Repository navigation
What's Changed
- MLEBNodeFDLaplacian: coarsen without the EB factory by @WeiqunZhang in #5970
- ParmParse: keep text after a closing parenthesis in the same token by @WeiqunZhang in #5984
- SUNDIALS MRI: set fixed fast step in advance() by @WeiqunZhang in #5986
- EB least squares: align grad_eb 2D/3D stencil selection by @WeiqunZhang in #5987
- EB 2D: fix centroid-to-face-centroid interpolation on the -1 side by @WeiqunZhang in #5989
- Use -fno-math-errno in non-debug builds by @WeiqunZhang in #5990
- EB: fix face-centered data build with odd eb2.max_grid_size by @WeiqunZhang in #5991
- HIP: use default OpenMP runtime; no RDC by default with OpenMP by @WeiqunZhang in #5994
- CMake: make AMReX_TP_PROFILE take effect by @WeiqunZhang in #5995
- MLMG: AlgMG bottom solver and multigrid types by @WeiqunZhang in #5973
- FFT Poisson test: no FP traps by default in optimized clang builds by @WeiqunZhang in #6004
- HypreMLABecLap: zero the device fine-to-coarse values before accumulating by @drummerdoc in #6006
- EB 2D: avoid covered cells in centroid-to-face-centroid stencil by @WeiqunZhang in #5997
- MLMG: support maxorder = 4 in AlgMG, hypre IJ and PETSc by @WeiqunZhang in #6003
- AlgMG: faster CPU setup by @WeiqunZhang in #6002
- Fortran multifab by @Andris-R in #5985
- EB2: classify 2D edges by level-set sign for edge centroids by @WeiqunZhang in #5993
- AlgMG test: fewer GPU variations, run serially by @WeiqunZhang in #6005
- Conduit Blueprint: include runtime SoA particle components by @jmusser304 in #5656
- Arena, SpGEMM: recover from GPU out-of-memory by @WeiqunZhang in #6000
- Add optional GPU_MIN_BLOCKS to cap registers via launch_bounds by @ju1s04 in #5980
- MLEBNodeFDLaplacian: skip GPU bottom solver in 1D by @ax3l in #6009
- MLEBABecLap::Fapply: classify tiles on a grown box for phi on centroids by @WeiqunZhang in #5999
- AlgMG: deterministic PMIS tie-breaker by @WeiqunZhang in #6008
- Add ParallelForStrided; use it in GPU Gauss-Seidel smoothers by @WeiqunZhang in #5996
- Fix nodal GPU smoother issues by @WeiqunZhang in #5988
- Handle MSVC Debug CRT Reports: No Dialog Boxes by @ax3l in #6013
- CArena: Thread-Safe Usage Counters by @ax3l in #6011
- MLEBNodeFDLaplacian: converge on stretched cells by @WeiqunZhang in #6012
- Bump anchore/sbom-action from 0.24.2 to 0.24.3 by @dependabot[bot] in #6019
- TinyProfiler: Thread-Safe
CurrentName()by @ax3l in #6010 - Morton: clamp coordinates so round-off at xmax cannot wrap to zero by @WeiqunZhang in #6025
- ParallelForStrided: Fix unused variable warning by @WeiqunZhang in #6022
- MLTerrainPoisson: cell-centered terrain-following Poisson operator by @WeiqunZhang in #6030
- PETSc: check return codes of PETSc calls by @WeiqunZhang in #6029
- SUNDIALS: grow per-thread storage on re-initialization by @WeiqunZhang in #6028
- MPMD: do not free MPI_COMM_NULL in Finalize by @WeiqunZhang in #6026
- ABecLaplacian_C test: use a single-precision tolerance in all solves by @WeiqunZhang in #6024
- MLMG: interpAssign zeros ghost cells without valid data by @WeiqunZhang in #6023
- NodeEBFDLap test: tighter tolerance for the bottom-solver comparison by @WeiqunZhang in #6021
- MLEBNodeFDLaplacian: fused one-color Gauss-Seidel launches on GPU by @WeiqunZhang in #6020
- MLALaplacian: fused one-color Gauss-Seidel launches on GPU by @WeiqunZhang in #6017
- MLEBABecLap: launch half the box per color on GPU by @WeiqunZhang in #6016
- MLCurlCurl: launch one color at a time on GPU by @WeiqunZhang in #6015
- CI: HIP 2D with WARN_ERROR by @ax3l in #6032
- Use Warp-aggregated atomics in amrex::DenseBins by @atmyers in #6031
- Parser: free scan buffer and parser when amrex_parser_new aborts by @WeiqunZhang in #6045
