PETSc

Documentation: Changes: 3.10

• Home
• Download
• Features
• Documentation
  • Manual pages and Users Manual
  • Citing PETSc
  • Tutorials
  • Installation
  • SAWs
  • Changes
    • dev
    • 3.15
    • 3.14
    • 3.13
    • 3.12
    • 3.11
    • 3.10
    • 3.9
    • 3.8
    • 3.7
    • 3.6
    • 3.5
    • 3.4
    • 3.3
    • 3.2
    • 3.1
    • 3.0.0
    • 2.3.3
    • 2.3.2
    • 2.3.1
    • 2.3.0
    • 2.2.1
    • 2.2.0
    • 2.1.6
    • 2.1.5
    • 2.1.3
    • 2.1.2
    • 2.1.1
    • 2.1.0
    • 2.0.29
    • 2.0.28
    • 2.0.24
    • 2.0.22
    • 2.0.18-21
    • 2.0.17
    • 2.0.16
    • 2.0.15
  • Bug Reporting
  • Code Management
  • FAQ
  • License
  • Linear Solver Table
  • Nonlinear Solver Table
  • TAO Solver Table
• Applications/Publications
• Miscellaneous
• External Software
• Developers Site

General:

• Removed --with-errorchecking configure option.
• Added new command line option -viewfromoptions to disable XXXViewFromOptions() calls.

Configure/Build:

• Removed --with-viewfromoptions configure option in favor of the new command line option -viewfromoptions.
• Following options are renamed to have more consistant names:

            --with-packages-dir          -> --with-packages-download-dir
            --with-external-packages-dir -> --with-packages-build-dir
            --package-dirs               -> --with-packages-search-path
            --search-dirs                -> --with-executables-search-path

IS:

PetscDraw:

PF:

Vec:

VecScatter:

PetscSection:

Mat:

• MatNullSpaceTest() now respects the options prefix of the provided Mat.
• Added MatMatTransposeSolve() to MUMPS interface.
• Added -mat_mumps_use_omp_threads [m] option to better support OpenMP in MUMPS.
• Added MatPartitioningViewFromOptions().
• Added MatPartitioningApplyND() to compute a nested dissection ordering of a matrix.
• Deprecated MatISGetMPIXAIJ() in favour of MatConvert().
• Added new matrix types for limited-memory variable-metric (MATLMVM) approximations of Jacobians.
• Added MATAIJSELL, a sub-type of AIJ that uses SELL format internally for SELL-optimized operations.

PC:

• PCKSPGetKSP() now raises an error if called on a PC which is not of type PCKSP
• Added PCKSPSetKSP()
• Added new PC based on the new LMVM matrices.

KSP:

• The KSP viewing options have changed. Now
  • -ksp_compute_eigenvalues becomes -ksp_view_eigenvalues ascii
  • -ksp_plot_eigenvalues becomes -ksp_view_eigenvalues draw
  • -ksp_plot_eigencontours becomes -ksp_view_eigenvalues draw::draw_contour
  • -ksp_compute_eigenvalues_explicitly becomes -ksp_view_eigenvalues_explicit ascii
  • -ksp_plot_eigenvalues_explicitly becomes -ksp_view_eigenvalues_explicit draw
  • -ksp_compute_singularvalues becomes -ksp_view_singularvalues ascii
  • -ksp_final_residual becomes -ksp_view_final_residual
• KSPLSQR
  • Computes now an estimate of the operator's norm in the same way as Saunders and MATLAB, and uses it in the type-specific stopping criterion KSPLSQRConvergedDefault(); this now gives the same number of iterations as MATLAB's lsqr().
  • Added -ksp_lsqr_exact_mat_norm and KSPLSQRSetExactMatNorm() to optionally compute and use the exact Frobenius norm instead of the estimate.
  • KSPLSQRMonitorDefault() and -ksp_lsqr_monitor now additionally print the (estimate of) the operator's norm.
  • Allow setting -ksp_convergence_test lsqr (it's a default for KSPLSQR, though).
  • KSPLSQRSetStandardErrorVec() is replaced by KSPLSQRSetComputeStandardErrorVec() which now specifies whether standard error vector should be computed instead of setting the vector;
  • likewise, option -ksp_lsqr_set_standard_error is replaced by -ksp_lsqr_compute_standard_error.
  • Renamed KSPLSQRGetArnorm() to KSPLSQRGetNorms() and removed its rhs_norm argument.
  • Renamed KSPLSQRDefaultConverged() to KSPLSQRConvergedDefault() merged with KSPConvergedLSQR().

SNES:

• SNESSetTolerances() and -snes_max_funcs now accept -1 to indicate unlimited number of function evaluations.

SNESLineSearch:

TAO:

• Added bound-constrained Newton-Krylov methods (BNLS, BNTR, BNTL).
• All quasi-Newton methods updated to use the new LMVM Mat objects.
• Added a bound-constrained quasi-Newton line search (BQNLS) method that re-uses BNLS infrastructure to replace BLMVM on next release.
• Bound-constrained conjugate gradient method expanded with new CG update types and nonlinear preconditioning.
• Removed -tao_type test.
• Gradient testing can now be accessed with the options -tao_test_gradient, -tao_test_gradient_view, and Hessian testing can be done with -tao_test_hessian, -tao_test_hessian_view (also works for matrix-free Hessian).

TS:

• Added symplectic time integrators for separable Hamiltonian systems (-ts_type basicsymplectic).

DM/DA:

• Renamed DMDAGetReducedDMDA() to DMDACreateCompatibleDMDA().
• Added DMGetCompatibility() and implementation for DMDA

DMPlex:

• Added an interface DM and boundary label to DMPlexCreateHybridMesh()
• More refined control of DMView() when PetscViewerType=PETSCVIEWERHDF5. PetscViewerFormat can now be one of {PETSC_VIEWER_HDF5_VIZ, PETSC_VIEWER_HDF5_XDMF, PETSC_VIEWER_HDF5_PETSC, PETSC_VIEWER_DEFAULT}.
• Parallel mesh load (i.e. mesh distributed right from beginning) with DMLoad(), PetscViewerType=PETSCVIEWERHDF5 and PetscViewerFormat=PETSC_VIEWER_HDF5_XDMF, or DMPlexCreateFromFile() and -dm_plex_create_from_hdf5_xdmf.

PetscViewer:

Added argument to PetscViewerVTKAddField() to allow skipping of DM check as fields are added

SYS:

AO:

Sieve:

Fortran: