[PATCH] D113350: [X86][Costmodel] `getReplicationShuffleCost()`: implement cost model for 32/64 bit-wide elements with AVX512F

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added inline comments.


================
Comment at: llvm/lib/Target/X86/X86TargetTransformInfo.cpp:3654
+  std::pair<InstructionCost, MVT> ReplicatedVecTyLegalization =
+      TLI->getTypeLegalizationCost(DL, ReplicatedVecTy);
+
----------------
RKSimon wrote:
> We're not using the cost at all. just the legalized type?
Could you please clarify, are you asking to change this to
```
MVT LegalReplicatedVecTy =
      TLI->getTypeLegalizationCost(DL, ReplicatedVecTy).second;
```
, or asking why we don't use the cost?
I'm mainly not sure why that cost is there or how i should use it here.
We don't need to legalize the input, it already as,
we just need to know how it's laid out (how many ?MM vectors), i believe?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D113350/new/

https://reviews.llvm.org/D113350