optimizing: add block-scoped constructor fence merging pass
Introduce a new "Constructor Fence Redundancy Elimination" pass.
The pass currently performs local optimization only, i.e. within instructions
in the same basic block.
All constructor fences preceding a publish (e.g. store, invoke) get
merged into one instruction.
==============
OptStat#ConstructorFenceGeneratedNew: 43825
OptStat#ConstructorFenceGeneratedFinal: 17631 <+++
OptStat#ConstructorFenceRemovedLSE: 164
OptStat#ConstructorFenceRemovedPFRA: 9391
OptStat#ConstructorFenceRemovedCFRE: 16133 <---
Removes ~91.5% of the 'final' constructor fences in RitzBenchmark:
(We do not distinguish the exact reason that a fence was created, so
it's possible some "new" fences were also removed.)
==============
Test: art/test/run-test --host --optimizing 476-checker-ctor-fence-redun-elim
Bug: 36656456
Change-Id: I8020217b448ad96ce9b7640aa312ae784690ad99
diff --git a/compiler/optimizing/code_sinking.cc b/compiler/optimizing/code_sinking.cc
index 6c3a9fd..b558eb1 100644
--- a/compiler/optimizing/code_sinking.cc
+++ b/compiler/optimizing/code_sinking.cc
@@ -64,6 +64,11 @@
// A fence with "0" inputs is dead and should've been removed in a prior pass.
DCHECK_NE(0u, ctor_fence->InputCount());
+ // TODO: this should be simplified to 'return true' since it's
+ // potentially pessimizing any code sinking for inlined constructors with final fields.
+ // TODO: double check that if the final field assignments are not moved,
+ // then the fence is not moved either.
+
return ctor_fence->GetAssociatedAllocation() != nullptr;
}