OSDN Git Service
Andreas Gampe [Tue, 7 Apr 2015 00:17:34 +0000 (17:17 -0700)]
ART: Fix noreturn for Mac
Change-Id: I59b60340a90672a4de4d3bdf092fa504394d5892
Andreas Gampe [Mon, 6 Apr 2015 21:44:50 +0000 (21:44 +0000)]
Merge "ART: Make unknown variant "mips64r6" go away"
Andreas Gampe [Mon, 6 Apr 2015 21:40:36 +0000 (14:40 -0700)]
ART: Make unknown variant "mips64r6" go away
It's the standard variant, accept it just like "default."
Change-Id: I9b600345e3e1ebfb508ffe22ddde17757be2172e
Andreas Gampe [Mon, 6 Apr 2015 21:34:51 +0000 (21:34 +0000)]
Merge "ART: Fix no-return in sigchainlib"
Andreas Gampe [Mon, 6 Apr 2015 21:33:54 +0000 (14:33 -0700)]
ART: Fix no-return in sigchainlib
Ignore warning in sigchain_dummy.
Change-Id: I2c9ed93c0125d06b6a3d4011a794b062cf1c4093
Andreas Gampe [Mon, 6 Apr 2015 21:21:05 +0000 (21:21 +0000)]
Merge "ART: Enable more Clang warnings"
Andreas Gampe [Mon, 6 Apr 2015 16:35:22 +0000 (09:35 -0700)]
ART: Enable more Clang warnings
Change-Id: Ie6aba02f4223b1de02530e1515c63505f37e184c
Andreas Gampe [Mon, 6 Apr 2015 21:04:16 +0000 (21:04 +0000)]
Merge "ART: Enable Clang's -Wdeprecated"
Andreas Gampe [Sat, 4 Apr 2015 04:28:42 +0000 (21:28 -0700)]
ART: Enable Clang's -Wdeprecated
Replace throw() with noexcept.
Add default copy constructors and copy assignment constructors for
cases with destructors, as the implicit definition is deprecated.
Change-Id: Ice306a3f510b072b00bec4d4360f7c8055135c9d
Andreas Gampe [Mon, 6 Apr 2015 20:36:03 +0000 (20:36 +0000)]
Merge "ART: Add Clang's -Wused-but-marked-unused"
Mathieu Chartier [Mon, 6 Apr 2015 17:59:37 +0000 (17:59 +0000)]
Merge "Refactor and improve GC root handling"
Mathieu Chartier [Fri, 3 Apr 2015 18:21:55 +0000 (11:21 -0700)]
Refactor and improve GC root handling
Changed GcRoot to use compressed references. Changed root visiting to
use virtual functions instead of function pointers. Changed root visting
interface to be an array of roots instead of a single root at a time.
Added buffered root marking helper to avoid dispatch overhead.
Root marking seems a bit faster on EvaluateAndApplyChanges due to batch
marking. Pause times unaffected.
Mips64 is untested but might work, maybe.
Before:
MarkConcurrentRoots: Sum: 67.678ms 99% C.I. 2us-664.999us Avg: 161.138us Max: 671us
After:
MarkConcurrentRoots: Sum: 54.806ms 99% C.I. 2us-499.986us Avg: 136.333us Max: 602us
Bug:
19264997
Change-Id: I0a71ebb5928f205b9b3f7945b25db6489d5657ca
Vladimir Marko [Mon, 6 Apr 2015 09:01:34 +0000 (09:01 +0000)]
Merge "Pass linker patches around as const."
Mathieu Chartier [Sun, 5 Apr 2015 20:21:11 +0000 (20:21 +0000)]
Merge "Delete missed ClampGrowthLimit CHECK"
Andreas Gampe [Sat, 4 Apr 2015 02:41:34 +0000 (19:41 -0700)]
ART: Add Clang's -Wused-but-marked-unused
Add detection of wrong unused annotations. Fix our codebase.
Change-Id: I85cc20f2eac71c1ec6c5c7cd6efb08454a629634
Mathieu Chartier [Sat, 4 Apr 2015 02:08:10 +0000 (19:08 -0700)]
Delete missed ClampGrowthLimit CHECK
Bug:
20043461
Change-Id: If4a0bb91c64571945756151be162f3626534781c
Mingyao Yang [Sat, 4 Apr 2015 01:27:19 +0000 (01:27 +0000)]
Merge "Add missing RecordPcInfo's for intrinsics_x86."
