OSDN Git Service

android-x86/art.git
8 years agoRevert "Disable JDWP tests with read barriers."
Hiroshi Yamauchi [Tue, 19 Apr 2016 18:23:15 +0000 (11:23 -0700)]
Revert "Disable JDWP tests with read barriers."

This reverts commit f2a5c9c5669664ad008ed48f511e8f101d4bc299.

Bug: 25800335

8 years agoFix a deadlock between debugger and GC.
Hiroshi Yamauchi [Tue, 19 Apr 2016 18:14:06 +0000 (11:14 -0700)]
Fix a deadlock between debugger and GC.

Avoid a deadlock between GC and debugger's SuspendAllForDebugger where
GC gets suspended during GC.

Fix the jdwp tests with the CC collector, but the deadlock is NOT
specific to the CC collector.

Bug: 25800335
Bug: 12687968
Change-Id: I8f2869872bce4692d943020072e35ebf5a4e68a5

8 years agoMerge "Disable stripped-dex test on target."
Richard Uhler [Thu, 14 Apr 2016 21:19:56 +0000 (21:19 +0000)]
Merge "Disable stripped-dex test on target."

8 years agoDisable stripped-dex test on target.
Richard Uhler [Thu, 14 Apr 2016 20:31:30 +0000 (13:31 -0700)]
Disable stripped-dex test on target.

Because the targets don't have the required zip command.

Bug: 27599626

Change-Id: I5ba075165d96d831a4b8e0912cff1319071c60a5

8 years agoMerge "ART: Make iget receiver mismatch hard verifier error"
Andreas Gampe [Thu, 14 Apr 2016 20:09:00 +0000 (20:09 +0000)]
Merge "ART: Make iget receiver mismatch hard verifier error"

8 years agoART: Make iget receiver mismatch hard verifier error
Andreas Gampe [Thu, 14 Apr 2016 17:55:04 +0000 (10:55 -0700)]
ART: Make iget receiver mismatch hard verifier error

If the classes are resolved, and still not match, this should be
a verify error.

Bug: 28187158
Change-Id: I89c996ae15865674f21cf32ec378d37bac34861b

8 years agoMerge "Restore fallback behavior in the no-image stripped-dex case."
Richard Uhler [Thu, 14 Apr 2016 16:11:40 +0000 (16:11 +0000)]
Merge "Restore fallback behavior in the no-image stripped-dex case."

8 years agoMerge "ART: Save a bit of memory in production LICM"
David Brazdil [Thu, 14 Apr 2016 15:59:02 +0000 (15:59 +0000)]
Merge "ART: Save a bit of memory in production LICM"

8 years agoMerge "ART: Update DCHECKs in SsaLivenessAnalysis::AddBackEdgeUses"
David Brazdil [Thu, 14 Apr 2016 15:56:40 +0000 (15:56 +0000)]
Merge "ART: Update DCHECKs in SsaLivenessAnalysis::AddBackEdgeUses"

8 years agoART: Save a bit of memory in production LICM
David Brazdil [Thu, 14 Apr 2016 15:36:15 +0000 (16:36 +0100)]
ART: Save a bit of memory in production LICM

LICM uses a bit vector to DCHECK that inner loops are visited before
outer loops. Allocate the bit vector only if kIsDebugBuild.

Change-Id: Ic2641607ebe55ad8974de2ebe03973f0bf182e42

8 years agoART: Update DCHECKs in SsaLivenessAnalysis::AddBackEdgeUses
David Brazdil [Thu, 14 Apr 2016 12:47:24 +0000 (13:47 +0100)]
ART: Update DCHECKs in SsaLivenessAnalysis::AddBackEdgeUses

Graph linearization in the presence of irreducible loops is not
guaranteed to generate a linear order where all blocks of a loop are
adjacent, first block is the header and last block is one of the back
edges.

These assumptions are made when inserting synthesized uses at the back
edges to aid the register allocator. Not meeting them will result in
the algorithm's early termination and back-edge uses not being added.

This patch updates the DCHECKs so the compiler does not fail in such
circumstances.

Bug: 27615840
Bug: 27624868

Change-Id: I63632e8819ea3644d5c6fdfea00b66128bf22c24

8 years agoMerge "Fix initialization of field after conflicting merges."
Nicolas Geoffray [Thu, 14 Apr 2016 13:34:36 +0000 (13:34 +0000)]
Merge "Fix initialization of field after conflicting merges."

8 years agoFix initialization of field after conflicting merges.
Nicolas Geoffray [Thu, 14 Apr 2016 13:17:29 +0000 (14:17 +0100)]
Fix initialization of field after conflicting merges.

Change-Id: I5c887d5cbcd94f51c3566e964d217a906dbaa9ea

8 years agoMerge "Don't use JACK alpha anymore."
Nicolas Geoffray [Thu, 14 Apr 2016 12:39:05 +0000 (12:39 +0000)]
Merge "Don't use JACK alpha anymore."

