OSDN Git Service
Sebastien Hertz [Thu, 28 Aug 2014 12:41:50 +0000 (14:41 +0200)]
Don't hold any lock when visiting classes from JDWP
Computes reference type ids of all loaded classes without holding the class
linker lock. Because computing the JDWP reference type id can cause thread
suspension, we can't hold any lock. This is detected in debug build (using
libartd.so) and causes an abort.
Also adds missing thread safety annotations related to ObjectRegistry::lock_.
Bug:
17305632
Bug:
16720689
Change-Id: Id2cbd56624499c09e4755aaef7a9b15d2bdf4c91
Brian Carlstrom [Wed, 27 Aug 2014 23:40:53 +0000 (23:40 +0000)]
Merge "Support booting without functioning boot.oat/art patchoat." into lmp-dev
Mathieu Chartier [Wed, 27 Aug 2014 21:46:42 +0000 (21:46 +0000)]
Merge "Avoid using NanoTime for compiler and verifier" into lmp-dev
Mathieu Chartier [Wed, 27 Aug 2014 21:10:58 +0000 (14:10 -0700)]
Avoid using NanoTime for compiler and verifier
Target timings before with -j2 --compiler-filter=interpret-only:
Approximate speedup 3%.
Before:
real 0m36.852s
real 0m36.050s
real 0m36.246s
After:
real 0m35.667s
real 0m35.779s
real 0m35.458s
Bug:
16853450
Change-Id: Id996172d8283639fe293ddfc26d01fc1a0beed0b
Alex Light [Sat, 23 Aug 2014 00:49:35 +0000 (17:49 -0700)]
Support booting without functioning boot.oat/art patchoat.
Bug:
17000769
(cherry picked from commit
84d7605f93f1e6e86a16e02017e305c90e93117a)
Change-Id: I89c26a905af12ea288742368c2c038afd57a879a
Dave Allison [Wed, 27 Aug 2014 19:42:39 +0000 (19:42 +0000)]
Merge "Handle nested signals" into lmp-dev
Mathieu Chartier [Tue, 19 Aug 2014 21:29:46 +0000 (14:29 -0700)]
Reduce interpret-only compile time.
Before:
39.04user 5.18system 0:29.24elapsed 151%CPU (0avgtext+0avgdata 164176maxresident)k
38.87user 5.16system 0:29.14elapsed 151%CPU (0avgtext+0avgdata 164144maxresident)k
After:
36.26user 3.25system 0:27.00elapsed 146%CPU (0avgtext+0avgdata 162592maxresident)k
36.25user 3.28system 0:26.28elapsed 150%CPU (0avgtext+0avgdata 162688maxresident)k
Disabled implicit stack protection for the compiler, this reduces page faults.
Added support for not timing every method compilation and verification. NanoTime is
slow and adds ~2 seconds of real time. This is currently enabled since people want
to know which methods are slow to compile.
Bug:
16853450
(cherry picked from commit
8e219ae27624116b6d23e858fb21e93342f81d66)
Change-Id: I349ffb3f36db8c437137387aa6914dc17d743f09
Dave Allison [Tue, 26 Aug 2014 18:07:58 +0000 (11:07 -0700)]
Handle nested signals
This allows for signals to be raised inside the ART signal handler.
This can occur when the JavaStackTraceHandler attempts to generate
a stack trace and something goes wrong.
It also fixes an issue where the fault manager was not being
correctly shut down inside the signal chaining code. In this
case the signal handler was not restored to the original.
Bug:
17006816
Bug:
17133266
Change-Id: I9c25bf4f6921e6a107aefbdf47d2c0db9f41508f
Mathieu Chartier [Tue, 26 Aug 2014 18:01:37 +0000 (18:01 +0000)]
Merge "Fix race condition in TransitionCollector." into lmp-dev
Mathieu Chartier [Sat, 23 Aug 2014 20:56:03 +0000 (13:56 -0700)]
Fix race condition in TransitionCollector.
There was a minor race condition that could happen if multiple
threads called TransitionCollector at the same time. Starting out
in CMS:
Thread 1: TransitionCollector(SS).
Thread 2: TransitionCollector(CMS) sees that it is already CMS so
has copying_transition == false. But then thread 2 changes to SS.
Thread 1 resumes but incorrectly doesn't check the
disable_moving_gc_count_ possibly resulting in errors if
disable_moving_gc_count_ != 0 due to JNI since we are going from
SS -> CMS and are goign to move objects.
