[PATCH] D25485: [DAG] optimize negation of bool

Simon Dardis via llvm-commits llvm-commits at lists.llvm.org
Mon Oct 17 15:32:39 PDT 2016

sdardis added a comment.


Looking at the existing test output for division by i1 leads me to believe they're correct but perhaps not optimal in the sense of "division by zero => undefined behaviour".  Taking that view, division of an i1 by an i1 should yield the numerator always, as for MIPS division by zero yields an undefined result and hence division by an i1 should be folded away (unless -mcheck-zero-division is active).

I'll investigate that issue in a bit, but it's an optimization issue, not a correctness issue.



More information about the llvm-commits mailing list