Introduce MPIU_Allreduce() to detect cases where all processes that share an MPI comm do not call the same sequence of MPI_Allreduce()
This is a bug that can be difficult to track down.

Suggested-by

Introduce MPIU_Allreduce() to detect cases where all processes that share an MPI comm do not call the same sequence of MPI_Allreduce()
This is a bug that can be difficult to track down.

Suggested-by: Eric Chamberland <Eric.Chamberland@giref.ulaval.ca>
Suggested-by: Patrick Lacasse <placasse@giref.ulaval.ca>

show more ...

Merge branch 'master' into stefano_zampini/feature-pcbddc-saddlepoint

Conflicts:
src/ksp/pc/impls/bddc/bddcprivate.c

Merge branch 'master' into tisaac/dmforest

Get 3.6.2 patches

* master: (1360 commits)
remove nasty memory leak with CUDA code when using block Jacobi with multiple blocks per process
print abs

Merge branch 'master' into tisaac/dmforest

Get 3.6.2 patches

* master: (1360 commits)
remove nasty memory leak with CUDA code when using block Jacobi with multiple blocks per process
print abs in Fortran test example in order to have same result with real and complex builds
changed outdate mpirun and -np to mpiexec and -n
force sqrt to be taken in the precision of PetscReal in fortran examples
continuation line too soon for some compilers
stop configure if user selected C++ clanguage, single precision, complex combination since we do not support it
fixes for Fortran examples to compile cleanly in single precision that do not involve complex numbers
fixes to Fortran examples to compile cleanly in single precision complex
fix previous mistake
must cast 1. to PetscScalar before dividing by a PetscScalar (C++ complex class is terrible)
must cast 0.0 to PetscScalar before comparison with a PetscScalar
changed order of suggested locations of downloadable PETSc to Barry's preference
Increase patchlevel to 3.6.2
remove debug print statement from two commits before
remove debug print statements I accidently left in the last commit
PetscDraw: Fix dummy definition of PetscDrawCollective{Begin|End}()
remove cleanupMumps flag from MatDestroy_Mumps() it should always free the memory it allocated Bug detected while adding features to libMesh
improved viewing of fractional counts in xml nested log viewing fixed bug in output of doubles to xml nested log files
chombo: build only libs - not examples
fix typo. 'gcc-4.8 --with-cxx-dialect=C++11' gives a warning with this line - even though its in a #ifdef thats never enabled.
...

show more ...

Merge branch 'master' into knepley/feature-label-empty-stratum

* master: (151 commits)
moved PCTELESCOPE comment into users manual since no one will see the comment
message in makefile about hyp

Merge branch 'master' into knepley/feature-label-empty-stratum

* master: (151 commits)
moved PCTELESCOPE comment into users manual since no one will see the comment
message in makefile about hypre bugs causing test crashes
remove use of some unitialized variables, remove use of "dummy" entries in Vec and IS since Dave May thought they could not be of size zero. Now three telescope tests give no valgrind errors
code sent unitialized data via MPI and allocated too large space for creating indices
Force generateetags.py to use /usr/local/bin/ctags before using ctags because the ctags provided with Apple XCode gets confused by function usage in .tex files and does not make proper tags for TSSolve() and many many others
Optimized version of snes/examples/tutorials/ex1f.F produced incorrect results when lx_v() and lf_v() were declared as of dimension 1 with optimzed builds only. Cannot explain exactly why this happens but since we are using VecGetArray() from Fortran we are gambling with the compiler regardless
make the PetscRandom object used in GAMG to be a public object that can be controlled via the options database. This allows the same example to produce the same results with real and complex numbers by using the option -pc_gamg_random_no_imaginary_part
added parallel tester for examples. Crude but it reduces a basic alltests on my laptop from 25 minutes to 7 1/2 minutes.
remove unneccessary if
fix the problem for ex9 which involves multiple TSSolve calls
shortened the time of several test examples
decrease dramatically the time used by several test examples that take more than a minute on may machine with debug mode
Some of the vectors created in MatNullSpaceCreateRigidBody() were not normalized to 1
fix strange PCTelescope struct name of P which caused duplicate symbol error on some systems cleanup use of PetscMalloc and PetscFree for telescope for modern PETSc standards replace semiredundat to telescope in makefile that runs test examples
SNESUpdateCheckJacobian: make relative norm output syntactically correct
DMCompositeGetGlobalISs: document Fortran interface
fix warnings on bsd. Also fix usage of PetscFunctionBegin/Return
Making an even bigger mess, I cannot explain why this is not in the branch
hypre: update with "Bug fix for Parasails; src/ksp/ksp/examples/tutorials/ex10.c runex10_16"
decrease duration of ex22 by factor of 10 to decrease testing time
...

show more ...

