OSDN Git Service

android-x86/art.git
7 years agoARM/ARM64: Improve comments for kSaveEverything assembly.
Vladimir Marko [Mon, 22 Aug 2016 15:02:12 +0000 (16:02 +0100)]
ARM/ARM64: Improve comments for kSaveEverything assembly.

Addresses post-submit comments for
    https://android-review.googlesource.com/255290

Test: Rely on TreeHugger (just adds comments).
Bug: 30212852
Change-Id: I6ccc004774e18bbd83e3940e908cfc6c521142db

7 years agoMerge "ART: Enforce arena alignment in Realloc"
Treehugger Robot [Sat, 20 Aug 2016 05:19:15 +0000 (05:19 +0000)]
Merge "ART: Enforce arena alignment in Realloc"

7 years agoART: Enforce arena alignment in Realloc
Andreas Gampe [Sat, 20 Aug 2016 03:22:19 +0000 (20:22 -0700)]
ART: Enforce arena alignment in Realloc

Enforce the expected alignment in Realloc when the given buffer
can be expanded.

Add debug checks to prove alignment guarantees.

Change-Id: I7dbaf3ca0896f714cff07ff08f067a7c088a6c8d
Test: m test-art-host

7 years agoMerge "ART: Added DCHECKS to find alignment breakage."
Treehugger Robot [Sat, 20 Aug 2016 00:26:36 +0000 (00:26 +0000)]
Merge "ART: Added DCHECKS to find alignment breakage."

7 years agoART: Added DCHECKS to find alignment breakage.
Christina Wadsworth [Fri, 19 Aug 2016 22:58:05 +0000 (15:58 -0700)]
ART: Added DCHECKS to find alignment breakage.

Change-Id: I47aed866ec3c3858dc3ee27cb84cfaddf44009d1

7 years agoMerge "Revert "Improve the thread flip.""
Treehugger Robot [Fri, 19 Aug 2016 22:12:40 +0000 (22:12 +0000)]
Merge "Revert "Improve the thread flip.""

7 years agoMerge "ART: Fixing build breakage (sharpening DCHECK)."
Mathieu Chartier [Fri, 19 Aug 2016 21:53:57 +0000 (21:53 +0000)]
Merge "ART: Fixing build breakage (sharpening DCHECK)."

7 years agoMerge "ART: Fixing build breakage."
Mathieu Chartier [Fri, 19 Aug 2016 21:53:20 +0000 (21:53 +0000)]
Merge "ART: Fixing build breakage."

7 years agoART: Fixing build breakage (sharpening DCHECK).
Christina Wadsworth [Fri, 19 Aug 2016 21:38:01 +0000 (14:38 -0700)]
ART: Fixing build breakage (sharpening DCHECK).

Change-Id: I7ba19808dde0712739a278075da57f724166a233

7 years agoART: Fixing build breakage.
Christina Wadsworth [Fri, 19 Aug 2016 20:28:19 +0000 (13:28 -0700)]
ART: Fixing build breakage.

Caused by Ic654a1fb9c1ae127dde59290bf36a23edb55ca8e.

Change-Id: Ia0c0fbcb6c8b141c5ee30a89ab6bc1c653916698

7 years agoMerge "Fix building tests with partial arch codegen support"
Colin Cross [Fri, 19 Aug 2016 20:16:29 +0000 (20:16 +0000)]
Merge "Fix building tests with partial arch codegen support"

7 years agoMerge "Implemented first version of java fuzz testing script."
Treehugger Robot [Fri, 19 Aug 2016 20:15:03 +0000 (20:15 +0000)]
Merge "Implemented first version of java fuzz testing script."

7 years agoImplemented first version of java fuzz testing script.
Aart Bik [Wed, 17 Aug 2016 23:51:12 +0000 (16:51 -0700)]
Implemented first version of java fuzz testing script.

Test: run_java_fuzz_test.py

BUG=30610121

Change-Id: I2a802476bcb0986e2891748ad85f8feac21656a8

7 years agoRevert "Improve the thread flip."
Hiroshi Yamauchi [Fri, 19 Aug 2016 17:56:11 +0000 (17:56 +0000)]
Revert "Improve the thread flip."

This reverts commit e61e6fdd404b8f04bd33fc3ea49e38e92e31ad69.

ThreadStress failing.

Bug: 29517059
Bug: 12687968
Change-Id: I0ee8ef04f77c5f9378e4bd21c5dedb4e435a2dc6

7 years agoFix building tests with partial arch codegen support
Colin Cross [Thu, 18 Aug 2016 20:45:24 +0000 (13:45 -0700)]
Fix building tests with partial arch codegen support

Add conditionals around more code that is only used for codegen for
specific architectures, and move a few more files into the
architecture-specific codegen lists.

