[llvm-commits] [llvm] r115205 - in /llvm/trunk: lib/Transforms/Scalar/GVN.cpp test/Transforms/GVN/load-pre-licm.ll

Owen Anderson resistor at mac.com
Thu Sep 30 15:02:02 PDT 2010


On Sep 30, 2010, at 2:52 PM, Bob Wilson wrote:

> In the case when EnableFullLoadPre is true, we should still exclude the case where LI is the only definition.  I think you should keep this loop and just change the condition to "(e == 1)".

Is that really distinct from the fact that EnableFullLoadPRE just isn't worthwhile in general?  I'd rather just get rid of the option, since it seems pretty clear that we're not going to turn it on.

--Owen



More information about the llvm-commits mailing list