8 years agoDon't use JACK alpha anymore.
Nicolas Geoffray [Thu, 14 Apr 2016 12:32:01 +0000 (13:32 +0100)]
Don't use JACK alpha anymore.

Change-Id: Id37d90446be5d63cb0cf0131e34b562147e88335

8 years agoMerge "Fix copy-paste nit"
Vladimir Marko [Thu, 14 Apr 2016 10:48:16 +0000 (10:48 +0000)]
Merge "Fix copy-paste nit"

8 years agoMerge "Use dex cache from compilation unit in HInstructionBuilder."
Vladimir Marko [Thu, 14 Apr 2016 10:30:39 +0000 (10:30 +0000)]
Merge "Use dex cache from compilation unit in HInstructionBuilder."

8 years agoFix copy-paste nit
Goran Jakovljevic [Thu, 14 Apr 2016 09:46:58 +0000 (11:46 +0200)]
Fix copy-paste nit

Pair ENTRY and END macros.

Change-Id: I17dcc61cdd02b2c26b9920e5385b6cb1420b18dd

8 years agoMerge "Remove the JIT from the instrumentation framework."
Nicolas Geoffray [Thu, 14 Apr 2016 09:47:04 +0000 (09:47 +0000)]
Merge "Remove the JIT from the instrumentation framework."

8 years agoUse dex cache from compilation unit in HInstructionBuilder.
Vladimir Marko [Wed, 13 Apr 2016 18:29:26 +0000 (19:29 +0100)]
Use dex cache from compilation unit in HInstructionBuilder.

Avoid calling costly ClassLinker::FindDexCache() from
HInstructionBuilder, the dex cache is already available in
the compilation unit. Compiling Nexus 5 boot image on host
under perf(1) shows that the time spent in FindDexCache()
is reduced from about 2% to well under 0.2%, 90% of the
remaining hits coming from ReferenceTypePropagation which
doesn't have access to the compilation unit.

Change-Id: Id9df61c79db9a67e3f12c857f32a4ca80e5420c3

8 years agoMerge "Revert "Remove the no-longer-needed F/I and D/J alias.""
Aart Bik [Wed, 13 Apr 2016 22:34:17 +0000 (22:34 +0000)]
Merge "Revert "Remove the no-longer-needed F/I and D/J alias.""

8 years agoMerge "Allow private methods in interfaces."
Treehugger Robot [Wed, 13 Apr 2016 22:05:51 +0000 (22:05 +0000)]
Merge "Allow private methods in interfaces."

8 years agoRevert "Remove the no-longer-needed F/I and D/J alias."
Aart Bik [Wed, 13 Apr 2016 21:17:17 +0000 (21:17 +0000)]
Revert "Remove the no-longer-needed F/I and D/J alias."

This reverts commit 2f52064dcfe5ebce5a998d30766ca079a366c920.

Reason:

Arrays.sort() returns wrong result on double[] and this CL is the most likely suspect. Rolling back to buy some time for careful analysis and debugging.

Change-Id: I58223c42e95c2287520eef863fbcb738b0736d4d

8 years agoAllow private methods in interfaces.
Alex Light [Tue, 12 Apr 2016 22:50:55 +0000 (15:50 -0700)]
Allow private methods in interfaces.

Private methods may be generated in interfaces during compilation of
some default methods.  Change the verifier to allow these methods.

Bug: 27999840

Change-Id: Ib8120a8f6cb036021334d9af0ed78ae372974ecb

8 years agoRestore fallback behavior in the no-image stripped-dex case.
Richard Uhler [Mon, 4 Apr 2016 20:30:16 +0000 (13:30 -0700)]
Restore fallback behavior in the no-image stripped-dex case.

Add --strip-dex option to run-test to test this case.

Bug: 27599626
Change-Id: I010543f48c39adbda007e2e05dda62ac1d70715c

8 years agoMerge "Add imgdiag support for diff against zygote"
Mathieu Chartier [Wed, 13 Apr 2016 19:11:29 +0000 (19:11 +0000)]
Merge "Add imgdiag support for diff against zygote"

8 years agoAdd imgdiag support for diff against zygote
Mathieu Chartier [Fri, 8 Apr 2016 21:08:37 +0000 (14:08 -0700)]
Add imgdiag support for diff against zygote

Adds option --zygote-diff-pid. If this option is passed in then it
tries to differentiate objects that are private dirty by comparing
them against the zygote image contents if they are different from
the boot image contents.

Bug: 27906566

(cherry picked from commit 1398cf23ab47d250ea3eea39812e29a748b70777)

Change-Id: I99b1552d046e65dd9da0d1bdb8cb309aa0f9b0db

8 years agoMerge "Move Assemblers to the Arena."
Vladimir Marko [Wed, 13 Apr 2016 16:55:11 +0000 (16:55 +0000)]
Merge "Move Assemblers to the Arena."

