OSDN Git Service

android-x86/art.git
9 years agoMerge "Handle another untyped dex CONST issue"
buzbee [Fri, 27 Feb 2015 21:29:54 +0000 (21:29 +0000)]
Merge "Handle another untyped dex CONST issue"

9 years agoMerge "Fix jit tests"
Mathieu Chartier [Fri, 27 Feb 2015 20:55:16 +0000 (20:55 +0000)]
Merge "Fix jit tests"

9 years agoFix jit tests
Mathieu Chartier [Fri, 27 Feb 2015 18:39:48 +0000 (10:39 -0800)]
Fix jit tests

Missing image suffix was missing causing tests to fail.
Also added missing phony targets.

Bug: 19524713
Change-Id: Ib40d39b1fc16e1adfdc01744a199ae9b0b562664

9 years agoMerge "Fix JIT direct call to interpreter bridge"
Mathieu Chartier [Fri, 27 Feb 2015 02:58:48 +0000 (02:58 +0000)]
Merge "Fix JIT direct call to interpreter bridge"

9 years agoFix JIT direct call to interpreter bridge
Mathieu Chartier [Fri, 27 Feb 2015 02:28:07 +0000 (18:28 -0800)]
Fix JIT direct call to interpreter bridge

FormulaEvaluationActions.EvaluateAndApplyChanges on hammerhead eng:
Before: 3550ms
After: 2540

Bug: 17950037
Change-Id: If7acaea20d4adb5418d00ef13192e2d4c29032af

9 years agoMerge "Change card cache mod-union table to use bitmaps"
Mathieu Chartier [Thu, 26 Feb 2015 23:04:02 +0000 (23:04 +0000)]
Merge "Change card cache mod-union table to use bitmaps"

9 years agoChange card cache mod-union table to use bitmaps
Mathieu Chartier [Fri, 23 Jan 2015 21:18:53 +0000 (13:18 -0800)]
Change card cache mod-union table to use bitmaps

Previously used card sets, using bitmaps saves memory and slightly
increases performance.

Added mod union table test.

Performance EvaluateAndApplyChanges (minimal changes):

Before (card cache image mu table):
UpdateAndMarkImageModUnionTable: Avg: 524.320us
ImageModUnionClearCards: Avg: 54.580us
Native PSS: ~67500kB

After (card cache image mu table):
UpdateAndMarkImageModUnionTable: Avg: 515.600us
ImageModUnionClearCards: Avg: 53.780us
Native PSS: ~66014kB

Native PSS was higher before since the mod_union_table->SetCards()
which happens pre zygote fork was allocating a large amount of
std::nodes.

Bug: 11859910

Change-Id: I956b7e51d5572feec1393ffa618b7b7d8c147b28

9 years agoMerge "Add standard gc log for transition collections."
Hiroshi Yamauchi [Thu, 26 Feb 2015 21:12:25 +0000 (21:12 +0000)]
Merge "Add standard gc log for transition collections."

9 years agoAdd standard gc log for transition collections.
Hiroshi Yamauchi [Thu, 26 Feb 2015 20:53:45 +0000 (12:53 -0800)]
Add standard gc log for transition collections.

Factor out the gc log logic into a function.

Note we still don't log in background as before but do for things like
OOME compactions in foreground.

Change-Id: Ifd35a796cf26aba2fba83bf07e72902357d91f8b

9 years agoMerge "Add missing JIT stuff in make files"
Mathieu Chartier [Thu, 26 Feb 2015 19:30:35 +0000 (19:30 +0000)]
Merge "Add missing JIT stuff in make files"

9 years agoAdd missing JIT stuff in make files
Mathieu Chartier [Thu, 26 Feb 2015 18:58:08 +0000 (10:58 -0800)]
Add missing JIT stuff in make files

Was causing core.oat to not be created.

Bug: 19524713
Change-Id: I5da005fd36f9682623fdda9931280997b59c2681

9 years agoMerge "Update expectations for volantis."
Nicolas Geoffray [Thu, 26 Feb 2015 15:11:58 +0000 (15:11 +0000)]
Merge "Update expectations for volantis."

9 years agoUpdate expectations for volantis.
Nicolas Geoffray [Thu, 26 Feb 2015 14:53:03 +0000 (14:53 +0000)]
Update expectations for volantis.

Change-Id: Icd4aecd9b7a274febfd99bad6432bb9d46f9e1d3

9 years agoMerge "ARM: Fix LIR flags in encoding map."
Vladimir Marko [Thu, 26 Feb 2015 12:17:57 +0000 (12:17 +0000)]
Merge "ARM: Fix LIR flags in encoding map."

