search for: oanything

..., at 1:44 PMPDT, Bill Wendling wrote: > > -Os runs these passes more than -O2: -domfrontier -lcssa -loop-unroll Are you sure? That looks wrong; I don't think -Os should be doing loop unrolling. (Perhaps it should be on at -O2, as well. gcc's loop unrolling is not turned on by -Oanything, you have to use the switch explicitly, and that's how it's documented. But I think that's because its loop unrolling heuristics aren't all that good, and I don't think we should feel constrained to follow this if ours shows a win at -O2.)

On Wed, Oct 1, 2008 at 1:19 PM, Villmow, Micah <Micah.Villmow at amd.com> wrote: > LLVM seems to be generating way too complex of branching based on the > short-circuit optimization. The code in question is as follows: > > define void @ test_fc_while_and(float %x, float %y, float addrspace(11)* > %result) nounwind { > > entry: > > %tobool3 = fcmp une float

LLVM seems to be generating way too complex of branching based on the short-circuit optimization. The code in question is as follows: define void @ test_fc_while_and(float %x, float %y, float addrspace(11)* %result) nounwind { entry: %tobool3 = fcmp une float %x, 0.000000e+000 ; <i1> [#uses=1] %tobool24 = fcmp une float %y, 0.000000e+000 ;