8 years agoMove Assemblers to the Arena.
Vladimir Marko [Wed, 13 Apr 2016 10:59:46 +0000 (11:59 +0100)]
Move Assemblers to the Arena.

And clean up some APIs to return std::unique_ptr<> instead
of raw pointers that don't communicate ownership.

Change-Id: I3017302307a0253d661240750298802fb0d9585e

8 years agoRemove the JIT from the instrumentation framework.
Nicolas Geoffray [Tue, 12 Apr 2016 15:33:24 +0000 (16:33 +0100)]
Remove the JIT from the instrumentation framework.

This was slowing down the interpreter for no reason.
Also, call AddSamples for invoke-static and invoke-direct.

Change-Id: I7b5641097f7741dd32feb1ce6af739fd27fb37c2

8 years agoMerge "Remove the no-longer-needed F/I and D/J alias."
Aart Bik [Wed, 13 Apr 2016 16:15:35 +0000 (16:15 +0000)]
Merge "Remove the no-longer-needed F/I and D/J alias."

8 years agoMerge "Revert "Revert "ART: Improve JitProfile perf in x86 mterp"""
Bill Buzbee [Wed, 13 Apr 2016 12:37:16 +0000 (12:37 +0000)]
Merge "Revert "Revert "ART: Improve JitProfile perf in x86 mterp"""

8 years agoMerge "Add option to tune sample collection based on thread sensitivity"
Calin Juravle [Wed, 13 Apr 2016 11:44:36 +0000 (11:44 +0000)]
Merge "Add option to tune sample collection based on thread sensitivity"

8 years agoAdd option to tune sample collection based on thread sensitivity
Calin Juravle [Thu, 7 Apr 2016 16:09:25 +0000 (17:09 +0100)]
Add option to tune sample collection based on thread sensitivity

Bug: 28065407
Bug: 27865109

Change-Id: Icdb89f8f8874a41c07e73185523d18e8956620d3

8 years agoMerge "Allocate code generators on the arena."
Vladimir Marko [Wed, 13 Apr 2016 10:41:20 +0000 (10:41 +0000)]
Merge "Allocate code generators on the arena."

8 years agoMerge "Revert "Ignore libcore.java.lang.OldSystemTest#test_gc failures with CC.""
Treehugger Robot [Wed, 13 Apr 2016 10:26:35 +0000 (10:26 +0000)]
Merge "Revert "Ignore libcore.java.lang.OldSystemTest#test_gc failures with CC.""

8 years agoMerge "ART: Simplify unstarted-runtime ceil and floor"
Andreas Gampe [Wed, 13 Apr 2016 04:21:53 +0000 (04:21 +0000)]
Merge "ART: Simplify unstarted-runtime ceil and floor"

8 years agoART: Simplify unstarted-runtime ceil and floor
Andreas Gampe [Wed, 13 Apr 2016 01:07:36 +0000 (18:07 -0700)]
ART: Simplify unstarted-runtime ceil and floor

Just use the C functions like libcore does. Add tests.

Bug: 28132336
Change-Id: Ic5f469e8504c9f2b7756280558fd1610ed8015ba

8 years agoRemove the no-longer-needed F/I and D/J alias.
Aart Bik [Tue, 12 Apr 2016 22:05:45 +0000 (15:05 -0700)]
Remove the no-longer-needed F/I and D/J alias.

Rationale:
Now that our HIR is type clean (yeah!), we no longer have
to conservatively assume F/I and D/J are aliased. This
enables more accurate side effects analysis, with improvements
in all clients, such a LICM.

BUG=22538329

Change-Id: Iba1fb09ff063f31b5893f588aa6d0c5ab3b42f39

8 years agoRevert "Ignore libcore.java.lang.OldSystemTest#test_gc failures with CC."
Hiroshi Yamauchi [Wed, 13 Apr 2016 01:22:29 +0000 (18:22 -0700)]
Revert "Ignore libcore.java.lang.OldSystemTest#test_gc failures with CC."

This reverts commit e693ba17e8b758a1894dd1b53f733471afc5976d.

Bug: 26155567

8 years agoMerge "Fix InsertDexFileInToClassLoader to handle null class loaders"
Mathieu Chartier [Wed, 13 Apr 2016 01:19:08 +0000 (01:19 +0000)]
Merge "Fix InsertDexFileInToClassLoader to handle null class loaders"

8 years agoFix InsertDexFileInToClassLoader to handle null class loaders
Mathieu Chartier [Tue, 12 Apr 2016 23:54:48 +0000 (16:54 -0700)]
Fix InsertDexFileInToClassLoader to handle null class loaders

Maybe used by legacy apps. This functionality is planned for
deprecation.

