[PATCH] D113212: [DAGCombiner] add fold for vselect of signbit

[Sanjay Patel via Phabricator via llvm-commits]

spatel marked an inline comment as done.
spatel added inline comments.


================
Comment at: llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp:9566
+    return DAG.getNode(ISD::AND, DL, VT, Sra, N1);
+  }
+
----------------
RKSimon wrote:
> Would it be useful to have the inverse of this if the target hasAndNot() ?
Yes - I hinted at this in the patch description and plan to follow-up with that and possibly a couple of other patterns based on the scalar folds. 

There are potentially several edge/weird cases though (my recent codegen patches for saturating sub were trying to avoid regressions on this set of folds). So I was planning to try these one at a time.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D113212/new/

https://reviews.llvm.org/D113212