9 years agoMerge "JDWP: fix thread state on event suspension"
Sebastien Hertz [Thu, 26 Feb 2015 06:17:12 +0000 (06:17 +0000)]
Merge "JDWP: fix thread state on event suspension"

9 years agoARM: Fix LIR flags in encoding map.
Ningsheng Jian [Wed, 4 Feb 2015 06:13:45 +0000 (14:13 +0800)]
ARM: Fix LIR flags in encoding map.

Also correct memory reference type for PC relative load.

Change-Id: I7a5258f2ed718448dc3e6e7fda6569b3f0c2fe46

9 years agoMerge "Fix method_verifier_test with the SS collector."
Hiroshi Yamauchi [Thu, 26 Feb 2015 02:12:01 +0000 (02:12 +0000)]
Merge "Fix method_verifier_test with the SS collector."

9 years agoMerge "Change already inline error to warning"
Mathieu Chartier [Thu, 26 Feb 2015 01:41:57 +0000 (01:41 +0000)]
Merge "Change already inline error to warning"

9 years agoFix method_verifier_test with the SS collector.
Hiroshi Yamauchi [Wed, 25 Feb 2015 20:52:34 +0000 (12:52 -0800)]
Fix method_verifier_test with the SS collector.

Rename art_default_gc_type to ART_DEFAULT_GC_TYPE.

Bug: 19514492

Change-Id: I3c7920e7a84327c5908462e51e21199b89fb187d

9 years agoChange already inline error to warning
Mathieu Chartier [Thu, 26 Feb 2015 01:17:37 +0000 (17:17 -0800)]
Change already inline error to warning

There is a race condition in the JIT where the JIT and another
running thread can both verify the same method at the same time.
This causes the inline method to be added twice and the LOG(ERROR)
was causing occasional test failures.

Bug: 17950037
Bug: 18713034
Change-Id: I7f94384106ca5c621a5cf7d2ba36910d923d110c

9 years agoMerge "Fix JIT for vmdebug test 99"
Mathieu Chartier [Thu, 26 Feb 2015 00:34:11 +0000 (00:34 +0000)]
Merge "Fix JIT for vmdebug test 99"

9 years agoMerge "[MIPS] Fix mips32 trampoline problem."
Andreas Gampe [Thu, 26 Feb 2015 00:00:40 +0000 (00:00 +0000)]
Merge "[MIPS] Fix mips32 trampoline problem."

9 years agoMerge "Fix argument register allocation for mips32."
Andreas Gampe [Wed, 25 Feb 2015 23:58:29 +0000 (23:58 +0000)]
Merge "Fix argument register allocation for mips32."

9 years ago[MIPS] Fix mips32 trampoline problem.
Nikola Veljkovic [Fri, 20 Feb 2015 16:21:15 +0000 (17:21 +0100)]
[MIPS] Fix mips32 trampoline problem.

Reserve argument slots (for $a0-$a3) on the stack in the
InvokeTrampoline() function, for entrypoints that contain direct
reference to native implementation. Called function may use this
space to store $a0-$a3 regs.

This fixes the Settings app crash on mips32, when navigating to:
Settings->Apps->(swipe) On SD card.

Change-Id: If53ce822fd3ef6ef9839c89b556f76fac0792190

9 years agoFix JIT for vmdebug test 99
Mathieu Chartier [Wed, 25 Feb 2015 21:22:57 +0000 (13:22 -0800)]
Fix JIT for vmdebug test 99

Test was flaky due to JIT re-compiliation after deoptimization
resulting in some invalid PC offsets.

Bug: 17950037
Change-Id: I276c84c918579259ce47ef873892c3c5dcf0c977

9 years agoFix argument register allocation for mips32.
Douglas Leung [Sat, 14 Feb 2015 00:53:03 +0000 (16:53 -0800)]
Fix argument register allocation for mips32.

Change-Id: Ic21b0371d7cafb9b67217b92bb2edb735a43d1b0

9 years agoMerge "ART: oatdump enhancements"
Andreas Gampe [Wed, 25 Feb 2015 19:51:12 +0000 (19:51 +0000)]
Merge "ART: oatdump enhancements"

9 years agoART: oatdump enhancements
Anestis Bechtsoudis [Mon, 23 Feb 2015 06:32:57 +0000 (22:32 -0800)]
ART: oatdump enhancements

New features list includes:
- Class filter option to limit classes search space
- Method filter is applied only against the method
  name, instead of the entire signature. Can be
  combined with class filter for maximum efficiency.
- Bulk dump of class and method names list only.
  Can be combined with filters to limit results.
- Export embedded dex files from input oat files
  to filesystem (symlinks not supported as utils
  functions are utilized for os & fs operations).
- addr2instr option to locate the in-range method
  implementation and limit disassemble dumps. Input
  relative addr is added to oat executable offset to
  calculate the search offset. If method has been
  successfully located, code is dumped and program
  aborts further analysis of the input file. Methods
  located before the target address just print their
  signature, although skip all disassemble and other
  info. Calculated search offset is also printed as
  part of the initial header info.
- Little-endian dex instructions bytecode is printed
  in the same line before the instruction string.

