/***************************************************************************//**
*
* @ingroup double_Tile_Async
*
* PLASMA_dsygst_Tile_Async - reduces a complex Hermitian-definite
* generalized eigenproblem to standard form.
* If PlasmaItype == 1, the problem is A*x = lambda*B*x, and A is
* overwritten by inv(U**T)*A*inv(U) or inv(L)*A*inv(L**T)
* If PlasmaItype == 2 or 3, the problem is A*B*x = lambda*x or B*A*x
* = lambda*x, and A is overwritten by U*A*U**T or L**T*A*L. B must
* have been previously factorized as U**T*U or L*L**T by
* PLASMA_DPOTRF.
* ONLY PlasmaItype == 1 and PlasmaLower supported!
* Non-blocking equivalent of PLASMA_dsygst_Tile().
* May return before the computation is finished.
* Allows for pipelining of operations ar runtime.
*
*******************************************************************************
*
* @param[in] sequence
* Identifies the sequence of function calls that this call belongs to
* (for completion checks and exception handling purposes).
*
* @param[out] request
* Identifies this function call (for exception handling purposes).
*
*******************************************************************************
*
* @sa PLASMA_dsygst
* @sa PLASMA_dsygst_Tile
* @sa PLASMA_chegst_Tile_Async
* @sa PLASMA_dsygst_Tile_Async
* @sa PLASMA_ssygst_Tile_Async
* @sa PLASMA_dsygv_Tile_Async
*
******************************************************************************/
int PLASMA_dsygst_Tile_Async(PLASMA_enum itype, PLASMA_enum uplo,
PLASMA_desc *A,
PLASMA_desc *B,
PLASMA_sequence *sequence, PLASMA_request *request)
{
PLASMA_desc descA = *A;
PLASMA_desc descB = *B;
plasma_context_t *plasma;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_fatal_error("PLASMA_dsygst_Tile", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
if (sequence == NULL) {
plasma_fatal_error("PLASMA_dsygst_Tile", "NULL sequence");
return PLASMA_ERR_UNALLOCATED;
}
if (request == NULL) {
plasma_fatal_error("PLASMA_dsygst_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_dsygst_Tile", "invalid first descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if (plasma_desc_check(&descB) != PLASMA_SUCCESS) {
plasma_error("PLASMA_dsygst_Tile", "invalid second descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Check input arguments */
if (descA.nb != descA.mb) {
plasma_error("PLASMA_dsygst_Tile", "only square tiles supported");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/*
* Transform Hermitian-definite generalized eigenproblem
* to standard form
*/
plasma_dynamic_call_6(plasma_pdsygst,
PLASMA_enum, itype,
PLASMA_enum, uplo,
PLASMA_desc, descA,
PLASMA_desc, descB,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
return PLASMA_SUCCESS;
}
/***************************************************************************//**
*
* @ingroup PLASMA_Complex64_t_Tile_Async
*
* PLASMA_zlaswp_Tile_Async - performs a series of row interchanges
* on the matrix A. One row interchange is initiated for each of
* rows K1 through K2 of A.
* Non-blocking equivalent of PLASMA_zlaswp_Tile().
* May return before the computation is finished.
* Allows for pipelining of operations ar runtime.
*
*******************************************************************************
*
* @param[in] sequence
* Identifies the sequence of function calls that this call belongs to
* (for completion checks and exception handling purposes).
*
* @param[out] request
* Identifies this function call (for exception handling purposes).
*
*******************************************************************************
*
* @sa PLASMA_zlaswp
* @sa PLASMA_zlaswp_Tile
* @sa PLASMA_claswp_Tile_Async
* @sa PLASMA_dlaswp_Tile_Async
* @sa PLASMA_slaswp_Tile_Async
* @sa PLASMA_zgetrf_Tile_Async
*
******************************************************************************/
int PLASMA_zlaswp_Tile_Async(PLASMA_desc *A, int K1, int K2, int *IPIV, int INCX,
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_zlaswp_Tile", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
if (sequence == NULL) {
plasma_fatal_error("PLASMA_zlaswp_Tile", "NULL sequence");
return PLASMA_ERR_UNALLOCATED;
}
if (request == NULL) {
plasma_fatal_error("PLASMA_zlaswp_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_zlaswp_Tile", "invalid first descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if ( (K1 != 1) || (K2 != descA.m) ) {
plasma_error("PLASMA_zlaswp_Tile", "invalid K1 or K2 (1..M is the only interval supported right now)");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
plasma_dynamic_call_3(
plasma_pzbarrier_tl2pnl,
PLASMA_desc, descA,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
/* swap */
plasma_dynamic_call_5(
plasma_pzlaswp,
PLASMA_desc, descA,
int *, IPIV,
int, INCX,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
plasma_dynamic_call_3(
plasma_pzbarrier_pnl2tl,
PLASMA_desc, descA,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
return PLASMA_SUCCESS;
}
/***************************************************************************//**
*
**/
int plasma_alloc_ibnb_tile(int M, int N, PLASMA_enum func, int type, PLASMA_desc **desc)
{
int status;
int IB, NB, MT, NT;
plasma_context_t *plasma;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_fatal_error("plasma_alloc_ibnb_tile", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
/* Tune NB & IB depending on M & N; Set IBNBSIZE */
status = plasma_tune(func, M, N, 0);
if (status != PLASMA_SUCCESS) {
plasma_error("plasma_alloc_ibnb_tile", "plasma_tune() failed");
return PLASMA_ERR_UNEXPECTED;
}
/* Set MT & NT & allocate */
NB = PLASMA_NB;
IB = PLASMA_IB;
MT = (M%NB==0) ? (M/NB) : (M/NB+1);
NT = (N%NB==0) ? (N/NB) : (N/NB+1);
/* Size is doubled for RH QR to store the reduction T */
if ((plasma->householder != PLASMA_FLAT_HOUSEHOLDER) &&
((func == PLASMA_FUNC_SGELS) ||
(func == PLASMA_FUNC_DGELS) ||
(func == PLASMA_FUNC_CGELS) ||
(func == PLASMA_FUNC_ZGELS) ||
(func == PLASMA_FUNC_SGESVD) ||
(func == PLASMA_FUNC_DGESVD) ||
(func == PLASMA_FUNC_CGESVD) ||
(func == PLASMA_FUNC_ZGESVD)))
NT *= 2;
/* Allocate and initialize descriptor */
*desc = (PLASMA_desc*)malloc(sizeof(PLASMA_desc));
if (*desc == NULL) {
plasma_error("plasma_alloc_ibnb_tile", "malloc() failed");
return PLASMA_ERR_OUT_OF_RESOURCES;
}
**desc = plasma_desc_init(type, IB, NB, IB*NB, MT*IB, NT*NB, 0, 0, MT*IB, NT*NB);
/* Allocate matrix */
if (plasma_desc_mat_alloc(*desc)) {
plasma_error("plasma_alloc_ibnb_tile", "malloc() failed");
return PLASMA_ERR_OUT_OF_RESOURCES;
}
/* Check that everything is ok */
status = plasma_desc_check(*desc);
if (status != PLASMA_SUCCESS) {
plasma_error("plasma_alloc_ibnb_tile", "invalid descriptor");
return status;
}
return PLASMA_SUCCESS;
}
/***************************************************************************//**
*
* @ingroup PLASMA_Complex32_t_Tile_Async
*
* PLASMA_cpotrf_Tile_Async - Computes the Cholesky factorization of a symmetric
* positive definite or Hermitian positive definite matrix.
