[x86] avoid flipping sign bits for vector icmp by using known bits

If we know that both operands of an unsigned integer vector comparison are non-negative, 
then it's safe to directly use a signed-compare-greater-than instruction (the only non-equality
integer vector compare predicate provided by SSE/AVX).

We're intentionally not changing the condition code to signed in order to preserve the
existing transforms that use min/max/psubus below here.

This should solve PR33276:
https://bugs.llvm.org/show_bug.cgi?id=33276

Differential Revision: https://reviews.llvm.org/D33862

llvm-svn: 304909

1 files changed, 7 insertions, 1 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 246f2e91eb7..0ecd23e9334 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -17184,7 +17184,13 @@ static SDValue LowerVSETCC(SDValue Op, const X86Subtarget &Subtarget,
               Cond == ISD::SETGE || Cond == ISD::SETUGE;
   bool Invert = Cond == ISD::SETNE ||
                 (Cond != ISD::SETEQ && ISD::isTrueWhenEqual(Cond));
-  bool FlipSigns = ISD::isUnsignedIntSetCC(Cond);
+
+  // If both operands are known non-negative, then an unsigned compare is the
+  // same as a signed compare and there's no need to flip signbits.
+  // TODO: We could check for more general simplifications here since we're
+  // computing known bits.
+  bool FlipSigns = ISD::isUnsignedIntSetCC(Cond) &&
+                   !(DAG.SignBitIsZero(Op0) && DAG.SignBitIsZero(Op1));
 
   // Special case: Use min/max operations for SETULE/SETUGE
   MVT VET = VT.getVectorElementType();