Bug:
17189964
Change-Id: I4f0d002717516b81355d66f26e0e8ebe3958348d
Stephen Kyle [Fri, 22 Aug 2014 14:03:07 +0000 (15:03 +0100)]
ART: Reject field accesses to non-reference vregs.
The verifier will now reject any iget/iput insts
that do not have a reference object in vB.
Bug:
17207857
Signed-off-by: Stuart Monteith <stuart.monteith@arm.com>
(cherry picked from commit
695c5982eeddee18c643cc2f94c9d78fa75599ab)
Change-Id: I3790b8687673643bb142549d74f36a3e2978bf13
Mathieu Chartier [Tue, 26 Aug 2014 17:15:20 +0000 (17:15 +0000)]
Merge "Map heap maps at a lower address." into lmp-dev
Andreas Gampe [Tue, 26 Aug 2014 17:10:28 +0000 (17:10 +0000)]
Merge "ART: kSuper also has IncompatibleClassChangeError" into lmp-dev
Andreas Gampe [Tue, 26 Aug 2014 17:04:14 +0000 (17:04 +0000)]
Merge "ART: Fix IncompatibleClassChange check for ArtMethod" into lmp-dev
Sebastien Hertz [Tue, 26 Aug 2014 13:43:17 +0000 (13:43 +0000)]
Merge "Reduce lock contention when debugging" into lmp-dev
buzbee [Tue, 26 Aug 2014 12:58:44 +0000 (12:58 +0000)]
Merge "Quick compiler: fix DCHECK" into lmp-dev
Brian Carlstrom [Tue, 26 Aug 2014 06:07:30 +0000 (23:07 -0700)]
CheckAndClearResolveException should allow LinkageError
Bug:
16565964
Change-Id: Ibf236c627519a5cdef9479b7bc21e931d5474abf
Andreas Gampe [Tue, 26 Aug 2014 03:55:01 +0000 (20:55 -0700)]
ART: kSuper also has IncompatibleClassChangeError
When resolving methods, kSuper can also have
IncompatibleClassChangeError.
Bug:
17266767
Change-Id: I5fafe03ad578f605825a3d0c89f1254a4a385b76
Andreas Gampe [Tue, 26 Aug 2014 03:46:31 +0000 (20:46 -0700)]
ART: Fix IncompatibleClassChange check for ArtMethod
In case of kSuper we should complain sometimes.
Bug:
17265201
Change-Id: Ifad903a750495963c7128296713fbd7c4add71cd
Mathieu Chartier [Tue, 26 Aug 2014 03:30:33 +0000 (03:30 +0000)]
Merge "Reduce GC log spam." into lmp-dev
Mathieu Chartier [Tue, 26 Aug 2014 02:46:57 +0000 (19:46 -0700)]
Reduce GC log spam.
Homogeneous compaction and collector transitions now print to
VLOG(gc) instead of LOG(INFO).
Bug:
17166556
Change-Id: I5d54acf36a9b2763922e6329596a2f191d45535c
Mathieu Chartier [Mon, 25 Aug 2014 23:32:41 +0000 (16:32 -0700)]
Map heap maps at a lower address.
We now map the heap mmaps at a lower address in case the app needs
larger continuous address space.
Bug:
16502380
Change-Id: I2cc11d0c207c0eae8db0c5025a48b11119c5a802
buzbee [Mon, 25 Aug 2014 16:34:03 +0000 (09:34 -0700)]
Quick compiler: fix DCHECK
A DCHECK intended to verify the soundness of the mapping between
a wide Dalvik value and a pair of temp registers incorrectly fired
when the temp register pair was associated with a wide temporary
(i.e. - when (SReg() == INVALID_SREG)).
In this particular situation, the wideness flag is meaningful only when
there is an associated valid vreg. Code rearranged to only perform
the DCHECK and reset the partner's wideness bit when we're dealing with
a live wide value.
b/
16484538
Change-Id: I3b8a7bebc5d918c65ea56ae8db0a8a5cd8003386
Alex Light [Thu, 21 Aug 2014 16:45:00 +0000 (09:45 -0700)]
Reduce log spam by changing a LOG(INFO) to VLOG(startup)
Bug:
17166556
(cherry picked from commit
b6cabc1345ec307559a6e85141fb69b7caa9413c)
Change-Id: I9975ba07ad8f119a6e367e2770a60040a462149b
Alex Light [Tue, 19 Aug 2014 21:23:13 +0000 (14:23 -0700)]
Support running without a boot image.