Tests: ART_HOST_CODEGEN_ARCHS="x86_64 mips" m -j ART_TARGET_CODEGEN_ARCHS=svelte test-art-host
Bug: 30928847
Change-Id: I0444d15e1cafe4c9b13ff78718c3b13b544270e7

7 years agoMerge "ART: Implement a fixed size string dex cache"
Mathieu Chartier [Fri, 19 Aug 2016 17:33:03 +0000 (17:33 +0000)]
Merge "ART: Implement a fixed size string dex cache"

7 years agoMerge "Add support for Baker read barriers in SystemArrayCopy intrinsics."
Roland Levillain [Fri, 19 Aug 2016 15:07:31 +0000 (15:07 +0000)]
Merge "Add support for Baker read barriers in SystemArrayCopy intrinsics."

7 years agoAdd support for Baker read barriers in SystemArrayCopy intrinsics.
Roland Levillain [Fri, 19 Aug 2016 11:02:34 +0000 (12:02 +0100)]
Add support for Baker read barriers in SystemArrayCopy intrinsics.

Benchmarks (ARM64) score variations on Nexus 5X with CPU
cores clamped at 960000 Hz (aosp_bullhead-userdebug build):
- Ritzperf - average (lower is better):       -3.03% (slightly better)
- CaffeineMark - average (higher is better):  +1.26% (slightly better)
- DeltaBlue (lower is better):               -10.50% (better)
- Richards - average (lower is better):       -3.36% (slightly better)
- SciMark2 - average (higher is better):      +0.26% (virtually unchanged)

Details about Ritzperf benchmarks with meaningful variations
(lower is better):
- FormulaEvaluationActions.EvaluateAndApplyChanges: -13.26% (better)
- FormulaEvaluationActions.EvaluateCascadingSums:   -10.94% (better)
- FormulaEvaluationActions.EvaluateComplexFormulas: -15.50% (better)
- FormulaEvaluationActions.EvaluateFibonacci:       -10.41% (better)
- FormulaEvaluationActions.EvaluateLargeSums:        +6.02% (worse)

Boot image code size variation on Nexus 5X
(aosp_bullhead-userdebug build):
- total ARM64 framework Oat files size change:
  107047632 bytes -> 107154128 bytes (+0.10%)
- total ARM framework Oat files size change:
  90932028 bytes -> 91009852 bytes (+0.09%)

Test: ART host and target (ARM, ARM64) tests + Nexus 5X boot.
Bug: 29516905
Bug: 29506760
Bug: 12687968
Change-Id: I85431368d09965687a0301ae2eb3c991f276ce5d

7 years agoMerge "Improve the thread flip."
Treehugger Robot [Fri, 19 Aug 2016 01:29:50 +0000 (01:29 +0000)]
Merge "Improve the thread flip."

7 years agoMerge "Color spill slots in gc regalloc"
Treehugger Robot [Fri, 19 Aug 2016 00:30:44 +0000 (00:30 +0000)]
Merge "Color spill slots in gc regalloc"

7 years agoMerge "Log optimization passes in verbose mode"
Aart Bik [Thu, 18 Aug 2016 23:35:09 +0000 (23:35 +0000)]
Merge "Log optimization passes in verbose mode"

7 years agoART: Implement a fixed size string dex cache
Christina Wadsworth [Thu, 18 Aug 2016 17:37:42 +0000 (10:37 -0700)]
ART: Implement a fixed size string dex cache

Previously, the string dex cache was dex_file->NumStringIds() size, and
@ruhler found that only ~1% of that cache was ever getting filled. Since
many of these string dex caches were previously 100,000+ indices in
length, we're wasting a few hundred KB per app by storing null pointers.
The intent of this project was to reduce the space the string dex cache
is using, while not regressing on time that much. This is the first of a
few CLs, which implements the new fixed size array and disables the
compiled code so it always goes slow path. In four other CLs, I
implemented a "medium path" that regresses from the previous "fast path"
only a bit in assembly in the entrypoints. @vmarko will introduce new
compiled code in the future so that we ultimately won't be regressing on
time at all. Overall, space savings have been confirmed as on the order
of 100 KB per application.

A 4-5% slow down in art-opt on Golem, and no noticeable slow down in the
interpreter. The opt slow down should be diminished once the new
compiled code is introduced.

Test: m test-art-host

Bug: 20323084

Change-Id: Ic654a1fb9c1ae127dde59290bf36a23edb55ca8e

7 years agoMerge "dex_instruction: k45cc HasVarArgs"
Treehugger Robot [Thu, 18 Aug 2016 10:56:54 +0000 (10:56 +0000)]
Merge "dex_instruction: k45cc HasVarArgs"

7 years agodex_instruction: k45cc HasVarArgs
Narayan Kamath [Mon, 15 Aug 2016 15:08:18 +0000 (16:08 +0100)]
dex_instruction: k45cc HasVarArgs

Change-Id: I3be92269bb78c96ec41ab533c5792bd369c02178
Test: make test-art-host
bug: 30550796

7 years agoMerge "Fix optimizing compiler not building for svelte"
Treehugger Robot [Thu, 18 Aug 2016 06:07:42 +0000 (06:07 +0000)]
Merge "Fix optimizing compiler not building for svelte"

7 years agoImprove the thread flip.
Hiroshi Yamauchi [Tue, 2 Aug 2016 23:03:06 +0000 (16:03 -0700)]
Improve the thread flip.