* Non-blocking equivalent of PLASMA_cpotrf_Tile().
* May return before the computation is finished.
* Allows for pipelining of operations ar runtime.
*
*******************************************************************************
*
* @param[in] sequence
* Identifies the sequence of function calls that this call belongs to
* (for completion checks and exception handling purposes).
*
* @param[out] request
* Identifies this function call (for exception handling purposes).
*
*******************************************************************************
*
* @sa PLASMA_cpotrf
* @sa PLASMA_cpotrf_Tile
* @sa PLASMA_cpotrf_Tile_Async
* @sa PLASMA_dpotrf_Tile_Async
* @sa PLASMA_spotrf_Tile_Async
* @sa PLASMA_cpotrs_Tile_Async
*
******************************************************************************/
int PLASMA_cpotrf_Tile_Async(PLASMA_enum uplo, 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_cpotrf_Tile", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
if (sequence == NULL) {
plasma_fatal_error("PLASMA_cpotrf_Tile", "NULL sequence");
return PLASMA_ERR_UNALLOCATED;
}
if (request == NULL) {
plasma_fatal_error("PLASMA_cpotrf_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_cpotrf_Tile", "invalid descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Check input arguments */
if (descA.nb != descA.mb) {
plasma_error("PLASMA_cpotrf_Tile", "only square tiles supported");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if (uplo != PlasmaUpper && uplo != PlasmaLower) {
plasma_error("PLASMA_cpotrf_Tile", "illegal value of uplo");
return plasma_request_fail(sequence, request, -1);
}
/* Quick return */
/*
if (max(N, 0) == 0)
return PLASMA_SUCCESS;
*/
plasma_parallel_call_4(plasma_pcpotrf,
PLASMA_enum, uplo,
PLASMA_desc, descA,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
return PLASMA_SUCCESS;
}
/***************************************************************************//**
*
* @ingroup float_Tile_Async
*
* PLASMA_splgsy_Tile_Async - Generate a random hermitian matrix by tiles.
* Non-blocking equivalent of PLASMA_splgsy_Tile().
* May return before the computation is finished.
* Allows for pipelining of operations ar runtime.
*
*******************************************************************************
*
* @param[in] sequence
* Identifies the sequence of function calls that this call belongs to
* (for completion checks and exception handling purposes).
*
* @param[out] request
* Identifies this function call (for exception handling purposes).
*
*******************************************************************************
*
* @sa PLASMA_splgsy
* @sa PLASMA_splgsy_Tile
* @sa PLASMA_cplgsy_Tile_Async
* @sa PLASMA_dplgsy_Tile_Async
* @sa PLASMA_splgsy_Tile_Async
* @sa PLASMA_splgsy_Tile_Async
* @sa PLASMA_splgsy_Tile_Async
*
******************************************************************************/
int PLASMA_splgsy_Tile_Async( float bump,
PLASMA_desc *A,
unsigned long long int seed,
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_splgsy_Tile", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
if (sequence == NULL) {
plasma_fatal_error("PLASMA_splgsy_Tile", "NULL sequence");
return PLASMA_ERR_UNALLOCATED;
}
if (request == NULL) {
plasma_fatal_error("PLASMA_splgsy_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_splgsy_Tile", "invalid descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Check input arguments */
if (descA.nb != descA.mb) {
plasma_error("PLASMA_splgsy_Tile", "only square tiles supported");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Quick return */
if (min( descA.m, descA.n ) == 0)
return PLASMA_SUCCESS;
plasma_parallel_call_5(plasma_psplgsy,
float, bump,
PLASMA_desc, descA,
unsigned long long int, seed,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
return PLASMA_SUCCESS;
}
/***************************************************************************//**
*
* @ingroup PLASMA_Complex64_t_Tile_Async
*
* PLASMA_zgetrf_nopiv_Tile_Async - Computes the tile LU factorization of a
* matrix. Non-blocking equivalent of PLASMA_zgetrf_nopiv_Tile(). May return
* before the computation is finished. Allows for pipelining of operations ar
* runtime.
*
*******************************************************************************
*
* @param[in,out] A
* On entry, the M-by-N matrix to be factored.
* On exit, the tile factors L and U from the factorization.
*
* @param[in] sequence
* Identifies the sequence of function calls that this call belongs to
* (for completion checks and exception handling purposes).
*
* @param[out] request
* Identifies this function call (for exception handling purposes).