Bug:
17000769
(cherry picked from commit
64ad14dbe2225441fb7734bf6d89358d96692eea)
Change-Id: I6404d5050c8a2f4ee6e70d58532eb25ee9de248e
Mathieu Chartier [Mon, 25 Aug 2014 17:43:17 +0000 (17:43 +0000)]
Merge "Fix GC memory overhead accounting." into lmp-dev
Mathieu Chartier [Sat, 23 Aug 2014 01:01:19 +0000 (01:01 +0000)]
Merge "Add fallback for non moving space being full." into lmp-dev
Mathieu Chartier [Fri, 22 Aug 2014 21:53:04 +0000 (14:53 -0700)]
Add fallback for non moving space being full.
When the non moving space becomes full, we disable moving GC and
make the main space the new non moving space.
Also added a runtime option for changing the non moving space size:
-XX:NonMovingSpaceCapacity.
Bug:
17189964
Change-Id: I04d00d0f457f6c1f5724bf32932b6b6ce4d341b1
Ian Rogers [Fri, 22 Aug 2014 21:48:42 +0000 (21:48 +0000)]
Merge "Using the correct time for wait_start_ms when lock contention logging is enabled" into lmp-dev
Xin Guan [Fri, 22 Aug 2014 16:55:37 +0000 (11:55 -0500)]
Using the correct time for wait_start_ms when lock contention logging is enabled
Bug:
17208457
Change-Id: I35555e0acf3b37bb3776a62c64d59ae93403b237
Mathieu Chartier [Thu, 21 Aug 2014 19:21:48 +0000 (12:21 -0700)]
Fix GC memory overhead accounting.
There was some missing null checks.
Bug:
16238192
Change-Id: Iaf8d752db5f21e76f668c0066a063239ff374eee
Andreas Gampe [Fri, 22 Aug 2014 17:14:12 +0000 (17:14 +0000)]
Merge "ART: Allow uninitialized runtime in oat code" into lmp-dev
Calin Juravle [Fri, 22 Aug 2014 16:55:31 +0000 (16:55 +0000)]
Merge "Always setup the native bridge library" into lmp-dev
Mathieu Chartier [Fri, 22 Aug 2014 16:48:58 +0000 (16:48 +0000)]
Merge "Fix testing wrong variable for non zygote." into lmp-dev
Mathieu Chartier [Fri, 22 Aug 2014 01:28:20 +0000 (18:28 -0700)]
Fix testing wrong variable for non zygote.
For non zygote we want to only have a single non moving space to
let dex2oat use RosAlloc for all allocations.
Fixed stale low memory logic in parsed options.
Bug:
17189964
Change-Id: I85a565002fd18ae30105df084796521abd60ea12
Sebastien Hertz [Tue, 19 Aug 2014 13:33:43 +0000 (15:33 +0200)]
Reduce lock contention when debugging
Uses a ReaderWriterMutex for the breakpoint lock to reduce contention during
debugging session.
Also adds missing thread safety annotations on fields and methods related to
instrumentation and debugging.
Bug:
16814665
Bug:
11667502
Change-Id: I056cdafa91109e0c83806c8d8df75c37ade0a354
Calin Juravle [Fri, 22 Aug 2014 11:53:59 +0000 (12:53 +0100)]
Always setup the native bridge library
... even if the string is empty. This will initialize the native bridge
library but mark it as unavailable.
- also, rename native_bridge_library_path to
native_bridge_library_filename to be closer to the actual meaning (it's
just the filename without any path).
Bug:
16404669.
Change-Id: I94628639691459d48d1fbf0841f36b68d51818e7
Andreas Gampe [Fri, 22 Aug 2014 06:10:47 +0000 (23:10 -0700)]
ART: Allow uninitialized runtime in oat code
Oatdump does not initialize a runtime. Avoid a segfault.
Bug:
17194559
Change-Id: I1a1bd3fc7762ff4714251617a975e832522fa879
Daniel Mihalyi [Thu, 21 Aug 2014 13:57:25 +0000 (15:57 +0200)]
Fixed wrong field usage in Instrumentation::RemoveListener
Signed-off-by: Daniel Mihalyi <daniel.mihalyi@mattakis.com>
Bug:
16814665
(cherry picked from commit
6644521cfaff718a55e3602efb3b54bf19a0cd1a)
Change-Id: Ie68c70ba6ee2e0bd918717c8659c057f98e66ff7
Daniel Mihalyi [Mon, 18 Aug 2014 16:45:31 +0000 (18:45 +0200)]
Optimized instrumentation listener handling
Some instrumentation listener lists may be modified while iterating
over the list to deliver an instrumentation event. Therefore the
previous implementation copied the list of listeners before starting
the iteration.
