OSDN Git Service
Mathieu Chartier [Mon, 18 Jul 2016 18:11:45 +0000 (11:11 -0700)]
Do allocation fence before pushing on allocation stack
Heap::VisitObjects relies on having valid classes for objects in
the allocation stack. If the writes reorder, the thread calling
VisitObjects could see the free list pointer instead of the class
of the object. I believe this is causing crashes in VisitObjects.
Bug:
28790624
Test: Volantis booted
Change-Id: I0f2d4097de1ef3f5caf670ecc977d4d6837872ca
Roland Levillain [Mon, 18 Jul 2016 15:40:34 +0000 (15:40 +0000)]
Merge "MIPS64: Highest/Lowest Bit Intrinsic Support"
Roland Levillain [Mon, 18 Jul 2016 15:07:19 +0000 (15:07 +0000)]
Merge "Fix the build with respect to new VIXL."
Roland Levillain [Mon, 18 Jul 2016 15:03:05 +0000 (16:03 +0100)]
Fix the build with respect to new VIXL.
Test: Build ART for ARM64.
Change-Id: I2a9ebf145c61db9f8ceec6616963bac3ad5a7eb9
Roland Levillain [Mon, 18 Jul 2016 14:28:05 +0000 (14:28 +0000)]
Merge "Fixes to build against new VIXL interface."
Nicolas Geoffray [Mon, 18 Jul 2016 12:16:46 +0000 (12:16 +0000)]
Merge "Add a new control flow simplifier."
Treehugger Robot [Mon, 18 Jul 2016 11:19:14 +0000 (11:19 +0000)]
Merge "ARM64: Improve code generated to spill/restore for slow paths."
Alexandre Rames [Fri, 15 Jul 2016 16:41:13 +0000 (17:41 +0100)]
ARM64: Improve code generated to spill/restore for slow paths.
Aligning the accesses allows generating better code.
Before:
add x16, sp, #0x44 (68)
stp x0, x1, [x16, #-16]
After:
stp x0, x1, [sp, #56]
Change-Id: I3e20ad3fa59d00aee4b4d14ea9d59c7cd546509e
Nicolas Geoffray [Fri, 15 Jul 2016 09:46:17 +0000 (10:46 +0100)]
Add a new control flow simplifier.
Run it in the dead code elimination phase, as it relates to
creating dead branches.
From 0.04 to 0.07% less code size framework/gms/docs/fb (70K saved on fb)
3%-5% runtime performance improvements on Richards/DeltaBlue/Ritz.
Compile-time is mixed, so in the noise (from 2% slower to 1% faster).
test:611-checker-simplify-if
Change-Id: Ife8b7882d57b5481f5ca9dc163beba655d7e78bf
Treehugger Robot [Sat, 16 Jul 2016 04:46:46 +0000 (04:46 +0000)]
Merge "ART: Replace ScopedFd with FdFile"
Andreas Gampe [Sat, 16 Jul 2016 00:17:34 +0000 (17:17 -0700)]
ART: Replace ScopedFd with FdFile
FdFile can now be used like ScopedFd. Remove ScopedFd.
Bug:
21192156
Test: m test-art-host
Test: m test-art-target (shamu)
Change-Id: I32115fa8b2b8bb5aa5d1886eae63522f80ce836b
Treehugger Robot [Fri, 15 Jul 2016 16:40:28 +0000 (16:40 +0000)]
Merge "ART: disassembler_x86 doesn't recognize NOPs"
Scott Wakeling [Fri, 24 Jun 2016 15:19:36 +0000 (16:19 +0100)]
Fixes to build against new VIXL interface.
- Fix namespace usage and use of deprecated functions.
- Link all dependants to new libvixl-arm64 target for now.
Change-Id: Iee6f299784fd663fc2a759f3ee816fdbc511e509
Andreas Gampe [Fri, 15 Jul 2016 00:45:22 +0000 (00:45 +0000)]
Merge changes from topic 'cherry_pinner'
* changes:
ART: Fix build break
Add API for getting location of odex or oat file
Andreas Gampe [Thu, 14 Jul 2016 23:50:52 +0000 (16:50 -0700)]
ART: Fix build break
Build break introduced by mismatch from merges.
