matrixpacked 0.1.0

Packed structured matrices with BLAS and LAPACK integration
Documentation

matrixpacked

CI

Packed triangular, symmetric, Hermitian, and positive-definite matrices backed directly by traditional BLAS/LAPACK packed-column storage.

The crate stores n(n+1)/2 scalars, supports owned matrices and zero-copy borrowed views, and exposes packed matrix-vector operations, factorizations, solves, refinement, diagnostics, eigensolvers, and format conversions.

use matrixpacked::{PackedMatrixError, PackedSPD};

fn main() -> Result<(), PackedMatrixError> {
    let a = PackedSPD::from_vec(2, vec![4.0_f64, 1.0, 3.0])?;
    let factor = a.cholesky()?;
    let x = factor.solve_vector(&[6.0, 7.0])?;
    assert_eq!(x.len(), 2);
    Ok(())
}

Running numerical operations such as this solve requires a native BLAS/LAPACK provider in the final application.

Installation

For packed storage and provider-free APIs:

[dependencies]

matrixpacked = "0.1"

Choose one native provider below when using linked numerical operations.

Native backend

blas and lapack are bindings, so numerical executables need a native provider. The crate selects none by default. Enable the repository-supported bundled provider for your platform, or link compatible symbols in the final application:

# Linux

matrixpacked = { version = "0.1", features = ["openblas-static"] }



# Windows

matrixpacked = { version = "0.1", features = ["intel-mkl-static"] }

Select only one bundled provider. The supported Windows workflow uses MKL; the non-vcpkg OpenBLAS source build is not assumed to work there.

Cargo unifies features across the dependency graph. If different dependencies enable both bundled provider features, both providers are resolved; the crate does not reject that combination at compile time. Linking two native providers can produce conflicting symbols and is not a supported configuration, so the final application should ensure that only one provider feature is selected.

Optional nalgebra support

Enable nalgebra-interop for owned nalgebra::DMatrix conversion, explicit triangle extraction, tolerance-aware structural validation, and nalgebra Cholesky-backed SPD/HPD validation:

matrixpacked = { version = "0.1", features = ["nalgebra-interop"] }

Packed storage cannot be exposed as a zero-copy rectangular nalgebra view. Structured conversion is pure Rust; triangular packed/full conversion currently requires a native LAPACK provider.

Supported Rust version

The minimum supported Rust version for matrixpacked 0.1.0 is Rust 1.89. This version covers the core crate and the optional nalgebra-interop feature.

Development commands

# Provider-free compilation and documentation

cargo check --all-targets --no-default-features

cargo doc --no-deps --features nalgebra-interop


# Linux linked suite

cargo test --all-targets --features openblas-static


# Windows linked suite

cargo test --all-targets --features intel-mkl-static

Runner scripts are scripts_run_lapack_examples.{sh,bat} and scripts_run_nonlapack_examples.{sh,bat}. The batch scripts use --intel-mkl-static; the shell LAPACK runner accepts --openblas-static.

Stability

matrixpacked is pre-1.0. Minor releases may include API changes while the interface evolves; user-visible changes will be recorded in the changelog.

Guides and status

License

Licensed under either of Apache License, Version 2.0 or the MIT license at your option.