[PATCH] D73021: [ARM] Basic gather scatter cost model

[Anna Welker via Phabricator via llvm-commits]

anwel added inline comments.


================
Comment at: llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp:901
+
+  // And (aligned) i32 gather will not need to be vectorized.
+  if (EltSize == 32)
----------------
nit: An ?


================
Comment at: llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp:918
+                         ->getPrimitiveSizeInBits();
+    // Scale need to be correct (which is only relevant for i16s).
+    if (Scale != 8 && Scale != EltSize)
----------------
nit: needs


================
Comment at: llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp:922-924
+    if (auto ZExt = dyn_cast<ZExtInst>(GEP->getOperand(1)))
+      if (ZExt->getOperand(0)->getType()->getScalarSizeInBits() <= EltSize)
+        return VectorCost;
----------------
The gather/scatter pass can also handle a sext here, in the manner that it will just leave it the way it is and use the result if it's of the correct type. That increases `VectorCost` by the cost of the sext, but doesn't necessarily mean that building a gather is more expensive than expanding.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D73021/new/

https://reviews.llvm.org/D73021