[PATCH] D22726: [DAGCombine] Match shift amount by value rather than relying on common sub-expressions.

[Hal Finkel via llvm-commits]

hfinkel added a subscriber: hfinkel.

================
Comment at: lib/CodeGen/SelectionDAG/DAGCombiner.cpp:4867
@@ -4860,3 +4866,3 @@
     }
   }
 
----------------
bryant wrote:
> bryant wrote:
> > RKSimon wrote:
> > > It's be great if you can generalise this to work with larger types - i128 for instance. 
> > that would change the intent of the original code. would it still belong in this patch?
> Also, `APInt::getAllOnesValue` only goes up to 2**64 - 1. So a patch for that would be needed beforehand?
> Also, APInt::getAllOnesValue only goes up to 2**64 - 1.

What do you mean by this? APInt::getAllOnesValue, AFAIK, handles all bit widths.



Repository:
  rL LLVM

https://reviews.llvm.org/D22726