cpucblk_dmpi_rhs_bwd.c

Go to the documentation of this file.

/**
 *
 * @file cpucblk_dmpi_rhs_bwd.c
 *
 * Precision dependent routines to manag communications for the solve part.
 *
 * @copyright 2015-2023 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
 *                      Univ. Bordeaux. All rights reserved.
 *
 * @version 6.3.2
 * @author Mathieu Faverge
 * @author Alycia Lisito
 * @date 2023-12-11
 *
 * @generated from /builds/solverstack/pastix/kernels/cpucblk_zmpi_rhs_bwd.c, normal z -> d, Wed Dec 13 12:09:17 2023
 *
 **/
#include "common/common.h"
#include "common/pastixdata.h"
#include <lapacke.h>
#include "blend/solver.h"
#include "blend/solver_comm_matrix.h"
#include "kernels.h"
#include "pastix_dcores.h"
#include "pastix_dlrcores.h"
 
#if defined( PASTIX_WITH_MPI )
/**
 *******************************************************************************
 *
 * @brief Asynchronously sends the rhs associated to a cblk->lcolidx to the remote node.
 *
 *******************************************************************************
 *
 * @param[in] solvmtx
 *          The solver matrix holding the communicator.
 *
 * @param[in] rhsb
 *          The pointer to the rhs data structure that holds the vectors of the
 *          right hand side.
 *
 * @param[in] cblk
 *          The cblk which defines the part to sent.
 *
 *******************************************************************************/
void
cpucblk_disend_rhs_bwd( SolverMatrix *solvmtx,
                        pastix_rhs_t  rhsb,
                        SolverCblk   *cblk )
{
    MPI_Request         request;
    double *b;
    pastix_int_t        colnbr = cblk_colnbr(cblk);
    pastix_int_t        size   = colnbr * rhsb->n;
    pastix_int_t        idx    = - cblk->bcscnum - 1;
    int                 rc;
 
    assert( colnbr <= solvmtx->colmax );
    assert( cblk->cblktype & CBLK_RECV );
 
#if defined (PASTIX_DEBUG_MPI)
    fprintf( stderr, "[%2d] RHS Bwd: Post Isend cblk %ld to %2d at index %ld of size %ld\n",
             solvmtx->clustnum, (long)cblk->gcblknum, cblk->ownerid,
             (long)cblk->lcolidx, (long)colnbr );
#endif
 
    /* Get the pointer to the right hand side */
    b = rhsb->b;
    b += cblk->lcolidx;
 
    /* Pack the data into a temporary buffer if non-contiguous */
    assert( rhsb->cblkb[ idx ] == NULL );
    if ( rhsb->n > 1 ) {
        rhsb->cblkb[ idx ] = malloc( size * sizeof(double) );
 
        rc = LAPACKE_dlacpy_work( LAPACK_COL_MAJOR, 'A', colnbr, rhsb->n,
                                  b, rhsb->ld, rhsb->cblkb[ idx ], colnbr );
        assert( rc == 0 );
 
        b = rhsb->cblkb[ idx ];
    }
 
    rc = MPI_Isend( b, size, PASTIX_MPI_DOUBLE,
                    cblk->ownerid, cblk->gcblknum, solvmtx->solv_comm, &request );
    assert( rc == MPI_SUCCESS );
 
    solverCommMatrixAdd( solvmtx, cblk->ownerid, size * sizeof(double) );
 
    /* Register the request to make it progress */
    pastix_atomic_lock( &(solvmtx->reqlock) );
 
    assert( solvmtx->reqidx[ solvmtx->reqnum ] == -1 );
    assert( solvmtx->reqnum >= 0 );
    assert( solvmtx->reqnum < solvmtx->reqnbr );
 
    solvmtx->reqtab[ solvmtx->reqnum ] = request;
    solvmtx->reqidx[ solvmtx->reqnum ] = cblk - solvmtx->cblktab;
    solvmtx->reqnum++;
 
    pastix_atomic_unlock( &(solvmtx->reqlock) );
 