Some minor re-orders have been also taken place for
more targeted results.

Change-Id: I3116ee3c99c258718f46faea8ea4295da6ae2bf7
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
9 years agoMerge "ART: Print maps directly to log"
Andreas Gampe [Wed, 25 Feb 2015 18:12:27 +0000 (18:12 +0000)]
Merge "ART: Print maps directly to log"

9 years agoJDWP: fix thread state on event suspension
Sebastien Hertz [Tue, 24 Feb 2015 13:56:59 +0000 (14:56 +0100)]
JDWP: fix thread state on event suspension

Before suspending event thread, we ensure its state is kSuspended
so the debugger sees it as RUNNING instead of WAIT.

Bug: 19103406
Change-Id: I58b4141a958bf6a3c360f6994967fad0078ea373

9 years agoMerge "ART: Nano optimization of LiveInterval"
David Brazdil [Wed, 25 Feb 2015 12:14:17 +0000 (12:14 +0000)]
Merge "ART: Nano optimization of LiveInterval"

9 years agoART: Nano optimization of LiveInterval
David Brazdil [Wed, 25 Feb 2015 11:57:05 +0000 (11:57 +0000)]
ART: Nano optimization of LiveInterval

Shuffling the order of conditions in the FirstIntersectionWith method
of LiveInterval class can save a couple of comparisons. Even though
this is a tiny optimization, it can save some compilation time since
the method is heavily used during register allocation.

Change-Id: I1817bd95db2d0eb96cc06fb2e9e06ac1cea784fe

9 years agoMerge "ART: Update description of failing libcore tests"
David Brazdil [Wed, 25 Feb 2015 09:29:55 +0000 (09:29 +0000)]
Merge "ART: Update description of failing libcore tests"

9 years agoART: Update description of failing libcore tests
David Brazdil [Wed, 25 Feb 2015 09:16:34 +0000 (09:16 +0000)]
ART: Update description of failing libcore tests

One of the tests that was described as failing on volantis fails on
hammerhead too. It was moved to a new category.

Change-Id: Ie58266d51d3aa180a36a460035ea7f016c0b7e3d

9 years agoART: Print maps directly to log
Andreas Gampe [Tue, 24 Feb 2015 23:50:19 +0000 (15:50 -0800)]
ART: Print maps directly to log

Do not read proc maps into a string before printing them later back
to the log. In low-memory situations this can cause a bad_alloc.

External bug: http://b.android.com/153990
Bug: 19494774

Change-Id: Ie63d8788afe8c9da65b30b2f89c50d3dbb820755

9 years agoMerge "ART: Fix "unused parameters""
Andreas Gampe [Wed, 25 Feb 2015 04:56:43 +0000 (04:56 +0000)]
Merge "ART: Fix "unused parameters""

9 years agoART: Fix "unused parameters"
Andreas Gampe [Wed, 25 Feb 2015 04:55:16 +0000 (20:55 -0800)]
ART: Fix "unused parameters"

GCC 4.8 decides that parameters for functions implemented with
"= default" are unused. This currently only impacts x86, but remove
the parameter names anyways.

Change-Id: I01865faa81af68c4c0e0b1cb1fb19e88ef548769

9 years agoMerge "Print thread name for attempting to unregister unattached thread"
Mathieu Chartier [Wed, 25 Feb 2015 02:57:41 +0000 (02:57 +0000)]
Merge "Print thread name for attempting to unregister unattached thread"

9 years agoMerge "Delete bad DCHECK"
Mathieu Chartier [Wed, 25 Feb 2015 01:53:10 +0000 (01:53 +0000)]
Merge "Delete bad DCHECK"

9 years agoDelete bad DCHECK
Mathieu Chartier [Tue, 24 Feb 2015 02:24:29 +0000 (18:24 -0800)]
Delete bad DCHECK

The declaring class of a super field could be in a different dex file.
Fixes sdk build.

