO;
       options.IterRefine        = SLU_DOUBLE;
       options.Trans             = NOTRANS;
       options.SolveInitialized  = NO;
       options.RefineInitialized = NO;
       options.PrintStat         = YES;
       options->num_lookaheads    = 10;
       options->lookahead_etree   = NO;
       options->SymPattern        = NO;
       options.DiagInv           = NO;
     */
    set_default_options_dist (&options);
    options.Algo3d = YES;
    options.IterRefine = SLU_SINGLE;
#if 0
    options.RowPerm = NOROWPERM;
    options.IterRefine = NOREFINE;
    options.ColPerm = NATURAL;
    options.Equil = NO;
    options.ReplaceTinyPivot = YES;
#endif

    if (rowperm != -1) options.RowPerm = rowperm;
    if (colperm != -1) options.ColPerm = colperm;
    if (lookahead != -1) options.num_lookaheads = lookahead;
    if (ir != -1) options.IterRefine = ir;

    if (!iam) {
	print_options_dist(&options);
	fflush(stdout);
    }

    // matrix is on 3D process grid
    m = A.nrow;
    n = A.ncol;

    /* Initialize ScalePermstruct and LUstruct. */
    sScalePermstructInit (m, n, &ScalePermstruct);
    sLUstructInit (n, &LUstruct);

    /* Initialize the statistics variables. */
    PStatInit (&stat);

    /* Call the linear equation solver. */
    psgssvx3d (&options, &A, &ScalePermstruct, b, ldb, nrhs, &grid,
               &LUstruct, &SOLVEstruct, berr, &stat, &info);

    if ( info ) {  /* Something is wrong */
        if ( iam==0 ) {
	    printf("ERROR: INFO = %d returned from psgssvx3d()\n", info);
	    fflush(stdout);
	}
    } else {
        /* Check the accuracy of the solution. */
        if ( !iam ) printf("\tSolve the first system:\n");
        psinf_norm_error (iam, ((NRformat_loc *) A.Store)->m_loc,
                              nrhs, b, ldb, xtrue, ldx, grid.comm);
    }

    if ( grid.zscp.Iam == 0 ) { // process layer 0
	PStatPrint (&options, &stat, &(grid.grid2d)); /* Print 2D statistics.*/
    }
    PStatFree (&stat);
    fflush(stdout);

   /* ------------------------------------------------------------
     2. NOW SOLVE ANOTHER SYSTEM WITH THE SAME A BUT DIFFERENT
     RIGHT-HAND SIDE,  WE WILL USE THE EXISTING L AND U FACTORS IN
     LUSTRUCT OBTAINED FROM A PREVIOUS FATORIZATION.
     ------------------------------------------------------------*/
    options.Fact = FACTORED; /* Indicate the factored form of A is supplied. */
    PStatInit(&stat); /* Initialize the statistics variables. */

    nrhs = 1;
    psgssvx3d (&options, &A, &ScalePermstruct, b1, ldb, nrhs, &grid,
               &LUstruct, &SOLVEstruct, berr, &stat, &info);

    if ( info ) {  /* Something is wrong */
        if ( iam==0 ) {
            printf("ERROR: INFO = %d returned from psgssvx3d()\n", info);
            fflush(stdout);
        }
    } else {
        /* Check the accuracy of the solution. */
        if ( !iam ) printf("\tSolve the system with a different B:\n");
        psinf_norm_error (iam, ((NRformat_loc *) A.Store)->m_loc,
                            nrhs, b1, ldb, xtrue, ldx, grid.comm);
    }

    /* ------------------------------------------------------------
       DEALLOCATE STORAGE.
       ------------------------------------------------------------ */
    if ( grid.zscp.Iam == 0 ) { // process layer 0

	PStatPrint (&options, &stat, &(grid.grid2d)); /* Print 2D statistics.*/
    }
        sSolveFinalize (&options, &SOLVEstruct);
        sDestroy_LU (n, &(grid.grid2d), &LUstruct);

    sDestroy_A3d_gathered_on_2d(&SOLVEstruct, &grid);

    Destroy_CompRowLoc_Matrix_dist (&A);
    SUPERLU_FREE (b);
    SUPERLU_FREE (b1);
    SUPERLU_FREE (b2);
    SUPERLU_FREE (xtrue);
    SUPERLU_FREE (berr);
    sScalePermstructFree (&ScalePermstruct);
    sLUstructFree (&LUstruct);
    PStatFree (&stat);
    fclose(fp);

    /* ------------------------------------------------------------
       RELEASE THE SUPERLU PROCESS GRID.
       ------------------------------------------------------------ */
out:
    superlu_gridexit3d (&grid);

    /* ------------------------------------------------------------
       TERMINATES THE MPI EXECUTION ENVIRONMENT.
       ------------------------------------------------------------ */
    MPI_Finalize ();

#if ( DEBUGlevel>=1 )
    CHECK_MALLOC (iam, "Exit main()");
#endif

}


int
cpp_defs ()
{
    printf (".. CPP definitions:\n");
#if ( PRNTlevel>=1 )
    printf ("\tPRNTlevel = %d\n", PRNTlevel);
#endif
#if ( DEBUGlevel>=1 )
    printf ("\tDEBUGlevel = %d\n", DEBUGlevel);
#endif
#if ( PROFlevel>=1 )
    printf ("\tPROFlevel = %d\n", PROFlevel);
#endif
    printf ("....\n");
    return 0;
}