Amesos2

• The interface to SuperLU_DIST now also works for the CUDA-enabled variant of the library. https://github.com/trilinos/Trilinos/pull/12524

Framework

• Began using semantic versioning for Trilinos with 15.1.0 release.

Ifpack2

• BlockRelaxation can now generate blocks using a Zoltan2. https://github.com/trilinos/Trilinos/pull/12728

Kokkos & Kokkos Kernels

• Inclusion of version 4.2.1 of Kokkos and Kokkos Kernels https://github.com/trilinos/Trilinos/pull/12707

MueLu

• The reformulated Maxwell solver (RefMaxwell) was generalized to also work for grad-div / Darcy flow problems. https://github.com/trilinos/Trilinos/pull/12142

• In an effort to consolidate the old non-Kokkos code path with the newer Kokkos code path, the following factories were deprecated and should be removed from input decks: NullspaceFactory_kokkos, SaPFactory_kokkos, UncoupledAggregationFactory_kokkos. https://github.com/trilinos/Trilinos/pull/12720 https://github.com/trilinos/Trilinos/pull/12740

Panzer

• MiniEM can now also assemble and solve Darcy problems using first or higher order mixed finite elements. https://github.com/trilinos/Trilinos/pull/12142

PyTrilinos2

• New package that auto-generates Python interfaces for Trilinos packages. Currently, most of Tpetra is exposed. We are planning on adding other packages. https://github.com/trilinos/Trilinos/pull/12332

ROL

• An auto-generated Python interface was added. A standalone Python package can be downloaded from rol.sandia.gov https://github.com/trilinos/Trilinos/pull/12770

Teko

• Block Jacobi and Gauss-Seidel methods allow now to specify preconditioners for the iterative solves of the diagonal blocks. https://github.com/trilinos/Trilinos/pull/12675

Tpetra

• Tpetra will now assume by default that the MPI library is GPU aware, unless automatic detection or the user indicates otherwise. https://github.com/trilinos/Trilinos/pull/12517

• Reject unrecognized TPETRA_* environment variable. Misspelled or removed environment variables are no longer silently ignored. https://github.com/trilinos/Trilinos/pull/12722

• In order to allocate in shared host/device space (i.e. CudaUVMSpace, HIPManagedSpace or SYCLSharedUSMSpace) by default, please use the CMake options KokkosKernels_INST_MEMSPACE_CUDAUVMSPACE=ON Tpetra_ALLOCATE_IN_SHARED_SPACE=ON https://github.com/trilinos/Trilinos/pull/12622