MAGMA
2.7.1
Matrix Algebra for GPU and Multicore Architectures
|
Functions | |
magma_int_t | magma_cgetrf (magma_int_t m, magma_int_t n, magmaFloatComplex *A, magma_int_t lda, magma_int_t *ipiv, magma_int_t *info) |
CGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_cgetrf2_mgpu (magma_int_t ngpu, magma_int_t m, magma_int_t n, magma_int_t nb, magma_int_t offset, magmaFloatComplex_ptr d_lAT[], magma_int_t lddat, magma_int_t *ipiv, magmaFloatComplex_ptr d_lAP[], magmaFloatComplex *W, magma_int_t ldw, magma_queue_t queues[][2], magma_int_t *info) |
CGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_cgetrf_gpu_expert (magma_int_t m, magma_int_t n, magmaFloatComplex_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info, magma_int_t nb, magma_mode_t mode) |
CGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_cgetrf_gpu (magma_int_t m, magma_int_t n, magmaFloatComplex_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
magma_cgetrf_gpu_expert with mode = MagmaHybrid. More... | |
magma_int_t | magma_cgetrf_native (magma_int_t m, magma_int_t n, magmaFloatComplex_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
magma_cgetrf_gpu_expert with mode = MagmaNative. More... | |
magma_int_t | magma_cgetrf_m (magma_int_t ngpu, magma_int_t m, magma_int_t n, magmaFloatComplex *A, magma_int_t lda, magma_int_t *ipiv, magma_int_t *info) |
CGETRF_m computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_cgetrf_mgpu (magma_int_t ngpu, magma_int_t m, magma_int_t n, magmaFloatComplex_ptr d_lA[], magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
CGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_dgetrf (magma_int_t m, magma_int_t n, double *A, magma_int_t lda, magma_int_t *ipiv, magma_int_t *info) |
DGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_dgetrf2_mgpu (magma_int_t ngpu, magma_int_t m, magma_int_t n, magma_int_t nb, magma_int_t offset, magmaDouble_ptr d_lAT[], magma_int_t lddat, magma_int_t *ipiv, magmaDouble_ptr d_lAP[], double *W, magma_int_t ldw, magma_queue_t queues[][2], magma_int_t *info) |
DGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_dgetrf_gpu_expert (magma_int_t m, magma_int_t n, magmaDouble_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info, magma_int_t nb, magma_mode_t mode) |
DGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_dgetrf_gpu (magma_int_t m, magma_int_t n, magmaDouble_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
magma_dgetrf_gpu_expert with mode = MagmaHybrid. More... | |
magma_int_t | magma_dgetrf_native (magma_int_t m, magma_int_t n, magmaDouble_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
magma_dgetrf_gpu_expert with mode = MagmaNative. More... | |
magma_int_t | magma_dgetrf_m (magma_int_t ngpu, magma_int_t m, magma_int_t n, double *A, magma_int_t lda, magma_int_t *ipiv, magma_int_t *info) |
DGETRF_m computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_dgetrf_mgpu (magma_int_t ngpu, magma_int_t m, magma_int_t n, magmaDouble_ptr d_lA[], magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
DGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_sgetrf (magma_int_t m, magma_int_t n, float *A, magma_int_t lda, magma_int_t *ipiv, magma_int_t *info) |
SGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_sgetrf2_mgpu (magma_int_t ngpu, magma_int_t m, magma_int_t n, magma_int_t nb, magma_int_t offset, magmaFloat_ptr d_lAT[], magma_int_t lddat, magma_int_t *ipiv, magmaFloat_ptr d_lAP[], float *W, magma_int_t ldw, magma_queue_t queues[][2], magma_int_t *info) |
SGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_sgetrf_gpu_expert (magma_int_t m, magma_int_t n, magmaFloat_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info, magma_int_t nb, magma_mode_t mode) |
SGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_sgetrf_gpu (magma_int_t m, magma_int_t n, magmaFloat_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
magma_sgetrf_gpu_expert with mode = MagmaHybrid. More... | |
magma_int_t | magma_sgetrf_native (magma_int_t m, magma_int_t n, magmaFloat_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
magma_sgetrf_gpu_expert with mode = MagmaNative. More... | |
magma_int_t | magma_sgetrf_m (magma_int_t ngpu, magma_int_t m, magma_int_t n, float *A, magma_int_t lda, magma_int_t *ipiv, magma_int_t *info) |
SGETRF_m computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_sgetrf_mgpu (magma_int_t ngpu, magma_int_t m, magma_int_t n, magmaFloat_ptr d_lA[], magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
SGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_xhsgetrf_gpu (magma_int_t m, magma_int_t n, magmaFloat_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info, magma_mp_type_t enable_tc, magma_mp_type_t mp_algo_type) |
XHSGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_hgetrf_gpu (magma_int_t m, magma_int_t n, magmaFloat_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
HGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_xshgetrf_gpu (magma_int_t m, magma_int_t n, magmaFloat_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info, magma_mp_type_t enable_tc, magma_mp_type_t mp_algo_type) |
XSHGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_htgetrf_gpu (magma_int_t m, magma_int_t n, magmaFloat_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
HTGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_zgetrf (magma_int_t m, magma_int_t n, magmaDoubleComplex *A, magma_int_t lda, magma_int_t *ipiv, magma_int_t *info) |
ZGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_zgetrf2_mgpu (magma_int_t ngpu, magma_int_t m, magma_int_t n, magma_int_t nb, magma_int_t offset, magmaDoubleComplex_ptr d_lAT[], magma_int_t lddat, magma_int_t *ipiv, magmaDoubleComplex_ptr d_lAP[], magmaDoubleComplex *W, magma_int_t ldw, magma_queue_t queues[][2], magma_int_t *info) |
ZGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_zgetrf_gpu_expert (magma_int_t m, magma_int_t n, magmaDoubleComplex_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info, magma_int_t nb, magma_mode_t mode) |
ZGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_zgetrf_gpu (magma_int_t m, magma_int_t n, magmaDoubleComplex_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
magma_zgetrf_gpu_expert with mode = MagmaHybrid. More... | |
magma_int_t | magma_zgetrf_native (magma_int_t m, magma_int_t n, magmaDoubleComplex_ptr dA, magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
magma_zgetrf_gpu_expert with mode = MagmaNative. More... | |
magma_int_t | magma_zgetrf_m (magma_int_t ngpu, magma_int_t m, magma_int_t n, magmaDoubleComplex *A, magma_int_t lda, magma_int_t *ipiv, magma_int_t *info) |
ZGETRF_m computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t | magma_zgetrf_mgpu (magma_int_t ngpu, magma_int_t m, magma_int_t n, magmaDoubleComplex_ptr d_lA[], magma_int_t ldda, magma_int_t *ipiv, magma_int_t *info) |
ZGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges. More... | |
magma_int_t magma_cgetrf | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloatComplex * | A, | ||
magma_int_t | lda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
CGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
This version does not require work space on the GPU passed as input. GPU memory is allocated in the routine.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
It uses 2 queues to overlap communication and computation.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | A | COMPLEX array, dimension (LDA,N) On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. Higher performance is achieved if A is in pinned memory, e.g. allocated using magma_malloc_pinned. |
[in] | lda | INTEGER The leading dimension of the array A. LDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_cgetrf2_mgpu | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magma_int_t | nb, | ||
magma_int_t | offset, | ||
magmaFloatComplex_ptr | d_lAT[], | ||
magma_int_t | lddat, | ||
magma_int_t * | ipiv, | ||
magmaFloatComplex_ptr | d_lAP[], | ||
magmaFloatComplex * | W, | ||
magma_int_t | ldw, | ||
magma_queue_t | queues[][2], | ||
magma_int_t * | info | ||
) |
CGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm. Use two buffer to send panels.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in] | nb | INTEGER The block size used for the matrix distribution. |
[in] | offset | INTEGER The first row and column indices of the submatrix that this routine will factorize. |
[in,out] | d_lAT | COMPLEX array of pointers on the GPU, dimension (ngpu). On entry, the M-by-N matrix A distributed over GPUs (d_lAT[d] points to the local matrix on d-th GPU). It uses a 1D block column cyclic format (with the block size nb), and each local matrix is stored by row. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | lddat | INTEGER The leading dimension of the array d_lAT[d]. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[in] | d_lAP | COMPLEX array of pointers on the GPU, dimension (ngpu). d_lAP[d] is the workspace on d-th GPU. Each local workspace must be of size (3+ngpu)*nb*maxm, where maxm is m rounded up to a multiple of 32 and nb is the block size. |
[in] | W | COMPLEX array, dimension (ngpu*nb*maxm). It is used to store panel on CPU. |
[in] | ldw | INTEGER The leading dimension of the workspace w. |
[in] | queues | magma_queue_t queues[d] points to the queues for the d-th GPU to execute in. Each GPU require two queues. |
[out] | info | INTEGER
|
magma_int_t magma_cgetrf_gpu_expert | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloatComplex_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info, | ||
magma_int_t | nb, | ||
magma_mode_t | mode | ||
) |
CGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | dA | COMPLEX array on the GPU, dimension (LDDA,N). On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array A. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
[in] | mode | magma_mode_t
|
magma_int_t magma_cgetrf_gpu | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloatComplex_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
magma_cgetrf_gpu_expert with mode = MagmaHybrid.
