Selectively invert ICMP operands for better address optimization

src/IceTargetLoweringX86BaseImpl.h

 //===- subzero/src/IceTargetLoweringX86BaseImpl.h - x86 lowering -*- C++ -*-==//
 //
 //                        The Subzero Code Generator
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 ///
 /// \file
@@ skipping 458 matching lines @@
   // Run this early so it can be used to focus optimizations on potentially hot
   // code.
   // TODO(stichnot,ascull): currently only used for regalloc not
   // expensive high level optimizations which could be focused on potentially
   // hot code.
   Func->computeLoopNestDepth();
   Func->dump("After loop nest depth analysis");
 
   // Address mode optimization.
   Func->getVMetadata()->init(VMK_SingleDefs);
+  Func->invertICMP();
   Func->doAddressOpt();
   Func->materializeVectorShuffles();
 
   // Find read-modify-write opportunities. Do this after address mode
   // optimization so that doAddressOpt() doesn't need to be applied to RMW
   // instructions as well.
   findRMW();
   Func->dump("After RMW transform");
 
   // Argument lowering
@@ skipping 7543 matching lines @@
         emitGlobal(*Var, SectionSuffix);
       }
     }
   } break;
   }
 }
 } // end of namespace X86NAMESPACE
 } // end of namespace Ice
 
 #endif // SUBZERO_SRC_ICETARGETLOWERINGX86BASEIMPL_H