Merge remote-tracking branch 'origin/release'

win: fix c++ build

/cygdrive/e/balay/petsc/lib/petsc/bin/win32fe/win32fe_cl -c -GF -MT -GR -EHsc -Zc:preprocessor -Zc:__cplusplus -GF -MT -GR -EHsc -Zc:preprocessor -Zc:__cplusplus -Z7 -Zm200 -Od

win: fix c++ build

/cygdrive/e/balay/petsc/lib/petsc/bin/win32fe/win32fe_cl -c -GF -MT -GR -EHsc -Zc:preprocessor -Zc:__cplusplus -GF -MT -GR -EHsc -Zc:preprocessor -Zc:__cplusplus -Z7 -Zm200 -Od -std:c++20 -TP -I/cygdrive/e/balay/petsc/include -I/cygdrive/e/balay/petsc/arch-ci-mswin-test1/include -I/cygdrive/c/PROGRA~2/MICROS~3/MPI/Include -I/cygdrive/c/PROGRA~2/MICROS~3/MPI/Include/x64 /cygdrive/e/balay/petsc/src/vec/vec/interface/vector.c -o arch-ci-mswin-test1/obj/src/vec/vec/interface/vector.o
vector.c
E:\balay\petsc\src\vec\vec\interface\vector.c(2066): error C2124: divide or mod by zero

show more ...

Kokkos: return execution space instead of reference to simplify the code

A Kokkos execution space is a shared pointer. We don't need to reference it

Merge branch 'release'

Merge branch 'tisaac/fix-vecgetlocal-offload' into 'release'

Vec: Avoid device sync on VecGetLocalVector()

See merge request petsc/petsc!8161

Fix bug in computation of Frobenius norm of vectors in parallel.

Reported-by: Venkata Narayana Sarma Dhavala <narayanasarma.ed@gmail.com>

It was discovered by using CIVL, a code correctness verific

Fix bug in computation of Frobenius norm of vectors in parallel.

Reported-by: Venkata Narayana Sarma Dhavala <narayanasarma.ed@gmail.com>

It was discovered by using CIVL, a code correctness verification tool developed by Stephen Siegel at U. Delaware

show more ...

make: add explicit '% : %.c' target so that LINK.c target is in sync with LINK.F target '% : %.F90'

with multiple source compile into a singe binary:
- fixes Fortran example build: currently broken,

make: add explicit '% : %.c' target so that LINK.c target is in sync with LINK.F target '% : %.F90'

with multiple source compile into a singe binary:
- fixes Fortran example build: currently broken, due to duplicate obj-files in link command
- can now use consitent obj file dependencies irrespective of C-main or Fortran-main sources
- cab avoid duplicate compiles of same sources [esp with SEPARATE_COMPILE_LINK=1]

show more ...

Vec: Avoid device sync on VecGetLocalVector()

VecKokkos: allocate memory using device malloc when host and device share the same memory

This is preferred on AMD MI300A, since "Memory allocations should be made using the hipMalloc function, whic

VecKokkos: allocate memory using device malloc when host and device share the same memory

This is preferred on AMD MI300A, since "Memory allocations should be made using the hipMalloc function, which gives great binding and page size settings automatically."
see https://hpc.llnl.gov/documentation/user-guides/using-el-capitan-systems/introduction-and-quickstart/pro-tips

show more ...

Merge branch 'jczhang/2025-01-21/fix-hypre-cpu' into 'main'

Run Hypre on host with host petsc matrix types even when Hypre is configured with device

See merge request petsc/petsc!8088

Merge branch 'jczhang/2025-01-30/feature-support-AMD-MI300A' into 'main'

Add support of AMD MI300A

Closes #1703

See merge request petsc/petsc!8110

Hypre: run hypre on the CPU when petsc is run on CPU

Kokkos: add support of AMD MI300A

* Use HostMirrorMemorySpace instead of HostSpace to fix compile errors on MI300A

* Replace Kokkos::HostSpace with HostMirrorMemorySpace to fix compile errors on MI

Kokkos: add support of AMD MI300A

* Use HostMirrorMemorySpace instead of HostSpace to fix compile errors on MI300A

* Replace Kokkos::HostSpace with HostMirrorMemorySpace to fix compile errors on MI300A, since the latter is what Kokkos::DualView use for its host view

* Fix a subtle bug in KokkosDualViewSync() w.r.t MI300A. Suppose we want to sync a petsc VecKokkos v on host. On MI300A, the host copy v_h and the device copy v_d share the memory. So in the old code, we used if (v_dual.need_sync_host()) to skip the device to host memory copy. But I should not skip the exec.fence(). As the device might still have kernels writing v_d, we still need to sync the device/stream to make v_d ready for use on CPU (via v_h).

show more ...

Merge branch 'jolivet/partially-revert-8099' into 'main'

Partially revert !8099 for non-blocking collectives and peer-to-peer communications

See merge request petsc/petsc!8103

sf: Add fallthrough comment to KDTreeCreate

This line should fallthrough correctly, but requires a comment at the
end of the line to signal that the fallthrough is intended.

Partially revert !8099 for non-blocking collectives

Merge branch 'jolivet/remove-unneeded-mpicast' into 'main'

Remove unneeded PetscMPIIntCast() for routines using PetscCount

Closes #1661

See merge request petsc/petsc!8099

Remove unneeded PetscMPIIntCast() for routines using PetscCount

Fix #1661

Fix -Wstrict-prototypes

petsc/src/vec/is/utils/kdtree.c:60:51: warning: a function declaration without a prototype is deprecated in all versions of C [-Wstrict-prototypes]
60 | static PetscErrorC

Fix -Wstrict-prototypes

petsc/src/vec/is/utils/kdtree.c:60:51: warning: a function declaration without a prototype is deprecated in all versions of C [-Wstrict-prototypes]
60 | static PetscErrorCode PetscKDTreeRegisterLogEvents()
| ^
| void
1 warning generated.

show more ...

Merge remote-tracking branch 'origin/release'

Improve some manual pages in KSP/SNES

Merge branch 'jrwrigh/kd_tree' into 'main'

Add PetscKDTree

See merge request petsc/petsc!7954

KDStem: Use uint8_t for axis and char for boolean

sf: Use KDTree for PetscSFSetGraphFromCoordinates

IS: Add PetscKDTree

Co-authored-by: Barry Smith <bsmith@mcs.anl.gov>