Merge branch 'jed/mat-assembly-perf' of bitbucket:petsc/petsc

VecAssembly and MatAssembly now use a scalable exchange pattern based on
PetscCommBuildTwoSided. This feature can be controlled with th

Merge branch 'jed/mat-assembly-perf' of bitbucket:petsc/petsc

VecAssembly and MatAssembly now use a scalable exchange pattern based on
PetscCommBuildTwoSided. This feature can be controlled with the options

-vec_assembly_bts 0 or 1 (default 0)
-matstash_bts 0 or 1 (default 0)

The rationale is that the new implementation with scalable data
structures can be slightly slower than the old version at small process
counts. The default here could be changed to depend on the process
count (leading to possibly-confusing scaling performance diagnostics) or
the implementation could learn to take a fast path.

* 'jed/mat-assembly-perf' of bitbucket:petsc/petsc: (49 commits)
MatStash: fix -Wsign-compare by using size_t for loop index when max is also size_t
Sys BuildTwoSided test: fix for non-POD std::complex
MatStash: cast to satisfy non-structural MPI type tag check
mpiuni: fix compile error /sandbox/petsc/petsc.clone-2/arch-linux-uni/lib/libpetsc.so: undefined reference to `MPI_Type_create_resized'
Vec: Silence compiler warning
Vec: fix typo in comment
Sys: fix C89 compiler warning
VecStash BTS: fix block stash InsertMode accounting
VecStash BTS: fix indexing bug counting sends to rank 0
MatStash BTS: fix memory leak on MAT_SUBSET_OFF_PROC_ENTRIES
VecAssemblyEnd_MPI_BTS: fix donotstash code path
VecAssemblyEnd_MPI_BTS: fix C++ conversion to InsertMode
Sys: fix datatypes test using MPI_Type_create_resized
MatStash BTS: work around lack of offsetof() for non-POD (std::complex)
MatAssembly: move check for InsertMode consistency into MatStashScatter impls
MatStash BTS: fix memory leak on reassembly with MAT_SUBSET_OFF_PROC_ENTRIES
MatStash BTS: small simplification to row ownership calculation
MatStash BTS: add MAT_SUBSET_OFF_PROC_ENTRIES, impl with neighbor-only comm
MatStash: initial BTS (BuildTwoSided) implementation
MatStash: add extension point for new BTS implementation
...

show more ...

Merge branch 'matpartitioning-hierach' into pcgasm-increaseoverlap

Got an example ( /src/mat/examples/tests) from matpartitioning-hierach

Conflicts:
src/mat/impls/adj/mpi/mpiadj.c

Merge branch 'mat-increaseoverlap' of ssh://bitbucket.org/Fande-Kong/petsc into pr332/master/Fande-Kong/mat-increaseoverlap-scalable

Merge remote-tracking branch 'petsc/jed/mat-assembly-perf' into matpartitioning-hierach

To make the interface PetscCommBuildTwoSided(MPI_Comm comm,PetscMPIInt count,MPI_Datatype dtype,PetscInt nto,c

Merge remote-tracking branch 'petsc/jed/mat-assembly-perf' into matpartitioning-hierach

To make the interface PetscCommBuildTwoSided(MPI_Comm comm,PetscMPIInt count,MPI_Datatype dtype,PetscInt nto,const PetscMPIInt *toranks,const void *todata,PetscInt *nfrom,PetscMPIInt **fromranks,void *fromdata) compatible with Jed's new changes.
DataType of 'nto' and 'nfrom' is changed to PetscMPIInt.

Conflicts:
include/petscsys.h
src/sys/utils/sorti.c
src/vec/vec/examples/tests/makefile

show more ...

Merge remote-tracking branch 'petsc/jed/mat-assembly-perf' into mat-increaseoverlap

To make the interface PetscCommBuildTwoSided(MPI_Comm comm,PetscMPIInt count,MPI_Datatype dtype,PetscInt nto,const

Merge remote-tracking branch 'petsc/jed/mat-assembly-perf' into mat-increaseoverlap

To make the interface PetscCommBuildTwoSided(MPI_Comm comm,PetscMPIInt count,MPI_Datatype dtype,PetscInt nto,const PetscMPIInt *toranks,const void *todata,PetscInt *nfrom,PetscMPIInt **fromranks,void *fromdata) compatible with Jed's new changes.
DataType of 'nto' and 'nfrom' changed to PetscMPIInt.

Conflicts:
include/petscsys.h
src/sys/utils/sorti.c
src/vec/vec/examples/tests/makefile

show more ...

