Enable class unloading in dex2oat for quicken multidex
Enable class unloading for the quicken compilation filter. This
reduces RAM used by dex2oat when compiling multidex apps with the
quicken filter.
Unit test is already in dex2oat_test.
Peak dex2oat PSS according to procrank for a large app:
243247K -> 147706K for quicken filter.
Bug: 63467744
Test: test-art-host
Test: test/testrunner/testrunner.py --interpreter --host -j40
(cherry picked from commit 6a2b67beea9d1a59f7fc6d7f58c52ad1773da1c6)
Change-Id: I476ee76b7c9efb57238673a117dcb94bbc09f430
diff --git a/dex2oat/dex2oat.cc b/dex2oat/dex2oat.cc
index 64268d9..80a194a 100644
--- a/dex2oat/dex2oat.cc
+++ b/dex2oat/dex2oat.cc
@@ -1769,11 +1769,11 @@
bool ShouldCompileDexFilesIndividually() const {
// Compile individually if we are not building an image, not using any compilation, and are
// using multidex.
- // This means extract, and verify, will use the individual compilation mode (to reduce RAM used
- // by the compiler).
+ // This means extract, verify, and quicken, will use the individual compilation mode (to reduce
+ // RAM used by the compiler).
return !IsImage() &&
dex_files_.size() > 1 &&
- !CompilerFilter::IsAnyCompilationEnabled(compiler_options_->GetCompilerFilter());
+ !CompilerFilter::IsAotCompilationEnabled(compiler_options_->GetCompilerFilter());
}
// Set up and create the compiler driver and then invoke it to compile all the dex files.