Computation is hybrid, part on CPU (panels), part on GPU (matrix updates).
magma_int_t magma_cgetrf_native | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloatComplex_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
magma_cgetrf_gpu_expert with mode = MagmaNative.
Computation is done only on the GPU, not on the CPU.
magma_int_t magma_cgetrf_m | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magmaFloatComplex * | A, | ||
magma_int_t | lda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
CGETRF_m computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
This version does not require work space on the GPU passed as input. GPU memory is allocated in the routine. The matrix may exceed the GPU memory.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
Note: The factorization of big panel is done calling multiple-gpu-interface. Pivots are applied on GPU within the big panel.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | A | COMPLEX array, dimension (LDA,N) On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. Higher performance is achieved if A is in pinned memory, e.g. allocated using magma_malloc_pinned. |
[in] | lda | INTEGER The leading dimension of the array A. LDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_cgetrf_mgpu | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magmaFloatComplex_ptr | d_lA[], | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
CGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | d_lA | COMPLEX array of pointers on the GPU, dimension (ngpu). On entry, the M-by-N matrix A distributed over GPUs (d_lA[d] points to the local matrix on d-th GPU). It uses 1D block column cyclic format with the block size of nb, and each local matrix is stored by column. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array d_lA. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_dgetrf | ( | magma_int_t | m, |
magma_int_t | n, | ||
double * | A, | ||
magma_int_t | lda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
DGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
This version does not require work space on the GPU passed as input. GPU memory is allocated in the routine.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
It uses 2 queues to overlap communication and computation.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | A | DOUBLE PRECISION array, dimension (LDA,N) On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. Higher performance is achieved if A is in pinned memory, e.g. allocated using magma_malloc_pinned. |
[in] | lda | INTEGER The leading dimension of the array A. LDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_dgetrf2_mgpu | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magma_int_t | nb, | ||
magma_int_t | offset, | ||
magmaDouble_ptr | d_lAT[], | ||
magma_int_t | lddat, | ||
magma_int_t * | ipiv, | ||
magmaDouble_ptr | d_lAP[], | ||
double * | W, | ||
magma_int_t | ldw, | ||
magma_queue_t | queues[][2], | ||
magma_int_t * | info | ||
) |
DGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm. Use two buffer to send panels.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in] | nb | INTEGER The block size used for the matrix distribution. |
[in] | offset | INTEGER The first row and column indices of the submatrix that this routine will factorize. |
[in,out] | d_lAT | DOUBLE PRECISION array of pointers on the GPU, dimension (ngpu). On entry, the M-by-N matrix A distributed over GPUs (d_lAT[d] points to the local matrix on d-th GPU). It uses a 1D block column cyclic format (with the block size nb), and each local matrix is stored by row. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | lddat | INTEGER The leading dimension of the array d_lAT[d]. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[in] | d_lAP | DOUBLE PRECISION array of pointers on the GPU, dimension (ngpu). d_lAP[d] is the workspace on d-th GPU. Each local workspace must be of size (3+ngpu)*nb*maxm, where maxm is m rounded up to a multiple of 32 and nb is the block size. |
[in] | W | DOUBLE PRECISION array, dimension (ngpu*nb*maxm). It is used to store panel on CPU. |
[in] | ldw | INTEGER The leading dimension of the workspace w. |
[in] | queues | magma_queue_t queues[d] points to the queues for the d-th GPU to execute in. Each GPU require two queues. |
[out] | info | INTEGER
|
magma_int_t magma_dgetrf_gpu_expert | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaDouble_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info, | ||
magma_int_t | nb, | ||
magma_mode_t | mode | ||
) |
DGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | dA | DOUBLE PRECISION array on the GPU, dimension (LDDA,N). On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array A. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
[in] | mode | magma_mode_t
|
magma_int_t magma_dgetrf_gpu | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaDouble_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
magma_dgetrf_gpu_expert with mode = MagmaHybrid.