This new implementation only copies the list of instrumentation
listeners when the list is changed. Instances of the list are
reference counted using std::shared_ptr<>.
Bug:
16814665
(cherry picked from commit
ca1d06cfa2f2b8d2be4390644e126cb68cdbb5ba)
Change-Id: Ib2e6b980de85b75f1c8f4a8825bdc7767154663a
Dave Allison [Fri, 22 Aug 2014 01:09:18 +0000 (01:09 +0000)]
Merge "Fix fugu build" into lmp-dev
Dave Allison [Fri, 22 Aug 2014 00:22:38 +0000 (17:22 -0700)]
Fix fugu build
Bug:
17188805
Change-Id: Ibbfc759b42ec0f08a1beb35bf52d3ca92b8fe1ea
Brian Carlstrom [Thu, 21 Aug 2014 20:43:54 +0000 (20:43 +0000)]
Merge "Make a couple of map checks debug only." into lmp-dev
Narayan Kamath [Thu, 21 Aug 2014 16:38:09 +0000 (17:38 +0100)]
Make a couple of map checks debug only.
This cost us close to 80ms in app startup times.
The checks that a reused region was within an already
existent map has been demoted to a debug check.
A couple of other negative checks have been removed
outright because one of them was superflous and the other
wasn't guaranteed to be correct.
bug:
16828525
Change-Id: I00f76de06df0ea4ced40fdcb7825248d4b662045
Dave Allison [Thu, 21 Aug 2014 16:45:04 +0000 (09:45 -0700)]
Remove profiler log line
This removes a LOG(INFO) saying that the profile is disabled.
Bug:
17166556
Change-Id: I4ddb4b0f792eb60ed5c927e17d06cff4ae0c073f
Mathieu Chartier [Thu, 21 Aug 2014 16:42:31 +0000 (16:42 +0000)]
Merge "Check pending exception result in AllocObjectWithAllocator." into lmp-dev
Dave Allison [Thu, 21 Aug 2014 16:38:39 +0000 (16:38 +0000)]
Merge "Fix fault handler to unregister on shutdown" into lmp-dev
Dave Allison [Thu, 21 Aug 2014 16:38:14 +0000 (16:38 +0000)]
Merge "Add supported architectures to signal test." into lmp-dev
Mathieu Chartier [Thu, 21 Aug 2014 01:19:23 +0000 (18:19 -0700)]
Check pending exception result in AllocObjectWithAllocator.
Possible previous bug:
Allocation fails due to OOM and the collector transitions.
This caused us to incorrectly retry the allocation with a pending
exception. We now return null if there is a pending exception.
Bug:
17164348
Change-Id: I22eab472afb2fdea6e800963ccb35ec0755ba0e6
Dave Allison [Thu, 21 Aug 2014 00:38:41 +0000 (17:38 -0700)]
Fix fault handler to unregister on shutdown
This fixes a problem with the fault handler where it wasn't
unregistering itself during shutdown of the runtime.
Bug:
17133266
Change-Id: If4758b14ded0fba194897191b1a8d8d7b0b3e6e5
Dave Allison [Wed, 20 Aug 2014 21:29:39 +0000 (14:29 -0700)]
Add supported architectures to signal test.
This adds ARM, AARCH64, i386 and x86_64 to the supported architectures
in 004-SignalTest
Bug:
16948053
Change-Id: I82b3f053e9340b9e2e8fc04775d03fb0592fa06e
Mathieu Chartier [Wed, 20 Aug 2014 19:09:38 +0000 (12:09 -0700)]
Add null check to CheckVirtualMethod
There was a runtime SIGSEGV that should have been a check jni
failure.
Also added regression test.
Bug:
16320699
Change-Id: If6c8e73959cefb24e4703f1562cdddb343d86630
buzbee [Wed, 20 Aug 2014 12:54:11 +0000 (12:54 +0000)]
Merge "Quick compiler: fix Arm64 inline locking code" into lmp-dev
buzbee [Tue, 19 Aug 2014 23:47:06 +0000 (16:47 -0700)]
Quick compiler: fix Arm64 inline locking code
The Arm64 monitor enter code was incorrectly comparing the
thread ID to zero rather than the lock value to determine if the
inline fast path was doable. As a result, we ended up always
using the out-of-line case.