Bug: 27954959

(cherry picked from commit f2bf9d640e37b72be8b4f6016d4aa95a0e27b7b4)

Change-Id: I1ae2cf1e33f2f1b237a41c8cc50d2a814a52a95a

8 years agoMerge "ART: Flag to fail thread creation"
Andreas Gampe [Tue, 12 Apr 2016 22:27:18 +0000 (22:27 +0000)]
Merge "ART: Flag to fail thread creation"

8 years agoART: Flag to fail thread creation
Andreas Gampe [Mon, 11 Apr 2016 15:42:26 +0000 (08:42 -0700)]
ART: Flag to fail thread creation

Add a flag to mark when the zygote is not allowed to create threads.

Bug: 27248115
Bug: 28149511

(cherry picked from commit 415d8070e37c20dfb7e6dc37e74fdb5fffc2022e)

Change-Id: I1dc3620d9e7d0054c672b993d89459fc4b353dfc

8 years agoMerge "ART: Add Makefile target for boot image"
Andreas Gampe [Tue, 12 Apr 2016 20:46:41 +0000 (20:46 +0000)]
Merge "ART: Add Makefile target for boot image"

8 years agoAllocate code generators on the arena.
Vladimir Marko [Tue, 12 Apr 2016 17:51:43 +0000 (18:51 +0100)]
Allocate code generators on the arena.

Change-Id: If8cf0ee43711f6e13171443e3c057ff370ccfbaa

8 years agoMerge "Assembly region TLAB allocation fast path for arm64."
Hiroshi Yamauchi [Tue, 12 Apr 2016 17:33:34 +0000 (17:33 +0000)]
Merge "Assembly region TLAB allocation fast path for arm64."

8 years agoART: Add Makefile target for boot image
Andreas Gampe [Mon, 11 Apr 2016 18:57:18 +0000 (11:57 -0700)]
ART: Add Makefile target for boot image

Add a phony Makefile target that depends on the dexopted boot image.
This allows to test out changes easily, for example, for compile-time
initialization:

 m art-boot-image ART_BOOT_IMAGE_EXTRA_ARGS=--dump-init-failures=fails.txt

Bug: 27265238
Change-Id: I9d54c562a2e623b23b7c1dea9cdbe303786c95dd

8 years agoMerge "Small changes in ARM and x86-64 SystemArrayCopy intrinsics."
Roland Levillain [Tue, 12 Apr 2016 16:18:08 +0000 (16:18 +0000)]
Merge "Small changes in ARM and x86-64 SystemArrayCopy intrinsics."

8 years agoSmall changes in ARM and x86-64 SystemArrayCopy intrinsics.
Roland Levillain [Tue, 12 Apr 2016 14:42:57 +0000 (15:42 +0100)]
Small changes in ARM and x86-64 SystemArrayCopy intrinsics.

Have these intrinsics share a more uniform style with the
ARM64 SystemArrayCopy intrinsic.

Also make some changes/improvements in:
- art::IntrinsicOptimizations
- art::arm64::GenSystemArrayCopyAddresses

Change-Id: Ieeb224795229580f8e5f7219c586d04786d8c705

8 years agoMerge "ART: Speed up HGraph::PopulateIrreducibleRecursive"
David Brazdil [Tue, 12 Apr 2016 14:19:33 +0000 (14:19 +0000)]
Merge "ART: Speed up HGraph::PopulateIrreducibleRecursive"

8 years agoART: Speed up HGraph::PopulateIrreducibleRecursive
David Brazdil [Tue, 5 Apr 2016 16:15:19 +0000 (17:15 +0100)]
ART: Speed up HGraph::PopulateIrreducibleRecursive

Populating an irreducible loop can potentially traverse all possible
paths through the HGraph, leading to an exponential algorithm.
This patch adds a bit vector of nodes whose membership in the loop
has been decided and need not be revisited again.

Bug: 27856014
Change-Id: I3696f08c846e6f40e5de44cb771811bac7e3e08a

8 years agoMerge "Remove recursive ALWAYS_INLINE"
Vladimir Marko [Tue, 12 Apr 2016 13:58:51 +0000 (13:58 +0000)]
Merge "Remove recursive ALWAYS_INLINE"

8 years agoMerge "unstarted_runtime: add cutout for Math.floor"
Sergio Giro [Tue, 12 Apr 2016 13:39:24 +0000 (13:39 +0000)]
Merge "unstarted_runtime: add cutout for Math.floor"

8 years agounstarted_runtime: add cutout for Math.floor
Sergio Giro [Mon, 11 Apr 2016 19:49:20 +0000 (20:49 +0100)]
unstarted_runtime: add cutout for Math.floor

Needed in order to make BouncyCastleProvider initializable in
compile time

(cherry picked from commit a8908ef14cf2d2e0d05e41e319da5d6909325ae2)
Bug: 28108158

Change-Id: Id6b4e3d4dde45354562603f41134d8d21da2b423

8 years agoMerge "Revert "Update expectations for 031-class-attributes.""
Narayan Kamath [Tue, 12 Apr 2016 12:52:34 +0000 (12:52 +0000)]
Merge "Revert "Update expectations for 031-class-attributes.""

