[PATCH] D32563: Add LiveRangeShrink pass to shrink live range within BB.

[Quentin Colombet via llvm-commits]

> On May 17, 2017, at 11:34 AM, Dehao Chen via Phabricator <reviews at reviews.llvm.org> wrote:
> 
> danielcdh added a comment.
> 
> Sorry about the regression. As Quentin pointed out, moving instructions close to phi should not be a problem as the algorithm checks the moved instruction uses >1 single-use defs. Looking at the example of the regression case, the transformation looks reasonable if we only consider shrinking live-ranges. Could you provide an IR that I can try reproduce the issue and see why register allocation is worse?

It is worse, because REG_SEQUENCEs imply more constrain on the definition than the x independent input arguments of that sequence. 

> 
> During the mean time, I can prepare a patch to move this to x86-only, or revert the patch if necessary.

Moving this to x86-only sounds sensible.

> 
> 
> https://reviews.llvm.org/D32563
> 
> 
>