[PATCH] D137913: [X86] Rewrite `getScalarizationOverhead()`

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added inline comments.


================
Comment at: llvm/lib/Target/X86/X86TargetTransformInfo.cpp:4442
 
-        // For vXf32 cases, insertion into the 0'th index in each v4f32
-        // 128-bit vector is free.
-        // NOTE: This assumes legalization widens vXf32 vectors.
-        if (MScalarTy == MVT::f32)
-          for (unsigned i = 0, e = cast<FixedVectorType>(Ty)->getNumElements();
-               i < e; i += 4)
-            if (DemandedElts[i])
-              Cost--;
+        APInt WidenedDemandedElts = DemandedElts.zext(NumLegalElts);
+        auto *LaneTy =
----------------
I'm a bit worried about legalization-induced padding,
that we obviously don't demand yet don't care about,
but i will look into that afterwards.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D137913/new/

https://reviews.llvm.org/D137913