*
*******************************************************************************
*
* @sa PLASMA_zgetrf_nopiv
* @sa PLASMA_zgetrf_nopiv_Tile
* @sa PLASMA_cgetrf_nopiv_Tile_Async
* @sa PLASMA_dgetrf_nopiv_Tile_Async
* @sa PLASMA_sgetrf_nopiv_Tile_Async
* @sa PLASMA_zgetrs_Tile_Async
*
******************************************************************************/
int PLASMA_zgetrf_nopiv_Tile_Async(PLASMA_desc *A,
PLASMA_sequence *sequence,
PLASMA_request *request)
{
PLASMA_desc descA;
plasma_context_t *plasma;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_fatal_error("PLASMA_zgetrf_nopiv_Tile", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
if (sequence == NULL) {
plasma_fatal_error("PLASMA_zgetrf_nopiv_Tile", "NULL sequence");
return PLASMA_ERR_UNALLOCATED;
}
if (request == NULL) {
plasma_fatal_error("PLASMA_zgetrf_nopiv_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(A) != PLASMA_SUCCESS) {
plasma_error("PLASMA_zgetrf_nopiv_Tile", "invalid first descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
} else {
descA = *A;
}
/* Check input arguments */
if (descA.nb != descA.mb) {
plasma_error("PLASMA_zgetrf_nopiv_Tile", "only square tiles supported");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
plasma_dynamic_call_3(plasma_pzgetrf_nopiv,
PLASMA_desc, descA,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
return PLASMA_SUCCESS;
}
/***************************************************************************//**
*
* @ingroup PLASMA_Complex64_t
*
* PLASMA_zgesvd - computes the singular value decomposition (SVD) of a complex
* M-by-N matrix A, optionally computing the left and/or right singular
* vectors. The SVD is written
*
* A = U * SIGMA * transpose(V)
*
* where SIGMA is an M-by-N matrix which is zero except for its
* min(m,n) diagonal elements, U is an M-by-M orthogonal matrix, and
* V is an N-by-N orthogonal matrix. The diagonal elements of SIGMA
* are the singular values of A; they are real and non-negative, and
* are returned in descending order. The first min(m,n) columns of
* U and V are the left and right singular vectors of A.
*
* Note that the routine returns V**T, not V.
* Not LAPACK Compliant for now!
* Note: Only PlasmaNoVec supported!
*******************************************************************************
*
* @param[in] jobu
* Specifies options for computing all or part of the matrix U.
* Intended usage:
* = PlasmaVec: all M columns of U are returned in array U;
* = PlasmaNoVec: no columns of U (no left singular vectors) are
* computed.
* Note: Only PlasmaNoVec supported!
*
* @param[in] jobvt
* Specifies options for computing all or part of the matrix V**H.
* Intended usage:
* = PlasmaVec: all M columns of U are returned in array U;
* = PlasmaNoVec: no columns of U (no left singular vectors) are
* computed.
* Note: Only PlasmaNoVec supported!
*
* @param[in] M
* The number of rows of the matrix A. M >= 0.
*
* @param[in] N
* The number of columns of the matrix A. N >= 0.
*
* @param[in,out] A
* On entry, the M-by-N matrix A.
* On exit,
* if JOBU = 'O', A is overwritten with the first min(m,n)
* columns of U (the left singular vectors,
* stored columnwise);
* if JOBVT = 'O', A is overwritten with the first min(m,n)
* rows of V**H (the right singular vectors,
* stored rowwise);
* if JOBU .ne. 'O' and JOBVT .ne. 'O', the contents of A
* are destroyed.
*
* @param[in] LDA
* The leading dimension of the array A. LDA >= max(1,M).
*
* @param[out] S
* The double precision singular values of A, sorted so that S(i) >= S(i+1).
*
* @param[out] U
* (LDU,M) if JOBU = 'A' or (LDU,min(M,N)) if JOBU = 'S'.
* If JOBU = 'A', U contains the M-by-M unitary matrix U;
* if JOBU = 'S', U contains the first min(m,n) columns of U
* (the left singular vectors, stored columnwise);
* if JOBU = 'N' or 'O', U is not referenced.
*
* @param[in] LDU
* The leading dimension of the array U. LDU >= 1; if
* JOBU = 'S' or 'A', LDU >= M.
*
* @param[out] VT
* If JOBVT = 'A', VT contains the N-by-N unitary matrix
* V**H;
* if JOBVT = 'S', VT contains the first min(m,n) rows of
* V**H (the right singular vectors, stored rowwise);
* if JOBVT = 'N' or 'O', VT is not referenced.
*
* @param[in] LDVT
* The leading dimension of the array VT. LDVT >= 1; if
* JOBVT = 'A', LDVT >= N; if JOBVT = 'S', LDVT >= min(M,N).
*
* @param[in, out] descT
* On entry, descriptor as return by PLASMA_Alloc_Workspace_zgesvd
* On exit, contains auxiliary factorization data.