8 years agoRevert "Update expectations for 031-class-attributes."
Narayan Kamath [Tue, 12 Apr 2016 12:38:09 +0000 (13:38 +0100)]
Revert "Update expectations for 031-class-attributes."

This reverts commit 32580ebd54bfeef62d5769439d336ca33f5592b0.

The original commit tracked OpenJDK behaviour, which we've decided
to revert because several apps depend on the value of Package.toString.

bug: 28057303

8 years agoRemove recursive ALWAYS_INLINE
Goran Jakovljevic [Tue, 12 Apr 2016 11:11:16 +0000 (13:11 +0200)]
Remove recursive ALWAYS_INLINE

GCC (mips32 Buildbot) is complaining about recursive inlining.

Clean up after
    https://android-review.googlesource.com/#/c/214918/.

Change-Id: I221c53c32403e5e433e30cfee7b2ae36680f39e6

8 years agoMerge changes from topic 'image_layout'
Mathieu Chartier [Tue, 12 Apr 2016 02:23:10 +0000 (02:23 +0000)]
Merge changes from topic 'image_layout'

* changes:
  Fix relocate in place to handle arbitrary app image layout
  Revert "Revert "Reduce dirty image pages by improving binning""
  Revert "Reduce dirty image pages by improving binning"
  Reduce dirty image pages by improving binning

8 years agoFix relocate in place to handle arbitrary app image layout
Mathieu Chartier [Mon, 11 Apr 2016 19:03:48 +0000 (12:03 -0700)]
Fix relocate in place to handle arbitrary app image layout

Previously the relocation had bugs where it would not work correctly
if the iftable was after the class since the object arrays would not
yet be updated to be the relocated address when we went through them
to fixup the ArtMethod pointers.

Also fixes a bug where the superclass of a class could not be
updated when we walk through it for visiting instance fields of
large objects.

Changed RelocateInPlace to use a simplier single pass approach.

Bug: 27906566
Bug: 22858531

(cherry picked from commit 564289c753cbe3521f9523b350bd713c1fe2fbd4)

Change-Id: I97affab6ff353dfdc4d1bf31df69ceb96a0c7a1a

8 years agoRevert "Revert "Reduce dirty image pages by improving binning""
Mathieu Chartier [Sat, 9 Apr 2016 02:01:05 +0000 (19:01 -0700)]
Revert "Revert "Reduce dirty image pages by improving binning""

This reverts commit 8ace610a222892f7b700e4f95e50fa6315ab85c0.

(cherry picked from commit 80c563ba136fd8da8791cf0f1a5f8f7026816ead)

Change-Id: I5b312f41d8b96eb189ea8b9d3a3ab861c5c9d515

8 years agoRevert "Reduce dirty image pages by improving binning"
Mathieu Chartier [Sat, 9 Apr 2016 01:54:36 +0000 (18:54 -0700)]
Revert "Reduce dirty image pages by improving binning"

This reverts commit 794bff5f622c79fd7bd9899e700cf052a375c675.

(cherry picked from commit 8ace610a222892f7b700e4f95e50fa6315ab85c0)

Change-Id: Id53d64a04aa8766a2cb283a2f7b8354590213b13

8 years agoReduce dirty image pages by improving binning
Mathieu Chartier [Thu, 7 Apr 2016 21:16:26 +0000 (14:16 -0700)]
Reduce dirty image pages by improving binning

Change ordering from dirty to less dirty since ArtFields are clean
and they are last. Add separate binning for DexCache and instances
of java lang object (probably a lock object).

Business card .art RAM (on shamu):
              PSS     Private Dirty
.art mmap     1011      600
.art mmap      990      560

Bug: 27906566

(cherry picked from commit 794bff5f622c79fd7bd9899e700cf052a375c675)

Change-Id: I31c9ff41c3d713fc3b14573c2b731325ac6f366f

8 years agoAssembly region TLAB allocation fast path for arm64.
Hiroshi Yamauchi [Fri, 8 Apr 2016 00:18:24 +0000 (17:18 -0700)]
Assembly region TLAB allocation fast path for arm64.

This is for the CC collector.

Share the common fast path code with the tlab fast path code.

Speedup (on N9):
    BinaryTrees:  1235 -> 443 ms (-64%)
    MemAllocTest: 1647 -> 766 ms (-53%)

Bug: 9986565
Bug: 12687968
Change-Id: I67049cc0b4d6508934f07d039d421ee162b330bf

8 years agoMerge "Don't abort during app image loading with no boot image"
Mathieu Chartier [Mon, 11 Apr 2016 21:12:11 +0000 (21:12 +0000)]
Merge "Don't abort during app image loading with no boot image"

8 years agoDon't abort during app image loading with no boot image
Mathieu Chartier [Mon, 11 Apr 2016 20:56:45 +0000 (13:56 -0700)]
Don't abort during app image loading with no boot image

Fail gracefully instead. Fixes test 119.

