Quasar > Linear algebra > BLAS

BLAS


An overview of the different functions and classes:

blas_amaxDetermines the smallest index of the element with the maximum magnitude of the specified input vector, using the cuBLAS/BLAS library.
blas_aminDetermines the smallest index of the element with the minimum magnitude of the specified input vector using the cuBLAS/BLAS library.
blas_asumCalculates the sum of the absolute values of the input matrix, using the cuBLAS/BLAS library
blas_axpyCalculates the matrix scaling and addition operation using the cuBLAS/BLAS library: : Y = alpha * X + Y
blas_cholCalculates the Cholesky decomposition of the Hermitian positive definite matrix A, using the cuSolver library: :A = L^H * L
blas_cholsolveSolves a linear system of equations using a previous Cholesky decomposition obtained using , using the cuSolver library: :A * X = B where X and B are vectors/matrices (corresponding to multiple right-handed sides)
blas_dotCalculates the dot-product of two vectors or matrices using the cuBLAS/BLAS library: : y = sum(A.*conj(B))
blas_geamCalculates the following matrix-matrix addition/transposition using the cuBLAS/BLAS library: :C = alpha*op(A) + beta*op(B) where : op(A) = At ? transpose(A) : A : op(B) = Bt ? transpose(B) : B
blas_gemmEfficiently calculates the matrix-matrix multiplication and addition using the cuBLAS/BLAS library:
blas_gemvEfficiently calculates the matrix-vector multiplication using the cuBLAS/BLAS library:
blas_gesvdCalculates the singular value decomposition of the matrix A, using the cuSolver library: : A = U * S * VH
blas_hermtransposeCalculates the Hermitian transpose of the matrix X using the cuBLAS/BLAS library.
blas_luPerforms an LU decomposition of the matrix A using the library cuSolver. The result of the operation is stored in the matrices A and P. The LU decomposition is defined as follows: :perm(P) * A = L * U where perm(P) is a permutation matrix (the vector P specifies the permutations to be applied in sequence to the rows of an identity matrix, the i-th permutation is given by [i,P[i]]).
blas_lubatchedPerforms an LU decomposition of a stack of K matrices A_i using the library cuSolver. The result of the operation is stored in the matrices A_i and P_i. The LU decomposition is defined as follows: :perm(P_i) * A_i = L_i * U_i, i = 0, ..., K-1 where perm(P_i) is a permutation matrix (the vector P_i specifies the permutations to be applied in sequence to the rows of an identity matrix, the i-th permutation is given by [i,P[i]]).
blas_lumatinvbatchedCalculates the matrix inverse of a stacked set of input matrices, using a LU factorization previously performed with the function .
blas_lusolveSolves a linear system of multiple right-hand sides using the cuSolver library: : op(A)*X = B where : op(A) = At ? transpose(A) : A
blas_lusolvebatchedSolves multiple linear systems with multiple right-hand sides using the cuSolver library: : op(A_i)*X_i = B_i, i = 0, ..., K-1 where : op(A) = At ? transpose(A) : A
blas_matinvbatchedCalculates the matrix inverse of a stacked set of input matrices, using the cuBLAS library.
blas_nrm2Calculates the Euclidean norm of the vector (or matrix) A, using the cuBLAS/BLAS library: : y = sqrt(sum(abs(A).^2)) In case of a matrix, the Euclidean norm is calculated over all elements of the matrix.
blas_qCalculates the Q factor of a QR matrix obtained using the function .
blas_qrCalculates the QR factorization of an m x n matrix A using the cuSolver library: :A = Q * R
blas_scalScales the matrix X using the cuBLAS/BLAS library: : X = X * alpha
blas_transposeCalculates the transpose of the matrix X using the cuBLAS/BLAS library. The function is identical to the function, but may have different performance characteristics.
blas_trsvSolves a triangular system of equations with a single right-hand side using the cuBLAS / BLAS library: : op(A)X = B where : op(A) = At ? transpose(A) : A
blas_versionReturns the version of the currently installed cuBLAS library.