Add SF implemented in Kokkos

Put more cuda related stuff to sfcuda.cu

Rename sfpackcuda.cu to sfcuda.cu

Merge branch 'jacobf/7-2020/feature/timsort' into 'master'

Feature: PetscTimSort[WithArray]

See merge request petsc/petsc!3019

PetscTimSort[WithArray]() implements Tim Peters hybrid sort algorithm
(https://bugs.python.org/file4451/timsort.txt) which takes advantage of already ordered
consecutive elements in data to sort "rea

PetscTimSort[WithArray]() implements Tim Peters hybrid sort algorithm
(https://bugs.python.org/file4451/timsort.txt) which takes advantage of already ordered
consecutive elements in data to sort "real world data" faster.

show more ...

Fix several typos in error strings

define VecCreate{Seq|MPI}ViennaCLWithArrays

define VecCreate{Seq|MPI}CUDAWithArrays

bugfix for VecGetSubVector() when local n = 0 in a process
Reported-by: Aidan Hamilton <aidan@udel.edu>

Fix ISRestoreIndicesF90() for zero length array.

https://gitlab.com/petsc/petsc/-/issues/607

Commit-type: bug-fix
/spend 30m
Reported-by: Fabian Jakub

Support VecLoad() for HDF5 when the vector sizes have not been set before the call to VecLoad()

Fixes the problem that the vector CPU array make have been allocated with cudaMallocHost() but when Ve

Support VecLoad() for HDF5 when the vector sizes have not been set before the call to VecLoad()

Fixes the problem that the vector CPU array make have been allocated with cudaMallocHost() but when VecReplace() is called
it tried to free the memory with PetscFree()

Reported-by: Sajid Ali <sajidsyed2021@u.northwestern.edu>
/spend 2.5h

show more ...

doc: fix build errors

Turn on checkbadSource test to generate an error when found; fix all source code that causes errors

Commit-type: portability-fix, testing-fix, style-fix, feature, maintainability
/spend 1.5h

Vec: Replace 0 -> NULL for pointers

Add support of CUDA/HIP initialization at the same time

adds missing interfaces to VECVIENNACL

Added routines:
- VecCreate{Seq|MPI}ViennaCLWithArray
- VecViennaCL{Place|Reset}Array
- VecBindToCPU_MPIViennaCL

Tests added:
- ex57.cxx

ex57: requires PETSC

adds missing interfaces to VECVIENNACL

Added routines:
- VecCreate{Seq|MPI}ViennaCLWithArray
- VecViennaCL{Place|Reset}Array
- VecBindToCPU_MPIViennaCL

Tests added:
- ex57.cxx

ex57: requires PETSC_HAVE_VIENNACL_NO_CUDA for compile (and cxx)

show more ...

introduce Vec_ViennaCL.GPUarray_allocated

Merge remote-tracking branch 'origin/maint'

cudavecimpl.h move to include/petsc/private as its used by private/sfimpl.h

Reported-by: Nidish <nb25@rice.edu>

Fix offload flag in VecCreateSeqCUDAWithArray

address review comments:

- get the rvalues of cl_* objects
- cast cl_* objects to uintptr

Merge remote-tracking branch 'origin/maint'

fix bug with using VecLoad for subvector

modified: src/vec/vec/interface/rvector.c
new file: src/vec/vec/tests/ex51.c
modified: src/vec/vec/tests/makefile
new file: src/vec/vec/tests/out

fix bug with using VecLoad for subvector

modified: src/vec/vec/interface/rvector.c
new file: src/vec/vec/tests/ex51.c
modified: src/vec/vec/tests/makefile
new file: src/vec/vec/tests/output/ex51_1.out
new file: src/vec/vec/tests/output/ex51_2.out

show more ...

Do lazy petsc cuda initialization and validation

So that we can avoid overburdening the GPU in CI.
See more at https://gitlab.com/petsc/petsc/-/issues/360#note_373723821

Merge remote-tracking branch 'origin/maint'