Computation is hybrid, part on CPU (panels), part on GPU (matrix updates).
magma_int_t magma_dgetrf_native | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaDouble_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
magma_dgetrf_gpu_expert with mode = MagmaNative.
Computation is done only on the GPU, not on the CPU.
magma_int_t magma_dgetrf_m | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
double * | A, | ||
magma_int_t | lda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
DGETRF_m computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
This version does not require work space on the GPU passed as input. GPU memory is allocated in the routine. The matrix may exceed the GPU memory.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
Note: The factorization of big panel is done calling multiple-gpu-interface. Pivots are applied on GPU within the big panel.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | A | DOUBLE PRECISION array, dimension (LDA,N) On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. Higher performance is achieved if A is in pinned memory, e.g. allocated using magma_malloc_pinned. |
[in] | lda | INTEGER The leading dimension of the array A. LDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_dgetrf_mgpu | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magmaDouble_ptr | d_lA[], | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
DGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | d_lA | DOUBLE PRECISION array of pointers on the GPU, dimension (ngpu). On entry, the M-by-N matrix A distributed over GPUs (d_lA[d] points to the local matrix on d-th GPU). It uses 1D block column cyclic format with the block size of nb, and each local matrix is stored by column. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array d_lA. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_sgetrf | ( | magma_int_t | m, |
magma_int_t | n, | ||
float * | A, | ||
magma_int_t | lda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
SGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
This version does not require work space on the GPU passed as input. GPU memory is allocated in the routine.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
It uses 2 queues to overlap communication and computation.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | A | REAL array, dimension (LDA,N) On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. Higher performance is achieved if A is in pinned memory, e.g. allocated using magma_malloc_pinned. |
[in] | lda | INTEGER The leading dimension of the array A. LDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_sgetrf2_mgpu | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magma_int_t | nb, | ||
magma_int_t | offset, | ||
magmaFloat_ptr | d_lAT[], | ||
magma_int_t | lddat, | ||
magma_int_t * | ipiv, | ||
magmaFloat_ptr | d_lAP[], | ||
float * | W, | ||
magma_int_t | ldw, | ||
magma_queue_t | queues[][2], | ||
magma_int_t * | info | ||
) |
SGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm. Use two buffer to send panels.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in] | nb | INTEGER The block size used for the matrix distribution. |
[in] | offset | INTEGER The first row and column indices of the submatrix that this routine will factorize. |
[in,out] | d_lAT | REAL array of pointers on the GPU, dimension (ngpu). On entry, the M-by-N matrix A distributed over GPUs (d_lAT[d] points to the local matrix on d-th GPU). It uses a 1D block column cyclic format (with the block size nb), and each local matrix is stored by row. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | lddat | INTEGER The leading dimension of the array d_lAT[d]. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[in] | d_lAP | REAL array of pointers on the GPU, dimension (ngpu). d_lAP[d] is the workspace on d-th GPU. Each local workspace must be of size (3+ngpu)*nb*maxm, where maxm is m rounded up to a multiple of 32 and nb is the block size. |
[in] | W | REAL array, dimension (ngpu*nb*maxm). It is used to store panel on CPU. |
[in] | ldw | INTEGER The leading dimension of the workspace w. |
[in] | queues | magma_queue_t queues[d] points to the queues for the d-th GPU to execute in. Each GPU require two queues. |
[out] | info | INTEGER
|
magma_int_t magma_sgetrf_gpu_expert | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloat_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info, | ||
magma_int_t | nb, | ||
magma_mode_t | mode | ||
) |
SGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | dA | REAL array on the GPU, dimension (LDDA,N). On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array A. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
[in] | mode | magma_mode_t
|
magma_int_t magma_sgetrf_gpu | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloat_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
magma_sgetrf_gpu_expert with mode = MagmaHybrid.
Computation is hybrid, part on CPU (panels), part on GPU (matrix updates).
magma_int_t magma_sgetrf_native | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloat_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
magma_sgetrf_gpu_expert with mode = MagmaNative.
