ztrtri.c

Go to the documentation of this file.

#include "common.h"

/***************************************************************************/
int PLASMA_ztrtri(PLASMA_enum uplo, PLASMA_enum diag, int N,
                  PLASMA_Complex64_t *A, int LDA)
{
    int NB;
    int status;
    plasma_context_t *plasma;
    PLASMA_sequence *sequence = NULL;
    PLASMA_request request = PLASMA_REQUEST_INITIALIZER;
    PLASMA_desc descA;

    plasma = plasma_context_self();
    if (plasma == NULL) {
        plasma_fatal_error("PLASMA_ztrtri", "PLASMA not initialized");
        return PLASMA_ERR_NOT_INITIALIZED;
    }
    /* Check input arguments */
    if (uplo != PlasmaUpper && uplo != PlasmaLower) {
        plasma_error("PLASMA_ztrtri", "illegal value of uplo");
        return -1;
    }
    if (diag != PlasmaUnit && diag != PlasmaNonUnit) {
        plasma_error("PLASMA_ztrtri", "illegal value of diag");
        return -2;
    }
    if (N < 0) {
        plasma_error("PLASMA_ztrtri", "illegal value of N");
        return -3;
    }
    if (LDA < max(1, N)) {
        plasma_error("PLASMA_ztrtri", "illegal value of LDA");
        return -5;
    }
    /* Quick return */
    if (max(N, 0) == 0)
        return PLASMA_SUCCESS;

    /* Tune NB depending on M, N & NRHS; Set NBNB */
    status = plasma_tune(PLASMA_FUNC_ZPOSV, N, N, 0);
    if (status != PLASMA_SUCCESS) {
        plasma_error("PLASMA_ztrtri", "plasma_tune() failed");
        return status;
    }

    /* Set NT */
    NB   = PLASMA_NB;

    plasma_sequence_create(plasma, &sequence);

    if ( PLASMA_TRANSLATION == PLASMA_OUTOFPLACE ) {
        plasma_zooplap2tile( descA, A, NB, NB, LDA, N, 0, 0, N, N, plasma_desc_mat_free(&(descA)) );
    } else {
        plasma_ziplap2tile(  descA, A, NB, NB, LDA, N, 0, 0, N, N);
    }

    /* Call the tile interface */
    PLASMA_ztrtri_Tile_Async(uplo, diag, &descA, sequence, &request);

    if ( PLASMA_TRANSLATION == PLASMA_OUTOFPLACE ) {
        plasma_zooptile2lap( descA, A, NB, NB, LDA, N );
        plasma_dynamic_sync();
        plasma_desc_mat_free(&descA);
    } else {
        plasma_ziptile2lap( descA, A, NB, NB, LDA, N );
        plasma_dynamic_sync();
    }

    status = sequence->status;
    plasma_sequence_destroy(plasma, sequence);
    return status;
}

/***************************************************************************/
int PLASMA_ztrtri_Tile(PLASMA_enum uplo, PLASMA_enum diag, PLASMA_desc *A)
{
    plasma_context_t *plasma;
    PLASMA_sequence *sequence = NULL;
    PLASMA_request request = PLASMA_REQUEST_INITIALIZER;
    int status;

    plasma = plasma_context_self();
    if (plasma == NULL) {
        plasma_fatal_error("PLASMA_ztrtri_Tile", "PLASMA not initialized");
        return PLASMA_ERR_NOT_INITIALIZED;
    }
    plasma_sequence_create(plasma, &sequence);
    PLASMA_ztrtri_Tile_Async(uplo, diag, A, sequence, &request);
    plasma_dynamic_sync();
    status = sequence->status;
    plasma_sequence_destroy(plasma, sequence);
    return status;
}

/***************************************************************************/
int PLASMA_ztrtri_Tile_Async(PLASMA_enum uplo, PLASMA_enum diag, PLASMA_desc *A,
                             PLASMA_sequence *sequence, PLASMA_request *request)
{
    PLASMA_desc descA = *A;
    plasma_context_t *plasma;

    plasma = plasma_context_self();
    if (plasma == NULL) {
        plasma_fatal_error("PLASMA_ztrtri_Tile", "PLASMA not initialized");
        return PLASMA_ERR_NOT_INITIALIZED;
    }
    if (sequence == NULL) {
        plasma_fatal_error("PLASMA_ztrtri_Tile", "NULL sequence");
        return PLASMA_ERR_UNALLOCATED;
    }
    if (request == NULL) {
        plasma_fatal_error("PLASMA_ztrtri_Tile", "NULL request");
        return PLASMA_ERR_UNALLOCATED;
    }
    /* Check sequence status */
    if (sequence->status == PLASMA_SUCCESS)
        request->status = PLASMA_SUCCESS;
    else
        return plasma_request_fail(sequence, request, PLASMA_ERR_SEQUENCE_FLUSHED);

    /* Check descriptors for correctness */
    if (plasma_desc_check(&descA) != PLASMA_SUCCESS) {
        plasma_error("PLASMA_ztrtri_Tile", "invalid descriptor");
        return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
    }
    /* Check input arguments */
    if (descA.nb != descA.mb) {
        plasma_error("PLASMA_ztrtri_Tile", "only square tiles supported");
        return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
    }
    if (uplo != PlasmaUpper && uplo != PlasmaLower) {
        plasma_error("PLASMA_ztrtri_Tile", "illegal value of uplo");
        return plasma_request_fail(sequence, request, -1);
    }
    if (diag != PlasmaUnit && diag != PlasmaNonUnit) {
        plasma_error("PLASMA_ztrtri_Tile", "illegal value of diag");
        return plasma_request_fail(sequence, request, -2);
    }
    /* Quick return */
/*
    if (max(N, 0) == 0)
        return PLASMA_SUCCESS;
*/
    plasma_dynamic_call_5(plasma_pztrtri,
        PLASMA_enum, uplo,
        PLASMA_enum, diag,
        PLASMA_desc, descA,
        PLASMA_sequence*, sequence,
        PLASMA_request*, request);

    return PLASMA_SUCCESS;
}