[PATCH] D43582: [SLP] Generalization of stores vectorization.

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a comment.

A couple of final minor comments but this is looking almost ready.



================
Comment at: lib/Transforms/Vectorize/SLPVectorizer.cpp:5338
+  int E = Stores.size();
+  SmallVector<bool, 16> Tails(E, false);
+  SmallVector<int, 16> ConsecutiveChain(E, E + 1);
----------------
Better to use SmallBitVector or APInt ?


================
Comment at: lib/Transforms/Vectorize/SLPVectorizer.cpp:5382
+    unsigned EltSize = R.getVectorElementSize(Stores[0]);
+    if (MaxVecRegSize < EltSize)
+      continue;
----------------
Would it be better to use:
```
if ((MaxVecRegSize % EltSize) != 0)
```


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D43582/new/

https://reviews.llvm.org/D43582