[PATCH] D35722: [DAGCombine] Improve (sra (sra x, c1), c2) -> (sra x, (add c1, c2)) folding

[Andrew Zhogin via Phabricator via llvm-commits]

andrew.zhogin added a comment.

Isn't it a little regression for scalar types?

  N1.getValueType() == N0.getOperand(1).getValueType()

Instead of

  zeroExtendToMatch(c1, c2, 1 /* Overflow Bit */);

There is no vector analog of zeroExtendToMatch currently, right?

Btw, I don't think I have rights to accept the revision.


Repository:
  rL LLVM

https://reviews.llvm.org/D35722