Change-Id: I8258d64eeff6539afb52448595f5a6ec4c71a6bc

9 years agoPrint thread name for attempting to unregister unattached thread
Mathieu Chartier [Wed, 25 Feb 2015 01:20:30 +0000 (17:20 -0800)]
Print thread name for attempting to unregister unattached thread

Bug: 18713034
Change-Id: Ic0654c6569277e5bbd6d2582140595faaa36cded

9 years agoMerge "Revert "Revert "Add JIT"""
Mathieu Chartier [Wed, 25 Feb 2015 00:47:14 +0000 (00:47 +0000)]
Merge "Revert "Revert "Add JIT"""

9 years agoMerge "ART: Add -Wunused"
Andreas Gampe [Wed, 25 Feb 2015 00:26:04 +0000 (00:26 +0000)]
Merge "ART: Add -Wunused"

9 years agoART: Add -Wunused
Andreas Gampe [Fri, 20 Feb 2015 02:21:24 +0000 (18:21 -0800)]
ART: Add -Wunused

Until the global CFLAGS are fixed, add Wunused. Fix declarations
in the optimizing compiler.

Change-Id: Ic4553f08e809dc54f3d82af57ac592622c98e000

9 years agoRevert "Revert "Add JIT""
Mathieu Chartier [Tue, 24 Feb 2015 17:37:21 +0000 (09:37 -0800)]
Revert "Revert "Add JIT""

Added missing EntryPointToCodePointer.

This reverts commit a5ca888d715cd0c6c421313211caa1928be3e399.

Change-Id: Ia74df0ef3a7babbdcb0466fd24da28e304e3f5af

9 years agoMerge "JDWP: assert no pending exception when using JNI"
Sebastien Hertz [Tue, 24 Feb 2015 17:47:38 +0000 (17:47 +0000)]
Merge "JDWP: assert no pending exception when using JNI"

9 years agoMerge "ART: Disable libcore crypto test failing on target"
David Brazdil [Tue, 24 Feb 2015 17:27:17 +0000 (17:27 +0000)]
Merge "ART: Disable libcore crypto test failing on target"

9 years agoART: Disable libcore crypto test failing on target
David Brazdil [Tue, 24 Feb 2015 17:01:19 +0000 (17:01 +0000)]
ART: Disable libcore crypto test failing on target

Another test failing due to missing loop network device on hammerhead
and volantis bots. We disable it until the device configuration issue
is fixed.

Change-Id: I5b4610d706a8e4bdf178764f64f2d4359a83682d

9 years agoJDWP: assert no pending exception when using JNI
Sebastien Hertz [Wed, 29 Oct 2014 11:06:51 +0000 (12:06 +0100)]
JDWP: assert no pending exception when using JNI

Checks there is no pending exception when creating new JDWP ids using
JNI routines. We can create JDWP ids when executing code and report
an event like a breakpoint.

Bug: 17491155
Change-Id: I4e5851bf2261510b7776a709a2388e7ff586a77c

9 years agoMerge "not-int can also take non-int (byte and short) instructions."
Nicolas Geoffray [Tue, 24 Feb 2015 16:26:54 +0000 (16:26 +0000)]
Merge "not-int can also take non-int (byte and short) instructions."

9 years agoMerge "ART: Fix run-test 118"
Andreas Gampe [Tue, 24 Feb 2015 16:07:56 +0000 (16:07 +0000)]
Merge "ART: Fix run-test 118"

9 years agoART: Fix run-test 118
Andreas Gampe [Tue, 24 Feb 2015 16:06:34 +0000 (08:06 -0800)]
ART: Fix run-test 118

Need to strip error logging for no-prebuild config.

Change-Id: I870bbbf3b6c7e92996e40e45b1bfa58c7e80f709

9 years agonot-int can also take non-int (byte and short) instructions.
Nicolas Geoffray [Tue, 24 Feb 2015 16:02:06 +0000 (16:02 +0000)]
not-int can also take non-int (byte and short) instructions.

So we should use the result-type instead if the input type
for knowning what instruction to use.

Bug: 19454010
Change-Id: I88782ad27ae8c8e1b7868afede5057d26f14685a

9 years agoMerge "ART: Enable more libcore tests (crypto)"
David Brazdil [Tue, 24 Feb 2015 15:43:18 +0000 (15:43 +0000)]
Merge "ART: Enable more libcore tests (crypto)"

9 years agoMerge "Fix inlining in the presence of multiple returns."
Nicolas Geoffray [Tue, 24 Feb 2015 15:08:36 +0000 (15:08 +0000)]
Merge "Fix inlining in the presence of multiple returns."