- In addition to the threads that are suspended in FullSuspendCheck(),
  prioritize the resume of threads that are blocking for the thread
  flip at the JNI critical section entry and threads are about to
  transition to runnable (eg. blocking at the SOA entry from JNI).
- Shorten the length of the thread flip critical
  section (ThreadFlipBegin/End).
- Add some systrace scopes.

Bug: 29517059
Bug: 12687968
Test: test-art-host, Ritz EAAC, N9 libartd boot.
Change-Id: Idecec9936ae432c23d5f83321ba13339852018df

7 years agoMerge "Remove unused include."
Treehugger Robot [Thu, 18 Aug 2016 03:17:20 +0000 (03:17 +0000)]
Merge "Remove unused include."

7 years agoFix optimizing compiler not building for svelte
Wojciech Staszkiewicz [Thu, 18 Aug 2016 02:15:51 +0000 (19:15 -0700)]
Fix optimizing compiler not building for svelte

Fixes compiler not building when some of the codegen paths
are disabled.

Test: mmma -j art ART_TARGET_CODEGEN_ARCHS=svelte
m -j32 test-art-host

BUG=30928847

Change-Id: I52c78e8a4e507f74b1f2a39352970079721b737e

7 years agoRemove unused include.
Dmitriy Filchenko [Thu, 18 Aug 2016 00:58:08 +0000 (17:58 -0700)]
Remove unused include.

BUG: 29824203

Change-Id: Iea74f7b166c3e81c0d00c92442f27e162631e5ae

7 years agoLog optimization passes in verbose mode
Wojciech Staszkiewicz [Wed, 17 Aug 2016 00:31:19 +0000 (17:31 -0700)]
Log optimization passes in verbose mode

Introduce verbose logging of optimization passes run during
compilation.

Test: m test-art-host -j32
art -Xcompiler-option --runtime-arg -Xcompiler-option -verbose:compiler
-classpath classes.dex Test
Change-Id: Iae98ce9dcafc252f2d0eec138aa05b34e424bd2a

7 years agoMerge "Remove ART_DEFAULT_INSTRUCTION_SET_FEATURES"
Treehugger Robot [Thu, 18 Aug 2016 01:24:31 +0000 (01:24 +0000)]
Merge "Remove ART_DEFAULT_INSTRUCTION_SET_FEATURES"

7 years agoRemove ART_DEFAULT_INSTRUCTION_SET_FEATURES
Colin Cross [Wed, 17 Aug 2016 23:52:08 +0000 (16:52 -0700)]
Remove ART_DEFAULT_INSTRUCTION_SET_FEATURES

It is never used.

Test: mmma -j art
Change-Id: I014a8ab11d3a3241c075503bc2674279130cb86f

7 years agoMerge "Revert "Logging improvements for C++ / Java structure issues""
Treehugger Robot [Wed, 17 Aug 2016 23:58:47 +0000 (23:58 +0000)]
Merge "Revert "Logging improvements for C++ / Java structure issues""

7 years agoRevert "Logging improvements for C++ / Java structure issues"
Mathieu Chartier [Wed, 17 Aug 2016 22:43:52 +0000 (22:43 +0000)]
Revert "Logging improvements for C++ / Java structure issues"

Bug: 30913672

This reverts commit d833525cfe3ac2086e3156a8b3193c05974d6316.

Change-Id: I4e51591b49d54b8bb3e920fbc221c21aca0084fb

7 years agoMerge "Add newline for INTERNAL_FATAL automatically"
Mathieu Chartier [Wed, 17 Aug 2016 22:43:17 +0000 (22:43 +0000)]
Merge "Add newline for INTERNAL_FATAL automatically"

7 years agoColor spill slots in gc regalloc
Matthew Gharrity [Sat, 13 Aug 2016 02:07:11 +0000 (19:07 -0700)]
Color spill slots in gc regalloc

Coloring spill slots avoids pathologically large stack
sizes by reusing spill slots when possible.

Test: ART_TEST_OPTIMIZING_GRAPH_COLOR=true m test-art-host

Change-Id: I4b4aea859c78b0515758f8b057ee870dbbfc2300

7 years agoMerge "Track cumulative objects and bytes copied for CC"
Mathieu Chartier [Wed, 17 Aug 2016 20:22:24 +0000 (20:22 +0000)]
Merge "Track cumulative objects and bytes copied for CC"

7 years agoAdd newline for INTERNAL_FATAL automatically
Mathieu Chartier [Wed, 17 Aug 2016 17:28:48 +0000 (10:28 -0700)]
Add newline for INTERNAL_FATAL automatically

Before, the behavior was different compared to other log levels
that automatically added newlines.