    (void)rc;
}
 
/**
 *******************************************************************************
 *
 * @brief Handle a finished request on a fanin
 *
 *******************************************************************************
 *
 * @param[in] enums
 *          Enums needed for the solve.
 *
 * @param[in] solvmtx
 *          The solver matrix structure.
 *
 * @param[in] rhsb
 *          The pointer to the rhs data structure that holds the vectors of the
 *          right hand side.
 *
 * @param[inout] cblk
 *          The cblk concerned by the computation.
 *
 *******************************************************************************/
void
cpucblk_drequest_rhs_bwd_handle_send( const args_solve_t *enums,
                                      SolverMatrix       *solvmtx,
                                      pastix_rhs_t        rhsb,
                                      const SolverCblk   *cblk )
{
    pastix_int_t idx = - cblk->bcscnum - 1;
 
    assert( cblk->cblktype & CBLK_RECV );
    assert( enums->solve_step == PastixSolveBackward );
 
#if defined(PASTIX_DEBUG_MPI)
    {
        size_t cblksize = cblk_colnbr( cblk ) * rhsb->n * sizeof(double);
 
        fprintf( stderr, "[%2d] RHS Bwd: Isend for cblk %ld toward %2d ( %ld Bytes ) (DONE)\n",
                 solvmtx->clustnum, (long)cblk->gcblknum, cblk->ownerid, (long)cblksize );
    }
#endif
 
    if ( rhsb->cblkb[ idx ] ) {
        memFree_null( rhsb->cblkb[ idx ] );
    }
    (void)solvmtx;
    (void)enums;
}
 
/**
 *******************************************************************************
 *
 * @brief Handle a finished request on a recv cblk.
 *
 *******************************************************************************
 *
 * @param[in] enums
 *          Enums needed for the solve.
 *
 * @param[inout] solvmtx
 *          The solver matrix structure.
 *
 * @param[in] rhsb
 *          The pointer to the rhs data structure that holds the vectors of the
 *          right hand side.
 *
 * @param[inout] threadid
 *          Id of the thread calling this method.
 *
 * @param[inout] status
 *          Statuses of the completed request.
 *
 * @param[inout] recvbuf
 *          Reception buffer of the completed request.
 *
 *******************************************************************************/
static inline void
cpucblk_drequest_rhs_bwd_handle_recv( const args_solve_t *enums,
                                      SolverMatrix       *solvmtx,
                                      pastix_rhs_t        rhsb,
                                      int                 threadid,
                                      const MPI_Status   *status,
                                      double *recvbuf )
{
    SolverCblk   *cblk;
    int           src  = status->MPI_SOURCE;
    int           tag  = status->MPI_TAG;
    pastix_int_t  idx;
 
    assert( ( 0 <= src ) && ( src < solvmtx->clustnbr ) );
    assert( ( 0 <= tag ) && ( tag < solvmtx->gcblknbr ) );
 
    /*
     * Let's look for the local cblk
     */
    cblk = solvmtx->cblktab + solvmtx->gcbl2loc[ tag ];
    assert( cblk->cblktype & CBLK_FANIN );
 
#if defined(PASTIX_DEBUG_MPI)
    {
        pastix_int_t colnbr;
        pastix_int_t size;
        int          rc;
        int          count = 0;
 
        colnbr = cblk_colnbr( cblk );
        size   = colnbr * rhsb->n * sizeof(double);
 
        rc = MPI_Get_count( status, MPI_CHAR, &count );
        assert( rc == MPI_SUCCESS );
        assert( count == size );
 
        /* We can't know the sender easily, so we don't print it */
        fprintf( stderr, "[%2d] RHS Bwd : recv of size %d/%ld for cblk %ld (DONE)\n",
                 solvmtx->clustnum, count, (long)size, (long)cblk->gcblknum );
    }
#endif
 
    idx = - cblk->bcscnum - 1;
    rhsb->cblkb[ idx ] = recvbuf;
 
    cblk->threadid = threadid;
    if ( solvmtx->computeQueue ) {
        pastix_queue_t   *queue = solvmtx->computeQueue[ cblk->threadid ];
        const SolverBlok *blok  = solvmtx->bloktab + solvmtx->browtab[ cblk[1].brownum-1 ];
        const SolverCblk *fcbk  = solvmtx->cblktab + blok->lcblknm;
 
        pqueuePush1( queue, cblk - solvmtx->cblktab, - fcbk->priority );
    }
    else {
        solve_cblk_dtrsmsp_backward( enums, solvmtx, cblk, rhsb );
        /* Check it has been freed */
        assert( rhsb->cblkb[ idx ] == NULL );
    }
 