9 years agoFix inlining in the presence of multiple returns.
Nicolas Geoffray [Tue, 24 Feb 2015 13:35:38 +0000 (13:35 +0000)]
Fix inlining in the presence of multiple returns.

One return could actually return a phi, so doing a phi check for
knowing if the result phi was already created was bogus.

Bug: 19454010

Change-Id: Iee703a2d1071ae263092354465eda368e5d6770d

9 years agoMerge "ART: Disable one more libcore test failing on hammerhead"
David Brazdil [Tue, 24 Feb 2015 14:17:32 +0000 (14:17 +0000)]
Merge "ART: Disable one more libcore test failing on hammerhead"

9 years agoART: Disable one more libcore test failing on hammerhead
David Brazdil [Tue, 24 Feb 2015 14:13:05 +0000 (14:13 +0000)]
ART: Disable one more libcore test failing on hammerhead

One more test from the harmony package fails due to the locale issue
on hammerhead. This patch disables it.

Change-Id: I1f3941bfcc20d5b2e23fe5ae963676138eecccb8

9 years agoART: Enable more libcore tests (crypto)
David Brazdil [Tue, 24 Feb 2015 13:33:01 +0000 (13:33 +0000)]
ART: Enable more libcore tests (crypto)

This patch enables additional libcore tests run by the ART buildbots.
These are tests from the org.apache.harmony.crypto and
libcore.javax.crypto packages and take extra ~8 minutes to run on
target and ~2 minutes on host.

Change-Id: I7cd314eca654320104fee26d8c34437450814dc2

9 years agoMerge "ART: Disable libcore tests failing on bots"
David Brazdil [Tue, 24 Feb 2015 12:47:42 +0000 (12:47 +0000)]
Merge "ART: Disable libcore tests failing on bots"

9 years agoMerge "[optimizing] Add if-context sensitivity for null popagation."
Calin Juravle [Tue, 24 Feb 2015 12:39:53 +0000 (12:39 +0000)]
Merge "[optimizing] Add if-context sensitivity for null popagation."

9 years agoART: Disable libcore tests failing on bots
David Brazdil [Tue, 24 Feb 2015 12:36:03 +0000 (12:36 +0000)]
ART: Disable libcore tests failing on bots

Some tests in previous commits are failing on the testing devices.
Tests dependent on DateTimeFormatter fail on hammerhead due to
misconfigured locales and tests with a server running on localhost
fail due to the loop device being switched off on volantisg. These
problems will be further investigated but this patch disables them
for the time being.

Change-Id: I2dde178925de555f26a4d532be15487f499571b4

9 years ago[optimizing] Add if-context sensitivity for null popagation.
Calin Juravle [Mon, 23 Feb 2015 16:46:57 +0000 (16:46 +0000)]
[optimizing] Add if-context sensitivity for null popagation.

Change-Id: I3725b6c6a6cf44440c34a1bfb67e623531e665d6

9 years agoMerge "ART: Enable more libcore tests (harmony package)"
David Brazdil [Tue, 24 Feb 2015 10:52:45 +0000 (10:52 +0000)]
Merge "ART: Enable more libcore tests (harmony package)"

9 years agoART: Enable more libcore tests (harmony package)
David Brazdil [Tue, 24 Feb 2015 10:44:29 +0000 (10:44 +0000)]
ART: Enable more libcore tests (harmony package)

This patch enables additional libcore tests run by the ART buildbots.
These are tests from the org.apache.harmony.* packages and take extra
~5 minutes to run on target and ~2 minutes on host.

Change-Id: Idb9fc09ac7df460bd4337bc5151301adddea4b97

9 years agoMerge "ART: Enable more libcore tests (libcore.* package)"
David Brazdil [Tue, 24 Feb 2015 10:21:43 +0000 (10:21 +0000)]
Merge "ART: Enable more libcore tests (libcore.* package)"

9 years agoART: Enable more libcore tests (libcore.* package)
David Brazdil [Tue, 24 Feb 2015 10:12:06 +0000 (10:12 +0000)]
ART: Enable more libcore tests (libcore.* package)

This patch enables additional libcore tests run by the ART buildbots.
These are tests from the libcore.* packages and take extra ~3 minutes
to run on both host and target.

