Add Advanced SIMD instructions for Rounding Double Multiply Add/Subtract v8.1 extension
In the discussion on related feature in AArch64 backend (http://reviews.llvm.org/D7998), there were decisions taken:
- rename "rdma" subtarget feature to "v8_1a"
- not to introduce intrinsics, instead perform fusion "x = sqrdmlah(x.y.z) --> x = sqadd(x, sqrdmulh(y.z)"
This way, this revision is invalid, I will upload next one soon.