BlasSharp.OpenBlas 0.3.0

BlasSharp

BLAS/LAPACK bindings for .NET (OpenBLAS, MKL, Apple Accelerate)

English | 日本語

Overview

BlasSharp is a library that provides bindings to BLAS/LAPACK for .NET.

BLAS/LAPACK libraries optimized for various CPUs are provided by Intel, Apple, and others. By bringing these libraries into C#, you can perform linear algebra computations with optimal performance.

Usage

You can use the bindings by installing the package that corresponds to the desired library. Below is a sample using OpenBLAS:

using BlasSharp;
using static BlasSharp.OpenBlas.NativeMethods;

unsafe
{
    double[] A = [
        1.0, 2.0, 3.0,
        4.0, 5.0, 6.0,
        7.0, 8.0, 9.0
    ];

    double[] B = [
        1.0, 2.0, 3.0,
        4.0, 5.0, 6.0,
        7.0, 8.0, 9.0,
    ];

    double[] C = new double[9];

    fixed (double* a = A, b = B, c = C)
    {
        cblas_dgemm(CBLAS_ORDER.CblasColMajor,
            CBLAS_TRANSPOSE.CblasNoTrans,
            CBLAS_TRANSPOSE.CblasNoTrans,
            3, 3, 3, 1.0, a, 3, b, 3, 0.0, c, 3);
    }

    Console.WriteLine(string.Join(',', C));
}

IBlasOperations / ILapackOperations

The BlasSharp.Shared package provides IBlasOperations / ILapackOperations, which abstracts BLAS/LAPACK routines. Each package provides its own implementation.

public interface IBlasOperations : IBlasLevel1, IBlasLevel2, IBlasLevel3
{
}

public unsafe interface IBlasLevel1
{
    void Saxpy(int n, float alpha, float* x, int incX, float* y, int incY);
    void Daxpy(int n, double alpha, double* x, int incX, double* y, int incY);
    void Caxpy(int n, void* alpha, void* x, int incX, void* y, int incY);
    void Zaxpy(int n, void* alpha, void* x, int incX, void* y, int incY);

    ...
}

public interface ILapackOperations : ILapackDriverRoutines, ILapackComputationalRoutines, ILapackAuxiliaryRoutines
{
}

public unsafe interface ILapackDriverRoutines
{
    int Sgesv(int* n, int* nrhs, float* a, int* lda, int* ipiv, float* b, int* ldb, int* info);
    int Cgesv(int* n, int* nrhs, void* a, int* lda, int* ipiv, void* b, int* ldb, int* info);
    int Dgesv(int* n, int* nrhs, double* a, int* lda, int* ipiv, double* b, int* ldb, int* info);
    int Zgesv(int* n, int* nrhs, void* a, int* lda, int* ipiv, void* b, int* ldb, int* info);

    ...
}

OpenBLAS

You can use the OpenBLAS binaries and bindings by adding BlasSharp.OpenBlas to your project.

BlasSharp.OpenBlas supports multiple platforms:
win-x64, osx-x64, osx-arm64, linux-x64, and linux-arm64.

.NET CLI

dotnet add package BlasSharp.OpenBlas

Package Manager

Install-Package BlasSharp.OpenBlas

MKL

You can use the bindings for Intel's MKL by adding BlasSharp.Mkl to your project.

.NET CLI

dotnet add package BlasSharp.Mkl

Package Manager

Install-Package BlasSharp.Mkl

To use MKL, you also need to install the official Intel MKL NuGet package (e.g., intelmkl.devel.win-x64).
See the MKL download page for more details.

Accelerate (Apple)

You can use the bindings for Apple’s Accelerate framework by adding BlasSharp.AppleAccelerate to your project.

BlasSharp.AppleAccelerate currently supports only osx-arm64.

.NET CLI

dotnet add package BlasSharp.AppleAccelerate

Package Manager

Install-Package BlasSharp.AppleAccelerate

License

This library is released under the MIT License.