ART: Implement scalar loop peeling.

Implement scalar loop peeling for invariant exits elimination
(on arm64). If the loop exit condition is loop invariant then
loop peeling + GVN + DCE can eliminate this exit in the loop
body. Note: GVN and DCE aren't applied during loop optimizations.

Note: this functionality is turned off by default now.

Test: test-art-host, test-art-target, boot-to-gui.

Change-Id: I98d20054a431838b452dc06bd25c075eb445960c

1 files changed, 4 insertions, 9 deletions

diff --git a/compiler/optimizing/loop_optimization.h b/compiler/optimizing/loop_optimization.h
index 0120cffa56..f9a31a34d4 100644
--- a/compiler/optimizing/loop_optimization.h
+++ b/compiler/optimizing/loop_optimization.h
@@ -145,19 +145,14 @@ class HLoopOptimization : public HOptimization {
   // Performs optimizations specific to inner loop. Returns true if anything changed.
   bool OptimizeInnerLoop(LoopNode* node);
 
-  // Performs loop peeling/unrolling once (depends on the 'do_unrolling'); the transformation
-  // preserves the header and the loop info.
-  //
-  // Note: the function records copying information about blocks and instructions.
-  void PeelOrUnrollOnce(LoopNode* loop_node,
-                        bool do_unrolling,
-                        SuperblockCloner::HBasicBlockMap* bb_map,
-                        SuperblockCloner::HInstructionMap* hir_map);
-
   // Tries to apply loop unrolling for branch penalty reduction and better instruction scheduling
   // opportunities. Returns whether transformation happened.
   bool TryUnrollingForBranchPenaltyReduction(LoopNode* loop_node);
 
+  // Tries to apply loop peeling for loop invariant exits elimination. Returns whether
+  // transformation happened.
+  bool TryPeelingForLoopInvariantExitsElimination(LoopNode* loop_node);
+
   //
   // Vectorization analysis and synthesis.
   //