-
Notifications
You must be signed in to change notification settings - Fork 23
coverage
Mark Gates edited this page Jul 13, 2023
·
1 revision
Basic Linear Algebra | status |
---|---|
Level 1 BLAS | no |
Level 2 BLAS | use Level 3 |
Level 3 BLAS | yes |
optimizations for Level 2-2.5 | gemm, hemm, trsm when A >> C |
Level 2-like routines | add, scale, scale_row_col, set |
Matrix norms ️ | yes |
———————————— | ———————————— |
Linear systems (Ax = b) | status |
LU, partial pivoting (PP) | yes |
LU, communication avoiding (CA) | in progress (2022) |
LU, band (PP) | yes |
LU, non-pivoting | yes |
Cholesky | yes |
Cholesky, band | yes |
Symmetric indefinite (block Aasen) | prototype, CPU only |
Mixed precision (single-double) | yes (LU, Cholesky) |
Inverses | yes (LU, Cholesky) |
Condition estimate | no |
———————————— | ———————————— |
Least squares (Ax ≈ b) | status |
QR | yes |
optimizations for tall-skinny | yes |
Cholesky QR | yes |
LQ | yes |
Least squares solver | yes |
———————————— | ———————————— |
SVD and eigenvalues | status |
Singular value decomposition (SVD) | yes, values only |
Hermitian eigenvalues | yes, values only |
vectors | in progress (2022) |
Generalized Hermitian eigenvalues | yes, values only |
Polar decomposition (QDWH) | in progress |
Non-symmetric eigenvalues | no current plan |
———————————— | ———————————— |
Platforms | status |
MPI | yes (≥ 2.0) |
GPU-aware MPI | yes, experimental |
OpenMP | yes (≥ 4.5) |
CUDA for NVIDIA GPUs | yes |
HIP for AMD GPUs | yes |
OpenMP offload for Intel GPUs | in progress (2022) |
———————————— | ———————————— |
Compilers | status |
GNU g++ compiler | yes (≥ 7.3 tested; 6.4 fails) |
Clang compiler | yes (≥ 9.0 tested) |
Intel icpc compiler | yes |
IBM xlc compiler | no (C++17 not supported) |
Cray compiler | yes (clang v14 tested) |