Bug: 30913672

Test: test-art-host

Change-Id: I02147d8ac7ea38f3e93450ef5290cdfed7b5a2eb

7 years agoMerge "Add skeleton of a jvmti plugin for art."
Treehugger Robot [Wed, 17 Aug 2016 19:15:54 +0000 (19:15 +0000)]
Merge "Add skeleton of a jvmti plugin for art."

7 years agoMerge "Blacklist failing libcore tests."
Narayan Kamath [Wed, 17 Aug 2016 18:50:12 +0000 (18:50 +0000)]
Merge "Blacklist failing libcore tests."

7 years agoBlacklist failing libcore tests.
Narayan Kamath [Wed, 17 Aug 2016 18:14:01 +0000 (19:14 +0100)]
Blacklist failing libcore tests.

bug: 30917411
test: art/tools/buildbot_build.sh && art/tools/run-libcore-tests.sh
Change-Id: Ib5e9adb6e9222d607b6d79f0f7eebf3d313272d2

7 years agoAdd skeleton of a jvmti plugin for art.
Alex Light [Thu, 11 Aug 2016 22:35:28 +0000 (15:35 -0700)]
Add skeleton of a jvmti plugin for art.

Test: mma test-art-host-run-test-901-hello-ti-agent
Change-Id: If6807b6238d57471e4ba0dd75c717721246443f6

7 years agoTrack cumulative objects and bytes copied for CC
Mathieu Chartier [Wed, 17 Aug 2016 17:07:29 +0000 (10:07 -0700)]
Track cumulative objects and bytes copied for CC

Also print out these values when dumping GC performance info.

Bug: 12687968

Test: Test that values are reasonable after running EAAC.

Change-Id: Id04fadeaf52511560fd4b261f5287ea0a5dae9d4

7 years agoMerge "Logging improvements for C++ / Java structure issues"
Neil Fuller [Wed, 17 Aug 2016 13:19:28 +0000 (13:19 +0000)]
Merge "Logging improvements for C++ / Java structure issues"

7 years agoMerge "Adjustments for changes made to java.lang.Class"
Neil Fuller [Wed, 17 Aug 2016 12:17:06 +0000 (12:17 +0000)]
Merge "Adjustments for changes made to java.lang.Class"

7 years agoLogging improvements for C++ / Java structure issues
Neil Fuller [Wed, 17 Aug 2016 10:39:26 +0000 (11:39 +0100)]
Logging improvements for C++ / Java structure issues

When C++ and Java structures are different sizes or arranged
differently some log statements are useful but are hard to
pick out because they merge with other log lines. Adding
"\n" to the end of them makes it easier to read.

Change-Id: Ia6d78e0832c7814bf1fef42bb3d8c056f83766bd
Test: make build-art-host

7 years agoAdjustments for changes made to java.lang.Class
Neil Fuller [Tue, 16 Aug 2016 18:53:07 +0000 (19:53 +0100)]
Adjustments for changes made to java.lang.Class

Fix up the runtime after removal of the annotationType field
from java.lang.Class.

Test: make test-art-host / boot a device
Change-Id: I5f37567d2b5ec381b35e66fef29287abe307b840

7 years agoMerge "Always mark zygote large objects for CC"
Treehugger Robot [Wed, 17 Aug 2016 03:18:36 +0000 (03:18 +0000)]
Merge "Always mark zygote large objects for CC"

7 years agoAlways mark zygote large objects for CC
Mathieu Chartier [Tue, 16 Aug 2016 19:15:59 +0000 (12:15 -0700)]
Always mark zygote large objects for CC

Prevent needing to gray holders of zygote large objects.
System wide zygote space PSS after boot:
12644 kB -> 5571 kB for CC.

Also PSS reduction in zygote large objects themselves since their
gray bit would have been set each GC.
Overall LOS savings hard to measure, could be up to 316 * 4KB per
app since there are 316 zygote large objects.

Also clear mod-union tables for image spaces to prevent dirty
image pages if any of the image spaces point to zygote large objects.
System wide .art mmap: 37432 kB -> 34372 kB

System server before (N6P):
LOS shared dirty: 12888 kB
Zygote space shared dirty: 700 kB
Zygote space private dirty: 868 kB
.art private dirty: 1696 kB

After:
LOS shared dirty 13672 kB
Zygote space shared dirty: 1072 kB
Zygote space private dirty: 496 kB
.art private dirty: 1432 kB

Bug: 29516968

Test: test-art-host with baker CC, debug N6P phone booting

Change-Id: Ia37ce2c11217cf56885bd1d1dc084332fcbb7843

7 years agoMerge "jni: Fast path for @FastNative annotated java methods"
Treehugger Robot [Tue, 16 Aug 2016 21:51:02 +0000 (21:51 +0000)]
Merge "jni: Fast path for @FastNative annotated java methods"

7 years agojni: Fast path for @FastNative annotated java methods
Igor Murashkin [Fri, 29 Jul 2016 16:51:58 +0000 (09:51 -0700)]
jni: Fast path for @FastNative annotated java methods

Adds a faster path for java methods annotated with
dalvik.annotation.optimization.FastNative .

