[PATCH] D12082: [X86] Do not lower scalar sdiv/udiv to a shifts + mul sequence when optimizing for minsize

[Sanjay Patel via llvm-commits]

spatel added a comment.

Since we're making these virtual functions, can we have them return 'false' by default and eliminate the member variables (IntDivIsCheap, Pow2SDivIsCheap) in TargetLowering? It doesn't look like anything in tree even overrides these, so changes would be minimal.


http://reviews.llvm.org/D12082