Implement irreducible loop support in optimizing.

So we don't fallback to the interpreter in the presence of
irreducible loops.

Implications:
- A loop pre-header does not necessarily dominate a loop header.
- Non-constant redundant phis will be kept in loop headers, to
  satisfy our linear scan register allocation algorithm.
- while-graph optimizations, such as gvn, licm, lse, and dce
  need to know when they are dealing with irreducible loops.

Change-Id: I2cea8934ce0b40162d215353497c7f77d6c9137e

1 files changed, 6 insertions, 1 deletions

diff --git a/compiler/optimizing/bounds_check_elimination.cc b/compiler/optimizing/bounds_check_elimination.cc
index dc75ff1abc..da2f9cbed5 100644
--- a/compiler/optimizing/bounds_check_elimination.cc
+++ b/compiler/optimizing/bounds_check_elimination.cc
@@ -1295,8 +1295,13 @@ class BCEVisitor : public HGraphVisitor {
    */
   bool DynamicBCESeemsProfitable(HLoopInformation* loop, HBasicBlock* block) {
     if (loop != nullptr) {
+      // The loop preheader of an irreducible loop does not dominate all the blocks in
+      // the loop. We would need to find the common dominator of all blocks in the loop.
+      if (loop->IsIrreducible()) {
+        return false;
+      }
       // A try boundary preheader is hard to handle.
-      // TODO: remove this restriction
+      // TODO: remove this restriction.
       if (loop->GetPreHeader()->GetLastInstruction()->IsTryBoundary()) {
         return false;
       }