[PATCH] D122071: [SDAG] enable binop identity constant folds for multiplies

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: llvm/test/CodeGen/X86/midpoint-int-vec-128.ll:3580
   ret <16 x i8> %a10
 }
 
----------------
@lebedev.ri This code now optimizes considerably in IR: https://llvm.godbolt.org/z/edhMY1crG

Should these test cases be updated to match?

```
define <16 x i8> @vec128_i8_signed_reg_mem(<16 x i8> %a1, <16 x i8>* nocapture readonly %a2_addr) {
  %a2 = load <16 x i8>, <16 x i8>* %a2_addr, align 16
  %t3 = icmp slt <16 x i8> %a2, %a1
  %1 = tail call <16 x i8> @llvm.smin.v16i8(<16 x i8> %a2, <16 x i8> %a1)
  %2 = tail call <16 x i8> @llvm.smax.v16i8(<16 x i8> %a2, <16 x i8> %a1)
  %t7 = sub <16 x i8> %2, %1
  %t8 = lshr <16 x i8> %t7, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
  %3 = sub nsw <16 x i8> zeroinitializer, %t8
  %4 = select <16 x i1> %t3, <16 x i8> %3, <16 x i8> %t8
  %a10 = add nsw <16 x i8> %4, %a1
  ret <16 x i8> %a10
}
declare <16 x i8> @llvm.smin.v16i8(<16 x i8>, <16 x i8>)
declare <16 x i8> @llvm.smax.v16i8(<16 x i8>, <16 x i8>)
```


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D122071/new/

https://reviews.llvm.org/D122071