Bug:
28251566
(cherry picked from commit
a472e4fd6260740af230bab7be5740d8c7bac360)
Test: m test-art-host
Change-Id: I83a333dd7394af630c5a79153d4838eb3a910714
Philip Cuadra [Tue, 12 Jul 2016 23:37:40 +0000 (16:37 -0700)]
Add API for getting location of odex or oat file
Add an API for getting the file path of odex or oat file given a dex
path.
Bug
28251566
(cherry picked from commit
b4827ace453b9280060a826e8f22cc8c9b6edb7d)
Change-Id: Ibebaa20f15d8135b25d9eb5927b7979801ebf0b2
Treehugger Robot [Thu, 14 Jul 2016 21:09:55 +0000 (21:09 +0000)]
Merge "Revert "Revert "Dump more dex file data in oatdump"""
Mathieu Chartier [Thu, 14 Jul 2016 17:10:44 +0000 (10:10 -0700)]
Revert "Revert "Dump more dex file data in oatdump""
Delete runtime to fix leak before callin exit.
Bug:
29462018
This reverts commit
9c05578dd2306231437bd290c0f70abc2bb3b6d8.
Change-Id: Ica23ba0f2d07496d0e4a3288329945f612ac3b20
Treehugger Robot [Thu, 14 Jul 2016 19:44:38 +0000 (19:44 +0000)]
Merge "Integer.bitCount and Long.bitCount intrinsics for ARM"
Treehugger Robot [Thu, 14 Jul 2016 18:48:03 +0000 (18:48 +0000)]
Merge "ART: Add FdFile constructors"
xueliang.zhong [Tue, 5 Jul 2016 14:28:19 +0000 (15:28 +0100)]
Integer.bitCount and Long.bitCount intrinsics for ARM
Change-Id: I4ed3e779415be026c7d090b61a3e356b37c418e5
Roland Levillain [Thu, 14 Jul 2016 09:32:34 +0000 (09:32 +0000)]
Merge "Introduce more compact ReadBarrierMark slow-paths."
Nicolas Geoffray [Thu, 14 Jul 2016 09:24:45 +0000 (09:24 +0000)]
Merge "Revert "Dump more dex file data in oatdump""
Nicolas Geoffray [Thu, 14 Jul 2016 09:24:30 +0000 (09:24 +0000)]
Revert "Dump more dex file data in oatdump"
Breaks valgrind.
Bug:
29462018
This reverts commit
8e2c56252aa9527bd9a82bdd147fdc46cf5deb9c.
Change-Id: If58cedcee75dd0eda8571e90d63e080a4709d773
Nicolas Geoffray [Thu, 14 Jul 2016 07:56:16 +0000 (07:56 +0000)]
Merge "Fix a bug in ClassTableGet code generation for IMTs."
Andreas Gampe [Thu, 13 Aug 2015 23:44:54 +0000 (16:44 -0700)]
ART: Add FdFile constructors
Make Open protected, and expose constructors instead. Add a move
constructor and move assignment operator.
Add OS functions that return the FdFile non-pointer version.
Add tests.
Bug:
21192156
Test: m test-art-host
Test: m test-art-target (shamu)
Change-Id: I83e390edde7cd37c900e9d5c3e4d21da22981b3f
Treehugger Robot [Wed, 13 Jul 2016 21:35:59 +0000 (21:35 +0000)]
Merge "Dump more dex file data in oatdump"
Mathieu Chartier [Tue, 21 Jun 2016 22:14:20 +0000 (15:14 -0700)]
Dump more dex file data in oatdump
Dump some statistics for each dex file along side with strings loaded
from code and dex code bytes.
Sample output:
Cumulative dex file data
Num string ids: 202809
Num method ids: 320464
Num field ids: 162822
Num type ids: 68151
Num class defs: 48061
Unique strings loaded from dex code: 51049
Total strings loaded from dex code: 106651
Number of unique dex code items: 247929
Total number of dex code bytes:
11090574
Added content testing to oat dump test. No significant slowdown.