Merge branch 'master' into knepley/solkx

* master: (6933 commits)
Bib: Added reference
Includegraph: Updated for compatibility with new folder layout.
Add SNESLineSearchReason to fortran inclu

Merge branch 'master' into knepley/solkx

* master: (6933 commits)
Bib: Added reference
Includegraph: Updated for compatibility with new folder layout.
Add SNESLineSearchReason to fortran includes Also truncate names to fit 32 char fortran limit
fix compile warnings /usr/home/balay/petsc.clone-2/src/vec/vec/interface/vector.c:1944: warning: division by zero in '1.0e+0 / 0.' /usr/home/balay/petsc.clone-2/src/ksp/ksp/interface/itfunc.c:508: warning: 'vec_rhs' may be used uninitialized in this function
updated output for new trust region initial size
initial size of the trust region is set as a percent of the norm of the initial guess, it should not be related to the initial norm of the function (which is kind of nuts).
The default bounds for SNESVISetVariableBounds() in the manual page were reversed from correct values
Bib: Added refs
Bib: Update MPICH webpage
Plex: Forgot to close file
removed nonexistent ex46f from makefile test rule
Revert "fixed bad merge into master"
fixed bad merge into master
fix memory leak in mkl_pardiso fix some formatting in mkl_pardiso code
mv runex111 from TESTEXAMPLES_C to TESTEXAMPLES_DATAFILESPATH
Mat+Doc: More Fortran docs
Plex ex3: Run tests with the correct number of field components
Bib: Added TetGen cite
parmetis: make sure 'ldd libparmetis.so' points to the correct libmetis.so that its linked with.
small fix as reported by the nightly test
...

Conflicts:
config/builder.py

show more ...

Merge branch 'master' into shri/ts-is-for-differential-variables

Conflicts:
include/petsc-private/tsimpl.h
src/ts/impls/eimex/eimex.c
src/ts/interface/ts.c

Merge branch 'master' into jed/chem
Note merging a branch with good stuff in it for a year and a half is insane!

Conflicts:
src/ts/examples/tutorials/makefile
src/ts/interface/ts.c

BuildTwoSided: validate toranks[]

Merge branch 'jed/assembly-perf' into jed/mat-assembly-perf

Improve VecAssembly performance using BuildTwoSided

* jed/assembly-perf: (27 commits)
Vec: document when InsertMode cannot change betwe

Merge branch 'jed/assembly-perf' into jed/mat-assembly-perf

Improve VecAssembly performance using BuildTwoSided

* jed/assembly-perf: (27 commits)
Vec: document when InsertMode cannot change between reassemblies
VecAssembly BTS: allow reassembly with proper subsets of comm
BuildTwoSided: distinguish events BuildTwoSided vs BuildTwoSidedF
MPI: check for MPIX_Ibarrier (e.g., BG/Q)
VecAssembly BTS: enable by default, use -vec_assembly_bts 0 to disable
VecAssembly BTS: allow proper subsets with VEC_SUBSET_OFF_PROC_ENTRIES
VecAssembly BTS: VEC_SUBSET_OFF_PROC_ENTRIES elides global synchronization
VecAssembly BTS: elide sending empty scalar/block stashes
VecAssembly BTS: separate pending count for scalar and block stashes
VecAssembly BTS: use PetscCommBuildTwoSidedFReq and MPI_Waitsome
VecAssembly BTS: avoid copying data on receive by using pointer frames
Sys BuildTwoSidedFReq: generalization to return requests
VecAssembly: add more scalable implementation, -vec_assembly_bts
Vec_MPI: remove unused members conflated with VecStash
Sys: add PetscMergeMPIIntArray
Sys PetscMergeIntArrayPair: fix PetscFunctionBegin and comment
Sys BuildTwoSidedF_Ibarrier: less synchronous dynamic sparse data exchange
Sys BuildTwoSidedF: build two-sided network with functions in rendezvous
Sys BuildTwoSided test: extract datatype construction
Sys BuildTwoSided: update test for non-contiguous types and verbose output
...

show more ...

Merge branch 'master' into stefano_zampini/pcbddc-multilevel-improvement

Conflicts:
src/docs/website/documentation/changes/dev.html
src/ksp/ksp/examples/tutorials/ex59.c
src/ksp/pc/impls/bddc/bdd

Merge branch 'master' into stefano_zampini/pcbddc-multilevel-improvement

Conflicts:
src/docs/website/documentation/changes/dev.html
src/ksp/ksp/examples/tutorials/ex59.c
src/ksp/pc/impls/bddc/bddcprivate.c

show more ...