- STLtools: fix ASCII STL read message by @WeiqunZhang in #6044
- TimeIntegrator: initialize integrator_type by @WeiqunZhang in #6043
- Fix complete elliptic integral values at k = ±1 by @Robertboy18 in #6033
- Docs: fix stale class name, header path and SUNDIALS table names by @WeiqunZhang in #6052
- SundialsIntegrator: use ARKODE-wide API with SUNDIALS >= 7.1 and fix warnings by @WeiqunZhang in #6051
- Docs: default GPU arena size counts only the processes in one communicator by @WeiqunZhang in #6049
- Add Initialize overload for a plain function pointer callback by @WeiqunZhang in #6048
- ParmParse: document that queryarr/getarr never shrink the vector by @WeiqunZhang in #6042
- PlotFileData::get(level, varname): sync stream before freeing source FAB by @WeiqunZhang in #6047
- Parser: reject statement lists without a final value by @WeiqunZhang in #6046
- Parser: Speed up evaluation by @WeiqunZhang in #6050
- MLEBNodeFDLaplacian: RZ support for AlgMG and hypre by @WeiqunZhang in #6053
- ParticleContainerBase: preserve anisotropic ref ratios in Vector ctor by @atmyers in #6056
- Add GPU-synchronized TinyProfiler timers by @WeiqunZhang in #5669
- Particles: use ParallelContext sub-communicator consistently by @atmyers in #6057
- Rename template parameter I by @WeiqunZhang in #6063
- GMRES: add flexible GMRES option by @WeiqunZhang in #6054
- GPU-fused MLEBABecLap Fapply and Fsmooth by @ankithadas in #5622
- CI: oneAPI 2026.1 image for the Intel PVC job and oneMKL 2026 fixes by @WeiqunZhang in #6066
- MLTerrainPoisson: implement getFluxes by @WeiqunZhang in #6068
- FillPatch: do not share FPinfo across interpolaters with different CoarseBox by @WeiqunZhang in #6027
- Bump actions/upload-artifact from 7.0.1 to 7.0.2 by @dependabot[bot] in #6071
- Bump github/codeql-action/upload-sarif from 4.38.2 to 4.38.3 by @dependabot[bot] in #6074
- Bump github/codeql-action/init from 4.38.2 to 4.38.3 by @dependabot[bot] in #6070
- Bump github/codeql-action/analyze from 4.38.2 to 4.38.3 by @dependabot[bot] in #6072
- Bump github/codeql-action/autobuild from 4.38.2 to 4.38.3 by @dependabot[bot] in #6073
- Bump the ci-requirements group in /.github/workflows/requirements with 3 updates by @dependabot[bot] in #6075
- HypreIJIface: hypre error checks, non-finite residual check, ILUT keys by @drummerdoc in #6081
- EB: Performance optimize the main kernel of MLStateRedistribute by @skang67 in #6078
- SYCL: Replace deprecated nd_item::barrier with sycl::group_barrier by @WeiqunZhang in #6091
- MLEBNodeFDLaplacian, MLCurlCurl: clear m_needs_update in prepareForSolve by @WeiqunZhang in #6096
- AlgMG: gather the direct bottom matrix and rhs with in-place Allgatherv by @WeiqunZhang in #6089
- CI: group GitHub Actions bumps, ignore patch updates, skip GitLab for dependabot by @WeiqunZhang in #6088
- Cell-centered interpolation in 1D cylindrical coordinates by @WeiqunZhang in #6086
- EB: fix face-centered data outside the domain by @WeiqunZhang in #6085
- Select the GPU before MPI_Init by @WeiqunZhang in #6079
- Bump the github-actions group with 5 updates by @dependabot[bot] in #6097
- GitLab CI: skip push pipelines on newly created dependabot branches by @WeiqunZhang in #6098
- EB: Performance optimize MakeStateRedistUtils for SRD by @skang67 in #6094
- Backtrace: prefer llvm-addr2line and batch the lookups by @WeiqunZhang in #6099
- MLEBABecLap: opt-in race-free GPU Gauss-Seidel at cut cells under setDeterministic by @pourion in #6093
- Cross proj by @jbbel in #6100
- hypreSynchronize: really synchronize with hypre >= 2.33; HypreMLABecLap get_solution race by @drummerdoc in #6104
- EB: robust node classification near the surface; EB generation stress test by @WeiqunZhang in #6064