Bug: 22858531
Change-Id: If17acf5365f8a8fc0a2f6445c558960c62c8f948

8 years agoMerge "ART: Make mterp jit profiling race tolerant"
Calin Juravle [Mon, 11 Apr 2016 17:32:24 +0000 (17:32 +0000)]
Merge "ART:  Make mterp jit profiling race tolerant"

8 years agoART: Make mterp jit profiling race tolerant
buzbee [Mon, 11 Apr 2016 14:48:18 +0000 (07:48 -0700)]
ART:  Make mterp jit profiling race tolerant

The JIT profiling mechanism is intentionally non-precise to minimize
performance overhead.  In general, this is not a problem.   However,
the on-stack replacement mechanism assumes an order of method
compilation than can sometimes be violated if conditions are just
right.

This change allows compilation requests that were dropped due to
a race condition to eventually be re-issued.  It does this by allowing
the 16-bit hotness counter to wrap around.

Change-Id: I2ac8056af8c4f7f8cef3f2c3db70b0394c26a566

8 years agoMerge "MIPS: Improving art_quick_imt_conflict_trampoline"
Nicolas Geoffray [Mon, 11 Apr 2016 15:45:57 +0000 (15:45 +0000)]
Merge "MIPS: Improving art_quick_imt_conflict_trampoline"

8 years agoMerge "ART: Change prebuild run-test setup"
Andreas Gampe [Mon, 11 Apr 2016 15:29:56 +0000 (15:29 +0000)]
Merge "ART: Change prebuild run-test setup"

8 years agoMerge "Oatdump: Smarter StackMap lookup to improve performance."
Vladimir Marko [Mon, 11 Apr 2016 14:18:05 +0000 (14:18 +0000)]
Merge "Oatdump: Smarter StackMap lookup to improve performance."

8 years agoMerge "Revert "Revert "Revert "Revert "ART: Improve JitProfile perf in arm/arm64...
Calin Juravle [Mon, 11 Apr 2016 13:56:56 +0000 (13:56 +0000)]
Merge "Revert "Revert "Revert "Revert "ART: Improve JitProfile perf in arm/arm64 mterp"""""

8 years agoOatdump: Smarter StackMap lookup to improve performance.
Vladimir Marko [Fri, 8 Apr 2016 10:18:25 +0000 (11:18 +0100)]
Oatdump: Smarter StackMap lookup to improve performance.

Shaves off roughly 10% from the oatdump_test time.

Bug: 27824283
Change-Id: I6c2fd3b1db2723871641d43aeb543581d405cfcb

8 years agoMerge "Implement on-stack replacement for MIPS32 and MIPS64"
Nicolas Geoffray [Mon, 11 Apr 2016 08:24:18 +0000 (08:24 +0000)]
Merge "Implement on-stack replacement for MIPS32 and MIPS64"

8 years agoART: Change prebuild run-test setup
Andreas Gampe [Wed, 23 Mar 2016 17:54:21 +0000 (10:54 -0700)]
ART: Change prebuild run-test setup

Prebuild was meant to check whether the runtime correctly loads
pre-existing oat files. With the advent of patching images, this
has actually been reduced to "does the runtime run self-patchoat"
correctly.

We cannot produce a prebuilt oat file that does not have to be
patched, as the runtime will patch the image when it starts.

There are two options to make this more meaningful:

* In the prebuild setting, first run patchoat to explicitly create
  a valid dalvik-cache.

* Change the dex2oat output to be the odex file, which can be
  patched like a regular file.

This change shows the latter approach.

Change-Id: I5234e10d78f7ea6c7ad8598db67d3fad8ba42b67

8 years agoMerge "Make sure non-fatal errors aren't printed."
Alex Light [Fri, 8 Apr 2016 18:42:03 +0000 (18:42 +0000)]
Merge "Make sure non-fatal errors aren't printed."

8 years agoMerge "ART: Refactor imgdiag internals"
Andreas Gampe [Fri, 8 Apr 2016 18:33:25 +0000 (18:33 +0000)]
Merge "ART: Refactor imgdiag internals"

8 years agoMerge changes from topic 'rr_barrier'
Mathieu Chartier [Fri, 8 Apr 2016 17:50:05 +0000 (17:50 +0000)]
Merge changes from topic 'rr_barrier'

* changes:
  Store precice set of which constructors require barriers
  Revert "Revert "Check if we require barrier if we did not resolve classes""
  Revert "Revert "Change RequiresConstructorBarrier default to yes""

8 years agoStore precice set of which constructors require barriers
Mathieu Chartier [Thu, 7 Apr 2016 20:19:19 +0000 (13:19 -0700)]
Store precice set of which constructors require barriers

Fixes bugs where things in the boot image might not have been
calculated even though resolved_clases was true. This only occured
for app and test compiles though.

Fixes test 476-checker-ctor-memory-barrier which was failing due to
inlining something in the boot class path and getting a unexpected
barrier since the barrier defaults to enabled.