Computation is done only on the GPU, not on the CPU.
magma_int_t magma_sgetrf_m | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
float * | A, | ||
magma_int_t | lda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
SGETRF_m computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
This version does not require work space on the GPU passed as input. GPU memory is allocated in the routine. The matrix may exceed the GPU memory.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
Note: The factorization of big panel is done calling multiple-gpu-interface. Pivots are applied on GPU within the big panel.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | A | REAL array, dimension (LDA,N) On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. Higher performance is achieved if A is in pinned memory, e.g. allocated using magma_malloc_pinned. |
[in] | lda | INTEGER The leading dimension of the array A. LDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_sgetrf_mgpu | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magmaFloat_ptr | d_lA[], | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
SGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | d_lA | REAL array of pointers on the GPU, dimension (ngpu). On entry, the M-by-N matrix A distributed over GPUs (d_lA[d] points to the local matrix on d-th GPU). It uses 1D block column cyclic format with the block size of nb, and each local matrix is stored by column. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array d_lA. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_xhsgetrf_gpu | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloat_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info, | ||
magma_mp_type_t | enable_tc, | ||
magma_mp_type_t | mp_algo_type | ||
) |
XHSGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
It uses mixed precision FP32/FP16-w/o TensorCores factorization techniques.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | dA | REAL array on the GPU, dimension (LDDA,N). On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array A. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
[in] | enable_tc | MAGMA_MP_TYPE_T internal and expert API uses. enable/disable tensor cores |
[in] | mp_algo_type | MAGMA_MP_TYPE_T internal and expert API uses. |
magma_int_t magma_hgetrf_gpu | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloat_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
HGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
It uses mixed precision FP32/FP16 techniques.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | dA | REAL array on the GPU, dimension (LDDA,N). On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array A. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
More details can be found in Azzam Haidar, Stanimire Tomov, Jack Dongarra, and Nicholas J. Higham. 2018. Harnessing GPU tensor cores for fast FP16 arithmetic to speed up mixed-precision iterative refinement solvers. In Proceedings of the International Conference for High Performance Computing, Networking, Storage, and Analysis (SC '18). IEEE Press, Piscataway, NJ, USA, Article 47, 11 pages.
magma_int_t magma_xshgetrf_gpu | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloat_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info, | ||
magma_mp_type_t | enable_tc, | ||
magma_mp_type_t | mp_algo_type | ||
) |
XSHGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
It uses mixed precision FP32/FP16-w/o TensorCores factorization techniques.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | dA | REAL array on the GPU, dimension (LDDA,N). On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array A. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
[in] | enable_tc | MAGMA_MP_TYPE_T internal and expert API uses. enable/disable tensor cores |
[in] | mp_algo_type | MAGMA_MP_TYPE_T internal and expert API uses. |
More details can be found in Azzam Haidar, Stanimire Tomov, Jack Dongarra, and Nicholas J. Higham. 2018. Harnessing GPU tensor cores for fast FP16 arithmetic to speed up mixed-precision iterative refinement solvers. In Proceedings of the International Conference for High Performance Computing, Networking, Storage, and Analysis (SC '18). IEEE Press, Piscataway, NJ, USA, Article 47, 11 pages.