- HypreMLABecLap: IJ assembly mode with hypre or AlgMG backend by @drummerdoc in #6082
- Fix dangling reference in variadic amrex::min/max by @ax3l in #6109
- Clamp returns by value; raise EB small_volfrac default to 1e-12 by @WeiqunZhang in #6110
- CI: shorten slow ctests; fix hypre and GNU Multi-D builds with GCC 15 by @WeiqunZhang in #6111
- Bump certifi from 2026.7.22 to 2026.10.19 in /.github/workflows/requirements by @dependabot[bot] in #6114
- Vectorize the CPU reductions by @WeiqunZhang in #6105
- CI: skip clang-tidy on ccache misses already checked on development by @WeiqunZhang in #6113
- HypreMLABecLap: coarse/fine b coefficients across periodic boundaries by @drummerdoc in #6106
- Add GMG::CrossProj: cross-stencil projection with geometric multigrid by @WeiqunZhang in #6103
- algoim: use the 4D ParallelFor for surface integrals under CUDA by @WeiqunZhang in #6126
- EB: Performance-optimize MakeITracker by @skang67 in #6112
- GPU: free Elixir and AsyncArray memory via Gpu::freeAsync by @WeiqunZhang in #6130
- FillSignedDistance: zero implicit function is body, as in EB2 by @WeiqunZhang in #6115
- MLEBNodeFDLaplacian: only isolated hidden EB features stop coarsening by @WeiqunZhang in #6122
- MLMG: set the AlgMG tolerance from the reduction MLMG still needs by @WeiqunZhang in #6124
- Add Marc as code owner for Hypre by @WeiqunZhang in #6127
- Vectorize the BaseFab CPU reductions by @WeiqunZhang in #6131
- SignedDistance test: copy the data to the host before checking by @WeiqunZhang in #6133
- EB 2D: zero level set is body in the EB centroid of axis-aligned facets by @WeiqunZhang in #6135
- Arena: reuse MPI buffers in CPU builds, and check BArena allocations by @WeiqunZhang in #6138
- Bump the ci-requirements group in /.github/workflows/requirements with 2 updates by @dependabot[bot] in #6134
- Make the CPU reduction accumulator loops vectorize with GCC by @WeiqunZhang in #6132
- HypreMLNodeLap: composite (all AMR levels) nodal solve with hypre or AlgMG by @drummerdoc in #6083
- SYCL: enable the host path of the HOST_DEVICE launch functions and macros by @WeiqunZhang in #6139
- Add ParReducePerComp and ParReducePerBoxComp by @WeiqunZhang in #6137
- Smaller default arena on integrated GPUs by @WeiqunZhang in #6140
- VisMF: compute GPU header min/max with ParReducePerBoxComp by @WeiqunZhang in #6141
- FillPatchSingleLevel: time-interpolate only the patch, not the level by @WeiqunZhang in #6143
- FillBoundary: OpenMP pack/unpack over tags, not neighbor ranks by @WeiqunZhang in #6144
- FillBoundary: drop edge/corner-only messages for cross stencils by @WeiqunZhang in #6145
- FillPatch: keep FPinfo patch layouts' caches alive by @WeiqunZhang in #6147
- FluxRegister: restrict Reflux, CrseInit and CrseAdd to the touched cells by @WeiqunZhang in #6146
- FabArray: avoid host sync in build_arrays() by @BenWibking in #6148
- HypreMLABecLap: faster repeated setups in IJ mode; AlgMG keeps its hierarchy by @drummerdoc in #6142
- ParallelCopy/FillBoundary: each team member writes only its own FABs by @WeiqunZhang in #6151
- Document compile-time options for ParallelFor by @BenWibking in #6129
- FluxRegister::CrseAdd: document that shared faces are summed by @WeiqunZhang in #6152
- TinyProfiler: compare TimerKey names once by @WeiqunZhang in #6153
- Remove Github Team links from GOVERNANCE.md by @atmyers in #6155
- FFT: collective transposes with automatic strategy selection by @WeiqunZhang in #6150
- Update CHANGES.md for 26.11 by @WeiqunZhang in #6161
New Contributors
- @jmusser304 made their first contribution in #5656
- @ju1s04 made their first contribution in #5980
- @Robertboy18 made their first contribution in #6033
- @pourion made their first contribution in #6093
Full Changelog: 26.10...26.11