Phase 5: no_std, parallelism, cache, benchmarks

Cargo.toml

@@ -14,13 +14,31 @@ categories = ["mathematics", "science", "algorithms"]
 [lib]
 bench = false
 
+[features]
+default = ["std"]
+std = []
+parallel = ["dep:rayon", "std"]
+
 [dependencies]
-num-traits = "0.2"
+num-traits = { version = "0.2", default-features = false, features = ["libm"] }
+rayon = { version = "1", optional = true }
 
 [dev-dependencies]
 criterion = { version = "0.8", features = ["html_reports"] }
 approx = "0.5"
 
+[[bench]]
+name = "node_generation"
+harness = false
+
+[[bench]]
+name = "integration_1d"
+harness = false
+
+[[bench]]
+name = "cubature"
+harness = false
+
 [package.metadata.docs.rs]
 all-features = true
 rustdoc-args = ["--cfg", "docsrs"]

README.md

@@ -8,9 +8,12 @@
 
 A high-performance numerical quadrature (integration) library for Rust.
 
-- **Gauss-Legendre** -- O(1) per-node generation via the Bogaert (2014) algorithm
-- **Gauss-Kronrod** -- embedded pairs (G7-K15 through G25-K51) for error estimation
+- **Comprehensive** -- Gauss-Legendre, Gauss-Kronrod, Jacobi, Hermite, Laguerre, Chebyshev, Radau, Lobatto, Clenshaw-Curtis, tanh-sinh, oscillatory, Cauchy PV
+- **Adaptive integration** -- QUADPACK-style error-driven refinement with configurable GK pairs
+- **Multi-dimensional** -- tensor products, Smolyak sparse grids, Genz-Malik adaptive cubature, quasi-Monte Carlo (Sobol, Halton)
 - **Generic over `F: Float`** -- works with `f32`, `f64`, and AD types (e.g., [echidna](https://github.com/Entrolution/echidna))
+- **`no_std` compatible** -- works without the standard library (with `alloc`)
+- **Optional parallelism** -- `parallel` feature for rayon-based `_par` methods
 - **Precomputed rules** -- generate nodes/weights once, integrate many times with zero allocation
 
 ## Quick Start
@@ -22,37 +25,53 @@ Add to `Cargo.toml`:
 bilby = "0.1"
 ```
 
-```rust,ignore
+```rust
 use bilby::GaussLegendre;
 
 // Create a 10-point Gauss-Legendre rule
-let gl = GaussLegendre::new(10);
+let gl = GaussLegendre::new(10).unwrap();
 
-// Integrate x^2 over [0, 1]
-let result = gl.integrate(0.0, 1.0, |x| x * x);
+// Integrate x^2 over [0, 1] (exact result = 1/3)
+let result = gl.rule().integrate(0.0, 1.0, |x: f64| x * x);
 assert!((result - 1.0 / 3.0).abs() < 1e-14);
 ```
 
-## Roadmap
+## Features
 
-bilby is being developed in phases. See [ROADMAP.md](ROADMAP.md) for details.
+| Feature | Default | Description |
+|---------|---------|-------------|
+| `std` | Yes | Enables `std::error::Error` impl and `cache` module |
+| `parallel` | No | Enables rayon-based `_par` methods (implies `std`) |
 
-| Phase | Version | Content |
-|-------|---------|---------|
-| 0 | v0.1.0 | Gauss-Legendre (Bogaert), Gauss-Kronrod pairs, fixed-order integration |
-| 1 | v0.2.0 | Adaptive integration (QUADPACK-style), infinite domains, error control |
-| 2 | v0.3.0 | Classical rule families (Jacobi, Laguerre, Hermite, Chebyshev, Radau, Lobatto, Clenshaw-Curtis) |
-| 3 | v0.4.0 | Multi-dimensional (tensor product, sparse grids, adaptive cubature, quasi-Monte Carlo) |
-| 4 | v0.5.0 | Specialty methods (oscillatory, tanh-sinh, Cauchy principal value) |
-| 5 | v1.0.0 | Parallelism, caching, `no_std`, benchmarks |
+### `no_std` support
+
+bilby works in `no_std` environments (with `alloc`). Disable default features:
+
+```toml
+[dependencies]
+bilby = { version = "0.1", default-features = false }
+```
+
+### Parallelism
+
+Enable the `parallel` feature for parallel variants of integration methods:
+
+```toml
+[dependencies]
+bilby = { version = "0.1", features = ["parallel"] }
+```
+
+This provides `integrate_composite_par`, `integrate_par`, `integrate_box_par`,
+`MonteCarloIntegrator::integrate_par`, and `GaussLegendre::new_par`.
 
 ## Development
 
 ```bash
-cargo test          # Run tests
-cargo bench         # Run benchmarks
-cargo clippy        # Lint
-cargo fmt           # Format
+cargo test                         # Run tests (default features)
+cargo test --all-features          # Run tests including parallel
+cargo test --no-default-features   # Run tests in no_std mode
+cargo bench                        # Run benchmarks
+cargo clippy --all-features        # Lint
 ```
 
 ## License