[PATCH] D87188: [InstCombine] Canonicalize SPF to abs intrinc

[Roman Lebedev via Phabricator via cfe-commits]

lebedev.ri added a comment.

In D87188#2467915 <https://reviews.llvm.org/D87188#2467915>, @nlopes wrote:

> This patch regressed the following test of Transforms/InstCombine/abs-1.ll:
> (need to drop NSW in this case).
>
>   define i8 @nabs_canonical_3(i8 %x) {
>   %0:
>     %cmp = icmp slt i8 %x, 0
>     %neg = sub nsw i8 0, %x
>     %abs = select i1 %cmp, i8 %x, i8 %neg
>     ret i8 %abs
>   }
>   =>
>   define i8 @nabs_canonical_3(i8 %x) {
>   %0:
>     %1 = abs i8 %x, 1
>     %abs = sub nsw i8 0, %1
>     ret i8 %abs
>   }
>   Transformation doesn't verify!
>   ERROR: Target is more poisonous than source
>   
>   Example:
>   i8 %x = #x80 (128, -128)
>   
>   Source:
>   i1 %cmp = #x1 (1)
>   i8 %neg = poison
>   i8 %abs = #x80 (128, -128)
>   
>   Target:
>   i8 %1 = poison
>   i8 %abs = poison
>   Source value: #x80 (128, -128)
>   Target value: poison
>
> (same bug occurs with @nabs_nabs_x01 in Transforms/InstCombine/abs_abs.ll)

Thanks, fixed in rGf8079355c604fead0f8538548bd7eb51fcc81e31 <https://reviews.llvm.org/rGf8079355c604fead0f8538548bd7eb51fcc81e31>.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D87188/new/

https://reviews.llvm.org/D87188