[PATCH] D89116: [SVE][AArch64] Replace TypeSize comparisons with their integer equivalents

[Sander de Smalen via Phabricator via llvm-commits]

sdesmalen added inline comments.


================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:7460
   unsigned ResMultiplier =
-      VT.getScalarSizeInBits() / SmallestEltTy.getSizeInBits();
-  NumElts = VT.getSizeInBits() / SmallestEltTy.getSizeInBits();
+      VT.getScalarSizeInBits() / SmallestEltTy.getFixedSizeInBits();
+  uint64_t VTSize = VT.getFixedSizeInBits();
----------------
`SmallestEltTy.getScalarSizeInBits()` 


================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:7462
+  uint64_t VTSize = VT.getFixedSizeInBits();
+  NumElts = VTSize / SmallestEltTy.getFixedSizeInBits();
   EVT ShuffleVT = EVT::getVectorVT(*DAG.getContext(), SmallestEltTy, NumElts);
----------------
`SmallestEltTy.getScalarSizeInBits()`


================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:7478
     EVT EltVT = SrcVT.getVectorElementType();
-    unsigned NumSrcElts = VT.getSizeInBits() / EltVT.getSizeInBits();
+    unsigned NumSrcElts = VTSize / EltVT.getFixedSizeInBits();
     EVT DestVT = EVT::getVectorVT(*DAG.getContext(), EltVT, NumSrcElts);
----------------
`EltVT.getScalarSizeInBits()`


================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:7548
+    Src.WindowScale =
+        SrcEltTy.getFixedSizeInBits() / SmallestEltTy.getFixedSizeInBits();
     Src.WindowBase *= Src.WindowScale;
----------------
`SrcEltTy.getScalarSizeInBits() / SmallestEltTy.getScalarSizeInBits()`


================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:8322
 
-    if (ScalarVT.getSizeInBits() < 32 && ScalarVT.isInteger())
+    if (ScalarVT.getFixedSizeInBits() < 32 && ScalarVT.isInteger())
       ScalarVT = MVT::i32;
----------------
`ScalarVT.getScalarSizeInBits`


================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:10105
     return false;
-  unsigned NumBits1 = Ty1->getPrimitiveSizeInBits();
-  unsigned NumBits2 = Ty2->getPrimitiveSizeInBits();
+  uint64_t NumBits1 = Ty1->getPrimitiveSizeInBits().getFixedSize();
+  uint64_t NumBits2 = Ty2->getPrimitiveSizeInBits().getFixedSize();
----------------
Are you sure this shouldn't also work for vectors?

If not, then it's better to use `Ty1->getScalarSizeInBits()`



================
Comment at: llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp:458
   // the default cost for the extend.
-  if (DstVT.getSizeInBits() < SrcVT.getSizeInBits())
+  if (DstVT.getFixedSizeInBits() < SrcVT.getFixedSizeInBits())
     return Cost + getCastInstrCost(Opcode, Dst, Src, TTI::CastContextHint::None,
----------------
`DstVT.getScalarSizeInBits() < SrcVT.getScalarSizeInBits()`


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D89116/new/

https://reviews.llvm.org/D89116