Compile .NET to native k1om architecture

[ghost]

Greetings!

I would like to request support for native compilation on KNC Xeon PHI.

Thanks, coast

[jkotas]

https://en.wikipedia.org/wiki/Xeon_Phi

There are really two distinct problem:

• How to get the code compiled to native instructions. There are several strategies possible in the context of CoreRT. The easiest one to start with is using CppCodeGen. You can read more about CppCodeGen here: https://blogs.msdn.microsoft.com/alphageek/2016/10/13/native-compilation-why-jit-when-you-can-codegen/. @mancoast Would you like to give this a shot?

• What the programming model for the massively parallel GPU-like architectures should be. There are number of existing design discussions about that, like:

  https://github.com/dotnet/runtime/issues/77609 https://github.com/dotnet/roslyn/issues/14604 https://github.com/dotnet/roslyn/issues/14495

  Let's use the existing design discussions to discuss the programming model.

[ghost]

@jkotas I will attempt transpilation. "One strategy we are investigating is the ability to transpile .NET code to C++ code that could be compiled with an appropriate C++ compiler for the target platform."

I would like to utilize CppCodeGen with travis.yml Then we can utilize travis.yml with Intel Stack: https://github.com/nemequ/icc-travis EX. https://github.com/edanor/umesimd/blob/master/.travis.yml