*
*
*******************************************************************************
*
* @return
* \retval PLASMA_SUCCESS successful exit
* \retval <0 if -i, the i-th argument had an illegal value
*
*******************************************************************************
*
* @sa PLASMA_zgesvd_Tile
* @sa PLASMA_zgesvd_Tile_Async
* @sa PLASMA_cgesvd
* @sa PLASMA_dgesvd
* @sa PLASMA_sgesvd
*
******************************************************************************/
int PLASMA_zgesvd(PLASMA_enum jobu, PLASMA_enum jobvt, int M, int N,
PLASMA_Complex64_t *A, int LDA,
double *S,
PLASMA_Complex64_t *U, int LDU,
PLASMA_Complex64_t *VT, int LDVT,
PLASMA_desc *descT)
{
int NB, IB, IBNB, minMN, MT, NT, minMTNT;
int status;
plasma_context_t *plasma;
PLASMA_sequence *sequence = NULL;
PLASMA_request request = PLASMA_REQUEST_INITIALIZER;
PLASMA_desc descA, descU, descVT;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_fatal_error("PLASMA_zgesvd", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
/* Tune NB & IB depending on M & N; Set NBNB */
status = plasma_tune(PLASMA_FUNC_ZGESVD, M, N, 0);
if (status != PLASMA_SUCCESS) {
plasma_error("PLASMA_zgesvd", "plasma_tune() failed");
return status;
}
/* Set MT, NT */
NB = PLASMA_NB;
IB = PLASMA_IB;
IBNB = IB*NB;
MT = (M%NB==0) ? (M/NB) : (M/NB+1);
NT = (N%NB==0) ? (N/NB) : (N/NB+1);
minMN = min(M,N);
minMTNT = min(MT,NT);
/* Check input arguments */
if (jobu != PlasmaNoVec && jobu !=PlasmaVec) {
plasma_error("PLASMA_zgesvd", "illegal value of jobu");
return -1;
}
if (jobvt != PlasmaNoVec && jobvt != PlasmaVec) {
plasma_error("PLASMA_zgesvd", "illegal value of jobvt");
return -2;
}
if (M < 0) {
plasma_error("PLASMA_zgesvd", "illegal value of M");
return -3;
}
if (N < 0) {
plasma_error("PLASMA_zgesvd", "illegal value of N");
return -4;
}
if (LDA < max(1, M)) {
plasma_error("PLASMA_zgesvd", "illegal value of LDA");
return -6;
}
if (LDU < 1) {
plasma_error("PLASMA_zgesvd", "illegal value of LDU");
return -9;
}
if (LDVT < 1) {
plasma_error("PLASMA_zgesvd", "illegal value of LDVT");
return -11;
}
if ( (plasma_desc_check(descT) != PLASMA_SUCCESS) ||
( descT->m != MT*IB ) || (descT->n != NT*NB) ) {
plasma_error("PLASMA_zgesvd", "invalid T descriptor");
return -12;
}
/* Quick return */
if (min(M, N) == 0) {
return PLASMA_SUCCESS;
}
if (jobu == PlasmaVec) {
plasma_error("PLASMA_zgesvd", "computing the singular vectors is not supported in this version");
return -1;
}
if (jobvt == PlasmaVec) {
plasma_error("PLASMA_zgesvd", "computing the singular vectors is not supported in this version");
return -2;
}
plasma_sequence_create(plasma, &sequence);
if ( PLASMA_TRANSLATION == PLASMA_OUTOFPLACE ) {
plasma_zooplap2tile( descA, A, NB, NB, LDA, N, 0, 0, M, N, plasma_desc_mat_free(&(descA)) );
if (jobu == PlasmaVec){
plasma_zooplap2tile( descU, U, NB, NB, LDU, M, 0, 0, M, M, plasma_desc_mat_free(&(descA)); plasma_desc_mat_free(&(descU)));
}
/***************************************************************************//**
*
* @ingroup float
*
* PLASMA_ssytrd - reduces a complex Hermitian matrix A to real symmetric
* tridiagonal form S using a two-stage approach
* First stage: reduction to band tridiagonal form (unitary Q1);
* Second stage: reduction from band to tridiagonal form (unitary
* Q2). Let Q = Q1 * Q2 be the global unitary transformation; Q**T *
* A * Q = S.
* Not LAPACK compliant as A does not contain the T elements
* Note: Only PlasmaNoVec supported!
*
*******************************************************************************
*
* @param[in] jobz
* Intended usage:
* = PlasmaNoVec: computes eigenvalues only;
* = PlasmaVec: computes eigenvalues and eigenvectors.
* Note: Only PlasmaNoVec supported!
*
* @param[in] uplo
* Specifies whether the matrix A is upper triangular or
* lower triangular:
* = PlasmaUpper: Upper triangle of A is stored;
* = PlasmaLower: Lower triangle of A is stored.
*
* @param[in] N
* The order of the matrix A. N >= 0.
*
* @param[in,out] A
* On entry, the symmetric (or Hermitian) matrix A.
* If uplo = PlasmaUpper, the leading N-by-N upper triangular
* part of A contains the upper triangular part of the matrix
* A, and the strictly lower triangular part of A is not
* referenced.
* If uplo = PlasmaLower, the leading N-by-N lower triangular
* part of A contains the lower triangular part of the matrix
* A, and the strictly upper triangular part of A is not
* referenced.
* On exit, the lower triangle (if uplo = PlasmaLower) or the
* upper triangle (if uplo = PlasmaUpper) of A, including the
* diagonal, is destroyed.
*
* @param[in] LDA
* The leading dimension of the array A. LDA >= max(1,N).
*
* @param[out] D
* On exit, the diagonal elements of the tridiagonal matrix:
* D(i) = A(i,i).
*
* @param[out] E
* On exit, he off-diagonal elements of the tridiagonal matrix:
* E(i) = A(i,i+1) if uplo = PlasmaUpper, E(i) = A(i+1,i) if uplo = PlasmaLower.
*
* @param[in, out] descT
* On entry, descriptor as return by PLASMA_Alloc_Workspace_ssyev
* On exit, contains auxiliary factorization data.
*
* @param[out] Q
* On exit, if jobz = PlasmaVec and info = 0, the eigenvectors.
*
* @param[in] LDQ
* The leading dimension of the array Q. LDQ >= max(1,N).
*
*******************************************************************************
*
* @return
* \retval PLASMA_SUCCESS successful exit
* \retval <0 if -i, the i-th argument had an illegal value
* \retval >0 if INFO = i, the algorithm failed to converge; i
* off-diagonal elements of an intermediate tridiagonal
* form did not converge to zero.
