[PATCH] D93476: [LV][ARM] Inloop reduction cost modelling

[Sjoerd Meijer via Phabricator via llvm-commits]

SjoerdMeijer added inline comments.


================
Comment at: llvm/include/llvm/Analysis/TargetTransformInfo.h:1191
+  int getExtendedReductionCost(
+      bool IsMLA, bool IsUnsigned, Type *ResTy, VectorType *Ty,
+      TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput) const;
----------------
I am wondering if `IsMLA` is a bit too narrow as an interface, perhaps even unclear. If this is similar to `getArithmeticReductionCost` as mentioned in the comment, which takes an `opcode`, should this also take an `opcode` instead of `IsMLA`? The advantage we could describe costs for different type of reductions, or is this not useful/necessary?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D93476/new/

https://reviews.llvm.org/D93476