Intended to replace usage of fast JNI (registering with "!(FOO)BAR" descriptors).

Performance Microbenchmark Results (Angler):
* Regular JNI cost in nanoseconds: 115
* Fast JNI cost in nanoseconds: 60
* @FastNative cost in nanoseconds: 36

Summary: Up to 67% faster (vs fast jni) JNI transition cost

Change-Id: Ic23823ae0f232270c068ec999fd89aa993894b0e

7 years agoMerge "Update ahat README and manifest for release 0.7."
Treehugger Robot [Tue, 16 Aug 2016 19:53:15 +0000 (19:53 +0000)]
Merge "Update ahat README and manifest for release 0.7."

7 years agoMerge "Remove unused GetDalvikCacheFilenameOrDie."
Treehugger Robot [Tue, 16 Aug 2016 17:59:51 +0000 (17:59 +0000)]
Merge "Remove unused GetDalvikCacheFilenameOrDie."

7 years agoMerge "Welcome to JavaFuzz as our latest A[a]rt tools team member!"
Treehugger Robot [Tue, 16 Aug 2016 16:34:36 +0000 (16:34 +0000)]
Merge "Welcome to JavaFuzz as our latest A[a]rt tools team member!"

7 years agoWelcome to JavaFuzz as our latest A[a]rt tools team member!
Aart Bik [Tue, 9 Aug 2016 17:49:54 +0000 (10:49 -0700)]
Welcome to JavaFuzz as our latest A[a]rt tools team member!

Rationale:
JavaFuzz is tool for generating random Java programs with
the objective of fuzz testing the ART infrastructure. Each
randomly generated Java program can be run under various
modes of execution, such as using the interpreter, using
the optimizing compiler, using an external reference
implementation, or using various target architectures.
Any difference between the outputs (a divergence) may
indicate a bug in one of the execution modes.

Test: tbd

Bug=30610121

Change-Id: I92dcac35f5229996936d01a0ba7f5acf6dc7b433

7 years agoRemove unused GetDalvikCacheFilenameOrDie.
Richard Uhler [Mon, 15 Aug 2016 23:12:57 +0000 (16:12 -0700)]
Remove unused GetDalvikCacheFilenameOrDie.

Change-Id: I8987fa488a89865d4a5fb56dd00cc396ee33bcce
Test: m test-art-host-gtest-utils_test32
Test: m build-art-host-tests
Test: m build-art-target-tests

7 years agoMerge "Change gcstress runtest option"
Mathieu Chartier [Mon, 15 Aug 2016 22:28:19 +0000 (22:28 +0000)]
Merge "Change gcstress runtest option"

7 years agoMerge "ART: Refactor ImageSpace loading"
Treehugger Robot [Mon, 15 Aug 2016 21:32:30 +0000 (21:32 +0000)]
Merge "ART: Refactor ImageSpace loading"

7 years agoUpdate ahat README and manifest for release 0.7.
Richard Uhler [Mon, 15 Aug 2016 18:01:01 +0000 (11:01 -0700)]
Update ahat README and manifest for release 0.7.

Change-Id: Iccc5db7112cb92e24e9e294273a3ce2b0d4b7aba
Test: Launch ahat and verify it reports 0.7 as the version.

7 years agoART: Refactor ImageSpace loading
Andreas Gampe [Sat, 13 Aug 2016 04:53:32 +0000 (21:53 -0700)]
ART: Refactor ImageSpace loading

Make the code more straightforward so it is easier to reason
about and extend.

Also change return types to unique pointers so it's clear that
it is the caller's responsibility to free those when necessary.

Bug: 30832951
Test: m test-art-host
Test: device booting
Change-Id: I3216eb702e45357a48af5158dacbe40e79bd1ae9

7 years agoMerge "Clean up use of GetDalvikCache."
Treehugger Robot [Mon, 15 Aug 2016 17:12:02 +0000 (17:12 +0000)]
Merge "Clean up use of GetDalvikCache."

7 years agoMerge "Improve READ_BARRIER_MARK_REG for arm32"
Mathieu Chartier [Mon, 15 Aug 2016 16:50:11 +0000 (16:50 +0000)]
Merge "Improve READ_BARRIER_MARK_REG for arm32"

7 years agoMerge "Revert "Enable IntermediateAddress for primitive arrays with read barriers.""
Treehugger Robot [Mon, 15 Aug 2016 14:35:48 +0000 (14:35 +0000)]
Merge "Revert "Enable IntermediateAddress for primitive arrays with read barriers.""