*
*******************************************************************************
*
* @sa PLASMA_ssytrd_Tile
* @sa PLASMA_ssytrd_Tile_Async
* @sa PLASMA_chetrd
* @sa PLASMA_dsytrd
* @sa PLASMA_ssytrd
*
******************************************************************************/
int PLASMA_ssytrd(PLASMA_enum jobz, PLASMA_enum uplo, int N,
float *A, int LDA,
float *D,
float *E,
PLASMA_desc *descT,
float *Q, int LDQ)
{
int NB, IB, IBNB, NT;
int status;
plasma_context_t *plasma;
PLASMA_sequence *sequence = NULL;
PLASMA_request request = PLASMA_REQUEST_INITIALIZER;
PLASMA_desc descA, descQ;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_error("PLASMA_ssytrd", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
/* Tune NB & IB depending on N; Set NBNB */
status = plasma_tune(PLASMA_FUNC_SSYTRD, N, N, 0);
if (status != PLASMA_SUCCESS) {
plasma_error("PLASMA_ssytrd", "plasma_tune() failed");
return status;
}
/* Set NT */
NB = PLASMA_NB;
IB = PLASMA_IB;
IBNB = IB*NB;
NT = (N%NB==0) ? (N/NB) : (N/NB+1);
/* Check input arguments */
if (jobz != PlasmaNoVec && jobz != PlasmaVec) {
plasma_error("PLASMA_ssytrd", "illegal value of jobz");
return -1;
}
if (uplo != PlasmaLower && uplo != PlasmaUpper) {
plasma_error("PLASMA_ssytrd", "illegal value of uplo");
return -2;
}
if (N < 0) {
plasma_error("PLASMA_ssytrd", "illegal value of N");
return -3;
}
if (LDA < max(1, N)) {
plasma_error("PLASMA_ssytrd", "illegal value of LDA");
return -5;
}
if ( (plasma_desc_check(descT) != PLASMA_SUCCESS) ||
( descT->m != NT*IB ) || (descT->n != NT*NB) ) {
plasma_error("PLASMA_ssytrd", "invalid T descriptor");
return -8;
}
if (LDQ < max(1, N)) {
plasma_error("PLASMA_ssytrd", "illegal value of LDQ");
return -10;
}
/* Quick return */
if (N == 0)
return PLASMA_SUCCESS;
if (jobz == PlasmaVec) {
plasma_error("PLASMA_ssytrd", "computing the eigenvectors is not supported in this version");
return -1;
}
plasma_sequence_create(plasma, &sequence);
if ( PLASMA_TRANSLATION == PLASMA_OUTOFPLACE ) {
plasma_sooplap2tile( descA, A, NB, NB, LDA, N, 0, 0, N, N , plasma_desc_mat_free(&(descA)) );
if (jobz == PlasmaVec) {
plasma_sooplap2tile( descQ, Q, NB, NB, LDQ, N, 0, 0, N, N , plasma_desc_mat_free(&(descQ)) );
}
} else {
plasma_siplap2tile( descA, A, NB, NB, LDA, N, 0, 0, N, N );
if (jobz == PlasmaVec)
plasma_siplap2tile( descQ, Q, NB, NB, LDQ, N, 0, 0, N, N );
}
/* Call the tile interface */
PLASMA_ssytrd_Tile_Async(jobz, uplo, &descA, D, E, descT, &descQ, sequence, &request);
if ( PLASMA_TRANSLATION == PLASMA_OUTOFPLACE ) {
plasma_sooptile2lap( descA, A, NB, NB, LDA, N );
if (jobz == PlasmaVec) {
plasma_sooptile2lap( descQ, Q, NB, NB, LDQ, N );
}
plasma_dynamic_sync();
plasma_desc_mat_free(&descA);
if (jobz == PlasmaVec)
plasma_desc_mat_free(&descQ);
} else {
plasma_siptile2lap( descA, A, NB, NB, LDA, N );
if (jobz == PlasmaVec)
plasma_siptile2lap( descQ, Q, NB, NB, LDQ, N );
plasma_dynamic_sync();
}
status = sequence->status;
plasma_sequence_destroy(plasma, sequence);
return status;
}
/***************************************************************************//**
*
* @ingroup float_Tile_Async
*
* PLASMA_ssytrd_Tile_Async - Computes all eigenvalues and,
* optionally, eigenvectors of a complex Hermitian matrix A using a
* two-stage approach:
* First stage: reduction to band tridiagonal form;
* Second stage: reduction from band to tridiagonal form.
*
* May return before the computation is finished.
* Allows for pipelining of operations ar runtime.
*
*******************************************************************************
*
* @param[in] sequence
* Identifies the sequence of function calls that this call belongs to
* (for completion checks and exception handling purposes).
*
* @param[out] request
* Identifies this function call (for exception handling purposes).
*
*******************************************************************************
*
* @sa PLASMA_ssytrd
* @sa PLASMA_ssytrd_Tile
* @sa PLASMA_chetrd_Tile_Async
* @sa PLASMA_dsytrd_Tile_Async
* @sa PLASMA_ssytrd_Tile_Async
*
******************************************************************************/
int PLASMA_ssytrd_Tile_Async(PLASMA_enum jobz, PLASMA_enum uplo,
PLASMA_desc *A,
float *D,
float *E,
PLASMA_desc *T,
PLASMA_desc *Q,
PLASMA_sequence *sequence, PLASMA_request *request)
{
int NB, IB, IBNB, NT;
PLASMA_desc descA = *A;
PLASMA_desc descT = *T;
plasma_context_t *plasma;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_fatal_error("PLASMA_ssytrd_Tile_Async", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
if (sequence == NULL) {
plasma_fatal_error("PLASMA_ssytrd_Tile_Async", "NULL sequence");
return PLASMA_ERR_UNALLOCATED;
}
if (request == NULL) {
plasma_fatal_error("PLASMA_ssytrd_Tile_Async", "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);
/* Set NT & NTRHS */
NB = PLASMA_NB;
IB = PLASMA_IB;
IBNB = IB*NB;
NT = (descA.ln%NB==0) ? (descA.ln/NB) : (descA.ln/NB+1);
/* Check descriptors for correctness */
if (plasma_desc_check(&descA) != PLASMA_SUCCESS) {
plasma_error("PLASMA_ssytrd_Tile_Async", "invalid descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if (plasma_desc_check(&descT) != PLASMA_SUCCESS) {
plasma_error("PLASMA_ssytrd_Tile_Async", "invalid descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if ( (jobz == PlasmaVec) && (plasma_desc_check(Q) != PLASMA_SUCCESS) ) {
plasma_error("PLASMA_ssytrd_Tile_Async", "invalid descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Check input arguments */
if (jobz != PlasmaNoVec && jobz != PlasmaVec) {
plasma_error("PLASMA_ssytrd_Tile_Async", "illegal value of jobz");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if (descA.m != descA.n) {
plasma_error("PLASMA_ssytrd_Tile_Async", "matrix need to be square");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if (descA.nb != descA.mb) {
plasma_error("PLASMA_ssytrd_Tile_Async", "only square tiles supported");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if (jobz == PlasmaVec) {
plasma_error("PLASMA_ssytrd_Tile_Async", "computing the eigenvectors is not supported in this version");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if ( (jobz == PlasmaVec) && (Q->nb != Q->mb) ) {
plasma_error("PLASMA_ssytrd_Tile_Async", "only square tiles supported");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Reduction to tridiagonal form
* with a two-stage approach.