Mingyao Yang [Sat, 4 Apr 2015 00:56:54 +0000 (17:56 -0700)]
Add missing RecordPcInfo's for intrinsics_x86.
Change-Id: I5ad856e57f63e2bd350f62e7023911c276427edd
Andreas Gampe [Fri, 3 Apr 2015 21:42:06 +0000 (21:42 +0000)]
Merge "ART: Turn of native stack dump on x86 on-device"
Andreas Gampe [Fri, 3 Apr 2015 21:36:07 +0000 (14:36 -0700)]
ART: Turn of native stack dump on x86 on-device
Temporary workaround for bug.
Bug:
20040863
Change-Id: Ifdde68c66b94ebc2f13c7983f38700c8bcf41559
Hiroshi Yamauchi [Fri, 3 Apr 2015 21:06:51 +0000 (21:06 +0000)]
Merge "Fix the read barrier builds."
Hiroshi Yamauchi [Fri, 3 Apr 2015 20:35:16 +0000 (13:35 -0700)]
Fix the read barrier builds.
A check failure (!UseJit() in ArtMethod::Invoke()) in forced interpret
only mode in the eng build with JIT enabled by default on target.
A build failure (inl header file issues) in the USE_BAKER_READ_BARRIER
build.
Change-Id: Ib13632b10864ad8b4691b66971c0ab5d2ae1e675
Sebastien Hertz [Fri, 3 Apr 2015 18:12:22 +0000 (18:12 +0000)]
Merge "Use va_list argument to abort transaction"
Dmitriy Ivanov [Fri, 3 Apr 2015 17:49:53 +0000 (17:49 +0000)]
Merge "Delegate SIG_DFL on sigaction to libc."
Richard Uhler [Fri, 3 Apr 2015 17:47:14 +0000 (17:47 +0000)]
Merge "Store OatDexFile instead of OatFile in DexFile."
Dmitriy Ivanov [Fri, 3 Apr 2015 01:30:22 +0000 (18:30 -0700)]
Delegate SIG_DFL on sigaction to libc.
In the case when user registers SIG_DFL hanlder
sigchainlib's handler tends to go into infinite loop
because the function handling signals resets signal
using by calling to signchain's own implementation.
This cl fixes the problem by passing sigcations with
SIG_DFL to the next sigaction (usually libc's).
Bug:
19594886
Change-Id: I9eecf9afd1c7e6d1fe3cd1d4fc506383ecbebe04
Nicolas Geoffray [Fri, 3 Apr 2015 17:08:52 +0000 (17:08 +0000)]
Merge "Fix a bug when creating a HDeoptimization instruction."
Andreas Gampe [Fri, 3 Apr 2015 16:48:26 +0000 (16:48 +0000)]
Merge "ART: Avoid obvious segfault in arm & x86 fault handler"
Elliott Hughes [Fri, 3 Apr 2015 15:37:00 +0000 (15:37 +0000)]
Merge "Explicitly specify the C libraries that have mallinfo(3)."
Sebastien Hertz [Fri, 3 Apr 2015 14:07:05 +0000 (16:07 +0200)]
Use va_list argument to abort transaction
Creates AbortTransactionV taking a va_list argument and renames
AbortTransaction to AbortTransactionF which calls AbortTransactionV.
This fixes the compiler_driver_test under valgrind.
Change-Id: Ia1c57330091c055ae9e46585a944ce0b78864920
Nicolas Geoffray [Fri, 3 Apr 2015 10:02:38 +0000 (11:02 +0100)]
Fix a bug when creating a HDeoptimization instruction.
We need to copy the environment, instead of just pointing
to an existing one. Otherwise, if the instruction that initially
holds the environemnt gets removed from the graph, any update
to an instruction in that environment will not be reflected in it.
bug:
20058506
Change-Id: I2a62476d0851ecbc3707c0da395d8502ee437422
Andreas Gampe [Fri, 3 Apr 2015 09:17:06 +0000 (02:17 -0700)]
ART: Avoid obvious segfault in arm & x86 fault handler
These handlers need to read an instruction to determine where to
look for info. Don't try to read from pc=0.
Bug:
20040863
Change-Id: I38b56dc6dd806df22e608ee8d46c4091a738e4bc
Sebastien Hertz [Fri, 3 Apr 2015 08:00:37 +0000 (08:00 +0000)]
Merge "Revert "ART: Workaround for b/
20019689""
Sebastien Hertz [Thu, 2 Apr 2015 09:09:04 +0000 (09:09 +0000)]
Revert "ART: Workaround for b/
20019689"
This reverts commit
328dd4457632f59e67d08d9d086670b56141a4b1.
