[llvm] [RISCV][ISel] Add ISel support for experimental Zimop extension (PR #77089)

[Craig Topper via llvm-commits]

topperc wrote:

> > > Just a thought: can we just have only one intrinsic for `mop.r` and `mop.rr`? The number becomes an constant integer argument of these two intrinsics. If so, we can reduce some cases in legalization/lowering. And of course, we should change the ISel logics.
> > 
> > 
> > How would we represent the non-existent second source register for mop.r?
> 
> My expression is not clear. I meant one intrinsic for `mop.r` and another for `mop.rr`, there are two intrinsics. :-)

Thanks. I agree.

https://github.com/llvm/llvm-project/pull/77089