Change-Id: I89512836adb87b20ed5f9446d3b3ee245f9162dc

9 years agoMerge "JDWP: update thread synchronization"
Sebastien Hertz [Tue, 24 Feb 2015 09:28:49 +0000 (09:28 +0000)]
Merge "JDWP: update thread synchronization"

9 years agoMerge "Revert "Add JIT""
Nicolas Geoffray [Tue, 24 Feb 2015 08:11:23 +0000 (08:11 +0000)]
Merge "Revert "Add JIT""

9 years agoRevert "Add JIT"
Nicolas Geoffray [Tue, 24 Feb 2015 08:10:57 +0000 (08:10 +0000)]
Revert "Add JIT"

Sorry, run-test crashes on target:
0-05 12:15:51.633 I/DEBUG   (27995): Abort message: 'art/runtime/mirror/art_method.cc:349] Check failed: PcIsWithinQuickCode(reinterpret_cast<uintptr_t>(code), pc) java.lang.Throwable java.lang.Throwable.fillInStackTrace() pc=71e3366b code=0x71e3362d size=ad000000'
10-05 12:15:51.633 I/DEBUG   (27995):     r0 00000000  r1 0000542b  r2 00000006  r3 00000000
10-05 12:15:51.633 I/DEBUG   (27995):     r4 00000006  r5 b6f9addc  r6 00000002  r7 0000010c
10-05 12:15:51.633 I/DEBUG   (27995):     r8 b63fe1e8  r9 be8e1418  sl b6427400  fp b63fcce0
10-05 12:15:51.633 I/DEBUG   (27995):     ip 0000542b  sp be8e1358  lr b6e9a27b  pc b6e9c280  cpsr 40070010
10-05 12:15:51.633 I/DEBUG   (27995):

Bug: 17950037
This reverts commit 2535abe7d1fcdd0e6aca782b1f1932a703ed50a4.

Change-Id: I6f88849bc6f2befed0c0aaa0b7b2a08c967a83c3

9 years agoMerge "Delete bad DCHECK"
Mathieu Chartier [Tue, 24 Feb 2015 02:29:36 +0000 (02:29 +0000)]
Merge "Delete bad DCHECK"

9 years agoDelete bad DCHECK
Mathieu Chartier [Tue, 24 Feb 2015 02:24:29 +0000 (18:24 -0800)]
Delete bad DCHECK

The declaring class of a super field could be in a different dex file.
Fixes sdk build.

Change-Id: I6e1738603e71c3c673fa12d9aaa25c9eb84db406

9 years agoMerge "Add JIT"
Mathieu Chartier [Tue, 24 Feb 2015 01:29:46 +0000 (01:29 +0000)]
Merge "Add JIT"

9 years agoMerge "ART: Strip error message in run-test"
Andreas Gampe [Tue, 24 Feb 2015 01:17:13 +0000 (01:17 +0000)]
Merge "ART: Strip error message in run-test"

9 years agoART: Strip error message in run-test
Andreas Gampe [Tue, 24 Feb 2015 01:15:47 +0000 (17:15 -0800)]
ART: Strip error message in run-test

It seems the output is flaky - sometimes the LOG(ERROR) makes it
to output.txt, sometimes it doesn't. So instead always strip it.

Change-Id: I409b98eb9a5bf0c2e6c1df16ff3b8013a2a85cc5

9 years agoAdd JIT
Mathieu Chartier [Tue, 17 Feb 2015 18:38:49 +0000 (10:38 -0800)]
Add JIT

Currently disabled by default unless -Xjit is passed in.

The proposed JIT is a method JIT which works by utilizing interpreter
instrumentation to request compilation of hot methods async during
runtime.

JIT options:
-Xjit / -Xnojit
-Xjitcodecachesize:N
-Xjitthreshold:integervalue

The JIT has a shared copy of a compiler driver which is accessed
by worker threads to compile individual methods.

Added JIT code cache and data cache, currently sized at 2 MB
capacity by default. Most apps will only fill a small fraction of
this cache however.

Added support to the compiler for compiling interpreter quickened
byte codes.

Added test target ART_TEST_JIT=TRUE and --jit for run-test.

TODO:
Clean up code cache.
Delete compiled methods after they are added to code cache.
Add more optimizations related to runtime checks e.g. direct pointers
for invokes.
Add method recompilation.
Move instrumentation to DexFile to improve performance and reduce
memory usage.

Bug: 17950037

Change-Id: Ifa5b2684a2d5059ec5a5210733900aafa3c51bca

9 years agoMerge "Only detach shutdown thread if we attached it"
Mathieu Chartier [Tue, 24 Feb 2015 00:33:15 +0000 (00:33 +0000)]
Merge "Only detach shutdown thread if we attached it"

9 years agoOnly detach shutdown thread if we attached it
Mathieu Chartier [Mon, 23 Feb 2015 23:31:58 +0000 (15:31 -0800)]
Only detach shutdown thread if we attached it

Speculative fix to:
"Request to unregister unattached thread" bug.

