Crate apple_accelerate 

accelerate-rs

Safe Rust bindings for Apple’s Accelerate framework on macOS using a Swift bridge over the C APIs.

The GitHub repository is accelerate-rs; the published crates.io package is apple-accelerate.

Install

cargo add apple-accelerate

Enable the raw-ffi feature if you also want the underlying C declarations:

cargo add apple-accelerate --features raw-ffi

Quick start

use apple_accelerate::{
    add_f32, integrate, solve_linear_system_f32, sdot, QuadratureOptions,
};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let added = add_f32(&[1.0, 2.0, 3.0], &[4.0, 5.0, 6.0])?;
    assert_eq!(added, vec![5.0, 7.0, 9.0]);

    let dot = sdot(&[1.0, 2.0, 3.0], &[4.0, 5.0, 6.0])?;
    assert_eq!(dot, 32.0);

    // LAPACK matrices are column-major.
    let solution = solve_linear_system_f32(&[3.0, 1.0, 1.0, 2.0], 2, &[9.0, 8.0])?;
    assert!((solution[0] - 2.0).abs() < 1.0e-5);
    assert!((solution[1] - 3.0).abs() < 1.0e-5);

    let integral = integrate(|x| x * x, 0.0, 1.0, QuadratureOptions::default())?;
    assert!((integral.integral - (1.0 / 3.0)).abs() < 1.0e-9);
    Ok(())
}

v0.2.2 surface

• vDSP: FFT setup, biquad setup, vector arithmetic, reductions, and window generation
• vForce: element-wise transcendental and root functions over f32 slices
• BLAS: sdot, row-major sgemv, and row-major sgemm
• LAPACK: LU factorization and linear solves for column-major single-precision matrices
• BNNS: safe ReLU/sigmoid vector activations plus the existing thin unsafe filter owner
• Sparse: sparse-vector dot products and sparse-to-dense accumulation
• vImage: ARGB8888 rotate / box-convolve / scale and Planar8 contrast stretch
• simd: SIMD4 add / dot / length / normalize helpers
• Quadrature: one-dimensional adaptive numerical integration with Rust closures
• raw-ffi feature: re-exports exhaustive audited C declarations for vDSP, vForce, BLAS, LAPACK, BNNS, Sparse, vImage, and the remaining vecLib helper families surfaced by the SDK audit

Smoke examples

for ex in examples/*.rs; do
  cargo run --example "$(basename "$ex" .rs)"
done

The numbered examples cover every logical area of the crate.

Re-exports

pub use crate::blas::blas_order;

pub use crate::blas::blas_transpose;

pub use crate::blas::sdot;

pub use crate::blas::sgemm_row_major;

pub use crate::blas::sgemv_row_major;

pub use crate::bnns::graph_optimization_preference as bnns_graph_optimization_preference;

pub use crate::bnns::relu_f32 as bnns_relu_f32;

pub use crate::bnns::sigmoid_f32 as bnns_sigmoid_f32;

pub use crate::bnns::Filter as BnnsFilter;

pub use crate::bnns::GraphCompileOptions as BnnsGraphCompileOptions;

pub use crate::error::Error;

pub use crate::error::Result;

pub use crate::lapack::lu_decompose_f32;

pub use crate::lapack::solve_linear_system_f32;

pub use crate::lapack::LuDecompositionF32;

pub use crate::quadrature::integrate;

pub use crate::quadrature::Integrator as QuadratureIntegrator;

pub use crate::quadrature::Options as QuadratureOptions;

pub use crate::quadrature::QuadratureOutput;

pub use crate::simd::add_f32x4;

pub use crate::simd::dot_f32x4;

pub use crate::simd::length_f32x4;

pub use crate::simd::normalize_f32x4;

pub use crate::simd::Float4;

pub use crate::sparse::add_to_dense_f32 as sparse_add_to_dense_f32;

pub use crate::sparse::dot_dense_f32 as sparse_dot_dense_f32;

pub use crate::sparse::dot_sparse_f32 as sparse_dot_sparse_f32;

pub use crate::sparse::sparse_matrix_property;

pub use crate::sparse::sparse_status;

pub use crate::sparse::SparseIndex;

pub use crate::sparse::SparseMatrixF32;

pub use crate::vdsp::add_f32;

pub use crate::vdsp::add_f64;

pub use crate::vdsp::blackman_window;

pub use crate::vdsp::blackman_window_f64;

pub use crate::vdsp::dot_f32;

pub use crate::vdsp::dot_f64;

pub use crate::vdsp::fft_direction;

pub use crate::vdsp::fft_radix;

pub use crate::vdsp::hamming_window;

pub use crate::vdsp::hamming_window_f64;

pub use crate::vdsp::max_f32;

pub use crate::vdsp::max_f64;

pub use crate::vdsp::mean_f32;

pub use crate::vdsp::mean_f64;

pub use crate::vdsp::min_f32;

pub use crate::vdsp::min_f64;

pub use crate::vdsp::sub_f32;

pub use crate::vdsp::sub_f64;

pub use crate::vdsp::sum_f32;

pub use crate::vdsp::sum_f64;

pub use crate::vdsp::window_flags;

pub use crate::vdsp::BiquadSetup;

pub use crate::vdsp::FftSetup;

pub use crate::vforce::cos_f32;

pub use crate::vforce::exp_f32;

pub use crate::vforce::log_f32;

pub use crate::vforce::sin_f32;

pub use crate::vforce::sqrt_f32;

pub use crate::vimage::alpha_blend_argb8888;

pub use crate::vimage::box_convolve_argb8888;

pub use crate::vimage::clip_to_alpha_argb8888;

pub use crate::vimage::contrast_stretch_planar8;

pub use crate::vimage::convert_argb8888_to_planar8;

pub use crate::vimage::convert_planar8_to_argb8888;

pub use crate::vimage::premultiply_argb8888;

pub use crate::vimage::rotate_argb8888;

pub use crate::vimage::scale_argb8888;

pub use crate::vimage::unpremultiply_argb8888;

pub use crate::vimage::vimage_flags;

pub use crate::vimage::ImageBuffer;

Modules

blas

Safe wrappers for BLAS routines in Accelerate.

bnns

Safe wrappers for BNNS routines in Accelerate.

error

Error types returned by the Accelerate wrappers.

lapack

Safe wrappers for LAPACK routines in Accelerate.

quadrature

Safe wrappers for Accelerate quadrature integrators.

simd

Safe wrappers for four-lane SIMD helpers backed by Apple SIMD types.

sparse

Safe wrappers for sparse linear algebra routines in Accelerate.

vdsp

Safe wrappers for vDSP routines in Accelerate.

vforce

Safe wrappers for vForce vector functions in Accelerate.

vimage

Safe wrappers for vImage routines in Accelerate.