Scientific Computing in Rust Monthly #19

Welcome to this month's issue of Scientific Computing in Rust Monthly.

This year's Scientific Computing in Rust annual workshop is just over a week away. The schedule for the workshop is available at scientificcomputing.rs/2026/timetable and it's not too late to register at scientificcomputing.rs/2026/register.

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: hyperreal

crates.io ~ docs.rs ~ source hyperreal is a crate that provides exact rational arithmetic, symbolic real values, lazy computable real approximation, and conservative structural facts for the hyper ecosystem. If you'd like to hear more about hyperreal, be sure to attend this year's Scientific Computing in Rust annual workshop.

New releases

cog3pio v0.1.0

The first minor release of cog3pio - a Rust/Python library for reading GeoTIFF raster data - is out! Main highlight is the ability to do GPU-accelerated decoding of TIFF data via bindings to nvTIFF (see nvtiff-sys crate). The tensor data in CUDA memory can then be zero-copied across the Rust/Python boundary through the DLPack interchange protocol (using dlpark crate). Read more about the technical details and background motivation in this three-part blog series.

diffsol v0.14

diffsol is a Rust crate for solving ordinary differential equations (ODEs) and differential-algebraic equations (DAEs). The latest version includes:

• Hybrid ODE support - the classical example is the bouncing ball, but other examples are dosing schedules for pharmacological models or policy-based switching for epidemic models
• diffsol-c crate with dynamic dispatch & C FFI APIs for wrapping diffsol in other languages. pydiffsol is a Python wrapper that provides a nice example of how diffsol-c can be used in practice

ferreus_bbfmm v0.2.0

A black-box fast multipole method crate.

• 1D binary trees, 2D quadtrees, and 3D octrees
• Adaptive and non-adaptive tree structures
• Optimised low-rank M2L interactions using symmetry and compression
• Support for multiple right-hand sides
• User-defined kernels via traits
• Optional simultaneous evaluation of kernel values and gradients, new in this release

ferreus_rbf v0.2.2

A fast global radial basis function interpolator.

• Support for 1D, 2D, and 3D input domains
• Scales to datasets with more than 1,000,000 source points, subject to hardware
• Optional global trend transforms for capturing large-scale structure in the data
• Fast 3D isosurface extraction using a surface-following regularised marching tetrahedra method
• Optional simultaneous evaluation of RBF values and gradients
• Built on faer for linear algebra, avoiding complex external dependencies

ferreus_rmt v0.1.0

A new regularised marching tetrahedra isosurfacing crate, now used by ferreus_rbf.

• Regularised marching tetrahedra with vertex clustering
• Surface-following extraction to reduce evaluation counts compared with full-volume sampling
• Improved triangle quality compared with standard marching tetrahedra
• Manifold, self-intersection-free mesh generation
• Optional watertight extraction against an axis-aligned bounding box

oximo v0.3.0

oximo is a Rust algebraic modeling library for solving optimization problems, similar to Pyomo (Python) and JuMP (Julia).

At the moment, oximo supports building and solving LP/MILP models, with NLP/MINLP support planned for future releases. From a user perspective, only the main oximo crate is needed. However, the broader ecosystem is structured into multiple crates to simplify development and enable extensibility.

I would really appreciate any feedback, particularly on the API design. Having an ergonomic API with Rust's type system is hard, especially since I'm trying to avoid macros for now.

Events

Scientific Computing in Rust 2026, 8-10 July, held virtually

This year's Scientific Computing in Rust annual workshop is just over a week away.

The workshop features 10 minute talks on the use of Rust in a wide range of application areas, alongside tutorials, discussions and two longer invited talks. This year's invited talks will be given by:

• Tom Meltzer, who will talk about using the mdb debugging tool with MPI parallelised Rust codebases.
• Timothy Schmidt, who will talk about hyperreal and the hyper ecosystem.

You can register to attend by visiting scientificcomputing.rs/2026/register.

SciPy India x Rust Delhi Meetup, 22 August 2026, Noida, India

SciPy India and Rust Delhi are co-organising a technical meetup on scientific computing in Rust and Python. The sessions dig into real-world scientific problems and how people are solving them. The meetup runs from 2:00 PM to 5:00 PM at Essentia.dev in Noida, and the CFP is open to researchers, scientists, and developers (or anyone working at the intersection of Rust and Python) who want to present their work. You can submit a proposal or register at scipy.in/sci-py-rs.

Publications

Zhiqiu L., Agentics for Scientific Computing

Agentics is an open-source platform for executable computational-science challenges. This post discusses how Agentics can support Rust-based scientific-computing tools and research workflows by turning suitable problems into executable challenges where agents and humans can submit attempts, compare results, and preserve useful failures.

Trappitsch R., instrumentRs v0.2.0 design ideas

Blog post discussing design ideas for instrumentRs v0.2.0, a toolkit to develop drivers to control scientific instrumentation from Rust.

Help wanted

tenferro-rs

tenferro-rs is a Rust-native differentiable tensor stack for scientific computing, with dense tensors, eager and traced automatic differentiation (grad/vjp/jvp), NumPy/JAX-style einsum, linear algebra and FFT operation crates, and explicit CPU/CUDA/WebGPU backend control. The first crates were published to crates.io as v0.1 this month (a preview ahead of a stable 1.0), and the project leans on systematic validation: AD reference oracles, finite-difference checks, runnable examples, CI, and reproducible benchmarks. It is already the engine under tensor4all-rs, a Rust tensor-network stack.

We are looking for scientific-computing feedback and independent validation on real workloads: trying the examples and docs, reviewing AD and numerical behavior, checking backend limitations, and reviewing benchmark methodology in tenferro-benchmark. See the launch post for an overview.