Complain if PETSc is built with the wrong size index type

... at least until we can fix this properly

more PETSc pruning

Merge branch 'master' of github.com:PHASTA/phasta

bringing PETSc and svLS back to the master

make the residuals echoed in the log of PETSc and Native solver the same and switch petsc b vector construction to ADD_VALUES (let them do the global completion) likely fixed scalar solve bug as well

removed lots of unnecessary PETSc interface and lots of commented out stuff too

exploring pruning PETSc and comparing residuals more precisely

updated phasta.cc for petsc master (may brake older versions). Also added a dump of index sets for the rank specified in rankdump (static to file).

some compilers are more picky about f90 than others

PETSc: remove PETSC_VIEWER_ASCII_INDEX format; fixes -log_summary

PETSc-3.6.x's PetscLogView does not understand PETSC_VIEWER_ASCII_INDEX,
thus skipping the printing of -log_summary. Ken says
PETSC

PETSc: remove PETSC_VIEWER_ASCII_INDEX format; fixes -log_summary

PETSc-3.6.x's PetscLogView does not understand PETSC_VIEWER_ASCII_INDEX,
thus skipping the printing of -log_summary. Ken says
PETSC_VIEWER_ASCII_INDEX is not important for PHASTA at this time. I
will fix PetscLogView for PETSc-3.7 to handle this case.

show more ...

Broke IC leslib solve a few commits back but this fixes it. Also suspect svLS solve may still have issues.

a few more valgrind cleanups for the compressible code. This version runs duct cases with no errors

various cleanup's inspired by ASAN address and undefined in pursuit of the source of the PETSc failures on large cases

Should now compile for individual or combined IC solvers. Note flag changes for CMake. Some cleanup of old native GMRes solver for compressible flow

Extracts and prints communication info from ilwork array

make svLS disableable

Merge branch 'topic_remove_mpoint' into svls

build svLS

Kitware says HAVE_CXX is depreciated
(and it breaks the build on intel/pgi)

intermediate forces.dat file was missing the last step until this move of call to forces.dat

improved d2wall computation and export of petsc options to stdout for improved provenance of petsc command line and .petsrc inputs

driver to read ilwork from a single user specified posix file

fixed a missed merge of compressible set initial condition from file and added PETSc using pkg config and ISJED_BROWN option for MatAssembly improvement

remove unused "mpoint" memory allocator

allocate memory for linear solvers directly in itrdrv