Don't PRE compares.

CodeGenPrepare sinks compare instructions down to their uses to prevent
live flags and predicate registers across basic blocks.

PRE of a compare instruction prevents that, forcing the i1 compare
result into a general purpose register.  That is usually more expensive
than the redundant compare PRE was trying to eliminate in the first
place.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@153657 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Transforms/Scalar/GVN.cpp

@@ -2328,7 +2328,14 @@ bool GVN::performPRE(Function &F) {
           CurInst->mayReadFromMemory() || CurInst->mayHaveSideEffects() ||
           isa<DbgInfoIntrinsic>(CurInst))
         continue;
-      
+
+      // Don't do PRE on compares. The PHI would prevent CodeGenPrepare from
+      // sinking the compare again, and it would force the code generator to
+      // move the i1 from processor flags or predicate registers into a general
+      // purpose register.
+      if (isa<CmpInst>(CurInst))
+        continue;
+
       // We don't currently value number ANY inline asm calls.
       if (CallInst *CallI = dyn_cast<CallInst>(CurInst))
         if (CallI->isInlineAsm())