diff options
| author | 2015-07-22 17:50:37 +0100 | |
|---|---|---|
| committer | 2015-07-23 17:07:47 +0100 | |
| commit | d29e8487ff1774b6eb5f0e18d854415c1ee8f6b0 (patch) | |
| tree | 827ee7fedc275eb6801fcf200a36c06159e9f829 /compiler/dex/mir_optimization.cc | |
| parent | e6e38ce021ef5e0d326d76172307c000e0e6fab3 (diff) | |
ART: Fix Quick/Optimizing suspend check assumption mismatch.
Quick's SuspendCheckElimination (SCE) expects that every
method contains a suspend check and it eliminates suspend
checks in loops containing an invoke. Optimizing eliminates
the suspend check from leaf methods, so the combination of
a Quick-compiled loop calling an Optimizing-compiled leaf
method can lead to missing suspend checks and potentially
leading to ANRs.
Enable Quick's kLeafOptimization flag to remove suspend
checks from leaf methods and disable Quick's SCE. This
aligns the suspend check placement for the two backends
and avoids the broken combination.
Currently, all methods containing a try-catch are compiled
with Quick, so it's relatively easy to create a regression
test. However, this test will not be valid when Optimizing
starts supporting try-catch.
Bug: 22657404
Change-Id: I3bc40bf3f5c1e7d18704d1547b139e939950b770
Diffstat (limited to 'compiler/dex/mir_optimization.cc')
| -rw-r--r-- | compiler/dex/mir_optimization.cc | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/compiler/dex/mir_optimization.cc b/compiler/dex/mir_optimization.cc index 5bb0ce3ba5..80b7ac1e5b 100644 --- a/compiler/dex/mir_optimization.cc +++ b/compiler/dex/mir_optimization.cc @@ -1724,7 +1724,8 @@ void MIRGraph::StringChange() { bool MIRGraph::EliminateSuspendChecksGate() { - if ((cu_->disable_opt & (1 << kSuspendCheckElimination)) != 0 || // Disabled. + if (kLeafOptimization || // Incompatible (could create loops without suspend checks). + (cu_->disable_opt & (1 << kSuspendCheckElimination)) != 0 || // Disabled. GetMaxNestedLoops() == 0u || // Nothing to do. GetMaxNestedLoops() >= 32u || // Only 32 bits in suspend_checks_in_loops_[.]. // Exclude 32 as well to keep bit shifts well-defined. |