[PATCH] D42948: [SelectionDAG] Support some SimplifySetCC cases for comparing against vector splats of constants.

[Sanjay Patel via Phabricator via llvm-commits]

spatel added inline comments.


================
Comment at: test/CodeGen/X86/vector-compare-simplify.ll:29-31
+; CHECK-NEXT:    pcmpeqd {{.*}}(%rip), %xmm0
+; CHECK-NEXT:    pcmpeqd %xmm1, %xmm1
+; CHECK-NEXT:    pxor %xmm1, %xmm0
----------------
Should we account for this regression first (by adding x86-specific folds for vector compares with edge case constants)?


================
Comment at: test/CodeGen/X86/vector-compare-simplify.ll:78-80
-; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [2147483647,2147483647,2147483647,2147483647]
-; CHECK-NEXT:    pcmpgtd %xmm0, %xmm1
-; CHECK-NEXT:    movdqa %xmm1, %xmm0
----------------
Do we know why this didn't load-fold like the sgt_max test?


Repository:
  rL LLVM

https://reviews.llvm.org/D42948