This fix delivers a 6% improvement to DeltaBlue.
b/
16707253
Change-Id: I499530da84beecf1f3f5c5ebb666e3eb12599ac0
Andreas Gampe [Tue, 19 Aug 2014 19:16:29 +0000 (19:16 +0000)]
Merge "ART: Set default visibility to protected" into lmp-dev
Jeff Hao [Tue, 19 Aug 2014 17:17:36 +0000 (10:17 -0700)]
Check for constructor first when installing tracing stubs.
Addresses comments in
https://android-review.googlesource.com/#/c/104383/ made after the commit.
Bug:
16386215
Change-Id: I0e59b8507f601fd6508fbda8d24dd3c98ead1537
Mathieu Chartier [Wed, 20 Aug 2014 23:50:23 +0000 (23:50 +0000)]
Merge "Use homogeneous space compact by default if not low memory." into lmp-dev
Mathieu Chartier [Wed, 20 Aug 2014 21:13:18 +0000 (14:13 -0700)]
Use homogeneous space compact by default if not low memory.
Otherwise we use semispace as the background collector.
Bug:
16401001
Change-Id: I7b20e0888695521604758c98d09f0709fa2158ef
Brian Carlstrom [Tue, 19 Aug 2014 22:30:21 +0000 (22:30 +0000)]
Merge "Make apps able to run with a failing patchoat" into lmp-dev
Alex Light [Thu, 14 Aug 2014 21:16:26 +0000 (14:16 -0700)]
Make apps able to run with a failing patchoat
Bug:
17000769
(cherry picked from commit
9dcc4572949f6a8231a1b4ed859676ba6f411726)
Change-Id: I0a1a4dc7f5d4bb268530840302ecfb1555231e05
Dave Allison [Tue, 19 Aug 2014 22:08:31 +0000 (22:08 +0000)]
Merge "Workaround problem reading main stack on intel devices." into lmp-dev
Andreas Gampe [Mon, 18 Aug 2014 23:43:12 +0000 (16:43 -0700)]
ART: Fix ARM invoke stub, run Generic JNI in jni_compiler_test
On devices which support it, run all jni compiler tests also with
the Generic JNI infrastructure.
Fix art_quick_invoke_stub for ARM32 to enforce 16B stack alignment.
As the native convention does not require as much, the previous
code would keep it unaligned.
Bug:
17107929
(cherry picked from commit
6e498695b60f1532d2264ec6badb1cd6e10ecaa9)
Change-Id: I1481500ff7c627ab4e5bfbfd7245099a639aa049
Dave Allison [Mon, 18 Aug 2014 22:49:51 +0000 (15:49 -0700)]
Workaround problem reading main stack on intel devices.
This works around a problem where ART can't read the lowest page
of the main stack. The workaround is to add 4K to the stack
start address on the main stack on Intel.
Please see https://b2.corp.google.com/issues/
17111575 for
underlying cause.
Bug:
17031544
Change-Id: Ifc3216e10160bb1eec9d989fd57b10345bc89537
Andreas Gampe [Wed, 20 Aug 2014 22:24:11 +0000 (22:24 +0000)]
Merge "ART: Relax GetInstructionSetFromString" into lmp-dev
Andreas Gampe [Wed, 20 Aug 2014 00:28:06 +0000 (17:28 -0700)]
ART: Relax GetInstructionSetFromString
Do not abort on an unknown instruction set string. Instead return
kNone and let the caller handle this.
Also simplify the patchoat tool to use this.
Bug:
17136416
Change-Id: I24131914bcf91c04ae93179bf809a2907f1f2b7a
Vladimir Marko [Mon, 18 Aug 2014 10:42:51 +0000 (11:42 +0100)]
Disable load/store elimination. DO NOT MERGE
The optimization has several issues and it has been
effectively disabled by
https://android-review.googlesource.com/103511
Disable it completely to avoid wasting compilation time.
Bug:
16862804
Change-Id: I42f69ae64170dc44f5fd6455b6a51c9bd3f93672
Vladimir Marko [Wed, 20 Aug 2014 15:08:59 +0000 (15:08 +0000)]
Merge "X86: Fix alias info in GenInlinedIndexOf()." into lmp-dev
Vladimir Marko [Tue, 19 Aug 2014 14:00:34 +0000 (15:00 +0100)]
X86: Fix alias info in GenInlinedIndexOf().
For 32-bit X86, GenInlinedIndexOf() pushes and pops EDI. In
one branch it then calls Load32Disp() with adjusted stack
offset. That calculates wrong alias_info for the generated
insn. If left unfixed, this could confuse load hoisting.