No measurable increase in RAM usage.

Bug: 28005874

Change-Id: I4a417819aa129c95f4a83c38df1a66eb77824ea9

8 years agoMerge "ART: Fix race in on-stack replacement"
Bill Buzbee [Fri, 8 Apr 2016 17:28:49 +0000 (17:28 +0000)]
Merge "ART: Fix race in on-stack replacement"

8 years agoMake sure non-fatal errors aren't printed.
Alex Light [Fri, 8 Apr 2016 16:44:54 +0000 (09:44 -0700)]
Make sure non-fatal errors aren't printed.

During run-test setup `du` can print error messages if the folder it
is running on is modified during its execution. We silence these error
messages since they shouldn't affect the validity of the test.

Change-Id: I9f59e39c2a346414f025ce1fa79b453329080d30

8 years agoART: Fix race in on-stack replacement
buzbee [Thu, 7 Apr 2016 21:42:47 +0000 (14:42 -0700)]
ART: Fix race in on-stack replacement

The expected sequence of events for on-stack replacement is:

   1. Method goes warm, triggering enhanced profiling
   2. Method goes hot, triggering method compilation
   3. Method goes really hot, triggering an osr method compilation.
   4. Interpreter polls for the existence of an osr entry point,
      and transitons to compiled code if found.

We have a race problem if #2 and #3 happen closely together, and
the osr method compilation begins before the regular method
compilation.  In that case, the jit sees that the method is
already being compiled (the osr method - but it does not
distinguish the two), and discards the normal compilation request.
So, the osr version is compiled and the normal version is discarded.
In #4, the MaybeDoOnStackReplacement() check assumes that a normal
version of the compiled method must exist before doing an on-stack
replacement, so it keeps returning false.

This is why we were seeing sporadic timeout failures of
570-checker-osr when the mterp fast branch profiling was
introduced.  The branch profiling performance enhancements
greatly reduced the time between #2 and #3, increasing the liklihood
of losing the race.  Further, the new code clamped hotness to avoid
wrap-around.  The race existed (and likely occurred) in the previous
version, but because hotness counters were allowed to overflow and
wrap around you'd eventually hit the threshold a second time and
try again - masking the problem.

Tip 'o the hat to Serguei Katkov for identifying the problem.

A possible solution (taken in this CL) is to differentiate osr
compilations from normal method compilations.

Bug: 27939339
Change-Id: I71044516b35dc69de9fc2d2a445e33809ac650ed

8 years agoRevert "Revert "Revert "Revert "ART: Improve JitProfile perf in arm/arm64 mterp""""
Bill Buzbee [Mon, 4 Apr 2016 16:59:29 +0000 (16:59 +0000)]
Revert "Revert "Revert "Revert "ART: Improve JitProfile perf in arm/arm64 mterp""""

Bug: 28081559

This reverts commit 961ea9fe42edcc2c57469bf451d1ca421da5cd59.

Change-Id: I98a5bb8112646706ae7bd73bf6393cb956466be3

8 years agoMerge "Ignore libcore.java.lang.ThreadTest failures on device."
Roland Levillain [Fri, 8 Apr 2016 16:01:18 +0000 (16:01 +0000)]
Merge "Ignore libcore.java.lang.ThreadTest failures on device."

8 years agoIgnore libcore.java.lang.ThreadTest failures on device.
Roland Levillain [Fri, 8 Apr 2016 15:04:11 +0000 (16:04 +0100)]
Ignore libcore.java.lang.ThreadTest failures on device.

Since https://android-review.googlesource.com/#/c/209427
has been merged, libcore.java.lang.ThreadTest tests are
failing with this error on ARM64:

  java.lang.UnsatisfiedLinkError: dlopen failed: library "libnativehelper_compat_libc++.so" not found
      at java.lang.Runtime.loadLibrary0(Runtime.java:977)
      at java.lang.System.loadLibrary(System.java:1628)
      at libcore.java.lang.ThreadTest.<clinit>(ThreadTest.java:26)

Ignore those failures for now to make the build turn green
again, while we investigate.

Bug: 28082914
Change-Id: Ia8c3c776c99aace06701479d1440c828d39cbec3

8 years agoMerge "Fix combined checksum calculation to use isa."
Jeff Hao [Fri, 8 Apr 2016 03:16:37 +0000 (03:16 +0000)]
Merge "Fix combined checksum calculation to use isa."

8 years agoMerge "Combine checksum of all boot images."
Jeff Hao [Fri, 8 Apr 2016 03:16:24 +0000 (03:16 +0000)]
Merge "Combine checksum of all boot images."

8 years agoFix combined checksum calculation to use isa.
Jeff Hao [Fri, 8 Apr 2016 02:46:31 +0000 (19:46 -0700)]
Fix combined checksum calculation to use isa.