*/
/* Reduction to BAND tridiagonal form
*/
plasma_dynamic_call_5(plasma_pssyrbt,
PLASMA_enum, uplo,
PLASMA_desc, descA,
PLASMA_desc, descT,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
/*
* Build the Q of the first stage
*/
/* if (jobz == PlasmaVec){ */
/* /\* Initialize Q to Identity *\/ */
/* plasma_dynamic_call_6(plasma_pslaset, */
/* PLASMA_enum, PlasmaUpperLower, */
/* float, 0.0, */
/* float, 1.0, */
/* PLASMA_desc, descQ, */
/* PLASMA_sequence*, sequence, */
/* PLASMA_request*, request); */
/* /\* Accumulate the transformations from the first stage*\/ */
/* plasma_dynamic_call_6(plasma_psorgtr, */
/* PLASMA_enum, uplo, */
/* PLASMA_desc, descA, */
/* PLASMA_desc, descQ, */
/* PLASMA_desc, descT, */
/* PLASMA_sequence*, sequence, */
/* PLASMA_request*, request); */
/* } */
/* Set the V's to zero before the 2nd stage (bulge chasing) */
/*
*/
plasma_dynamic_call_5(plasma_pslaset2,
PLASMA_enum, uplo,
float, 0.0,
PLASMA_desc, uplo == PlasmaLower ? plasma_desc_submatrix(descA, descA.mb, 0, descA.m-descA.mb, descA.n-descA.nb)
: plasma_desc_submatrix(descA, 0, descA.nb, descA.m-descA.mb, descA.n-descA.nb),
PLASMA_sequence*, sequence,
PLASMA_request*, request);
/* Reduction from BAND tridiagonal to the final condensed form
*/
plasma_dynamic_call_7(plasma_pssbrdt,
PLASMA_enum, uplo,
PLASMA_desc, descA,
float*, D,
float*, E,
PLASMA_desc, descT,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
return PLASMA_SUCCESS;
}
/***************************************************************************//**
*
* @ingroup float
*
* PLASMA_ssygv - Computes all eigenvalues and, optionally,
* eigenvectors of a complex generalized Hermitian-definite
* eigenproblem of the form: A*x=(lambda)*B*x, A*Bx=(lambda)*x, or
* B*A*x=(lambda)*x.
* Here A and B are assumed to be Hermitian and B is also positive
* definite.
* Note: Only PlasmaNoVec supported!
*
*******************************************************************************
*
* @param[in] PlasmaItype
* Intended usage:
* = 1: A*x=(lambda)*B*x
* = 2: A*Bx=(lambda)*x
* = 3: B*A*x=(lambda)*x
*
* @param[in] jobz
* Intended usage:
* = PlasmaNoVec: computes eigenvalues only;
* = PlasmaVec: computes eigenvalues and eigenvectors.
* Note: Only PlasmaNoVec supported!
*
* @param[in] uplo
* Specifies whether the matrix A is upper triangular or
* lower triangular:
* = PlasmaUpper: Upper triangle of A and B are stored;
* = PlasmaLower: Lower triangle of A and B are stored.
*
* @param[in] N
* The order of the matrix A. N >= 0.
*
* @param[in,out] A
* On entry, the symmetric (or Hermitian) matrix A.
* If uplo = PlasmaUpper, the leading N-by-N upper triangular
* part of A contains the upper triangular part of the matrix
* A, and the strictly lower triangular part of A is not
* referenced.
* If uplo = PlasmaLower, the leading N-by-N lower triangular
* part of A contains the lower triangular part of the matrix
* A, and the strictly upper triangular part of A is not
* referenced.
* On exit, if jobz = PlasmaVec, then if return value = 0, A
* contains the matrix Z of eigenvectors.
* The eigenvectors are normalized as follows:
* if ITYPE = 1 or 2, Z**T*B*Z = I;
* if ITYPE = 3, Z**T*inv(B)*Z = I.
* If jobz = PlasmaNoVec, then on exit the lower triangle (if
* uplo = PlasmaLower) or the upper triangle (if uplo =
* PlasmaUpper) of A, including the diagonal, is destroyed.
*
* @param[in] LDA
* The leading dimension of the array A. LDA >= max(1,N).
*
* @param[in,out] B
* On entry, the symmetric (or Hermitian) positive definite
* matrix B.
* If uplo = PlasmaUpper, the leading N-by-N upper triangular
* part of B contains the upper triangular part of the matrix
* B, and the strictly lower triangular part of B is not
* referenced.
* If uplo = PlasmaLower, the leading N-by-N lower triangular
* part of B contains the lower triangular part of the matrix
* B, and the strictly upper triangular part of B is not
* referenced.
* On exit, if return value <= N, the part of B containing
* the matrix is overwritten by the triangular factor U or L
* from the Cholesky factorization B = U**T*U or B = L*L**T.
*
* @param[in] LDB
* The leading dimension of the array B. LDA >= max(1,N).
*
* @param[out] W
* On exit, if info = 0, the eigenvalues.
*
* @param[in, out] descT
* On entry, descriptor as return by PLASMA_Alloc_Workspace_ssygv
* On exit, contains auxiliary factorization data.
*
* @param[out] Q
* On exit, if jobz = PlasmaVec and info = 0, the eigenvectors.
*
* @param[in] LDQ
* The leading dimension of Q.
*
*******************************************************************************
*
* @return
* \retval PLASMA_SUCCESS successful exit
* \retval <0 if -i, the i-th argument had an illegal value
* \retval <=N if INFO = i, plasma_ssygv failed to converge; i
* off-diagonal elements of an intermediate tridiagonal
* form did not converge to zero.
* \retval >N if INFO = N + i, for 1 <= i <= N, then the leading
* minor of order i of B is not positive definite.
* The factorization of B could not be completed and
* no eigenvalues or eigenvectors were computed.