Bug:
20019689
Change-Id: Ia1087e45d1a1817b6ff4e45f8d74b84d34f078ee
Sebastien Hertz [Fri, 3 Apr 2015 07:49:44 +0000 (07:49 +0000)]
Merge "Use specific exception class to abort transaction"
Sebastien Hertz [Thu, 2 Apr 2015 09:11:19 +0000 (11:11 +0200)]
Use specific exception class to abort transaction
We used to throw a java.lang.InternalError when aborting a
transaction (when preinitializing image classes at compilation time).
We now use dedicated class dalvik.system.TransactionAbortError that
is only thrown by the compiler to abort a transaction. This class has
constructors taking a java.lang.Throwable "cause" so we can wrap
exceptions causing the transaction to abort (for instance class
java.lang.ClassNotFoundException) and give more information about the
cause of the transaction abort.
Bug:
20019689
Change-Id: I019a72a1c754d8bba6a7ad6bb0f02e4fd6668622
Elliott Hughes [Fri, 3 Apr 2015 03:55:07 +0000 (20:55 -0700)]
Explicitly specify the C libraries that have mallinfo(3).
Change-Id: Ib6d22134cb246e30452a7ed9acabbde42b3ef091
Andreas Gampe [Fri, 3 Apr 2015 02:03:34 +0000 (02:03 +0000)]
Merge "ART: Move trace options to a struct"
Andreas Gampe [Fri, 3 Apr 2015 01:51:05 +0000 (18:51 -0700)]
ART: Move trace options to a struct
Bundles the different options and moves them to a struct in the
implementation file. Will allow for easier change for streaming.
Change-Id: I486e0942614e714964eb233e2d6b4e3e12aa704f
Andreas Gampe [Thu, 2 Apr 2015 23:48:54 +0000 (23:48 +0000)]
Merge "Revert "ART: Valgrind hotfix for VIXL 1.9""
Andreas Gampe [Thu, 2 Apr 2015 23:10:46 +0000 (23:10 +0000)]
Merge "[optimizing] Implement x86/x86_64 math intrinsics"
Andreas Gampe [Thu, 2 Apr 2015 22:49:38 +0000 (22:49 +0000)]
Merge "Fix for issue
20051601 - ARM instruction alignment."
Dmitriy Ivanov [Thu, 2 Apr 2015 22:33:25 +0000 (22:33 +0000)]
Merge "Remove version-script option"
Dmitriy Ivanov [Wed, 1 Apr 2015 18:17:46 +0000 (11:17 -0700)]
Remove version-script option
Version script was used to limit number of symbols
exposed by main executable because of --export-dynamic
option. This was overthinking on my part which can
lead to unwanted side-effects like not working asan.
Change-Id: Iaab4d38f86ddf21357d26d21d35c9ed6e7ff989e
Vladimir Marko [Thu, 2 Apr 2015 20:25:21 +0000 (21:25 +0100)]
Pass linker patches around as const.
Change-Id: I0eabd713d29475db9eb6e186f331dbfb00e0cf6b
Vladimir Marko [Thu, 2 Apr 2015 19:51:49 +0000 (19:51 +0000)]
Merge "Split RelativePatcher::ReserveSpace() into two."
David Srbecky [Thu, 2 Apr 2015 19:38:40 +0000 (20:38 +0100)]
Fix for issue
20051601 - ARM instruction alignment.
The instructions do not have to be 4 byte aligned.
Change-Id: I58078a15190b835601094611ed01a599e2d99563
Vladimir Marko [Thu, 2 Apr 2015 18:45:06 +0000 (19:45 +0100)]
Split RelativePatcher::ReserveSpace() into two.
Instead of passing nullptr and MethodReference(nullptr, 0u)
to ReserveSpace() at the end, call a newly created function
ReserveSpaceEnd().
Change-Id: I38815fe9464b4e1a919878b6e8577614f1058d00
Vladimir Marko [Thu, 2 Apr 2015 19:07:17 +0000 (19:07 +0000)]
Merge "Add tests for Arm64RelativePatcher."