7 years agoRevert "Enable IntermediateAddress for primitive arrays with read barriers."
Roland Levillain [Mon, 15 Aug 2016 13:17:06 +0000 (13:17 +0000)]
Revert "Enable IntermediateAddress for primitive arrays with read barriers."

This CL breaks the angler-userdebug build with
`ART_USE_READ_BARRIER=true`.

Test: Build angler-userdebug with `ART_USE_READ_BARRIER=true`.
Bug: 30762467
Bug: 26601270
Bug: 12687968

This reverts commit 12ecf0800d465acdaa3deccd383ff8ed3428a183.

Change-Id: Ia2069ac9436d2336311dd8d0f183c02e587586ae

7 years agoMerge "ARM64: VIXL: Support a newer version of VIXL."
Roland Levillain [Mon, 15 Aug 2016 10:03:32 +0000 (10:03 +0000)]
Merge "ARM64: VIXL: Support a newer version of VIXL."

7 years agoChange gcstress runtest option
Mathieu Chartier [Fri, 12 Aug 2016 23:11:24 +0000 (16:11 -0700)]
Change gcstress runtest option

No longer passes down SS as the collector. Motivation is to allow
running gcstress with concurrent copying.

New usage is for SS:
mm test-art-host ART_TEST_GC_STRESS=true ART_DEFAULT_GC_TYPE=SS

Bug: 30807718

Test: mm test-art-host ART_TEST_GC_STRESS=true ART_DEFAULT_GC_TYPE=SS

Change-Id: I4416c1ca9ead027536acc9d24b1ec80d5a68c6a1

7 years agoImprove READ_BARRIER_MARK_REG for arm32
Mathieu Chartier [Fri, 12 Aug 2016 20:19:44 +0000 (13:19 -0700)]
Improve READ_BARRIER_MARK_REG for arm32

Use blocked register IP as scratch, avoid pushing in fast path.

Clean up slow path to not have simpler logic and one less memory
write.

Add simple fast path handling for region space TLAB object
allocation.

Test: test-art-target, N6P booting with CC baker

Bug: 30162165

Change-Id: I6594e42d3d6277ffe7bb79df09df8be6bee85eb5

7 years agoMerge "Fix heap poisoning"
Mathieu Chartier [Sat, 13 Aug 2016 00:06:30 +0000 (00:06 +0000)]
Merge "Fix heap poisoning"

7 years agoMerge "Make test work when running ndebug."
Treehugger Robot [Fri, 12 Aug 2016 23:16:45 +0000 (23:16 +0000)]
Merge "Make test work when running ndebug."

7 years agoFix heap poisoning
Mathieu Chartier [Fri, 12 Aug 2016 22:44:44 +0000 (15:44 -0700)]
Fix heap poisoning

Was wrong register.

Test: test-art-host-run-test CC table lookup + poisoning.

Bug: 30162165
Change-Id: If347a40963bcc2fb9e21c54b5405c852bcd3af08

7 years agoMerge "Revert "Revert "Iterative move coalescing for gc regalloc"""
Treehugger Robot [Fri, 12 Aug 2016 22:07:41 +0000 (22:07 +0000)]
Merge "Revert "Revert "Iterative move coalescing for gc regalloc"""

7 years agoMerge "ART: Remove zygote OTA code"
Treehugger Robot [Fri, 12 Aug 2016 21:58:25 +0000 (21:58 +0000)]
Merge "ART: Remove zygote OTA code"

7 years agoMake test work when running ndebug.
Alex Light [Fri, 12 Aug 2016 21:43:48 +0000 (14:43 -0700)]
Make test work when running ndebug.

Change-Id: Iea58ccef51fddaeae4b1fb7084a7956400a81dfe
Test: ART_TEST_RUN_TEST_NDEBUG=true mma -j40 test-art-host-run-test-900-hello-plugin

7 years agoMerge "Use TestCodeGeneratorX86 in codegen unit tests"
Treehugger Robot [Fri, 12 Aug 2016 20:54:05 +0000 (20:54 +0000)]
Merge "Use TestCodeGeneratorX86 in codegen unit tests"

7 years agoRevert "Revert "Iterative move coalescing for gc regalloc""
Matthew Gharrity [Fri, 12 Aug 2016 16:10:45 +0000 (16:10 +0000)]
Revert "Revert "Iterative move coalescing for gc regalloc""

This reverts commit 6f61ee5623a676ce983ccfa1aba9b2ae1237e163,
and fixes the arena allocator issues in the original CL.
Functionality from the original CL has not changed.

Test: m valgrind-test-art-host-gtest-register_allocator_test
Test: ART_TEST_OPTIMIZING_GRAPH_COLOR=true m test-art-host-run-test

Change-Id: Idd979f4e03f0c2800e0c07a0a7d392c8a4230aab

7 years agoART: Remove zygote OTA code
Andreas Gampe [Sat, 9 Jul 2016 17:51:12 +0000 (10:51 -0700)]
ART: Remove zygote OTA code

The renaming and relabeling of the A/B artifacts is now done in
a special-purpose script started by init, so as to not complicate
the logic here (e.g., name of the current slot).

