Crates.io | rust-blas |
lib.rs | rust-blas |
version | 0.2.0 |
source | src |
created_at | 2019-10-18 17:07:32.57135 |
updated_at | 2021-11-21 11:14:36.467279 |
description | BLAS bindings and wrappers, fork of rblas |
homepage | https://github.com/spearow/juice/tree/master/rust-blas |
repository | https://github.com/spearow/juice |
max_upload_size | |
id | 173767 |
size | 141,307 |
Rust bindings and wrappers for BLAS (Basic Linear Algebra Subprograms).
RBLAS wraps each external call in a trait with the same name (but capitalized).
This trait contains a single static method, of the same name. These traits are
generic over the four main types of numbers BLAS supports: f32
, f64
,
Complex32
, and Complex64
.
For example the functions cblas_saxpy
, cblas_daxpy
, cblas_caxypy
, and
cblas_zaxpy
are called with the function Axpy::axpy
.
Additionally, RBLAS introduces a few traits to shorten calls to these BLAS
functions: Vector
for types that implement vector-like characteristics and
Matrix
for types that implement matrix-like characteristics. The Vector
trait is already implemented by Vec
and []
types.
By default, the library links with blas
dynamically. To link to an alternate
implementation, like OpenBLAS, use the environment variable CARGO_BLAS
. If
you've already built the bindings, you may need to clean and build again.
export CARGO_BLAS=openblas
extern crate rblas;
use rblas::Dot;
fn main() {
let x = vec![1.0, -2.0, 3.0, 4.0];
let y = [1.0, 1.0, 1.0, 1.0, 7.0];
let d = Dot::dot(&x, &y[..x.len()]);
assert_eq!(d, 6.0);
}
#[macro_use]
extern crate rblas as blas;
use blas::math::Mat;
use blas::{Matrix, Vector};
use blas::math::Marker::T;
fn main() {
let x = vec![1.0, 2.0];
let xr = &x as &Vector<_>;
let i = mat![1.0, 0.0; 0.0, 1.0];
let ir = &i as &Matrix<_>;
assert!(xr + &x == 2.0 * xr);
assert!(ir * xr == x);
let dot = (xr ^ T) * xr;
assert!(dot == 5.0);
}