Crate unmtx_gpu

Micro neural matrix library for GPU is small library that operates on matrices.

This library uses GPU by the following computing platforms:

• OpenCL
• CUDA

If this library uses CUDA, this library can use the cuBLAS library to multiplication of matrices.

A frontend-backend architecture is used by this library. The frontend of this library can use one of two backends (OpenCL or CUDA). These backends allow to use GPU by the computing platforms. The frontend and the backend can have many instances. This library provides a high-level interfece to operations of matrices by the frontend and methods of a Matrix structure.

Examples

let a = matrix![
    [1.0, 2.0],
    [3.0, 4.0]
];
let x = matrix![
    [5.0],
    [6.0]
];
let b = matrix![
    [7.0],
    [8.0]
];
let c = a * x + b;
assert_eq!(vec![1.0 * 5.0 + 2.0 * 6.0 + 7.0, 3.0 * 5.0 + 4.0 * 6.0 + 8.0], c.elems());

Modules

cuda

A module that contains a CUDA backend.

opencl

A module that contains an OpenCL backend.

Macros

matrix

Creates a matrix from the arguments.

Structs

Frontend

A frontend structure.

Matrix

A matrix structure.

Enums

BackendArray

An enumeration of backend array.

Error

An error enumeration.

Traits

Backend

A backend trait.

Functions

finalize_default_backend

Finalizes a default backend.

get_default_backend

Returns a default backend.

initialize_default_backend_for_uninitialized

Initializes a default backend if the backend is uninitialized and returns the default backend.

set_default_backend

Sets a default backend.

set_default_backend_for_uninitialized

Sets a default backend if the default backend is uninitialized and returns the default backend.

unset_default_backend

Unsets a default backend.

Type Aliases

Result

A result type.