[PATCH] D150388: [CodeGen]Allow targets to use target specific COPY instructions for live range splitting

Matt Arsenault via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Jul 24 11:45:39 PDT 2023


arsenm added a comment.

In D150388#4529268 <https://reviews.llvm.org/D150388#4529268>, @vitalybuka wrote:

> In D150388#4529229 <https://reviews.llvm.org/D150388#4529229>, @arsenm wrote:
>
>> In D150388#4528824 <https://reviews.llvm.org/D150388#4528824>, @vitalybuka wrote:
>>
>>> @yassingh @arsenm
>>> The major target is broken for too long, we need either fix or revert, even if it's a chain of patches.
>>
>> I just posted about this issue:
>> https://discourse.llvm.org/t/subreg-to-reg-semantics-or-x86s-zext-implementation-is-broken/72250
>>
>> Either x86 is broken or the definition of SUBREG_TO_REG is broken
>
> So if we have not fix, just discussion, should we revert? Even if it's preexisted unknown bug, I don't think it's appropriate to keep regression in existing functionality, for sake of fixing previously broken or missing functionally. 
> Is it possible to workaround/revert some change with "if (x86) do old way" ?

Let me prepare a patch to insert explicit zeroing here, see if it fixes the issue and how badly it blows up the lit suite


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D150388/new/

https://reviews.llvm.org/D150388



More information about the llvm-commits mailing list