Partially reverts commit 754479f7aa056602de8282f4f03718b42c20b89b.

Bug: 25612095
Bug: 28069686
Change-Id: I4d965361000fd448e199c9af059b6d18bfe928bb
(cherry picked from commit ff2a8ef2ed6b3f21da26f85197170d7d2a9f5134)

7 years agoClean up use of GetDalvikCache.
Richard Uhler [Fri, 12 Aug 2016 16:05:13 +0000 (09:05 -0700)]
Clean up use of GetDalvikCache.

Switch the only caller of the non-void version of GetDalvikCache that set
create_if_absent true to use the void version of GetDalvikCache. Then no
callers of the non-void version of GetDalvikCache set create_if_absent to
true, so that argument is no longer needed. And without create_if_absent,
there is no case when the function will abort, so the 'OrDie" option is no
longer necessary.

Change-Id: I83f60dcc8a8f3b1f91fd2971e9028f6a7dd8384d
Test: m art-test-host
Test: m art-test-target

7 years agoMerge "ART: Reduce stack frame size limit on host"
Treehugger Robot [Fri, 12 Aug 2016 18:19:54 +0000 (18:19 +0000)]
Merge "ART: Reduce stack frame size limit on host"

7 years agoMerge "Optimize x86_64 TLAB allocation speed"
Mathieu Chartier [Fri, 12 Aug 2016 17:34:56 +0000 (17:34 +0000)]
Merge "Optimize x86_64 TLAB allocation speed"

7 years agoMerge changes I852b4daf,Id648eaed
Treehugger Robot [Fri, 12 Aug 2016 17:33:34 +0000 (17:33 +0000)]
Merge changes I852b4daf,Id648eaed

* changes:
  Add basic runtime-plugins support.
  Create a basic agent loading framework.

7 years agoMerge "ART: Configure stack guard size in the Makefile"
Treehugger Robot [Fri, 12 Aug 2016 17:11:01 +0000 (17:11 +0000)]
Merge "ART: Configure stack guard size in the Makefile"

7 years agoMerge "Fix duplicate checks in JNI macro assemblers."
Treehugger Robot [Fri, 12 Aug 2016 15:14:44 +0000 (15:14 +0000)]
Merge "Fix duplicate checks in JNI macro assemblers."

7 years agoMerge "Adjust spacing before NOLINT comments in ART."
Roland Levillain [Fri, 12 Aug 2016 13:45:57 +0000 (13:45 +0000)]
Merge "Adjust spacing before NOLINT comments in ART."

7 years agoFix duplicate checks in JNI macro assemblers.
Vladimir Marko [Fri, 12 Aug 2016 13:26:15 +0000 (14:26 +0100)]
Fix duplicate checks in JNI macro assemblers.

Instead of checking one register twice, check the other
register passed to the function.

Test: Run ART test suite on host and Nexus 5.
Bug: 30739460
Change-Id: If868ea14f5c192982488ed065b279a019b5b7cf4

7 years agoARM64: VIXL: Support a newer version of VIXL.
Artem Serov [Mon, 8 Aug 2016 14:11:24 +0000 (15:11 +0100)]
ARM64: VIXL: Support a newer version of VIXL.

Please note that compiling VIXL with -Wshadow is a known VIXL issue.

This will be resolved in a later version of VIXL, when we can drop
the deprecated API for getters and setters.

For more info take a look at VIXL_DEPRECATED in the VIXL source code.

Change-Id: Iea30b1a7b065f9b16a92c6cc7ebdc50ef068b348

7 years agoAdjust spacing before NOLINT comments in ART.
Roland Levillain [Thu, 11 Aug 2016 22:53:33 +0000 (23:53 +0100)]
Adjust spacing before NOLINT comments in ART.

Note that neither clang-tidy nor cpplint.py complain about
these style "issues", precisely because of the NOLINT
comments.

Test: WITH_TIDY=1 WITH_TIDY_CHECKS='-*,misc-macro-parentheses' mmma art
Change-Id: Id692fd394ffbd4fe208cbbe4407b4d5e208462bb

7 years agoMerge "ARM: Add vldm/vstm assembler support."
Vladimir Marko [Fri, 12 Aug 2016 11:23:40 +0000 (11:23 +0000)]
Merge "ARM: Add vldm/vstm assembler support."

7 years agoMerge "Initialize Heap's collector pointers to null."
Treehugger Robot [Fri, 12 Aug 2016 10:22:57 +0000 (10:22 +0000)]
Merge "Initialize Heap's collector pointers to null."