Bug: 18713034
Change-Id: I6e69088dd66970c704010397a11e3758cf107ff9

9 years agoMerge "ART: Allow the execution to stop if the compilation fails via an option"
Andreas Gampe [Mon, 23 Feb 2015 23:48:29 +0000 (23:48 +0000)]
Merge "ART: Allow the execution to stop if the compilation fails via an option"

9 years agoART: Allow the execution to stop if the compilation fails via an option
Jean Christophe Beyler [Fri, 12 Sep 2014 19:03:25 +0000 (12:03 -0700)]
ART: Allow the execution to stop if the compilation fails via an option

The current implementation continues the execution of the application if
dex2oat fails by relying on the interpreter.

This patch adds a -Xno-dex-file-fallback option to stop the default behavior.
This can be used two-fold.

First, one can enforce that a runtime only starts with a boot image. A
follow-up patch will ensure that dex2oat (for apps) and patchoat in general
request that mode and close gracefully otherwise.

Second, this can be used for testing and debugging purposes, as it ensures
that compiler failures & aborts are not silently ignored.

Add testing.

Bug: 19100590
Change-Id: Iaf07b5ccf00942ca8a8ec8687599320a3ddbc089
Signed-off-by: Jean Christophe Beyler <jean.christophe.beyler@intel.com>
9 years agoMerge "Optimizing: Remove redundant hash set copy in GVN"
David Brazdil [Mon, 23 Feb 2015 22:27:51 +0000 (22:27 +0000)]
Merge "Optimizing: Remove redundant hash set copy in GVN"

9 years agoMerge "Avoid unaligned accesses (SIGBUG/BUS_ADRALN) in IRT."
Hiroshi Yamauchi [Mon, 23 Feb 2015 19:30:29 +0000 (19:30 +0000)]
Merge "Avoid unaligned accesses (SIGBUG/BUS_ADRALN) in IRT."

9 years agoAvoid unaligned accesses (SIGBUG/BUS_ADRALN) in IRT.
Hiroshi Yamauchi [Mon, 23 Feb 2015 19:14:40 +0000 (11:14 -0800)]
Avoid unaligned accesses (SIGBUG/BUS_ADRALN) in IRT.

Pointers in IrtEntry aren't currently aligned under 64 bit builds. But
unaligned atomic stores (store exclusive) do not work on arm64 (causes
SIGBUG/BUS_ADRALN). Fix CC collector crashes caused by this.

Bug: 12687968
Change-Id: I1d2f5376778a9a1e5cfea876f1f57d7a88ad5445

9 years agoMerge "More checker tests for BCE."
Mingyao Yang [Mon, 23 Feb 2015 19:25:11 +0000 (19:25 +0000)]
Merge "More checker tests for BCE."

9 years agoMore checker tests for BCE.
Mingyao Yang [Tue, 10 Feb 2015 02:13:26 +0000 (18:13 -0800)]
More checker tests for BCE.

Also make sure the check on MonotonicValueRange narrow is more strict.
Plus some handling on array length division such as array.length/2.
Added checker tests for each case.

Change-Id: I9f32fc5f6ca1f3da8edec576de66b44d85a50bc6

9 years agoMerge "ART: Introduce NO_RETURN, Mark DoLongJump noreturn"
Andreas Gampe [Mon, 23 Feb 2015 18:41:53 +0000 (18:41 +0000)]
Merge "ART: Introduce NO_RETURN, Mark DoLongJump noreturn"

9 years agoART: Introduce NO_RETURN, Mark DoLongJump noreturn
Andreas Gampe [Mon, 23 Feb 2015 16:12:24 +0000 (08:12 -0800)]
ART: Introduce NO_RETURN, Mark DoLongJump noreturn

Add NO_RETURN macro that adds C++11 noreturn attribute. Mark
DoLongJump methods as noreturn.

Change-Id: Ifde4318e370493237050d4c1349285a0382df23f

9 years agoOptimizing: Remove redundant hash set copy in GVN
David Brazdil [Mon, 23 Feb 2015 18:02:02 +0000 (18:02 +0000)]
Optimizing: Remove redundant hash set copy in GVN

During the GVN analysis, a basic block inherits the set of movable
instructions from its dominator. If the block is the only successor
of the dominating block, there is no need for cloning of the parent
set (a very expensive operation).

Change-Id: I59e033b9e9e093984dc8e903e3a7be1cb3645cc2

9 years agoMerge "[optimizing] Fix float addition on x86."
Calin Juravle [Mon, 23 Feb 2015 16:05:21 +0000 (16:05 +0000)]
Merge "[optimizing] Fix float addition on x86."

