[PATCH] D129550: [LoopVectorize] Ensure the VPReductionRecipe is placed after all it's inputs

Tue Jul 12 08:00:30 PDT 2022

david-arm marked an inline comment as done.
david-arm added inline comments.

================
Comment at: llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll:26
+; CHECK-ORDERED-TF: %[[RDX]] = call float @llvm.vector.reduce.fadd.nxv8f32(float %[[VEC_PHI]], <vscale x 8 x float> %[[SEL]])
+; CHECK-ORDERED-TF: for.end
+; CHECK-ORDERED-TF: %[[PHI:.*]] = phi float [ %[[SCALAR:.*]], %for.body ], [ %[[RDX]], %middle.block ]
----------------
kmclaughlin wrote:
> I think the new check lines from `for.end` here and in the other tests below can be removed, since we will never reach the scalar epilogue?
I've left in most of them because I think the other CHECK lines want to show the reduction value being returned from the function. However, I did remove them from fadd_strict_interleave as they didn't add any value.

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D129550/new/

https://reviews.llvm.org/D129550