Bug:
17128502
(cherry picked from commit
74de63bb1cc275b411cae28a96f9b3a78b939bc2)
Change-Id: I5dc82b7aae9e9655e75843a952b8ebb04269f46b
Sebastien Hertz [Tue, 19 Aug 2014 09:05:47 +0000 (09:05 +0000)]
Merge "Fix JDWP event races" into lmp-dev
Sebastien Hertz [Mon, 18 Aug 2014 17:26:39 +0000 (19:26 +0200)]
Fix JDWP event races
Fix a race where the JDWP connection could be closed before the VM_DEATH event
is sent during runtime shutdown.
Fix potential race where we could wait forever for the JDWP thread to establish
connection.
Bug:
16720689
Change-Id: I227e0a15a2fd17d7bfe2a66a35d719d558fcd32d
Brian Carlstrom [Tue, 19 Aug 2014 00:42:52 +0000 (00:42 +0000)]
Merge "AttachCurrentThread should only warn on missing thread name if CheckJNI is enabled." into lmp-dev
Brian Carlstrom [Mon, 18 Aug 2014 22:38:34 +0000 (15:38 -0700)]
AttachCurrentThread should only warn on missing thread name if CheckJNI is enabled.
Bug:
17108693
(cherry picked from commit
3ea69c0abce1b81b3c45033867d49b00e6d6b709)
Change-Id: Ide8158cc3e12e604f4fa344cd881fe6f68476ab7
Mathieu Chartier [Mon, 18 Aug 2014 22:02:59 +0000 (15:02 -0700)]
Fix heap corruption and verification.
There was a faulty assumption that space End() was always aligned to
card boundaries. This was true for all spaces other than the image
and resulted in heap corruption when the last object of the image
space contained a reference to an object in another space.
Also fixed an error where we called the pre GC verification post GC.
Bug:
17080623
Change-Id: I041ee564518f53b79c6e8dc2ad782a3152577a4e
Andreas Gampe [Mon, 18 Aug 2014 05:06:15 +0000 (05:06 +0000)]
Merge "ART: Do not recursively abort when visiting locks in a bad state" into lmp-dev
Andreas Gampe [Sat, 16 Aug 2014 20:41:10 +0000 (13:41 -0700)]
ART: Do not recursively abort when visiting locks in a bad state
This avoids a nested abort in VisitLocks.
Bug:
17080621,
16382675
Change-Id: Id604976ac9dcac0e319fb25cab4d2cbc98d7ee24
Andreas Gampe [Sat, 16 Aug 2014 05:35:46 +0000 (22:35 -0700)]
ART: Avoid expensive class lookup in compiler driver
When we are not compiling multidex, avoid an expensive by-descriptor
lookup for classes when deciding whether to skip compilation.
Bug:
17072698
Change-Id: I19bcb2a2928207fcae0badd466cdc52945bd37c4
Nicolas Geoffray [Wed, 13 Aug 2014 02:40:45 +0000 (03:40 +0100)]
Fix bugs in the x86 and arm versions of generic JNI.
Also fix the run script of 116-nodex2oat to use the non-prebuild
script for target.
Bug:
17000769
(cherry-picked from commit
54accbca0b549b1b1ad3ef09655dad438bc1e104)
Change-Id: I439fcd710fb8bb408f3288ff8fb34fef23890adb
Nicolas Geoffray [Tue, 22 Jul 2014 09:48:00 +0000 (10:48 +0100)]
Execute an application even when dex2oat crashes.
Bug:
17000769
(cherry picked from commit
4fcdc94d22a4608e355aa8df36240181149d10e8)
Change-Id: Iccb1fec94fe64ce4c3097510952f275482b86aa9
Calin Juravle [Tue, 12 Aug 2014 19:55:20 +0000 (20:55 +0100)]
[ART] Move to system/core native bridge library
Some cleanups on the way:
- move NativeBridgeRuntimeCallbacks from a global to a field
- rename native_bridge_library_string to the more suggestive
navtive_bridge_library_path
- clean up imports
Bug:
16884833
(cherry picked from commit I73aab8e212860ba5aee9444d801806d3da326a41)
Change-Id: I67f037ea81d2a4ede4294afd8b84b7640e534a13
Ian Rogers [Fri, 15 Aug 2014 18:34:19 +0000 (18:34 +0000)]
Merge "Make Monitor::Wait more robust to spurious Inflate failures." into lmp-dev
Ian Rogers [Fri, 15 Aug 2014 18:09:28 +0000 (11:09 -0700)]
Make Monitor::Wait more robust to spurious Inflate failures.