9 years ago[optimizing] Fix float addition on x86.
Calin Juravle [Mon, 23 Feb 2015 15:53:39 +0000 (15:53 +0000)]
[optimizing] Fix float addition on x86.

Change-Id: Ic39aaae89b8e5184b98001ea67221a3564e9334a

9 years agoMerge "Each primitive kind now spills to different locations."
Nicolas Geoffray [Mon, 23 Feb 2015 15:17:07 +0000 (15:17 +0000)]
Merge "Each primitive kind now spills to different locations."

9 years agoMerge "Optimizing: Speed up HInstruction use removal"
David Brazdil [Mon, 23 Feb 2015 15:14:09 +0000 (15:14 +0000)]
Merge "Optimizing: Speed up HInstruction use removal"

9 years agoOptimizing: Speed up HInstruction use removal
David Brazdil [Tue, 17 Feb 2015 18:33:36 +0000 (18:33 +0000)]
Optimizing: Speed up HInstruction use removal

Similarly to a previous commit on HEnvironment use removal, this patch
adds links from instructions to their respective inputs' use lists for
contant-time removal at the cost of doubling the size of input lists
(from one pointer per entry to two). Manual testing shows that this
significantly reduces the time required to transform HGraph to SSA
form for some huge methods.

Change-Id: I8dc3e4b0c48a50ac1481eb55c31093b99f4dc29f

9 years agoEach primitive kind now spills to different locations.
Nicolas Geoffray [Mon, 23 Feb 2015 14:14:57 +0000 (14:14 +0000)]
Each primitive kind now spills to different locations.

Having different slots depending on the types greatly simplifies
the parallel move resolver. It also avoids doing FPU <-> Core
register swaps, and force backends to implement such a swap.

Change-Id: Ide9f0452e7ccf9efb8adddbcc246d44b937b253c

9 years agoHandle another untyped dex CONST issue
buzbee [Fri, 20 Feb 2015 22:07:27 +0000 (14:07 -0800)]
Handle another untyped dex CONST issue

The lack of typing for dex byte code instructions causes problems
for the Quick compiler's register allocation scheme.  In Quick,
Dalvik virtual registers can have multiple "home locations" depending
on type.  Because CONST operators are untyped and their result can
be used in multiple type contexts, the code generator must ensure
that any CONST #0 definitions of potential floating point uses
are also flushed to the associated reference view for the vreg
(if it exists) to support GC.  Previously, this issue was believed
to have been fixed by c/255389 (see internal b/7250540).

However, the earlier fix failed to account for const definitions
propogating to other vregs via MOVE.  This CL addresses that case by
applying the earlier workaround to MOVE operators.

Note that there are a few other places where we attempt to deal
with this general problem.  First, the verifier will disable
compilation for dex code sequences which have type mismatches on non-const
operations.  Second, during vreg analysis we detect type mismatches on
Phi insertion, and disable register promotion if there is a mismatch.

b.android.com/147187

Change-Id: I70f27a59e6f84a89777db91ceacaa94a42f5e952

9 years agoMerge "Implement a simple count based boot marker."
Narayan Kamath [Sat, 21 Feb 2015 11:19:21 +0000 (11:19 +0000)]
Merge "Implement a simple count based boot marker."

9 years agoImplement a simple count based boot marker.
Narayan Kamath [Mon, 16 Feb 2015 13:51:51 +0000 (13:51 +0000)]
Implement a simple count based boot marker.

We write the number of failed boots to the marker and only
prune the dalvik cache if the number of consecutive failed
boots is higher than a predefined threshold.

Note that the code is forgiving of errors related to boot
markers; we continue the boot process even if we're unable
to create / write or flush the boot marker.

bug: 19360096
Change-Id: Ia17c3b783318ddf43c9199d0f7f09c54a4176667

9 years agoMerge "ART: Add VALGRIND_LIB for gtests"
Andreas Gampe [Fri, 20 Feb 2015 21:19:05 +0000 (21:19 +0000)]
Merge "ART: Add VALGRIND_LIB for gtests"

9 years agoART: Add VALGRIND_LIB for gtests
Andreas Gampe [Fri, 20 Feb 2015 21:15:15 +0000 (13:15 -0800)]
ART: Add VALGRIND_LIB for gtests

Ensure that tests can be run without lunch-ing a target.

Change-Id: If9b4a6c56952418fde703145600717c84e11ef99

9 years agoMerge "Make UseTLAB a bool flag."
Hiroshi Yamauchi [Fri, 20 Feb 2015 19:35:57 +0000 (19:35 +0000)]
Merge "Make UseTLAB a bool flag."