[PATCH] D87850: [ConstantFold] Fold the comparison of bitcasted global values

Shimin Cui via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Oct 5 11:39:44 PDT 2020


scui added inline comments.


================
Comment at: llvm/test/Transforms/InstCombine/pr32686.ll:35
+  %lnot.ext = zext i1 %lnot to i32
+  %or = or i32 xor (i32 zext (i1 icmp ne (i32* bitcast (i8* @a2 to i32*), i32* @b) to i32), i32 2), %lnot.ext
+  store i32 %or, i32* @b, align 4
----------------
efriedma wrote:
> The "extra" version of this testcase isn't really helpful in this context; the version that folds will be folded away before instcombine can see it, so it's only testing constant folding itself.
> 
> If you want test coverage for additional constant expressions, please add them to the icmp folding test.
The added test cases in this file and the sccp tests are removed. One extra test case is added in the new test file.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D87850/new/

https://reviews.llvm.org/D87850



More information about the llvm-commits mailing list