Commits

Author Commit Message Labels Comments Date
BarryFSmith
fixed bug introduced with last commit, wrong name of KSP field
BarryFSmith
bug in KSPSolve() work vector came from vec_sol when it should have come from vec_rhs
BarryFSmith
removed seterrq for norm, dot and mdot when result is Nan or Inf
Jed Brown
MPIAIJ: Remove block size optimization This optimization cannot be determined locally, thus it is not safe without an additional reduction or different rules. Since this code is executed on each assembly, the additional reduction is not attractive. If the optimization is deemed important, there are several possible resolutions. 1. Guarantee full blocks at the matrix preallocation stage. This is complicated to implement for AIJ and …
Satish Balay
Increase patchlevel to 3.3-p6
Jed Brown
hdf5: remove bogus loop over dimension [petsc-maint 121765] This was leaking filespaces.
Satish Balay
add DESTDIR usage to installation instructions
Jed Brown
Declare variables if either xHSEQR or xTRSEN is available
Satish Balay
#148921 fix atoll() check for windows
Satish Balay
#147425 - initialize PETSC_NULL on fortran [so as to detect its incorrect usage]
Jed Brown
Fix non-compliant C89 Test --with-64-bit-indices 'CFLAGS=-std=c89 -pedantic -Wno-long-long'
BarryFSmith
Backport 56cb5e91a0a1 (fix for MatSOR_SeqAIJ_Inode) SOR sweep is performed. If you compare with MatSOR_SeqAIJ, you see that this case is handled by a while loop that is missing in the inode version. I added an equivalent while loop to the inode function.
Jed Brown
Backport 44a71c6d4b84 "Ensure that inode.ibdiagvalid gets set when diagonal has been invalidated"
Matt Knepley
Vec: Duplicate must preserve block size for local form of ghosted vectors
Michael Lange
Bug fix: Use MPIULong_Send/Recv in MatLoad_MPIBAIJ.
Satish Balay
change freebsd testbox from ver 7 to ver 9. but cmake hangs in configure? workarround it for now.
Satish Balay
fix integer overflow with -mat_view_info
Jed Brown
MatComposite matrices are implicitly preallocated
Satish Balay
Increase patchlevel to 3.3-p5
Jed Brown
merge
Jed Brown
Backport Fortran interface fix: matcreatempisbaij_ -> matcreatesbaij_
Jed Brown
Fix MatGetSubMatrix_MPIBAIJ_Private to use the compressed index sets [petsc-maint #141575] Recovers correct behavior for MatPermute
Satish Balay
fix autoclone of 'BuildSystem' to work with bitbucket repo organization
Satish Balay
download.html [web]: update with bitbucket urls
Satish Balay
ams,ctetgen: update tarball URLs to bitbucket
Matt Knepley
Mat: Doc fix
Matt Knepley
AO: Doc improvement
Jed Brown
Fix nasty bug in PetscTable reallocation The old code decided whether to reallocate up-front, before checking whether the new key was already in the hash, but then assumed that it was not when checking validity of the reallocated hash. The new code defers the check of hash size until after it has been determined that the key is not already in the hash. In the (common) case when the key is already in the hash, the new code is one less con…
Satish Balay
[petsc-maint #141119] MatGetRowMax error for parallel runs
Jed Brown
merge
  1. Prev
  2. Next