*
*******************************************************************************
*
* @sa PLASMA_ssygv_Tile
* @sa PLASMA_ssygv_Tile_Async
* @sa PLASMA_chegv
* @sa PLASMA_dsygv
* @sa PLASMA_ssygv
*
******************************************************************************/
int PLASMA_ssygv(PLASMA_enum itype, PLASMA_enum jobz, PLASMA_enum uplo, int N,
float *A, int LDA,
float *B, int LDB,
float *W,
PLASMA_desc *descT,
float *Q, int LDQ)
{
int NB, IB, IBNB, NT;
int status;
plasma_context_t *plasma;
PLASMA_sequence *sequence = NULL;
PLASMA_request request = PLASMA_REQUEST_INITIALIZER;
PLASMA_desc descA, descB, descQ;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_error("PLASMA_ssygv", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
/* Tune NB & IB depending on N; Set NBNB */
status = plasma_tune(PLASMA_FUNC_SSYGV, N, N, 0);
if (status != PLASMA_SUCCESS) {
plasma_error("PLASMA_ssygv", "plasma_tune() failed");
return status;
}
/* Set NT */
NB = PLASMA_NB;
IB = PLASMA_IB;
IBNB = IB*NB;
NT = (N%NB==0) ? (N/NB) : (N/NB+1);
/* Check input arguments */
if (itype != 1 && itype != 2 && itype != 3) {
plasma_error("PLASMA_ssygv", "Illegal value of itype");
return -1;
}
if (jobz != PlasmaNoVec && jobz != PlasmaVec) {
plasma_error("PLASMA_ssygv", "illegal value of jobz");
return -2;
}
if (uplo != PlasmaLower && uplo!= PlasmaUpper) {
plasma_error("PLASMA_ssygv", "only PlasmaLower supported");
return -3;
}
if (N < 0) {
plasma_error("PLASMA_ssygv", "illegal value of N");
return -4;
}
if (LDA < max(1, N)) {
plasma_error("PLASMA_ssygv", "illegal value of LDA");
return -6;
}
if (LDB < max(1, N)) {
plasma_error("PLASMA_ssygv", "illegal value of LDB");
return -8;
}
if ( (plasma_desc_check(descT) != PLASMA_SUCCESS) ||
( descT->m != NT*IB ) || (descT->n != NT*NB) ) {
plasma_error("PLASMA_ssygv", "invalid T descriptor");
return -10;
}
if (LDQ < max(1, N)) {
plasma_error("PLASMA_ssygv", "illegal value of LDQ");
return -12;
}
/* Quick return */
if (N == 0)
return PLASMA_SUCCESS;
if (jobz == PlasmaVec) {
plasma_error("PLASMA_ssygv", "computing the eigenvectors is not supported in this version");
return -1;
}
plasma_sequence_create(plasma, &sequence);
if ( PLASMA_TRANSLATION == PLASMA_OUTOFPLACE ) {
plasma_sooplap2tile( descA, A, NB, NB, LDA, N, 0, 0, N, N,
plasma_desc_mat_free(&(descA)) );
plasma_sooplap2tile( descB, B, NB, NB, LDB, N, 0, 0, N, N,
plasma_desc_mat_free(&(descA)); plasma_desc_mat_free(&(descB)) );
if (jobz == PlasmaVec) {
/* No need for conversion, it's just output */
plasma_sdesc_alloc( descQ, NB, NB, LDQ, N, 0, 0, N, N,
plasma_desc_mat_free(&(descA)); plasma_desc_mat_free(&(descB)); plasma_desc_mat_free(&(descQ)) );
}
/***************************************************************************//**
*
* @ingroup PLASMA_Complex64_t_Tile_Async
*
* PLASMA_zgetri_Tile_Async - Computes the inverse of a matrix using the LU
* factorization computed by PLASMA_zgetrf.
* This method inverts U and then computes inv(A) by solving the system
* inv(A)*L = inv(U) for inv(A).
* Non-blocking equivalent of PLASMA_zgetri_Tile().
* May return before the computation is finished.
* Allows for pipelining of operations at runtime.
*
*******************************************************************************
*
* @param[in] sequence
* Identifies the sequence of function calls that this call belongs to
* (for completion checks and exception handling purposes).
*
* @param[out] request
* Identifies this function call (for exception handling purposes).
*
*******************************************************************************
*
* @sa PLASMA_zgetri
* @sa PLASMA_zgetri_Tile
* @sa PLASMA_cgetri_Tile_Async
* @sa PLASMA_dgetri_Tile_Async
* @sa PLASMA_sgetri_Tile_Async
* @sa PLASMA_zgetrf_Tile_Async
*
******************************************************************************/
int PLASMA_zgetri_Tile_Async(PLASMA_desc *A, int *IPIV, PLASMA_desc *W,
PLASMA_sequence *sequence, PLASMA_request *request)
{
PLASMA_desc descA;
PLASMA_desc descW;
plasma_context_t *plasma;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_fatal_error("PLASMA_zgetri_Tile_Async", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
if (sequence == NULL) {
plasma_fatal_error("PLASMA_zgetri_Tile_Async", "NULL sequence");
return PLASMA_ERR_UNALLOCATED;
}
if (request == NULL) {
plasma_fatal_error("PLASMA_zgetri_Tile_Async", "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(A) != PLASMA_SUCCESS) {
plasma_error("PLASMA_zgetri_Tile_Async", "invalid A descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
} else {
descA = *A;
}
/* Check descriptors for correctness */
if (plasma_desc_check(W) != PLASMA_SUCCESS) {
plasma_error("PLASMA_zgetri_Tile_Async", "invalid W descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
} else {
descW = *W;
}
/* Check input arguments */
if (descA.nb != descA.mb) {
plasma_error("PLASMA_zgetri_Tile_Async", "only square tiles supported");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Quick return */
if (max(descA.m, 0) == 0)
return PLASMA_SUCCESS;
plasma_dynamic_call_5(plasma_pztrtri,
PLASMA_enum, PlasmaUpper,
PLASMA_enum, PlasmaNonUnit,
PLASMA_desc, descA,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
plasma_dynamic_call_9(plasma_pztrsmrv,
PLASMA_enum, PlasmaRight,
PLASMA_enum, PlasmaLower,
PLASMA_enum, PlasmaNoTrans,
PLASMA_enum, PlasmaUnit,
PLASMA_Complex64_t, (PLASMA_Complex64_t) 1.0,
PLASMA_desc, descA,
PLASMA_desc, descW,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
/* No need for barrier tile2row because of previous dependencies */
/* swap */
plasma_dynamic_call_5(
plasma_pzlaswpc,
PLASMA_desc, descA,
int *, IPIV,
int, -1,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
plasma_dynamic_call_3(
plasma_pzbarrier_row2tl,
PLASMA_desc, descA,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
return PLASMA_SUCCESS;
}
/***************************************************************************//**
*
* @ingroup PLASMA_Complex64_t_Tile_Async
*
* PLASMA_zlansy_Tile_Async - Non-blocking equivalent of PLASMA_zlansy_Tile().