    (void)src;
}
 
/**
 *******************************************************************************
 *
 * @brief Handle a finished request.
 *
 * If cblktype & CBLK_FANIN : Will deallocate the coeftab
 * If cblktype & CBLK_RECV  : Will add cblk and deallocate the coeftab
 *
 *******************************************************************************
 *
 * @param[in] enums
 *          Enums needed for the solve.
 *
 * @param[inout] solvmtx
 *          The solver matrix structure.
 *
 * @param[in] rhsb
 *          The pointer to the rhs data structure that holds the vectors of the
 *          right hand side.
 *
 * @param[in] threadid
 *          Id of the thread calling this method.
 *
 * @param[in] outcount
 *          Amount of finshed requests.
 *
 * @param[in] indexes
 *          Array of completed requests.
 *
 * @param[in] statuses
 *          Array of statuses for the completed requests.
 *
 *******************************************************************************
 *
 * @retval The amount of finished requests updated.
 *
 *******************************************************************************/
static inline int
cpucblk_drequest_rhs_bwd_handle( const args_solve_t *enums,
                                 SolverMatrix       *solvmtx,
                                 pastix_rhs_t        rhsb,
                                 int                 threadid,
                                 int                 outcount,
                                 const int          *indexes,
                                 const MPI_Status   *statuses )
{
    pastix_int_t i, reqid;
    int          nbrequest = outcount;
 
    for ( i = 0; i < outcount; i++ ) {
        reqid = indexes[i];
 
        /*
         * Handle the reception
         */
        if ( solvmtx->reqidx[reqid] == -1 ) {
            /* We're on a cblk recv, copy datas and restart communications */
            double *recvbuf;
            MPI_Status status;
            int        size;
 
            memcpy( &status, statuses + i, sizeof(MPI_Status) );
            MPI_Get_count( &status, PASTIX_MPI_DOUBLE, &size );
 
            MALLOC_INTERN( recvbuf, size, double );
            memcpy( recvbuf, solvmtx->rcoeftab, size * sizeof(double) );
 
            solvmtx->fanincnt--;
 
            /* Let's restart the communication */
            assert( solvmtx->fanincnt >= 0 );
            if ( solvmtx->fanincnt > 0 ) {
                MPI_Start( solvmtx->reqtab + reqid );
                nbrequest--;
            }
            else {
                MPI_Request_free( solvmtx->reqtab + reqid );
                solvmtx->reqtab[reqid] = MPI_REQUEST_NULL;
            }
 
            cpucblk_drequest_rhs_bwd_handle_recv( enums, solvmtx, rhsb,
                                                  threadid, &status, recvbuf );
        }
        /*
         * Handle the emission
         */
        else {
            SolverCblk *cblk = solvmtx->cblktab + solvmtx->reqidx[ reqid ];
            assert( cblk->cblktype & CBLK_RECV );
 
            cpucblk_drequest_rhs_bwd_handle_send( enums, solvmtx, rhsb, cblk );
 
#if !defined(NDEBUG)
            solvmtx->reqidx[ reqid ] = -1;
#endif
            solvmtx->recvcnt--;
        }
    }
 
    return nbrequest;
}
 
/**
 *******************************************************************************
 *
 * @ingroup kernel_fact
 * @brief Progress communications for one process
 *
 * If a communication is completed, it will be treated.
 * If cblktype & CBLK_FANIN : Will deallocate coeftab
 * If cblktype & CBLK_RECV  : Will add cblk to fcblk
 *
 *******************************************************************************
 *
 * @param[in] enums
 *          Enums needed for the solve.
 *
 * @param[inout] solvmtx
 *          The solver matrix structure.
 *
 * @param[in] rhsb
 *          The pointer to the rhs data structure that holds the vectors of the
 *          right hand side.
 *
 * @param[in] threadid
 *          Id of the thread calling this method.
 *
 *******************************************************************************/
void
cpucblk_dmpi_rhs_bwd_progress( const args_solve_t *enums,
                               SolverMatrix       *solvmtx,
                               pastix_rhs_t        rhsb,
                               int                 threadid )
{
    pthread_t  tid = pthread_self();
    int        outcount = 1;
    int        nbrequest, nbfree;
    int        indexes[ solvmtx->reqnbr ];
    MPI_Status statuses[ solvmtx->reqnbr ];
 
