Merge branch 'stefano_zampini/GPU-matdensecuda' [PR #1911]

* stefano_zampini/GPU-matdensecuda:
GPU: Initial implementation for SeqDense class on GPUs.

Merge branch 'karlrupp/aij-add-missing-settypes' [PR #1910]

* karlrupp/aij-add-missing-settypes:
Added missing type setting for MatTransposeSymbolic and MatPtAPSymbolic

Mat: get rid of the valid_GPU_matrix flag from the interface

A specific implementation may provide those operations on GPU
The previous interface would have invalidated the data

MatAssemblyEnd_MPIA

Mat: get rid of the valid_GPU_matrix flag from the interface

A specific implementation may provide those operations on GPU
The previous interface would have invalidated the data

MatAssemblyEnd_MPIAIJ now takes care of propagating the correct
flags to the blocks if we set its flag to OFFLOAD_CPU
This is to fix the case of MatFDColoring, which do not use the MatSetValues interface

show more ...

added missing settypes

Merge branch 'gropp/doc-fixes-1' [PR #1838]

Merge branch 'barry/no-batch-script'

No more batch submission script; this is all due to a very smart idea of Lisandro Dalcin

*) removed or fixed all remaining --know-xx options that required a batch submission script
*) removed all de

No more batch submission script; this is all due to a very smart idea of Lisandro Dalcin

*) removed or fixed all remaining --know-xx options that required a batch submission script
*) removed all details of --known-size-- that are no longer needed
*) cleaned up BAIJMKL source code, made it work with the rest of PETSc and removed is
dependence on a batch submission script, added test case for BAIJMKL
*) update the installation guide on batsch submissions and OpenMP
*) print MKL version, print OpenMP usage with mkl_sparse options

Commit-type: bug-fix, portability-fix, style-fix, feature, documentation, example
Thanks-to: Lisandro Dalcin <dalcinl@gmail.com>

$ ./configure --with-batch
===============================================================================
Configuring PETSc to compile on your system
===============================================================================
=============================================================================== ***** WARNING: Cannot determine if MPI_INT64_T works on your system in batch-mode! Assuming it does work. Run with --known-mpi-int64_t=0 if you know it does not work (very unlikely). Run with --known-mpi-int64_t=1 to remove this warning message. warning message ***** =============================================================================== =============================================================================== ***** WARNING: Cannot determine if MPI_C_DOUBLE_COMPLEX works on your system in batch-mode! Assuming it does work. Run with --known-mpi-c-double-complex=0 if you know it does not work (very unlikely). Run with --known-mpi-c-double-complex=1 to remove this warning message. warning message ***** =============================================================================== =============================================================================== ***** WARNING: Cannot verify that MPI is a shared library - in batch-mode! If MPI is a static library but linked into multiple shared libraries that the application uses, sometimes compiles work - but one might get run-time errors. If you know that the MPI library is shared - run with --known-mpi-shared-libraries=1 option to remove this warning message ***** =============================================================================== =============================================================================== ***** WARNING: Cannot determine if BLAS/LAPACK uses 32 bit or 64 bit integers in batch-mode! Assuming 32 bit integers. Run with --known-64-bit-blas-indices if you know they are 64 bit. Run with --known-64-bit-blas-indices=0 to remove this warning message. warning message ***** ===============================================================================

show more ...

Merge branch 'pr1835/al007/reset-prealloc-comm/master' [PR #1835]

* pr1835/al007/reset-prealloc-comm/master:
Change the communicator for MatResetPreallocation_SeqAIJ error, otherwise the user doesn'

Merge branch 'pr1835/al007/reset-prealloc-comm/master' [PR #1835]

* pr1835/al007/reset-prealloc-comm/master:
Change the communicator for MatResetPreallocation_SeqAIJ error, otherwise the user doesn't get any meaningful error message

show more ...

Merge branch 'maint'

fix MatMissingDiagonal() for seqaij and seqbaij when matrix is non-square

Merge from doc-fixes branch

If local size zero, do nothing

Change the communicator for MatResetPreallocation_SeqAIJ error

Otherwise the user doesn't get any meaningful error message

Fixes for doctext update

Merged in barry/cuda-multigrid-test (pull request #1763)

Various improvements for GPUs (mostly for performance and CUDA)

Merged in barry/optimize-aij-da (pull request #1762)

Non-numeric optimizations focused on AIJ, MatFDColoring, and DMCreateMatrix_DA_*AIJ, looking to improve performance in GPU environments

Non-numeric optimizations focused on AIJ, MatFDColoring, and DMCreateMatrix_DA_*AIJ, looking to improve performance in GPU environments

1) PetscCalloc*() now uses system calloc()
2) Merged some Pets

Non-numeric optimizations focused on AIJ, MatFDColoring, and DMCreateMatrix_DA_*AIJ, looking to improve performance in GPU environments

1) PetscCalloc*() now uses system calloc()
2) Merged some PetscMalloc*()
3) Eliminated unneeded PetscCalloc*()
4) Removed some memory allocations and copies in MatFDColoringSetUp(), added local variables for better compiler optimization
5) Added MatSetValues_SeqAIJ_SortedFull(), added MatSetOption(MAT_SORTED_FULL)
6) Optimized DMCreateMatrix_DA_*AIJ for nonperiodic case to automatically have sorted columns (faster MatSetValues() times)
7) Eliminated call to PetscMemzero() in PetscFree()

Commit-type: style-fix, feature

show more ...

Various improvements for GPUs (mostly for performance and CUDA)

1) Add VecPinToCPU() for CUDA vector and matrices
2) Move initialization of cuBLAS to PetscInitialize() since it takes 1/2 second and

Various improvements for GPUs (mostly for performance and CUDA)

1) Add VecPinToCPU() for CUDA vector and matrices
2) Move initialization of cuBLAS to PetscInitialize() since it takes 1/2 second and distorts timing with -log_view
3) Add logging for DMCreateMatrix (for large meshes this is very large)
4) Add VecGet/RestoreArrayWrite() to prevent unneeded copies from GPU (only implemented so far for CUDA);
added a small number of usages in the source so that snes tutorials ex19 does not do unneeded communication from the GPU
5) Automatically convert MAIJ matrices to AIJ for CUDA since they are not yet supported natively in PETSc's CUDA matrix implementation
6) Pinned objects should still use the CUDA/ViennaCL versions of Destroy to clean up the GPU stuff

Commit-type: feature

show more ...

Merge remote-tracking branch 'origin/barry/add-petscarrayxxx-ops' [PR #1771]

Add PetscArraycmp(), PetscArraycpy(), PetscArrayzero(), PetscArraymove(), update code to use new forms

Replace a couple multi-mallocs with a single malloc
Replace backwards loops in MatSetValues_*()

Add PetscArraycmp(), PetscArraycpy(), PetscArrayzero(), PetscArraymove(), update code to use new forms

Replace a couple multi-mallocs with a single malloc
Replace backwards loops in MatSetValues_*() with PetscArraymove()

Commit-type: style-fix, feature

show more ...

Merge remote-tracking branch 'origin/maint'

Merged in rmills/fix-doc-paradigm-spelling/maint (pull request #1765)

Fixed misspelling of "paradigm" throughout manual pages.

Fixed misspelling of "paradigm" throughout manual pages.

Merged in barry/update-collective-on (pull request #1744)

Update the use of Collective on in the manual pages to reflect the new style

Update the use of Collective on in the manual pages to reflect the new style

Commit-type: style-fix, documentation
Thanks-to: Patrick Sanan <patrick.sanan@gmail.com>