* May return before the computation is finished.
* Allows for pipelining of operations at runtime.
*
*******************************************************************************
*
* @param[in] sequence
* Identifies the sequence of function calls that this call belongs to
* (for completion checks and exception handling purposes).
*
* @param[out] request
* Identifies this function call (for exception handling purposes).
*
*******************************************************************************
*
* @sa PLASMA_zlansy
* @sa PLASMA_zlansy_Tile
* @sa PLASMA_clansy_Tile_Async
* @sa PLASMA_dlansy_Tile_Async
* @sa PLASMA_slansy_Tile_Async
*
******************************************************************************/
int PLASMA_zlansy_Tile_Async(PLASMA_enum norm, PLASMA_enum uplo, PLASMA_desc *A, double *value,
PLASMA_sequence *sequence, PLASMA_request *request)
{
PLASMA_desc descA;
double *work = NULL;
plasma_context_t *plasma;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_fatal_error("PLASMA_zlansy_Tile", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
if (sequence == NULL) {
plasma_fatal_error("PLASMA_zlansy_Tile", "NULL sequence");
return PLASMA_ERR_UNALLOCATED;
}
if (request == NULL) {
plasma_fatal_error("PLASMA_zlansy_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(A) != PLASMA_SUCCESS) {
plasma_error("PLASMA_zlansy_Tile", "invalid descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
} else {
descA = *A;
}
/* Check input arguments */
if (descA.nb != descA.mb) {
plasma_error("PLASMA_zlansy_Tile", "only square tiles supported");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if ( (norm != PlasmaMaxNorm) && (norm != PlasmaOneNorm)
&& (norm != PlasmaInfNorm) && (norm != PlasmaFrobeniusNorm) ) {
plasma_error("PLASMA_zlansy_Tile", "illegal value of norm");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if ( (uplo != PlasmaUpper) && (uplo != PlasmaLower) ) {
plasma_error("PLASMA_zlansy_Tile", "illegal value of uplo");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Quick return */
if ( descA.m == 0) {
*value = 0.0;
return PLASMA_SUCCESS;
}
if (PLASMA_SCHEDULING == PLASMA_STATIC_SCHEDULING) {
if (norm == PlasmaFrobeniusNorm) {
work = plasma_shared_alloc(plasma, 2*PLASMA_SIZE, PlasmaRealDouble );
} else {
work = plasma_shared_alloc(plasma, PLASMA_SIZE, PlasmaRealDouble );
}
}
plasma_parallel_call_7(plasma_pzlansy,
PLASMA_enum, norm,
PLASMA_enum, uplo,
PLASMA_desc, descA,
double*, work,
double*, value,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
if (work != NULL)
plasma_shared_free( plasma, work );
return PLASMA_SUCCESS;
}
/***************************************************************************//**
*
* @ingroup float_Tile_Async
*
* PLASMA_sgelqf_Tile_Async - Computes the tile LQ factorization of a matrix.
* Non-blocking equivalent of PLASMA_sgelqf_Tile().
* May return before the computation is finished.
* Allows for pipelining of operations ar runtime.
*
*******************************************************************************
*
* @param[in] sequence
* Identifies the sequence of function calls that this call belongs to
* (for completion checks and exception handling purposes).
*
* @param[out] request
* Identifies this function call (for exception handling purposes).
*
*******************************************************************************
*
* @sa PLASMA_sgelqf
* @sa PLASMA_sgelqf_Tile
* @sa PLASMA_cgelqf_Tile_Async
* @sa PLASMA_dgelqf_Tile_Async
* @sa PLASMA_sgelqf_Tile_Async
* @sa PLASMA_sgelqs_Tile_Async
*
******************************************************************************/
int PLASMA_sgelqf_Tile_Async(PLASMA_desc *A, PLASMA_desc *T,
PLASMA_sequence *sequence, PLASMA_request *request)
{
PLASMA_desc descA = *A;
PLASMA_desc descT = *T;
plasma_context_t *plasma;
plasma = plasma_context_self();
if (plasma == NULL) {
plasma_fatal_error("PLASMA_sgelqf_Tile", "PLASMA not initialized");
return PLASMA_ERR_NOT_INITIALIZED;
}
if (sequence == NULL) {
plasma_fatal_error("PLASMA_sgelqf_Tile", "NULL sequence");
return PLASMA_ERR_UNALLOCATED;
}
if (request == NULL) {
plasma_fatal_error("PLASMA_sgelqf_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_sgelqf_Tile", "invalid first descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
if (plasma_desc_check(&descT) != PLASMA_SUCCESS) {
plasma_error("PLASMA_sgelqf_Tile", "invalid second descriptor");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Check input arguments */
if (descA.nb != descA.mb) {
plasma_error("PLASMA_sgelqf_Tile", "only square tiles supported");
return plasma_request_fail(sequence, request, PLASMA_ERR_ILLEGAL_VALUE);
}
/* Quick return */
/*
if (min(M, N) == 0)
return PLASMA_SUCCESS;
*/
if (plasma->householder == PLASMA_FLAT_HOUSEHOLDER) {
plasma_parallel_call_4(plasma_psgelqf,
PLASMA_desc, descA,
PLASMA_desc, descT,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
}
else {
plasma_dynamic_call_5(plasma_psgelqfrh,
PLASMA_desc, descA,
PLASMA_desc, descT,
PLASMA_enum, PLASMA_RHBLK,
PLASMA_sequence*, sequence,
PLASMA_request*, request);
}
return PLASMA_SUCCESS;
}