magma_int_t magma_htgetrf_gpu | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaFloat_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
HTGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
It uses mixed precision FP32/FP16-TensorCores factorization techniques.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | dA | REAL array on the GPU, dimension (LDDA,N). On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array A. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_zgetrf | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaDoubleComplex * | A, | ||
magma_int_t | lda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
ZGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
This version does not require work space on the GPU passed as input. GPU memory is allocated in the routine.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
It uses 2 queues to overlap communication and computation.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | A | COMPLEX_16 array, dimension (LDA,N) On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. Higher performance is achieved if A is in pinned memory, e.g. allocated using magma_malloc_pinned. |
[in] | lda | INTEGER The leading dimension of the array A. LDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_zgetrf2_mgpu | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magma_int_t | nb, | ||
magma_int_t | offset, | ||
magmaDoubleComplex_ptr | d_lAT[], | ||
magma_int_t | lddat, | ||
magma_int_t * | ipiv, | ||
magmaDoubleComplex_ptr | d_lAP[], | ||
magmaDoubleComplex * | W, | ||
magma_int_t | ldw, | ||
magma_queue_t | queues[][2], | ||
magma_int_t * | info | ||
) |
ZGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm. Use two buffer to send panels.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in] | nb | INTEGER The block size used for the matrix distribution. |
[in] | offset | INTEGER The first row and column indices of the submatrix that this routine will factorize. |
[in,out] | d_lAT | COMPLEX_16 array of pointers on the GPU, dimension (ngpu). On entry, the M-by-N matrix A distributed over GPUs (d_lAT[d] points to the local matrix on d-th GPU). It uses a 1D block column cyclic format (with the block size nb), and each local matrix is stored by row. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | lddat | INTEGER The leading dimension of the array d_lAT[d]. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[in] | d_lAP | COMPLEX_16 array of pointers on the GPU, dimension (ngpu). d_lAP[d] is the workspace on d-th GPU. Each local workspace must be of size (3+ngpu)*nb*maxm, where maxm is m rounded up to a multiple of 32 and nb is the block size. |
[in] | W | COMPLEX_16 array, dimension (ngpu*nb*maxm). It is used to store panel on CPU. |
[in] | ldw | INTEGER The leading dimension of the workspace w. |
[in] | queues | magma_queue_t queues[d] points to the queues for the d-th GPU to execute in. Each GPU require two queues. |
[out] | info | INTEGER
|
magma_int_t magma_zgetrf_gpu_expert | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaDoubleComplex_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info, | ||
magma_int_t | nb, | ||
magma_mode_t | mode | ||
) |
ZGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | dA | COMPLEX_16 array on the GPU, dimension (LDDA,N). On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array A. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
[in] | mode | magma_mode_t
|
magma_int_t magma_zgetrf_gpu | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaDoubleComplex_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
magma_zgetrf_gpu_expert with mode = MagmaHybrid.
Computation is hybrid, part on CPU (panels), part on GPU (matrix updates).
magma_int_t magma_zgetrf_native | ( | magma_int_t | m, |
magma_int_t | n, | ||
magmaDoubleComplex_ptr | dA, | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
magma_zgetrf_gpu_expert with mode = MagmaNative.
Computation is done only on the GPU, not on the CPU.
magma_int_t magma_zgetrf_m | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magmaDoubleComplex * | A, | ||
magma_int_t | lda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
ZGETRF_m computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
This version does not require work space on the GPU passed as input. GPU memory is allocated in the routine. The matrix may exceed the GPU memory.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
Note: The factorization of big panel is done calling multiple-gpu-interface. Pivots are applied on GPU within the big panel.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | A | COMPLEX_16 array, dimension (LDA,N) On entry, the M-by-N matrix to be factored. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. Higher performance is achieved if A is in pinned memory, e.g. allocated using magma_malloc_pinned. |
[in] | lda | INTEGER The leading dimension of the array A. LDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|
magma_int_t magma_zgetrf_mgpu | ( | magma_int_t | ngpu, |
magma_int_t | m, | ||
magma_int_t | n, | ||
magmaDoubleComplex_ptr | d_lA[], | ||
magma_int_t | ldda, | ||
magma_int_t * | ipiv, | ||
magma_int_t * | info | ||
) |
ZGETRF computes an LU factorization of a general M-by-N matrix A using partial pivoting with row interchanges.
The factorization has the form A = P * L * U where P is a permutation matrix, L is lower triangular with unit diagonal elements (lower trapezoidal if m > n), and U is upper triangular (upper trapezoidal if m < n).
This is the right-looking Level 3 BLAS version of the algorithm.
[in] | ngpu | INTEGER Number of GPUs to use. ngpu > 0. |
[in] | m | INTEGER The number of rows of the matrix A. M >= 0. |
[in] | n | INTEGER The number of columns of the matrix A. N >= 0. |
[in,out] | d_lA | COMPLEX_16 array of pointers on the GPU, dimension (ngpu). On entry, the M-by-N matrix A distributed over GPUs (d_lA[d] points to the local matrix on d-th GPU). It uses 1D block column cyclic format with the block size of nb, and each local matrix is stored by column. On exit, the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored. |
[in] | ldda | INTEGER The leading dimension of the array d_lA. LDDA >= max(1,M). |
[out] | ipiv | INTEGER array, dimension (min(M,N)) The pivot indices; for 1 <= i <= min(M,N), row i of the matrix was interchanged with row IPIV(i). |
[out] | info | INTEGER
|