Mathieu Chartier [Thu, 2 Apr 2015 18:11:02 +0000 (18:11 +0000)]
Merge "Fix race with Heap::ClampGrowthLimit and GC"
Mathieu Chartier [Thu, 2 Apr 2015 17:08:03 +0000 (10:08 -0700)]
Fix race with Heap::ClampGrowthLimit and GC
Aded logic for handling the temp bitmap if the GC is running and
the live bitmap is clamped to the mark bitmap. This fixes the SIGABRT
from ClampGrowthLimit if the GC clamped the bitmaps at this point.
Also added locking of the heap_bitmap_lock_ so that added a lock so
that the temp bitmap doesn't change from underneath us.
Bug:
20043461
Change-Id: Ib427e40bcdf149de0408b4b53e6524f51463f0af
Vladimir Marko [Thu, 2 Apr 2015 14:28:45 +0000 (15:28 +0100)]
Add tests for Arm64RelativePatcher.
Change-Id: I9d2c21d323137ac143eabb8fdf6ca075bae45c51
Richard Uhler [Tue, 31 Mar 2015 22:57:54 +0000 (15:57 -0700)]
Store OatDexFile instead of OatFile in DexFile.
This requires moving OatDexFile out of the OatFile class so that
a forward class declaration can be used for OatDexFile.
Bug:
19071355
Change-Id: Ibda85b78d0577e9e81073090616fc0f2fa526be3
Andreas Gampe [Thu, 2 Apr 2015 16:41:58 +0000 (16:41 +0000)]
Merge "Update dexfuzz after optimizing default selection."
Serban Constantinescu [Thu, 2 Apr 2015 11:27:30 +0000 (12:27 +0100)]
Revert "ART: Valgrind hotfix for VIXL 1.9"
Fixed in external/vixl and upstream VIXL.
Change-Id: I7d1901da06a7c1517c88a2583bc668a3e23ef852
Vladimir Marko [Thu, 2 Apr 2015 15:46:31 +0000 (15:46 +0000)]
Merge "Add tests for Thumb2RelativePatcher."
Vladimir Marko [Wed, 1 Apr 2015 22:03:09 +0000 (23:03 +0100)]
Add tests for Thumb2RelativePatcher.
Also make the thumb2/arm64 thunk allocation precise instead
of eagerly allocating thunk space. This allows the calls to
use the maximum positive offset.
Change-Id: Ifa95b0bb00bd73eeab0c2905d21e2f3078f4b0a8
David Srbecky [Thu, 2 Apr 2015 14:12:01 +0000 (14:12 +0000)]
Merge "Write .debug_line section using the new DWARF library."
Vladimir Marko [Thu, 2 Apr 2015 12:46:16 +0000 (12:46 +0000)]
Merge "Refactor RelativePatcher out of OatWriter."
Vladimir Marko [Thu, 2 Apr 2015 12:44:49 +0000 (12:44 +0000)]
Merge "PC-relative loads from dex cache arrays for x86-64."
Stephen Kyle [Fri, 27 Mar 2015 18:50:10 +0000 (18:50 +0000)]
Update dexfuzz after optimizing default selection.
Now that the optimizing compiler is the default backend, the quick
compiler must be explicitly selected as an argument.
Change-Id: I5e6b8e6a79432118e0a91c9aea85bdd9cd7305dc
Vladimir Marko [Tue, 31 Mar 2015 20:49:49 +0000 (21:49 +0100)]
Refactor RelativePatcher out of OatWriter.
Move the relative patcher classes to compiler/linker/ and
compiler/linker/<arch>/ . Refactor them to avoid OatWriter
dependency so that they can be unit tested. Add tests for
x86 and x86-64.
Change-Id: I1b42baa9fc431378e4cce1399bec590c5b5a409f
Roland Levillain [Thu, 2 Apr 2015 11:36:15 +0000 (11:36 +0000)]
Merge "Move the "huge method" test into its own test case."
Roland Levillain [Thu, 2 Apr 2015 10:24:43 +0000 (11:24 +0100)]
Move the "huge method" test into its own test case.
Change-Id: Icd7c95d2a7a0cacf15f079eed38c5d1a8f138b3f
Vladimir Marko [Fri, 27 Mar 2015 18:18:36 +0000 (18:18 +0000)]
PC-relative loads from dex cache arrays for x86-64.