7 years agoMerge "Revert "Iterative move coalescing for gc regalloc""
Treehugger Robot [Fri, 12 Aug 2016 08:01:43 +0000 (08:01 +0000)]
Merge "Revert "Iterative move coalescing for gc regalloc""

7 years agoRevert "Iterative move coalescing for gc regalloc"
Andreas Gampe [Fri, 12 Aug 2016 06:33:15 +0000 (06:33 +0000)]
Revert "Iterative move coalescing for gc regalloc"

There are lifetime issues with allocators and coloring
iterations that got flagged by valgrind.

This reverts commit 465ed699e810868fe5bb39730e6d149a4734372d.

Change-Id: I9e08172321af61d109c116a4f0742fa809e8094b
Test: m test-art-host

7 years agoMerge "Remove kRestrictCompilationFiltersToImage"
Treehugger Robot [Fri, 12 Aug 2016 06:28:56 +0000 (06:28 +0000)]
Merge "Remove kRestrictCompilationFiltersToImage"

7 years agoMerge "Enable compiled-methods= dex2oat switch"
Treehugger Robot [Fri, 12 Aug 2016 06:16:40 +0000 (06:16 +0000)]
Merge "Enable compiled-methods= dex2oat switch"

7 years agoART: Reduce stack frame size limit on host
Andreas Gampe [Fri, 12 Aug 2016 04:11:59 +0000 (21:11 -0700)]
ART: Reduce stack frame size limit on host

Clang isn't that bad anymore.

Change-Id: Id9f3e3e34a743b43c6b535d2757be7869a8cf7ea

7 years agoART: Configure stack guard size in the Makefile
Andreas Gampe [Fri, 12 Aug 2016 01:35:58 +0000 (18:35 -0700)]
ART: Configure stack guard size in the Makefile

Configure the stack guard reserved size through defines set in
the Makefile. This can be used to increase the gap depending on
the configuration, which may be required to safely support Java
stack overflows in instrumented builds.

Bug: 30766843
Change-Id: Icdd3209e3a261ea8383ef8eab38153255339b04a
Test: m test-art-host

7 years agoMerge "Iterative move coalescing for gc regalloc"
Andreas Gampe [Fri, 12 Aug 2016 01:43:26 +0000 (01:43 +0000)]
Merge "Iterative move coalescing for gc regalloc"

7 years agoOptimize x86_64 TLAB allocation speed
Mathieu Chartier [Wed, 10 Aug 2016 01:37:09 +0000 (18:37 -0700)]
Optimize x86_64 TLAB allocation speed

Added assembly fast path code for resolved and initialized object
region TLAB allocations. Removed 3 instructions from TLAB fast path.

Added assembly fast path for array TLAB region allocators.

Should have more speedups for read barrier fast paths during
resolved and initialized alloc entrypoints.

Bug: 30162165

Test: test-art-host CC baker

Change-Id: I64dd06be5f18c8d6a5de0f15f0e2e7d488e99f18

7 years agoUse TestCodeGeneratorX86 in codegen unit tests
Matthew Gharrity [Fri, 12 Aug 2016 00:20:05 +0000 (17:20 -0700)]
Use TestCodeGeneratorX86 in codegen unit tests

The codegen unit tests are supposed to use special "test" code
generators when targeting ARM and x86 (due to differing calling
conventions between the C++ source code and the generated code),
yet TestCodeGeneratorX86 was not being used. This fixes that.

(The tests were only succeeding because the register allocator happened
to not assign the EBX register.)

Test: m test-art-host-gtest-codegen_test

Change-Id: Ia3dd6998c38e9ff27b8c2734457f86b3fed44ab4

7 years agoAdd basic runtime-plugins support.
Alex Light [Thu, 11 Aug 2016 17:48:03 +0000 (10:48 -0700)]
Add basic runtime-plugins support.

This allows one to pass shared-libraries on the command line that the
runtime will load as plugins. They have access to runtime code and can
install hooks to add functionality. Currently the only hook they can
touch is JavaVMExt::AddEnvironmentHook to register a callback for
GetEnv(). More hooks might be added in the future.

Test: ./test/run-test 900
Change-Id: I852b4daf5a3fa71e9888722bc07794632c0e5010

7 years agoCreate a basic agent loading framework.
Alex Light [Thu, 28 Jul 2016 17:07:45 +0000 (10:07 -0700)]
Create a basic agent loading framework.

Currently we only allow agents to be loaded at runtime startup, though
this is expected to change soon.

Test: ./test/run-test --host 900
Change-Id: Id648eaed4bbbe6fdef41d64922d023a4db0bfa54

7 years agoIterative move coalescing for gc regalloc
Matthew Gharrity [Fri, 22 Jul 2016 15:52:13 +0000 (08:52 -0700)]
Iterative move coalescing for gc regalloc

Implement iterative move coalescing for graph coloring
register allocation. Based on Andrew Appel's implementation
in "Modern Compiler Implementation in Java", modified to
support constraints such as pair intervals.

Test: ART_TEST_OPTIMIZING_GRAPH_COLOR=true m test-art-host

Change-Id: I8642297d3bd798a4fc4de4b356ac3304098471a5