https://bugs.freedesktop.org/show_bug.cgi?id=62357
--- Comment #10 from Roland Scheidegger <srol...@vmware.com> --- Hmm I just noticed this doesn't actually work. Instead of not executing the code after the conditional, the result will now always be as if the condition passed, so the original testcase now fails (and the IR clearly indicates llvm dropped all the code for the if condition (that is the comparison instructions are still there but that's just because the results of that are stored back to the register file). I can't quite see though why if this doesn't work when happening in main, how could it possibly work in a subroutine. I think there's something wrong with handling condition masks around function calls/rets. Can't quite see yet how this is all supposed to work together... -- You are receiving this mail because: You are the assignee for the bug.
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
