I need hardware support (it might be FPGA) at compile time to speed up compile-time computations. To be more specific compile-time training of a neural network. This might use OpenCL to greatly speed up compilation.
Would a compiler provide such an ability?
The best would be ability to call custom dynamic library function at compile-time.
I prefer C++. I see LLVM is moving forward pretty fast.
Does it provide something similar to enable it in Clang?