v0.96.22

Oceananigans v0.96.22

Diff since v0.96.21

Merged pull requests:

• Add a performance benchmark pipeline (#4375) (@simone-silvestri)
• Improve z-star tracer conservation tests (#4403) (@NoraLoose)
• Fused multiply add (fma) in stencil operations (#4407) (@simone-silvestri)
• Remove stretched advection (#4411) (@simone-silvestri)
• Don't invoke comm buffers on non-distributed grids and move to DistributedComputations (#4414) (@glwagner)
• Fix summary for PartialCellBottom (#4425) (@tomchor)
• Add CATKE and GPU dycore paper references (#4426) (@navidcy)
• Update CATKE docstring (#4427) (@navidcy)
• Remove trailing spaces part 2 (#4429) (@navidcy)
• Remove trailing white spaces part 3 (#4430) (@tomchor)
• Add operators functions for reciprocals of metrics (#4431) (@simone-silvestri)
• Add newton_div utility for fast (but approximate) mixed-precision computation of divisions (#4432) (@glwagner)
• Ensure user diffusivity NamedTuple have correct ordering (#4437) (@glwagner)
• Use tuple instead of vector in tilted_bottom_boundary_layer.jl (#4439) (@simone-silvestri)
• Compute the difference between the latest main and the PR (#4440) (@simone-silvestri)
• (0.96.22) Make NetCDFWriter work with GPU grids with stretched dimensions (#4441) (@tomchor)
• Improve benchmarks yml (#4442) (@simone-silvestri)
• Making condition size mismatch error a bit more helpful (#4443) (@taimoorsohail)
• Add Souza et al. 2025 + reorder refs chronologically (#4444) (@navidcy)
• CompatHelper: add new compat entry for MuladdMacro at version 0.2, (keep existing compat) (#4445) (@github-actions[bot])
• Use vertical spacing reciprocal in implicit solver (#4447) (@simone-silvestri)
• Adding the possibility of having dynamical particles (#4448) (@simone-silvestri)

Closed issues:

• Tracer conservation is broken in ZStar with Julia 1.11.4 (#4340)
• Improve and Verify Z-Star Tracer Conservation Tests (#4395)
• NetCDFWriter scalar indexing error on GPU (#4397)
• Tilted BBL example not working on GPU (#4438)