**If you install a math library to your .NET/C# project, LAPACK will be probably one of the key feature you expect from that: The routines provided by LAPACK (which actually means: “Linear Algebra Package”) cover a wide range of functionalities needed for nearly any numerical algorithm, in natural sciences, computer science, and social science.**

The LAPACK software library is written in FORTRAN code – until 2008 it was even written in FORTRAN 77. That’s why adding LAPACK functions to an enterprise software project written in Java or C#/.NET can be quite a demanding task: The implementation of native modules often causes problems regarding maintainability and steadiness of enterprise applications.

## Our LAPACK implementation for C#/.NET

ILNumerics offers a convenient implementation of LAPACK for C# and .NET: It provides software developers both the execution speed of highly optimized processor specific native code and the convenience of managed software frameworks. That allows our users to create powerful applications in a very short time.

For linear algebra functions ILNumerics uses the processor-optimized LAPACK library by the MIT and Intel’s MKL. `ILMath.Lapack`

is a concrete interface wrapper class that provides the native LAPACK functions. The LAPACK wrapper is initialized when a call to any static method of `ILMath`

is made. Once the corresponding binaries for your actual architecture have been found, consecutive calls will utilize them in a very efficient way.

The MKL is utilized (and needed) for all calls to any fft(A) function, for matrix decompositions (like for example linsolve, rank, svd, qr etc.). The only exception to that is ILMath.multiply – the general matrix multiplication. Matrix multiplication is just such an often needed feature, a math library simply could not go without. So we decided to implement ILMath.multiply() purely in managed code. The good thing: it is not really far behind the speed of the processor optimized version! If MKL binaries are found at runtime, those will be used, of course. But in the case of their absence, the managed version should work out just fast enough for the very most situations.

In most cases using this kind of .NET/C# LAPACK implementation means: faster results and more stable software applications. Learn more about Linear Equation Systems and other features of ILNumerics in our Documentation.