[PATCH] D75939: [x86][seses] Introduce SESES pass for LVI

Scott Constable via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu Mar 12 08:40:58 PDT 2020


sconstab added a comment.

In D75939#1918453 <https://reviews.llvm.org/D75939#1918453>, @zbrid wrote:

> In D75939#1917701 <https://reviews.llvm.org/D75939#1917701>, @andrew.w.kaylor wrote:
>
> > What is the intention of this set of patches in relation to D75938 <https://reviews.llvm.org/D75938>? It was unclear to me whether you intended to commit this implementation or were just offering it as an alternative for discussion.
>
>
> I offer these patches for discussion and will upstream in folks in the LLVM community would like them to be upstreamed. If this approach and the other approach are both desired by the LLVM community, then I will work with Intel to decide whether to merge the approaches into a single framework/pass or not.
>
> If you have thoughts on whether or not this should be upstreamed, please let me know and feel free to add to the discussion.
>
> For code review comments:
>
> - I will address these shortly.


I see it this way. We definitely need D75935 <https://reviews.llvm.org/D75935> to mitigate RET instructions, and D75934 <https://reviews.llvm.org/D75934> to mitigate indirect calls/jumps. The SESES patch (with D75944 <https://reviews.llvm.org/D75944>) and D75936 <https://reviews.llvm.org/D75936>+D75937 <https://reviews.llvm.org/D75937> each use very different approaches to mitigate loads. So the choice is D75939 <https://reviews.llvm.org/D75939>+D75944 <https://reviews.llvm.org/D75944> against D75936 <https://reviews.llvm.org/D75936>+D75937 <https://reviews.llvm.org/D75937>.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75939/new/

https://reviews.llvm.org/D75939





More information about the llvm-commits mailing list