MAGMA  2.2.0
Matrix Algebra for GPU and Multicore Architectures
Here is a list of all modules:
[detail level 12345]
 Allocate GPU device memoryImalloc, smalloc, etc
 Allocate CPU host memoryImalloc_cpu, smalloc_cpu, etc
 Allocate pinned CPU host memoryImalloc_pinned, smalloc_pinned, etc
 Communication CPU <=> GPU
 Constants and convertersMappings between LAPACK, MAGMA, CBLAS, cuBLAS, and clBLAS constants
 Device management
 Queue management
 Event management
 Error handling
 Miscellaneous utilities
 Print matrix
 Tuning (get_nb, etc.)Optimal block sizes vary with GPU and, to a lesser extent, CPU
 Internal routines============================================================
 Error handling
 Testing routines
 Thread management
 Timer utilities
 QR panel to q, q to panel
 GPU Kernels
 Dense linear algebra============================================================
 Linear system solversSolves \( Ax = b \)
 Orthogonal/unitary factorizations
 Singular Value Decomposition (SVD)
 MAGMA BLAS and Auxiliary
 Linear system solversSolves \( Ax = b \)
 Orthogonal/unitary factorizations
 MAGMA BLAS and Auxiliary
 Sparse linear systemsSolve \( Ax = b \)
 Sparse eigenvalues
 Sparse preconditioners
 GPU kernels for sparse LA
 Sparse BLAS
 Sparse auxiliary
 Sparse unfiled
 UnusedFunctions that LAPACK (or another package) have, which MAGMA does not have
 gerfs: Refine solution
 porfs: Refine solution
 hetrs: symmetric/Hermitian indefinite forward and back solves
 hetri: symmetric/Hermitian indefinite inverse
 herfs: Refine solution
 sysv: Solves Ax = b using symmetric indefinite factorization (driver)
 sytrf: Symmetric indefinite factorization (Bunch-Kaufman pivoting)
 sytrf: Symmetric indefinite factorization (Aasen)
 sytrs: Symmetric indefinite forward and back solves
 sytri: Symmetric indefinite inverse
 syrfs: Refine solution
 lasyf: Partial factorization; used by sytrf
 or/unmhr: Multiply by Q from Hessenberg reduction
 lahrd: Partial factorization; used by gehrd
 laex2: Merge two sets of eigenvalues.
 larf: Apply Householder reflector to general matrix
 larft: Generate T matrix for block of Householder reflectors
 lantr: Triangular matrix norm1, Frobenius, or Infinity norm; or largest element
 sy/heev: Solves using QR iteration (driver)
 sy/heevrx: Solves using MRRR (expert)
 sy/hegv: Solves using QR iteration (driver)
 sy/hegvrx: Solves using MRRR (expert)