Bug:
17062710
Change-Id: Ife5f6b335caacc70cab543cd568676d277d3beb6
Narayan Kamath [Fri, 15 Aug 2014 09:08:15 +0000 (09:08 +0000)]
Merge "More efficient stack walk in exception throwing." into lmp-dev
Mathieu Chartier [Fri, 15 Aug 2014 17:45:30 +0000 (17:45 +0000)]
Merge "Use handle in one case that spans a gc-point." into lmp-dev
Mingyao Yang [Wed, 13 Aug 2014 21:51:03 +0000 (14:51 -0700)]
Use handle in one case that spans a gc-point.
Bug:
16689428
(cherry picked from commit
38eecb0f4288a374c9b0b4b4df8793eb5fc6697c)
Change-Id: Id4a1bde9012e4dd333493e46997c200537c9cf55
Jeff Hao [Fri, 15 Aug 2014 00:18:52 +0000 (17:18 -0700)]
Prevent stubs from being installed in java.lang.reflect.Proxy.<init>.
This CL is a better fix for proxy tracing and undoes the changes in
https://android-review.googlesource.com/#/c/103025/
Bug:
16386215
(cherry picked from commit
db8a664e0b68c7c4d36270cd21dce8de1912d7f9)
Change-Id: Ic9e0ea2af7cb2da5d90c56aa009de92dba14cc47
Jeff Hao [Thu, 24 Jul 2014 23:26:09 +0000 (16:26 -0700)]
Fix proxy tracing and enable tests that now work with tracing.
Also updates proxy_test to generate an image for GetQuickOatCodeFor.
Bug:
16386215
(cherry picked from commit
f0a3f09c3d54646166a55c05a6b39c7dd504129c)
Change-Id: I138edbad9e1646db8590f2b1b73f2788d9710e68
Andreas Gampe [Thu, 14 Aug 2014 22:40:19 +0000 (22:40 +0000)]
Revert "Revert "ART: Fix mac build""
This reverts commit
350b03bca65f87895275d1ed4bb895d5afeeeff1.
Also fixes a fault_handler symbol.
Change-Id: I1315527400bffc8ded9f2a7e7fdb1c8a6c104b96
Dave Allison [Thu, 14 Aug 2014 17:02:48 +0000 (17:02 +0000)]
Revert "Revert "Reduce stack usage for overflow checks""
Fixes stack protection issue.
Fixes mac build issue.
This reverts commit
83b1940e6482b9d8feba5c492507735686650ea5.
Change-Id: I7ba17252882b23a740bcda2ea94aacf398255406
Hiroshi Yamauchi [Wed, 13 Aug 2014 18:12:22 +0000 (11:12 -0700)]
More efficient stack walk in exception throwing.
In the exception handling code, we currently walk down the stack
twice, once to get the stack height which we use to compute frame IDs
(the bottom frame is zero), and once more to find the catch block to
jump to.
For a deep stack, this could result in very slow exception
handling. That is, if have a lot of finally or catch blocks that we
end up jumping to in a deep stack, we need to do a lot of
catch/rethrow chains. Since we'd need to walk down to the bottom each
time to compute frames IDs in each catch/rethrow, we'd need to walk
down O(N^2) frames at the worst case.
Instead of frames IDs ((the bottom frame is zero), we will use the
frame depth (the top frame is zero) and no longer need to walk down
the stack just to get the stack height. We walk down O(N) frames.
This was what was happening with
code.google.gson.functional.CircularReferenceTest. With this change,
the test run time went from ~120s down to ~3s on N5 and it no longer
crashes due to the thread suspension timeout.
(cherry pick commit
649278cec7119cdd1bea3d0b710dbb2aa7c650b6)
Bug:
16800209
Change-Id: Ie815df1e3e8fb9d82e40685d4cc2b8838fd8aa07
Mathieu Chartier [Thu, 14 Aug 2014 01:07:31 +0000 (18:07 -0700)]
Use homogeneous space compaction if proper flag is set.
If the flag ART_USE_HSPACE_COMPACT is set then we use
hspace compaction.
Bug:
16401001
Change-Id: I74e1db764f9ff16c007fa3bd16cb2c9b468ec83c
Andreas Gampe [Wed, 13 Aug 2014 07:15:43 +0000 (00:15 -0700)]
ART: Set default visibility to protected
Set default visibility of symbols to protected. This allows the
linker to optimize internal calls and helps avoid plt calls.
