Right! We use Intel’s Math Kernel Library (MKL) inside NMath in order to take advantage of native, optimized code. In this way, we get extremely good performance, utilizing all cores. We chose to statically link the MKL libraries from within a CLI C++ kernel layer, which creates the dependency on VC++ 9.0.

In the future, we will likely also offer a C# kernel that calls MKL DLLs using PInvoke, for situations where a CLI C++ compiler is not available, such as Linux (Mono) and WinCE, although at some loss of performance. In this case, of course, there would be no dependency on VC++.