    /* Check if someone is already communicating or not */
    pthread_mutex_lock( &pastix_comm_lock );
    if ( pastix_comm_tid == (pthread_t)-1 ) {
        pastix_comm_tid = tid;
    }
    pthread_mutex_unlock( &pastix_comm_lock );
 
    if ( tid != pastix_comm_tid ) {
        pastix_yield();
        return;
    }
 
    /*
     * Let's register the number of active requests.
     * We now suppose that the current thread is working on the first nbrequest
     * active in the reqtab array. Additional requests can be posted during this
     * progression, but it will be with a larger index. Thus, we do not need to
     * protect every changes in these requests.
     * When this is done, the requests arrays is locked to be packed, and the
     * number of requests is updated for the next round.
     */
    pastix_atomic_lock( &(solvmtx->reqlock) );
    nbrequest = solvmtx->reqnum;
    pastix_atomic_unlock( &(solvmtx->reqlock) );
 
    while( (outcount > 0) && (nbrequest > 0) )
    {
        MPI_Testsome( nbrequest, solvmtx->reqtab, &outcount, indexes, statuses );
        nbfree = 0;
 
        /* Handle all the completed requests */
        if ( outcount > 0 ) {
            nbfree = cpucblk_drequest_rhs_bwd_handle( enums, solvmtx, rhsb, threadid,
                                                      outcount, indexes, statuses );
        }
 
        /*
         * Pack the request arrays, and update the number of active requests by
         * removing the completed ones
         */
        pastix_atomic_lock( &(solvmtx->reqlock) );
        if ( nbfree > 0 ) {
            cpucblk_dupdate_reqtab( solvmtx );
        }
        nbrequest = solvmtx->reqnum;
        pastix_atomic_unlock( &(solvmtx->reqlock) );
    }
 
    pastix_comm_tid = (pthread_t)-1;
    pastix_yield();
}
#endif /* defined(PASTIX_WITH_MPI) */
 
/**
 *******************************************************************************
 *
 * @brief Wait for incoming dependencies, and return when cblk->ctrbcnt has
 * reached 0.
 *
 *******************************************************************************
 *
 * @param[in] rank
 *          The rank of the current thread.
 *
 * @param[in] enums
 *          Enums needed for the solve.
 *
 * @param[inout] solvmtx
 *          The solver matrix structure.
 *
 * @param[inout] cblk
 *          The column block that contribute to fcblk.
 *
 * @param[in] rhsb
 *          The pointer to the rhs data structure that holds the vectors of the
 *          right hand side.
 *
 *******************************************************************************
 *
 * @return 1 if the cblk is a fanin, 0 otherwise
 *
 *******************************************************************************/
int
cpucblk_dincoming_rhs_bwd_deps( int                 rank,
                                const args_solve_t *enums,
                                SolverMatrix       *solvmtx,
                                SolverCblk         *cblk,
                                pastix_rhs_t        rhsb )
{
#if defined(PASTIX_WITH_MPI)
    if ( cblk->cblktype & CBLK_FANIN ) {
        /*
         * We are in the sequential case, we progress on communications and
         * return if nothing.
         */
        //cpucblk_dtestsome( side, solvmtx );
        return 1;
    }
 
    if ( cblk->cblktype & CBLK_RECV ) {
        return 1;
    }
 
    /* Make sure we receive every contribution */
    while( cblk->ctrbcnt > 0 ) {
        cpucblk_dmpi_rhs_bwd_progress( enums, solvmtx, rhsb, rank );
    }
#else
    assert( !(cblk->cblktype & (CBLK_FANIN | CBLK_RECV)) );
    do { pastix_yield(); } while( cblk->ctrbcnt > 0 );
#endif
 