Make almost all assembly stubs hidden, as -fvisibility does not
seem to apply to them. Change the assembly tests accordingly. Also
allows to clean up previous hacks to avoid plt calls.
Bug:
16974467
Change-Id: I9030dcf6116251f434f94a2b08e56e12085af652
Dave Allison [Fri, 25 Jul 2014 23:15:27 +0000 (16:15 -0700)]
Reduce stack usage for overflow checks
This reduces the stack space reserved for overflow checks to 12K, split
into an 8K gap and a 4K protected region. GC needs over 8K when running
in a stack overflow situation.
Also prevents signal runaway by detecting a signal inside code that
resulted from a signal handler invokation. And adds a max signal count to
the SignalTest to prevent it running forever.
Also reduces the number of iterations for the InterfaceTest as this was
taking (almost) forever with the --trace option on run-test.
Bug:
15435566
Change-Id: Id4fd46f22d52d42a9eb431ca07948673e8fda694
Conflicts:
compiler/optimizing/code_generator_x86_64.cc
runtime/arch/x86/fault_handler_x86.cc
runtime/arch/x86_64/quick_entrypoints_x86_64.S
Andreas Gampe [Thu, 14 Aug 2014 18:37:56 +0000 (18:37 +0000)]
Merge "ART: Fix run-test in case of relative TMPDIR" into lmp-dev
Andreas Gampe [Wed, 13 Aug 2014 22:40:22 +0000 (15:40 -0700)]
ART: Fix run-test in case of relative TMPDIR
Relative paths should be resolved to absolute paths, as we are
chdir-ing in the test.
Bug:
16499668
Change-Id: Ic7db35e21f519ced77c4ac468d8e909bcb40b24d
Mathieu Chartier [Wed, 13 Aug 2014 17:08:05 +0000 (10:08 -0700)]
Add allocation rate tracking to systrace.
Bug:
16956980
Change-Id: I7074f4ae0f75d1bb58e9860ded4e17ce6978d74e
Dave Allison [Wed, 16 Jul 2014 23:04:32 +0000 (16:04 -0700)]
Add implicit checks for x86_64 architecture.
This combines the x86 and x86_64 fault handlers into one. It also
merges in the change to the entrypoints for X86_64.
Replaces generic instruction length calculator with one that only
works with the specific instructions we use.
Bug:
16256184
Change-Id: I1e8ab5ad43f46060de9597615b423c89a836035c
Signed-off-by: Chao-ying Fu <chao-ying.fu@intel.com>
Andreas Gampe [Wed, 13 Aug 2014 06:24:16 +0000 (23:24 -0700)]
ART: Don't leave rex_ & length_ uninitialized
Don't leave fields uninitialized in the x86-64 utils assembler.
Bug:
16993295
(cherry picked from commit
181211de89df5615395a55114fe551b08e86bc4d)
Change-Id: I2d648991fc2a6a4d5f518eb9ef0bb10857f6318a
Sebastien Hertz [Wed, 13 Aug 2014 06:09:49 +0000 (06:09 +0000)]
Merge "Fix crash in VirtualMachine.AllThreads JDWP command" into lmp-dev
Ian Rogers [Wed, 13 Aug 2014 21:34:53 +0000 (21:34 +0000)]
Merge "Avoid use of std::string where we have const char*." into lmp-dev
Ian Rogers [Tue, 12 Aug 2014 09:30:58 +0000 (02:30 -0700)]
Avoid use of std::string where we have const char*.
Removing the ClassHelper caused std::string creation for all calls to
Class::GetDescriptor and a significant performance regression. Make the
std::string an out argument so the caller can maintain it and its life time
while allowing GetDescriptor to return the common const char* case.
Don't generate GC maps when compilation is disabled.
Remove other uses of std::string that are occuring on critical paths.
Use the cheaper SkipClass in CompileMethod in CompilerDriver.
Specialize the utf8 as utf16 comparison code for the common shorter byte
encoding.
Force a bit of inlining, remove some UNLIKELYs (they are prone to pessimizing
code), add some LIKELYs.
x86-64 host 1-thread interpret-only of 57 apks:
Before: 29.539s
After: 23.467s
Regular compile:
Before: 1m35.347s
After: 1m20.056s
Bug:
16853450
Change-Id: Ic705ea24784bee24ab80084d06174cbf87d557ad
Conflicts:
runtime/utils.cc