MatSetValues calls =%D => MatSetValues calls=%D
Remove legacy and cmake compile systems- Remove legacy and cmake compile systems from makefiles and config- Clean out various dead code from makefile and lib/petsc/conf/rules- Update lib/petsc/co
Remove legacy and cmake compile systems- Remove legacy and cmake compile systems from makefiles and config- Clean out various dead code from makefile and lib/petsc/conf/rules- Update lib/petsc/conf/test to build all examples, not just those beginning with ex, update examples makefiles to remove now unneeded rules- Deprecated make gnumake, replaced with make libsNote: retained clean-legacy since it is used by clean in rulesCommit-type: style-fix, featureThanks-to: Jed Brown <jed@jedbrown.org> and Satish Balay <balay@mcs.anl.gov>
show more ...
TAOSetUpdate: change the Update callbackIn SNES, the update function does not uses contexts, and thus it does not appear in the callback prototypeWith TAO, since we are explcitily passing the ctx
TAOSetUpdate: change the Update callbackIn SNES, the update function does not uses contexts, and thus it does not appear in the callback prototypeWith TAO, since we are explcitily passing the ctx via TAOSetUpdate, we have to give it back to the user when performing the call
Change CHKERRQ(ierr) after PetscInitialize() calls in examples to if (ierr) return ierr;Since the PETSc initialization may not have been completed in the failed PetscInitialize() it is notpropper
Change CHKERRQ(ierr) after PetscInitialize() calls in examples to if (ierr) return ierr;Since the PETSc initialization may not have been completed in the failed PetscInitialize() it is notpropper to call the PETSc error handling routinesCommit-type: bug-fix, testing-fix, exampleReported-by: jed Brown <jed@jedbrown.org>
Regularization changed to x-x_old, user hook added into optimization loops, eliminated explicit MatTransposeMatMult option in Hessian computation
test files updated to account for minor roundoff differences
Fixed a few more things. Added back in recycling. Added the ability for the user to set their own preconditioner with TaoBNCGSetH0.Removed the experimental Steffenson acceleration feature from this
Fixed a few more things. Added back in recycling. Added the ability for the user to set their own preconditioner with TaoBNCGSetH0.Removed the experimental Steffenson acceleration feature from this branch.Fixed everything in Todd's code walkthrough except turning the VecDots into VecDotBegin, VecDotEnd combinationsForgot to also remove xi from DiagBrdn. That is gone now too.Updated the TAO manual so far.Replaced VecDot with VecDotBegin and VecDotEnd in DiagBrdn.Addressing an uninitialized variable problem after merging with another branch.Changed the PetscMaxes back into zero checks; they were degrading the solvers.Whatever was broken is now fixed. Need to test all methods officially again. diagbrdn.c was reset to before Alp and Todd's comments were fixed.Fixed m=1 in diagbrdn.cTried changing the PetscMax bounds to 1.e-10 from 1.e-8Bring it back to 1.e-8 since it did a little worse. Worth tinkering with.Changed the last couple too-specific error checks on scaling terms in DiagBrdn.Changed data structure in bncg.h to put IS data type right after Vec.Removed alpha==-1.0 from DiagBrdn. Updated TAO manual to reflect name changes and added a parameter to it not added before.Added a KSPInitialize in BNCG_CreateHad to fix some logic with recycling.Minor bugfix.Fixed the recycle off-by-one bug from last commit.Fixed issues regarding complex numbers.Fix call to PetscIsInfOrNan.Managed to complete many of Todd's code walkthrough pt. 2 comments.Fixed things from Todd's code walkthrough. Primarily allowed more user options and added some comments on functionality.Changed default behavior of KouDai.Changed CG_Types to not have an underscore.
Updates to alt files for single precision and KNL
Tests now run to 1e-5 gatol at the lowest to ensure that output files are consistent across single and double precision
fix: small change to to max_it convergence check changed some iteration-limited test outputs
fix propagated throughout, convergence tolerance bug discovered and fixed in toy problem
indent level fix for output file
bug fixes for linux-matlab-ilp64 and cuda-single
test files updated for latest changed
Test outputs updated
Wrapping up for a pull request
turn off testing for many examples for !single since behavior is very different than doubleCommit-type: portability-fix, testing-fix
Convert TaoDefaultMonitor() to TaoMonitorDefault() to match other PETSc monitorsand have it format the output in a similar way with the name of the solver TAO.Commit-type: style-fix, feature, doc
Convert TaoDefaultMonitor() to TaoMonitorDefault() to match other PETSc monitorsand have it format the output in a similar way with the name of the solver TAO.Commit-type: style-fix, feature, documentation
Mark most TAO examples as not runable from complex.Commit-type: portability-fix, testing-fix
converted another directory over to new test harnessCommit-type: testing-fix
TAO: added matrix-free finite difference option for Hessian evaluation and VecLock{Push|Pop} around user callsAdded test for MFFD hessianFix several examples that improperly used VecGetArrayFix m
TAO: added matrix-free finite difference option for Hessian evaluation and VecLock{Push|Pop} around user callsAdded test for MFFD hessianFix several examples that improperly used VecGetArrayFix man pages and some coding style in tao code
Fix math calls to use the macros in petscmath.h
Remove the use and definition of __FUNCT__ throughout the codeSince all modern C/C++ compilers provide this functionality we no longer need to provide it manually in PETScTime: 1.5 hoursThanks-t
Remove the use and definition of __FUNCT__ throughout the codeSince all modern C/C++ compilers provide this functionality we no longer need to provide it manually in PETScTime: 1.5 hoursThanks-to: Andreas Mang <andreas@ices.utexas.edu>
Removed DMSetFromOptions() and DMSetUp() from DMDACreateNd()Time: 4 hours
more printdot so travis does not time out
12345678