rvv-intrinsic-doc VS rvv_example

Support for vector intrinsics as specified in version 0.11 of the RISC-V vector intrinsic specification, thanks Ju-Zhe Zhong from RiVAI for contributing most of implementation.

The docs at https://github.com/riscv-non-isa/rvv-intrinsic-doc makes me think there should be a vfmacc_vv_f32m1_ta. I'll raise an issue there if it really is missing and not just me misunderstanding / using outdated stuff.

If you're doing physics modeling or calculating light refractions or other stuff where you, a highly trained professional, have profiled and know that you'd benefit from V and the optimizer just isn't doing it, you can meet in the middle. The Vector Intrinsics are the contract between that highly trained professional and the compiler that aren't quite programming in assembly and handle the absurd avalanche (made easy by templates, but hard by C) of a templated/body that the optimizer can then easily inline without calling overhead and without blowing register allocation. It handles things like the generation of the 176 ways to find a maximum integer in an array. This is kind of the compromise between the autovectorization code not being able to decompose your nested Fortran loops or lighting effect in graphics or whatever but not quite reaching despair and coding it in assembly yourself. If you've ever programmed MMX, this will all look familiar.