[llvm] [SelectionDAG] optimize sdiv with positive divisor and positive magic (PR #189287)

[Takashi Idobe via llvm-commits]

Takashiidobe wrote:

> Interesting. "main" could avoid using x9 with a different optimization:
> 
> ```assembly
> sdiv3:                                  // @sdiv3
> 	mov	w8, #21846
> 	movk	w8, #21845, lsl #16
> 	smull	x8, w0, w8
> 	lsr	x8, x8, #32
> 	add	x0, x8, x8, lsr #31 # since x>>63 is x>>32>>31
> 	ret
> ```

Seems like good enough evidence to me to make this x86_64 only. I'll rewrite this diff and ask for reviews once i'm done. 

I can write a follow up issue for the lsr fold you suggested too. 

https://github.com/llvm/llvm-project/pull/189287