core_ztsmqr.c

Go to the documentation of this file.

#include "common.h"

/***************************************************************************/
#if defined(PLASMA_HAVE_WEAK)
#pragma weak CORE_ztsmqr = PCORE_ztsmqr
#define CORE_ztsmqr PCORE_ztsmqr
#endif
int CORE_ztsmqr(int side, int trans,
                int M1, int N1, int M2, int N2, int K, int IB,
                PLASMA_Complex64_t *A1, int LDA1,
                PLASMA_Complex64_t *A2, int LDA2,
                PLASMA_Complex64_t *V, int LDV,
                PLASMA_Complex64_t *T, int LDT,
                PLASMA_Complex64_t *WORK, int LDWORK)
{
    int i, i1, i3;
    int NQ, NW;
    int kb;
    int ic = 0;
    int jc = 0;
    int mi = M1;
    int ni = N1;

    /* Check input arguments */
    if ((side != PlasmaLeft) && (side != PlasmaRight)) {
        coreblas_error(1, "Illegal value of side");
        return -1;
    }

    /* NQ is the order of Q */
    if (side == PlasmaLeft) {
        NQ = M2;
        NW = IB;
    }
    else {
        NQ = N2;
        NW = M1;
    }

    if ((trans != PlasmaNoTrans) && (trans != PlasmaConjTrans)) {
        coreblas_error(2, "Illegal value of trans");
        return -2;
    }
    if (M1 < 0) {
        coreblas_error(3, "Illegal value of M1");
        return -3;
    }
    if (N1 < 0) {
        coreblas_error(4, "Illegal value of N1");
        return -4;
    }
    if ( (M2 < 0) || 
         ( (M2 != M1) && (side == PlasmaRight) ) ){
        coreblas_error(5, "Illegal value of M2");
        return -5;
    }
    if ( (N2 < 0) || 
         ( (N2 != N1) && (side == PlasmaLeft) ) ){
        coreblas_error(6, "Illegal value of N2");
        return -6;
    }
    if ((K < 0) || 
        ( (side == PlasmaLeft)  && (K > M1) ) ||
        ( (side == PlasmaRight) && (K > N1) ) ) {
        coreblas_error(7, "Illegal value of K");
        return -7;
    }
    if (IB < 0) {
        coreblas_error(8, "Illegal value of IB");
        return -8;
    }
    if (LDA1 < max(1,M1)){
        coreblas_error(10, "Illegal value of LDA1");
        return -10;
    }
    if (LDA2 < max(1,M2)){
        coreblas_error(12, "Illegal value of LDA2");
        return -12;
    }
    if (LDV < max(1,NQ)){
        coreblas_error(14, "Illegal value of LDV");
        return -14;
    }
    if (LDT < max(1,IB)){
        coreblas_error(16, "Illegal value of LDT");
        return -16;
    }
    if (LDWORK < max(1,NW)){
        coreblas_error(18, "Illegal value of LDWORK");
        return -18;
    }

    /* Quick return */
    if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0))
        return PLASMA_SUCCESS;

    if (((side == PlasmaLeft)  && (trans != PlasmaNoTrans))
        || ((side == PlasmaRight) && (trans == PlasmaNoTrans))) {
        i1 = 0;
        i3 = IB;
    }
    else {
        i1 = ((K-1) / IB)*IB;
        i3 = -IB;
    }

    for(i = i1; (i > -1) && (i < K); i += i3) {
        kb = min(IB, K-i);

        if (side == PlasmaLeft) {
            /*
             * H or H' is applied to C(i:m,1:n)
             */
            mi = M1 - i;
            ic = i;
        }
        else {
            /*
             * H or H' is applied to C(1:m,i:n)
             */
            ni = N1 - i;
            jc = i;
        }
        /*
         * Apply H or H' (NOTE: CORE_zparfb used to be CORE_ztsrfb)
         */
        CORE_zparfb(
            side, trans, PlasmaForward, PlasmaColumnwise,
            mi, ni, M2, N2, kb, 0,
            &A1[LDA1*jc+ic], LDA1,
            A2, LDA2,
            &V[LDV*i], LDV,
            &T[LDT*i], LDT,
            WORK, LDWORK);
    }
    return PLASMA_SUCCESS;
}

/***************************************************************************/
void QUARK_CORE_ztsmqr(Quark *quark, Quark_Task_Flags *task_flags,
                       int side, int trans,
                       int m1, int n1, int m2, int n2, int k, int ib, int nb,
                       PLASMA_Complex64_t *A1, int lda1,
                       PLASMA_Complex64_t *A2, int lda2,
                       PLASMA_Complex64_t *V, int ldv,
                       PLASMA_Complex64_t *T, int ldt)
{
    int ldwork = side == PlasmaLeft ? ib : nb;

    DAG_CORE_TSMQR;
    QUARK_Insert_Task(quark, CORE_ztsmqr_quark, task_flags,
        sizeof(PLASMA_enum),                &side,  VALUE,
        sizeof(PLASMA_enum),                &trans, VALUE,
        sizeof(int),                        &m1,    VALUE,
        sizeof(int),                        &n1,    VALUE,
        sizeof(int),                        &m2,    VALUE,
        sizeof(int),                        &n2,    VALUE,
        sizeof(int),                        &k,     VALUE,
        sizeof(int),                        &ib,    VALUE,
        sizeof(PLASMA_Complex64_t)*nb*nb,    A1,            INOUT,
        sizeof(int),                        &lda1,  VALUE,
        sizeof(PLASMA_Complex64_t)*nb*nb,    A2,            INOUT | LOCALITY,
        sizeof(int),                        &lda2,  VALUE,
        sizeof(PLASMA_Complex64_t)*nb*nb,    V,             INPUT,
        sizeof(int),                        &ldv,   VALUE,
        sizeof(PLASMA_Complex64_t)*ib*nb,    T,             INPUT,
        sizeof(int),                        &ldt,   VALUE,
        sizeof(PLASMA_Complex64_t)*ib*nb,    NULL,          SCRATCH,
        sizeof(int),                        &ldwork, VALUE,
        0);
}

/***************************************************************************/
#if defined(PLASMA_HAVE_WEAK)
#pragma weak CORE_ztsmqr_quark = PCORE_ztsmqr_quark
#define CORE_ztsmqr_quark PCORE_ztsmqr_quark
#endif
void CORE_ztsmqr_quark(Quark *quark)
{
    int side;
    int trans;
    int m1;
    int n1;
    int m2;
    int n2;
    int k;
    int ib;
    PLASMA_Complex64_t *A1;
    int lda1;
    PLASMA_Complex64_t *A2;
    int lda2;
    PLASMA_Complex64_t *V;
    int ldv;
    PLASMA_Complex64_t *T;
    int ldt;
    PLASMA_Complex64_t *WORK;
    int ldwork;

    quark_unpack_args_18(quark, side, trans, m1, n1, m2, n2, k, ib, 
                         A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
    CORE_ztsmqr(side, trans, m1, n1, m2, n2, k, ib, 
                A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
}