Merge branch 'master' into shri/ts-events

Conflicts:
src/ts/interface/ts.c

Merge branch 'master' into shri/feature-dm-circuit

Conflicts:
src/dm/impls/makefile

Merge branch 'master' into prbrune/snes-fascontinuation

Conflicts:
include/petscsnesfas.h

Merge branch 'master' into barry/wirth-fusion-materials
bad manual merge needs to be fixed
this is what happens when branches hang around too long

Conflicts:
src/ts/examples/tutorials/advection-dif

Merge branch 'master' into barry/wirth-fusion-materials
bad manual merge needs to be fixed
this is what happens when branches hang around too long

Conflicts:
src/ts/examples/tutorials/advection-diffusion-reaction/ex10.c

show more ...

Merge branch 'master' into knepley/feature-dmda-section

* master: (319 commits)
SNESGS: Fix for uninitialized variable warning in gssecant.c
DMPlex: Fix for orientation in Gmsh files - Fixed tes

Merge branch 'master' into knepley/feature-dmda-section

* master: (319 commits)
SNESGS: Fix for uninitialized variable warning in gssecant.c
DMPlex: Fix for orientation in Gmsh files - Fixed test output
do not use C99 struct initialization code in PETSc
Nightly Summary: Added time taken to each of the tests in the dashboard.
ASCII printing of floating point numbers must cast to (double) to support quad precision
configure: check fortranDatatypes only when fortran is enabled [this fixes configure crash with --with-fc=0]
don't use non-portable and also dependent on floating point precision system level math constants
picky compilers
fix a couple compiler warnings/errors (depending on compiler) for Matlab engine code
DMPlex: More fixup from bad rebase, moved to DMGet/SetCoordinateSection()
DM: Added DMGet/SetCoordinateSection()
Fuckup: Fix for merge that is still untangling bad rebase from Oct. 17th 2013 - Fixed const in declarations - Fixed merge from PetscFE - Fixed 2 mallocs
replace all echo in test/tutorial makefiles with printf since echo behaves differently on different systems
DMDA: Missing header - Damn rebase
DMPlex ex3: Added 2D Q_1 and 3D P_2 tests
DMDA: Added DMDAProjectFunction() and DMDAComputeL2Diff()
DMDA: Fixed 2D geometry - Should reuse DMPlex routines
DMDA: Fix bug in 2D closure operation for cells
DMDA: Make closure operations also return the size
DMDA: Changed DMDACreateSection() to emulate DMPlexCreateSection()
...

Conflicts:
config/builder.py
include/petscdmda.h
src/dm/dt/interface/dtfe.c
src/dm/impls/da/dageometry.c
src/dm/impls/da/dalocal.c
src/dm/impls/plex/plexgeometry.c

show more ...

Merged petsc/petsc into master

Merge branch 'master' into knepley/feature-plex-gmsh

* master: (163 commits)
do not use C99 struct initialization code in PETSc
Nightly Summary: Added time taken to each of the tests in the dash

Merge branch 'master' into knepley/feature-plex-gmsh

* master: (163 commits)
do not use C99 struct initialization code in PETSc
Nightly Summary: Added time taken to each of the tests in the dashboard.
ASCII printing of floating point numbers must cast to (double) to support quad precision
configure: check fortranDatatypes only when fortran is enabled [this fixes configure crash with --with-fc=0]
don't use non-portable and also dependent on floating point precision system level math constants
picky compilers
fix a couple compiler warnings/errors (depending on compiler) for Matlab engine code
DMPlex: More fixup from bad rebase, moved to DMGet/SetCoordinateSection()
DM: Added DMGet/SetCoordinateSection()
Fuckup: Fix for merge that is still untangling bad rebase from Oct. 17th 2013 - Fixed const in declarations - Fixed merge from PetscFE - Fixed 2 mallocs
replace all echo in test/tutorial makefiles with printf since echo behaves differently on different systems
DMDA: Missing header - Damn rebase
DMPlex ex3: Added 2D Q_1 and 3D P_2 tests
DMDA: Added DMDAProjectFunction() and DMDAComputeL2Diff()
DMDA: Fixed 2D geometry - Should reuse DMPlex routines
DMDA: Fix bug in 2D closure operation for cells
DMDA: Make closure operations also return the size
DMDA: Changed DMDACreateSection() to emulate DMPlexCreateSection()
DMDA: Added functions which emulate DMPlex functionality
DMDA: Now DMDAGetNumCells() returns the cells in each direction as well
...

Conflicts:
config/builder.py

show more ...

Merge branch 'madams/gamg-destroy' into prbrune/pcgamg-classicalinterpolationstrategies

Merge branch 'master' into madams/gamg-destroy

Conflicts:
src/ksp/pc/impls/gamg/gamg.c

Merge branch 'master' into barry/reduce-dmsetup-da-memoryusage

Conflicts:
src/dm/examples/tests/ex15.c
src/dm/examples/tutorials/ex3.c
src/dm/impls/da/da2.c
src/dm/impls/da/da3.c