Bug: 28054110

(cherry-picked from commit 8bbda8e683a1933755d0b8a68fed620cd63148ee)

Change-Id: I228db6ea1003e4e70d6a5fa7345d56409e70bd87

8 years agoCombine checksum of all boot images.
Jeff Hao [Thu, 7 Apr 2016 22:40:54 +0000 (15:40 -0700)]
Combine checksum of all boot images.

Allows a change to be detected in more than just the first image.
Combines checksums of all boot images using xor. A better hash is left
as TODO.

Bug: 28054110

(cherry-picked from commit 4f351aae00f43dd4e3290976f249d3664cc47091)

Change-Id: Ia557e2460ca582082368ef93508c04ce32c3036a

8 years agoMerge "ART: Fix interpreter single-step when doing OSR"
Bill Buzbee [Thu, 7 Apr 2016 23:47:55 +0000 (23:47 +0000)]
Merge "ART: Fix interpreter single-step when doing OSR"

8 years agoART: Fix interpreter single-step when doing OSR
buzbee [Thu, 7 Apr 2016 20:52:48 +0000 (13:52 -0700)]
ART: Fix interpreter single-step when doing OSR

Records an interpreter exit when in single-step mode and we've
finished processing the method via on-stack replacement.

Change-Id: I7a7d9ee9e4f3fbf7f829d9f54eca0e7c64395003

8 years agoMerge changes Ie23877ac,I262f6be8
Hiroshi Yamauchi [Thu, 7 Apr 2016 21:10:54 +0000 (21:10 +0000)]
Merge changes Ie23877ac,I262f6be8

* changes:
  Revert "Disable some image_test cases with concurrent collector."
  Fix image_test for the CC collector.

8 years agoRevert "Disable some image_test cases with concurrent collector."
Hiroshi Yamauchi [Thu, 7 Apr 2016 18:44:50 +0000 (11:44 -0700)]
Revert "Disable some image_test cases with concurrent collector."

This reverts commit d52765768b634c6f32a9bddd5c1269f26d32ea3a.

Bug: 27578460
Change-Id: Ie23877ac5332e05805c251749159e8f00af4ac8c

8 years agoFix image_test for the CC collector.
Hiroshi Yamauchi [Thu, 7 Apr 2016 18:31:26 +0000 (11:31 -0700)]
Fix image_test for the CC collector.

Increase the image address space reserve.

Bug: 27578460
Change-Id: I262f6be81c496a7a777013702e3da74ea4d619b4

8 years agoMerge "Remove AnnotationAccess and its remaining uses."
Jeff Hao [Thu, 7 Apr 2016 18:06:13 +0000 (18:06 +0000)]
Merge "Remove AnnotationAccess and its remaining uses."

8 years agoRevert "Revert "Check if we require barrier if we did not resolve classes""
Mathieu Chartier [Thu, 7 Apr 2016 17:52:52 +0000 (10:52 -0700)]
Revert "Revert "Check if we require barrier if we did not resolve classes""

This reverts commit a7ab4997f7263439561093ffbc7dea29181a47c5.

8 years agoRevert "Revert "Change RequiresConstructorBarrier default to yes""
Mathieu Chartier [Thu, 7 Apr 2016 17:19:48 +0000 (10:19 -0700)]
Revert "Revert "Change RequiresConstructorBarrier default to yes""

This reverts commit 0436ee6bd33a0b905cd2a7e333f7935da1bd5d86.

8 years agoMerge "Remove more Quick-related macros in tests."
Roland Levillain [Thu, 7 Apr 2016 16:13:10 +0000 (16:13 +0000)]
Merge "Remove more Quick-related macros in tests."

8 years agoMerge "Revert "Revert "Refactor HGraphBuilder and SsaBuilder to remove HLocals"""
David Brazdil [Thu, 7 Apr 2016 16:03:51 +0000 (16:03 +0000)]
Merge "Revert "Revert "Refactor HGraphBuilder and SsaBuilder to remove HLocals"""

8 years agoRevert "Revert "Refactor HGraphBuilder and SsaBuilder to remove HLocals""
David Brazdil [Thu, 7 Apr 2016 09:54:26 +0000 (09:54 +0000)]
Revert "Revert "Refactor HGraphBuilder and SsaBuilder to remove HLocals""

This patch merges the instruction-building phases from HGraphBuilder
and SsaBuilder into a single HInstructionBuilder class. As a result,
it is not necessary to generate HLocal, HLoadLocal and HStoreLocal
instructions any more, as the builder produces SSA form directly.

Saves 5-15% of arena-allocated memory (see bug for more data):
  GMS      20.46MB  =>  19.26MB  (-5.86%)
  Maps     24.12MB  =>  21.47MB  (-10.98%)
  YouTube  28.60MB  =>  26.01MB  (-9.05%)

This CL fixed an issue with parsing quickened instructions.

Bug: 27894376
Bug: 27998571
Bug: 27995065

Change-Id: I20dbe1bf2d0fe296377478db98cb86cba695e694