Don't break critical edges for single-bb loops, this helps with PR1877, though

it is only a partial fix.  This change is noise for most programs, but
speeds up Shootout-C++/matrix by 20%, Ptrdist/ks by 24%, smg2000 by 8%,
hexxagon by 9%, bzip2 by 9% (not sure I trust this), ackerman by 13%, etc.

OTOH, it slows down Shootout/fib2 by 40% (I'll update PR1877 with this info).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@45354 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Scalar/CodeGenPrepare.cpp b/lib/Transforms/Scalar/CodeGenPrepare.cpp
index f31183caf8cec8e25d7fc35215e6d2629707f44a..954e3286184f3eff34cdfd39cb043ed2ec97059d 100644 (file)
--- a/lib/Transforms/Scalar/CodeGenPrepare.cpp
+++ b/lib/Transforms/Scalar/CodeGenPrepare.cpp
@@ -37,8 +37,6 @@ using namespace llvm;
 namespace {
   cl::opt<bool> OptExtUses("optimize-ext-uses",
                            cl::init(true), cl::Hidden);
-  // LLCBETA option.
-  cl::opt<bool> DontHackBackedge("backedge-hack", cl::Hidden);
 }
 
 namespace {  
@@ -280,7 +278,7 @@ static void SplitEdgeNicely(TerminatorInst *TI, unsigned SuccNum, Pass *P) {
   // As a hack, never split backedges of loops.  Even though the copy for any
   // PHIs inserted on the backedge would be dead for exits from the loop, we
   // assume that the cost of *splitting* the backedge would be too high.
-  if (DontHackBackedge && Dest == TIBB)
+  if (Dest == TIBB)
     return;
   
   /// TIPHIValues - This array is lazily computed to determine the values of