NumSharp

NumPy for .NET — a native .NET array library with a NumPy-shaped API: NDArray, broadcasting, slicing views, dtype-aware np.* functions, unmanaged storage, and runtime-generated SIMD kernels. The compatibility target is NumPy 2.x; where behavior differs, NumPy is treated as the source of truth.

NumSharp.Core is 100% managed C# with no native dependency and no P/Invoke — every kernel is its own managed code. An optional OpenBLAS matrix-product backend ships separately as NumSharp.Interop.OpenBLAS.

Install

dotnet add package NumSharp

Quick start

using NumSharp;

var a = np.arange(12).reshape(3, 4);
var window = a[":, 1::2"];

Console.WriteLine(window);
Console.WriteLine(np.sum(window, axis: 0));

For Python readers, the shape is deliberately close:

import numpy as np

a = np.arange(12).reshape(3, 4)
print(a[:, 1::2].sum(axis=0))

Features

• NumPy-style NDArray — N-dimensional arrays with shape, strides, offsets, slicing, and view semantics (slices return views that share memory).
• Broadcasting — NumPy-style shape expansion without materializing repeated values.
• Dtype-aware operations — 15 core dtypes with NumPy-oriented promotion (NEP50) and conversion behavior.
• Broad np.* surface — creation, manipulation, math, reductions, comparisons, logic, linear algebra, FFT, random sampling, and .npy/.npz I/O.
• Generated IL + SIMD kernels — runtime-specialized kernels (V128/V256/V512) for supported dtype and layout combinations.

Links

• Documentation: https://scisharp.github.io/NumSharp/
• Getting started: https://scisharp.github.io/NumSharp/docs/NDArray.html
• API reference: https://scisharp.github.io/NumSharp/api/
• Coverage & support dashboard: https://scisharp.github.io/NumSharp/docs/coverage-support-dashboard.html
• Source: https://github.com/SciSharp/NumSharp

Licensed under the Apache License 2.0.