Change-Id: I6cfe22c7e69512b3c0f95b073aaa572db74ec189
Brian Carlstrom [Thu, 2 Apr 2015 07:51:06 +0000 (07:51 +0000)]
Merge "Switch to shared libcutils on host to fix odr-violation"
Brian Carlstrom [Thu, 2 Apr 2015 07:04:09 +0000 (00:04 -0700)]
Switch to shared libcutils on host to fix odr-violation
Change-Id: Ia8eb950806a09e12fec2956d51eff99852920bee
SUMMARY: AddressSanitizer: odr-violation: global atrace_is_ready at system/core/libcutils/trace-host.c:23:25
Mingyao Yang [Wed, 1 Apr 2015 23:15:52 +0000 (23:15 +0000)]
Merge "Revert "Revert "Deoptimization-based bce."""
Mingyao Yang [Wed, 1 Apr 2015 22:08:37 +0000 (22:08 +0000)]
Merge "CanThrow() of HNewArray should return true."
Mingyao Yang [Wed, 1 Apr 2015 22:07:15 +0000 (22:07 +0000)]
Merge "Clean up some RecordPcInfo in x86 slow paths."
Mingyao Yang [Tue, 31 Mar 2015 22:09:29 +0000 (15:09 -0700)]
CanThrow() of HNewArray should return true.
Change-Id: I9950f1c391dfeb26cf59cee769705d01d8e283d5
Andreas Gampe [Wed, 1 Apr 2015 21:54:08 +0000 (21:54 +0000)]
Merge "ART: Fix unstarted runtime"
Andreas Gampe [Wed, 1 Apr 2015 21:49:06 +0000 (14:49 -0700)]
ART: Fix unstarted runtime
Correctly delete local refs for dex/direct-byte-buffer.
Change-Id: Ia74dfafb959d69aa281af7316f393997e053a3b3
Andreas Gampe [Wed, 1 Apr 2015 21:29:46 +0000 (21:29 +0000)]
Merge "ART: Workaround for b/
20019689"
Andreas Gampe [Wed, 1 Apr 2015 21:28:09 +0000 (14:28 -0700)]
ART: Workaround for b/
20019689
Don't throw InternalError as wrapped for now, swallowing the
original exception.
Bug:
20019689
Change-Id: I77a56262304b09b326232bee0c7916bb73401d74
Nicolas Geoffray [Wed, 1 Apr 2015 21:27:40 +0000 (21:27 +0000)]
Merge "Opt compiler: Fix move from constant."
Mingyao Yang [Wed, 1 Apr 2015 21:03:04 +0000 (14:03 -0700)]
Revert "Revert "Deoptimization-based bce.""
This reverts commit
0ba627337274ccfb8c9cb9bf23fffb1e1b9d1430.
Change-Id: I1ca10d15bbb49897a0cf541ab160431ec180a006
Andreas Gampe [Wed, 1 Apr 2015 19:50:15 +0000 (19:50 +0000)]
Merge "ART: Set default prune counter to 10"
Calin Juravle [Wed, 1 Apr 2015 19:48:39 +0000 (19:48 +0000)]
Merge "Adds host execution to dexfuzz."
Roland Levillain [Wed, 1 Apr 2015 19:06:33 +0000 (19:06 +0000)]
Merge "Fix test 421-large-frame."
Roland Levillain [Wed, 1 Apr 2015 18:54:39 +0000 (19:54 +0100)]
Fix test 421-large-frame.
- Don't force the optimizing compiler to compile a huge
method by reducing the number of local variables and
instructions in Main.$opt$LargeFrame.
- Prevent the optimizing compiler from "optimizing too much"
(through constant folding and inlining), so as to preserve
a large stack frame in Main.$opt$LargeFrame.
- Keep the original huge method (without the $opt$ marker)
as Main.HugeFrame.
Bug:
20020727
Change-Id: I49bf8f03146020bd94bd392684e605d7bb6daf1e
Dmitriy Ivanov [Wed, 1 Apr 2015 18:25:34 +0000 (18:25 +0000)]
Merge "Fix formatting"
Dmitriy Ivanov [Wed, 1 Apr 2015 18:08:45 +0000 (11:08 -0700)]
Fix formatting
Change-Id: Id64aad5de1f911ac7e747788d586bac397f34f18
David Brazdil [Wed, 1 Apr 2015 17:42:59 +0000 (17:42 +0000)]
Merge "ART: Regression test for condition materialization bug"
David Brazdil [Wed, 1 Apr 2015 17:34:33 +0000 (18:34 +0100)]
ART: Regression test for condition materialization bug
This adds a regression test previously omitted in a hotfix.
