CUDA test example that requires DATAFILESPATH should only be run with real, double precision when DATAFILESPATH is available
test example requires parmetis and datafilespath so fix makefile to only run in that case
Merge branch 'master' into barry/always-mark-floatsNeeded to merge in master to resolve conflicts that will appear when this branch is merged with master
filter out clang++ message about .c files when compiling combined C and Fortran code
remove warning messages about long * vs long long * for MPI calls in building examples for 64 bit indices
Insure all %g output floating point numbers have a decimal point in them this is to make it easy to detect floating point processing in output for test purposes provided new diff operation that
Insure all %g output floating point numbers have a decimal point in them this is to make it easy to detect floating point processing in output for test purposes provided new diff operation that treats all floating point numbers in output as equal
show more ...
remove streams test from alltests because it clutters the nightly tests
added parallel tester for examples. Crude but it reduces a basic alltests on my laptop from25 minutes to 7 1/2 minutes.
remove last vestiges of threadcomm in testing
do package tests (e.g. superlu_dist) that require DATAFILESPATH only when data files are available and PETSc is built with 32 bit integer doubleFortran example did not free created communicator
handle examples that can only be tested with double precision and 32 bit integersthis reduces the number of error messages for other builds
gnumake: add in make --output-sync=recurse for gnumake version 4.0 or higher
Merge branch 'barry/test-trilinos-superlu'
updated list of PETSc enum objects that bfort will treat as native for generating Fortran stubs
fftw: add to nightly builds- Switch TESTEXAMPLES_FFTW_REAL to TESTEXAMPLES_FFTW- add in testexamples_FFTW_COMPLEX target for the complex tests[and enable FFTW_COMPLEX tests for fftw/complex buil
fftw: add to nightly builds- Switch TESTEXAMPLES_FFTW_REAL to TESTEXAMPLES_FFTW- add in testexamples_FFTW_COMPLEX target for the complex tests[and enable FFTW_COMPLEX tests for fftw/complex build]
Add SNESLineSearchReason to fortran includesAlso truncate names to fit 32 char fortran limitAnd fix typo MATOP_TRANSPOSE_MATM_ULT_NUMER
Add an example that independently uses Trilinos Amesos2 + SuperLU and PETSc + SuperLU to solve completely indendent linear systems in the same executable. This tests that there is no global state for
Add an example that independently uses Trilinos Amesos2 + SuperLU and PETSc + SuperLU to solve completely indendent linear systems in the same executable. This tests that there is no global state for the solver that would interfere between the packages.Also added simple test examples that initialize PETSc (with PETSc initializing MPI) plus Trilinos and Trilinos (with Trilinos initializing MPI) plus PETSc in the same executable.
stripped out all PETSc threadcomm code
updated PETSc directory layout to match standard packaging strategiesinclude/petsc finclude,private,mpiunilib/petsc confbin/petsc*
merge with master to get all the latest great stuff in master
hack example with ackward dependencies between superlu, superlu_dist and mumps installations to work for superlu_dist even if not using superlu
DMPlex: get nonconforming tests into test scripts
Merge branch 'barry/petsc-sa-refactor-rebase'
fix TSTrajectory for Fortranremove illegal declaration of variable within code
removed the PETSC_HAVE_PCBDDC nonsense. You are either part of the PETSc repository or not!Now it is always built and tested
1...<<4142434445464748