[PATCH] D63675: [InstCombine] Simplify icmp ult/uge (shl %x, C2), C1 iff C1 is power of two -> icmp eq/ne (and %x, (lshr -C1, C2)), 0.

[Sanjay Patel via Phabricator via llvm-commits]

spatel added inline comments.


================
Comment at: llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp:2041
+    }
+    // (X l<< C2) u</u>= C1 iff C1 is power of two -> X & (-C1 l>> C2) ==/!= 0
+    if (C.isPowerOf2() &&
----------------
lebedev.ri wrote:
> Like i commented in tests, it is not really possible to guarantely test those predicates.
> It seems like the predicates aren't being canonnicalized for vector compares, but that is likely a bug (CC @spatel)
> So it would be great if you could add vector test coverage for these, for now at least.
I haven't been following this set of patches closely, but lack of vector folds is still a fairly common problem. You can usually spot those limitations by searching for m_ConstantInt or dyn_cast to 'ConstantInt' rather than 'm_APInt'.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D63675/new/

https://reviews.llvm.org/D63675