Bug:
19999189
Change-Id: I02acb935960f22028bf877cd270b04e17515068f
David Brazdil [Wed, 1 Apr 2015 16:07:58 +0000 (16:07 +0000)]
Merge "ART: Materialize constants on environment use"
David Brazdil [Wed, 1 Apr 2015 16:00:59 +0000 (17:00 +0100)]
ART: Materialize constants on environment use
The decision whether to materialize constants assumed that they cannot
be used by an environment. This changed with introduction of the
boolean simplifier and manifested itself when environment uses became
included in liveness analysis. This patch fixes the issue.
This is a quick fix to unblock the build, test will follow.
Bug:
19999189
Change-Id: Iedb1e3a32073a75b9c0d041a88b8ae640e064d13
Mark Mendell [Wed, 1 Apr 2015 02:16:59 +0000 (22:16 -0400)]
[optimizing] Implement x86/x86_64 math intrinsics
Implement floor/ceil/round/RoundFloat on x86 and x86_64.
Implement RoundDouble on x86_64.
Add support for roundss and roundsd on both architectures. Support them
in the disassembler as well.
Add the instruction set features for x86, as the 'round' instruction is
only supported if SSE4.1 is supported.
Fix the tests to handle the addition of passing the instruction set
features to x86 and x86_64.
Add assembler tests for roundsd and roundss to x86_64 assembler tests.
Change-Id: I9742d5930befb0bbc23f3d6c83ce0183ed9fe04f
Signed-off-by: Mark Mendell <mark.p.mendell@intel.com>
David Srbecky [Mon, 30 Mar 2015 13:21:42 +0000 (14:21 +0100)]
Write .debug_line section using the new DWARF library.
Also simplify dex to java mapping and handle mapping
in prologues and epilogues.
Change-Id: I410f06024580f2a8788f2c93fe9bca132805029a
David Brazdil [Wed, 1 Apr 2015 09:03:28 +0000 (09:03 +0000)]
Merge "ART: Same skip-compilation heuristic in both backends"
Andreas Gampe [Wed, 1 Apr 2015 07:39:15 +0000 (07:39 +0000)]
Merge "ART: Valgrind hotfix for VIXL 1.9"
Andreas Gampe [Wed, 1 Apr 2015 07:36:42 +0000 (00:36 -0700)]
ART: Valgrind hotfix for VIXL 1.9
Make sure recommended_checkpoint_ is initialized in the VIXL macro
assembler by calling EmitLiteralPool with an empty pool.
Change-Id: I08589b8fb092a33a8f4aad824e91b5c16ff761b6
Andreas Gampe [Wed, 1 Apr 2015 03:51:24 +0000 (03:51 +0000)]
Merge "ARM64: Update to VIXL 1.9."
Andreas Gampe [Tue, 31 Mar 2015 21:54:03 +0000 (14:54 -0700)]
ART: Set default prune counter to 10
By default, prune cache at boot after ten failed boots.
Bug:
19983101
Change-Id: Ib88f807e0082d71292c14c7af38e02cca5a5602c
Mingyao Yang [Wed, 1 Apr 2015 00:03:08 +0000 (17:03 -0700)]
Clean up some RecordPcInfo in x86 slow paths.
Calling codegen->RecordPcInfo() is likely to miss the third argument which
is the slow path. And can cause deopt bugs later.
The change calls the slow path's version of RecordPcInfo() in slow paths
consistently.
Change-Id: I41605f1b06ee2c6d3d7ffd5aa0c1366940b5a029
Mathieu Chartier [Tue, 31 Mar 2015 23:40:11 +0000 (23:40 +0000)]
Merge "Only GC the first time we call PreZygoteFork"
Mathieu Chartier [Tue, 31 Mar 2015 23:28:53 +0000 (16:28 -0700)]
Only GC the first time we call PreZygoteFork
PreZygoteFork is called for every app launch, it is overkill to GC
each time since it takes 20-30ms and blocks app launch for that long.
Change-Id: I647c8ccca767ceca67a006c1d80a739c7860f0d0
Mathieu Chartier [Tue, 31 Mar 2015 23:20:41 +0000 (23:20 +0000)]
Merge "Visit image roots for hprof"
Mathieu Chartier [Tue, 31 Mar 2015 19:05:24 +0000 (12:05 -0700)]
Visit image roots for hprof
Bug:
19995360
(cherry picked from commit
0cab5e68f14ee403380664146db6dc7ddfc32064)
Change-Id: I17868bff2a701dc25291d41b9732c6b86f92be08
Nicolas Geoffray [Tue, 31 Mar 2015 22:07:29 +0000 (22:07 +0000)]
Merge "Fix lint error."