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, 3 insertions, 1 deletions

diff --git a/compiler/optimizing/induction_var_analysis.cc b/compiler/optimizing/induction_var_analysis.cc
index eef6cef5f0..37f2d79536 100644
--- a/compiler/optimizing/induction_var_analysis.cc
+++ b/compiler/optimizing/induction_var_analysis.cc
@@ -76,7 +76,9 @@ void HInductionVarAnalysis::Run() {
   // range analysis on outer loop while visiting inner loops.
   for (HReversePostOrderIterator it_graph(*graph_); !it_graph.Done(); it_graph.Advance()) {
     HBasicBlock* graph_block = it_graph.Current();
-    if (graph_block->IsLoopHeader()) {
+    // Don't analyze irreducible loops.
+    // TODO(ajcbik): could/should we remove this restriction?
+    if (graph_block->IsLoopHeader() && !graph_block->GetLoopInformation()->IsIrreducible()) {
       VisitLoop(graph_block->GetLoopInformation());
     }
   }