Scientific Computing in Rust Monthly #21
Welcome to this month's issue of Scientific Computing in Rust Monthly.
Scientific Computing in Rust Monthly can be viewed online at scientificcomputing.rs/monthly, you can sign up to receive an email copy at mailinglist.scientificcomputing.rs, or you can subscribe using the RSS feed at scientificcomputing.rs/monthly/rss.xml.
Items can be added to the next issue of the newsletter by opening a pull request to the Scientific Computing in Rust Monthly repository on GitHub. You can find more information about adding items here.
Crate of the month: CubeCL
CubeCL is a Rust language extension, a Just-in-Time compiler, and a set of runtimes for writing high-performance compute kernels.
Functions defined using #[cube] can be compiled on demand to CUDA, HIP, Metal, SPIR-V, WGSL, or CPU SIMD.
We recommend starting by looking at the examples in the CubeCL GitHub repository.
New releases
AXbit v0.1.0
AXbit is an experimental visual astrodynamics lab for SPICE/SBDB-based Solar System exploration, small-body trajectory propagation, and close-approach analysis and visualization. It is built with wgpu.
It combines NASA/JPL SPICE ephemerides and SBDB small-body data with numerical propagation and interactive 3D visualization. The first packaged release currently targets Apple Silicon Macs.
The project is still at an early stage: the feature set is limited, the visuals are not so stunning, and substantial work remains on numerical validation and accuracy.
AXbit is looking for contributors interested in numerical validation, astrodynamics and reference frames, low-level graphics and visualization, and the development of useful analysis tools.
delaunay v0.8.1
delaunay provides D-dimensional Delaunay triangulations and convex hulls with exact predicates, deterministic degeneracy handling, and explicit topology validation. Highlights of v0.8.1 include:
- Verified scientific checkpoints with SHA-256 digests, exact coordinate bits, topology and proof metadata, complete f-vectors, and Euler characteristics. Restoration rejects malformed or tampered evidence and revalidates the result through all five validation levels.
- A clearer proof-oriented API: Triangulation owns Levels 1–4, through valid geometric realization, while DelaunayTriangulation adds Level 5 Delaunay certification. Failure-atomic builders govern transitions between these layers.
- Faster validation: certified Euclidean point sets gain a linear-in-simplices Level 5 reporting path with exhaustive fallback diagnostics. Optimized Level 4 realization checks are independently cross-validated in 2D–5D.
- Improved diagnostics through named public error fields and more specific typed failures.
This release includes breaking API changes and raises the MSRV to Rust 1.98. See the full release notes.
et-rs et-k-rs et-abi v0.2.0
Pure-Rust tooling for the Esperanto ET-SoC-1 RISC-V accelerator: a host driver, a device-side kernel library, and the ABI they share.
| Crate | Library | Side | Role |
|-------|---------|------|------|
| et-abi/ | et_abi | shared (no_std) | Launch-argument structs plus shared constants (cache line, geometry), defined once so host and device cannot drift. |
| et-rs/ | et_soc1 | host (std) | Loads kernels, launches them on shires, moves results over DMA, decodes trace buffers. Pure Rust, no vendor C++ runtime. |
| et-k-rs/ | et_kernel | device (no_std) | Library for writing ET-SoC-1 compute kernels in Rust, plus demo kernels. |
et-rs and et-abi are host crates and form the workspace; default-members
makes bare cargo commands operate on et-rs. et-k-rs cross-compiles to
riscv64imac with its own target configuration, so it is excluded from the host
workspace and built from its own directory. Both et-rs and et-k-rs depend on
et-abi; neither depends on the other.
There is also an example/tutorial on GitHub to get you started: github.com/mathsDOTearth/test-et-rs
la-stack 0.4.5
la-stack provides fast, stack-allocated linear algebra for fixed dimensions in Rust, with no runtime dependencies by default and optional exact arithmetic. v0.4.5 focuses on numerical correctness and reproducible performance evidence:
- Determinant overflow is now surfaced even when the corresponding error-bound terms underflow.
- Checked determinant queries for D≥5 aggregate range proofs before a single scaled replay, reducing overhead while preserving existing error behavior.
- Performance reports are now retained as deterministic CSV with schema-versioned provenance, allowing published benchmarks to be validated and re-rendered from the measured artifacts.
- The MSRV is now Rust 1.98.0. The new algebraic floating-point operations are intentionally forbidden in repository-owned numerical code to preserve documented evaluation order, error bounds, exact fallbacks, and reproducibility.
The full changes are available in the v0.4.5 release notes.
oximo v0.6.0
oximo is an algebraic modeling library for solving optimization problems in Rust. This release expands its model interchange, solver result reporting, and solving stack with readers for LP, MPS, and NL files, automatic convex solver routing in POUNCE, more cross-backend result metadata, and Enzyme automatic differentiation support for WebAssembly.
PhastFT v0.4
PhastFT is a performant Fast Fourier Transform implementation in pure Rust with no unsafe code.
Full details of the latest release can be found in this Reddit post.
Events
RustConf, 8-11 September 2026, Montreal, Canada & Online
RustConf is turning 10! The original Rust programming language conference is taking place September 8–11 in Montréal, Canada, with both in-person and virtual attendance options.
The schedule is packed with technical talks, hands-on workshops, project updates, a Career Crawl, an UnConference, and lots of community discussions covering the latest developments across the Rust ecosystem.
Join us to see what the Rust community is building, dig into the topics defining the next decade of computing, and discover ideas you can bring back to your own projects.
Publications
Sergey "Shnatsel" Davidoff, Safe SIMD in Rust, even on the inside
M. S. Drehwald, M. Dominguez, K. Sala, A. Aspuru-Guzik, J. Doefert, GPU Offload in Rust: Portable, Safe, and Fast
R. Trappitsch, InstrumentRs: #[derive(Parameter)]
This is the fourth blog post in the instrumentRs series on designing a toolkit to write drivers for (scientific) instruments in Rust. This post discusses the implementation ans usage of a derive macro that makes makes instrument control with new types simple.
Help wanted
oximo
The project is looking for contributors interested in mathematical programming. There are good first issues available for contributors who would like to get started.
