Other operations are not really explained either, and this is no more complicated than looking up what DP2ADD does or why it exists.
My question is, if there is a consensus, @hverbeet, @zfigura, @fcasas, and last but not least @Mystral, I can surely adjust or expand on the comments. But is that really blocking?
In general I think we could do with a bit more documentation in that area, yes. For example, it's fairly obvious that HLSL_OP3_LERP takes 3 arguments, and it seems safe enough to assume it does a linear interpolation between two of them, but I don't think it's necessarily immediately obvious which argument is the start, end, or interpolation factor.