[PATCH] D145540: [InstCombine] Reduce absolute diff from min+max+sub

[Jun Zhang via Phabricator via llvm-commits]

junaire added inline comments.


================
Comment at: llvm/test/Transforms/InstCombine/sub-minmax.ll:1035
+  %min = call i8 @llvm.smin.i8(i8 %b, i8 %a)
+  %max = call i8 @llvm.smax.i8(i8 %b, i8 %a)
+  %ab = sub nsw i8 %max, %min
----------------
RKSimon wrote:
> shouldn't this be:
> ```
>  %max = call i8 @llvm.smax.i8(i8 %a, i8 %b)
> ```
> shouldn't this be:
> ```
>  %max = call i8 @llvm.smax.i8(i8 %a, i8 %b)
> ```

IIUC, we should test all the following cases?
1. max(a, b) - min(a, b)
2. max(b, a) - min(b, a)
3. max(a, b) - min(b, a)
4. max(b, a) - min(a, b)

However, I'm not certain it is worth doing so. I'll pick case 1 and case 3 then.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D145540/new/

https://reviews.llvm.org/D145540