Merge branch 'jczhang/feature-cusparse-ilu0' into 'main'

Support symbolic, numeric ILU0/ICC0 with natural ordering on device

See merge request petsc/petsc!5364

MATAIJCUSPARSE: support symbolic and numeric ILU0 and ICC0 with natural ordering on device

MATAIJCUSPARSE: function renaming to mimic cusparse APIs

Merge branch 'jczhang/fix-mpiaijcusparse-handle-stream' into 'main'

Mat_MPIAIJCUSPARSE does not need to have its own stream or cuSparse handle

See merge request petsc/petsc!5020

Mat: Mat_MPIAIJCUSPARSE does not need to have its own cuSparse handle or cuda stream

As Mat_SeqAIJCUSPARSE already has them. Otherwise, if we create a cuSparse handle in Mat_MPIAIJCUSPARSE,
and set

Mat: Mat_MPIAIJCUSPARSE does not need to have its own cuSparse handle or cuda stream

As Mat_SeqAIJCUSPARSE already has them. Otherwise, if we create a cuSparse handle in Mat_MPIAIJCUSPARSE,
and set it as diagonal A and offdiagonal B's handle, then in MatDestroy_SeqAIJCUSPARSE(), we may destroy
the same handle twice: one for for A, the other for B.

show more ...

Merge branch 'jczhang/feature-improve-mataij-coo' into 'main'

Do COO preallocation on MATAIJ and share the plan with other device matrix types

See merge request petsc/petsc!4831

Mat: support extended COO with mataijcusparse

Merge branch 'jacobf/2022-01-05/c99' into 'main'

Feature: C99 and C++11

See merge request petsc/petsc!4700

deprecate PETSC_STATIC_INLINE

SETERRQ[1-9]+ begone

Merge remote-tracking branch 'origin/release'

Merge branch 'jolivet/fix-hypre-convert' into 'release'

MatConvert_HYPRE_AIJ: fix MAT_REUSE_MATRIX

See merge request petsc/petsc!4824

Tranpose -> transpose

Merge branch 'jed/mat-coo-size_t' into 'main'

MatSetPreallocationCOO: use PetscCount for length argument

See merge request petsc/petsc!4789

Mat CUSPARSE: tidy error-prone PETSC_INTERN redeclaration

MatSetPreallocationCOO: use PetscCount for length argument

This is necessary to support assembly even when the vector size and
number of nonzeros is well below 2B. For example, with a scalar problem

MatSetPreallocationCOO: use PetscCount for length argument

This is necessary to support assembly even when the vector size and
number of nonzeros is well below 2B. For example, with a scalar problem
on Q_1 hexes, you have 8*8=64 entries per element and thus hit 2B around
33M elements. This change allows such problems to be solved without
promoting to 64-bit indices everywhere (a storage and bandwidth cost).

PetscCount is a new signed type equivalent to ptrdiff_t, meant for array
indices and sizes that may overflow int32_t even when solving a problem
whose dimension fits in int32_t.

show more ...

Merge branch 'rmills/cusparse-move-chkerrcusparse' into 'main'

Move CHKERRCUSPARSE() to petscdevice.h

See merge request petsc/petsc!4774

Move CHKERRCUSPARSE() to petscdevice.h.

Moving to be consistent with where the other CUDA error checking macros are located.

Thanks-to: Jacob Faibussowitsch <jacob.fai@gmail.com>

Merge remote-tracking branch 'origin/release'

Merge branch 'rmills/fix-cusparse-false-build-error/release' into 'release'

Change "cuSPARSE error" to "cuSPARSE errorcode".

See merge request petsc/petsc!4764

Change "cuSPARSE error" to "cuSPARSE errorcode".

This is needed to avoid a false positive in the egrep command in the top
level makefile when building with some CUDA versions. Some warnings
about de

Change "cuSPARSE error" to "cuSPARSE errorcode".

This is needed to avoid a false positive in the egrep command in the top
level makefile when building with some CUDA versions. Some warnings
about deprecated cuSPARSE functions print this part of the PETSc source,
and the occurrence of "error " causes incorrect logging of a build
error. Changing to "errorcode" avoids this issue, and is consistent with
terminology used in the cuSPARSE documentation.

Thanks-to: Satish Balay <balay@mcs.anl.gov>

show more ...

Merge branch 'adams/landau-cuda-opt' into 'main'

Landau optimizations + GPU assembly bug fix

See merge request petsc/petsc!4358

add hook for CPU cusparse solves

Merge branch 'stefanozampini/hypre-gpu' into 'main'

Interface HYPRE GPU solvers

See merge request petsc/petsc!4248

MATSEQAIJCUSPARSE: add accessors for device CSR pointers