Install and run a reference case

The crate is unpublished and versioned with the monorepo, so it is a git dependency.

[dependencies]
deep_causality_cfd = { git = "https://github.com/deepcausality-rs/deep_causality.git", branch = "main" }

Pin by branch, tag, or rev. The crate moves with the monorepo and has no release cadence of its own, so rev is the stable choice.

Features

deep_causality_cfd = { git = "...", features = ["parallel"] }

std is on by default. parallel opts into CPU parallelism through deep_causality_par, which is what makes sweeps and branch fan-outs concurrent.

Measured on the DEC solver, parallel reduces throughput on small and medium grids, where per-operator work does not cover the coordination cost. Benchmark at the target grid size before enabling it.

Reference run

cargo run --release -p avionics_examples --example plasma_blackout_corridor

Runtime is 40.9 s. The run marches a compressible flow, reacts its plasma chemistry, gates a Kalman filter on the result, forks the marched state into seventeen counterfactual worlds, and reports thirteen gate verdicts.

Use --release. Several examples carry wall-clock budgets that a debug build exceeds.

Run the verification suite

Every verification target is an example on the crate itself:

cargo run --release -p deep_causality_cfd --example qtt_sod
cargo run --release -p deep_causality_cfd --example dec_lid_cavity_re1000_verification trend

Each self-verifies and exits nonzero when its check fails. The validation status page lists what each target asserts and to what number.

Several target READMEs document an older -p avionics_examples invocation. The authoritative names are the [[example]] entries in deep_causality_cfd/Cargo.toml.

Where to go next

The worked examples are ordered by the solver and DSL surface each requires. The capability boundaries page records the measured limits, including several uses of the QTT marcher that measurement ruled out.

Worked examples using this

• Plasma blackout corridor