    (void)rank;
    (void)enums;
    (void)solvmtx;
    (void)rhsb;
 
    return 0;
}
 
/**
 *******************************************************************************
 *
 * @brief Release the dependencies of the given cblk after an update.
 *
 *******************************************************************************
 *
 * @param[in] enums
 *          Enums needed for the solve.
 *
 * @param[inout] solvmtx
 *          The solver matrix structure.
 *
 * @param[in] rhsb
 *          The pointer to the rhs data structure that holds the vectors of the
 *          right hand side.
 *
 * @param[in] cblk
 *          The column block that contribute to fcblk.
 *
 * @param[inout] fcbk
 *          The facing column block that is updated by cblk.
 *
 *******************************************************************************/
void
cpucblk_drelease_rhs_bwd_deps( const args_solve_t *enums,
                               SolverMatrix       *solvmtx,
                               pastix_rhs_t        rhsb,
                               const SolverCblk   *cblk,
                               SolverCblk         *fcbk )
{
    int32_t ctrbcnt;
    ctrbcnt = pastix_atomic_dec_32b( &(fcbk->ctrbcnt) );
    if ( !ctrbcnt ) {
#if defined(PASTIX_WITH_MPI)
        if ( fcbk->cblktype & CBLK_RECV ) {
            cpucblk_disend_rhs_bwd( solvmtx, rhsb, fcbk );
            return;
        }
#else
        (void)rhsb;
#endif
        if ( solvmtx->computeQueue ) {
            pastix_queue_t *queue = solvmtx->computeQueue[ cblk->threadid ];
            assert( fcbk->priority != -1 );
            pqueuePush1( queue, fcbk - solvmtx->cblktab, - fcbk->priority );
        }
    }
    (void)enums;
}
 
/**
 *******************************************************************************
 *
 * @brief  Waitall routine for current cblk request
 *
 * It may be possible that some cblk will not be deallocated with the static
 * scheduler. So a cleanup may be necessary.
 *
 *******************************************************************************
 *
 * @param[in] enums
 *          Enums needed for the solve.
 *
 * @param[in] sched
 *          Define which sched is used
 *          @arg PastixSchedSequential if sequential
 *          @arg PastixSchedStatic if multi-threaded static scheduler
 *          @arg PastixSchedDynamic if multi-threaded dynamic scheduler
 *          No other scheduler is supported.
 *
 * @param[inout] solvmtx
 *          The solver matrix structure.
 *
 * @param[in] rhsb
 *          The pointer to the rhs data structure that holds the vectors of the
 *          right hand side.
 *
 *******************************************************************************/
void
cpucblk_drequest_rhs_bwd_cleanup( const args_solve_t *enums,
                                  pastix_int_t        sched,
                                  SolverMatrix       *solvmtx,
                                  pastix_rhs_t        rhsb )
{
    if ( (sched != PastixSchedSequential) &&
         (sched != PastixSchedStatic)     &&
         (sched != PastixSchedDynamic) )
    {
        return;
    }
#if defined(PASTIX_WITH_MPI)
    pastix_int_t i;
    int          rc;
    SolverCblk  *cblk;
    int          reqnbr = solvmtx->reqnum;
    MPI_Status   status;
 
#if defined(PASTIX_DEBUG_MPI)
    fprintf( stderr, "[%2d] Wait for all pending communications\n",
             solvmtx->clustnum );
#endif
 
    for ( i=0; i<reqnbr; i++ ) {
        if ( solvmtx->reqtab[i] == MPI_REQUEST_NULL ) {
            assert( 0 /* MPI_REQUEST_NULL should have been pushed to the end */ );
            solvmtx->reqnum--;
            continue;
        }
 
        /* Make sure that we don't have an already cleaned request in dynamic */
        assert( solvmtx->reqidx[i] != -1 );
 
        rc = MPI_Wait( solvmtx->reqtab + i, &status );
        assert( rc == MPI_SUCCESS );
 
        cblk = solvmtx->cblktab + solvmtx->reqidx[i];
 
        /* We should wait only for recv */
        assert( cblk->cblktype & CBLK_RECV );
 
        cpucblk_drequest_rhs_bwd_handle_send( enums, solvmtx, rhsb, cblk );
 
        solvmtx->reqnum--;
    }
    assert( solvmtx->reqnum == 0 );
    (void)rc;
#else
    (void)enums;
    (void)solvmtx;
    (void)rhsb;
#endif
}