[PATCH] D101836: [LoopVectorize] Enable strict reductions when allowReordering() returns false

Mon May 24 07:37:22 PDT 2021

david-arm added inline comments.

================
Comment at: llvm/test/Transforms/LoopVectorize/AArch64/strict-fadd.ll:621
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %for.end, label %for.body
+
----------------
david-arm wrote:
> Are these new tests missing hints that the other tests seem to use? I just wondered if it was better to be consistent here that's all. The reason I mention this is because I was expecting the UNORDERED case to vectorise due to the `-hints-allow-reordering=true` flag.
I think I see now @kmclaughlin - you're testing the productisation of `-enable-strict-reductions` so you were adding some tests deliberately without hints, which also makes sense. In this case I'd also be happy if you left these tests as they are and just added some comments explaining why we expect the CHECK-UNORDERED case to not vectorise.

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D101836/new/

https://reviews.llvm.org/D101836