I'm the creator of the Pelemay series, which have been researched and developed to parallelize and optimize a series of map functions into native parallelized native code with SIMD instructions. I'd like to contribute these artifacts to Erlang. I began researching BeamAsm, to be applied to more efficient parallelization with SIMD instructions. I found it has amazing potential! Thanks to everyone involved in the development of BeamAsm.
I guess some BIF, including binary_match, has much opportunity and potential to be optimized by SIMD instructions, but have not been realized, yet.
Thus, I propose a new NIF function that wraps CpuInfo::hasFeature in AsmJit, and new configuring scripts that judge whether there are header files that define intrinsics or not. If they exist, we can describe BIF using SIMD instructions when they are available.