TEST: test-art-host
Bug:
29462018
Change-Id: I60effd3087d8c427eda4ee26431d5d77165b3939
Treehugger Robot [Wed, 13 Jul 2016 18:07:04 +0000 (18:07 +0000)]
Merge "Improve search for available spill slots in RA"
Treehugger Robot [Wed, 13 Jul 2016 17:55:52 +0000 (17:55 +0000)]
Merge "Refactored a few dexdump alloc/free into unique_ptr"
Vladimir Marko [Wed, 13 Jul 2016 16:37:34 +0000 (16:37 +0000)]
Merge "X86: Use memory to do array range checks"
Treehugger Robot [Wed, 13 Jul 2016 15:42:18 +0000 (15:42 +0000)]
Merge changes I91249ba8,Ic93812d9
* changes:
Use "" to indicate the oat filename could not be computed.
Compute and cache oat file status in OatFileAssistant.
Nicolas Geoffray [Wed, 13 Jul 2016 13:13:48 +0000 (14:13 +0100)]
Fix a bug in ClassTableGet code generation for IMTs.
Introduced by:
https://android-review.googlesource.com/#/c/244980/
test:566-polymorphic-inling for fixing x86 crash. Also
fixes a performance regression.
bug:
29188168
Change-Id: Id90cb819c88e7ba3db1cb3c50c517a112ab7d784
Mark Mendell [Tue, 12 Jul 2016 15:13:15 +0000 (11:13 -0400)]
X86: Use memory to do array range checks
Currently, an HBoundsCheck is fed by an HArrayLength, causing a load of
the array length, followed by a register compare.
Avoid the load when we can by comparing directly with the array length
in memory. Implement this by marking the HArrayLength as 'emitted at
use site', and then generating the code in the HBoundsCheck.
Only do this replacement when we are the only user of the ArrayLength
and it isn't visible to the environment.
Handle the special case where the array is 'null' and where an implicit
null check can't be eliminated.
This code moves the load of the length to the slow code for the failed
check, which is what we want.
Test: 609-checker-x86-bounds-check
Change-Id: I9cdb183301e048234bb0ffeda940eedcf4a655bd
Signed-off-by: Mark Mendell <mark.p.mendell@intel.com>
Roland Levillain [Wed, 13 Jul 2016 10:54:35 +0000 (11:54 +0100)]
Introduce more compact ReadBarrierMark slow-paths.
Replace entry point ReadBarrierMark with 32
ReadBarrierMarkRegX entry points, using register
number X as input and output (instead of the standard
runtime calling convention) to save two moves in Baker's
read barrier mark slow-path code.
Test: ART host and target (ARM, ARM64) tests.
Bug:
29506760
Bug:
12687968
Change-Id: I73cfb82831cf040b8b018e984163c865cc44ed87
Nicolas Geoffray [Wed, 13 Jul 2016 09:19:45 +0000 (09:19 +0000)]
Merge "Blacklist flaky test."
Nicolas Geoffray [Wed, 13 Jul 2016 09:18:48 +0000 (10:18 +0100)]
Blacklist flaky test.
bug:
30107038
Change-Id: Id976c6c7d764a44c1b30155714fb64b4441cdac2
Treehugger Robot [Wed, 13 Jul 2016 00:35:48 +0000 (00:35 +0000)]
Merge "Update the get_process_name call."
Aart Bik [Tue, 12 Jul 2016 22:53:13 +0000 (15:53 -0700)]
Refactored a few dexdump alloc/free into unique_ptr
Rationale: easier to read
Test: dexdump's output is well-tested
Change-Id: Ib5b36dce68d442e555599f9427bd9bfa7c6f3831
Treehugger Robot [Tue, 12 Jul 2016 18:32:29 +0000 (18:32 +0000)]
Merge changes Iafd12677,I40595d15
* changes:
profile_changed should not effect GetBestOatFile.
Make OpenImageSpace static and ArtFileName internal.
Matthew Gharrity [Mon, 11 Jul 2016 21:45:01 +0000 (14:45 -0700)]
Improve search for available spill slots in RA
Previously we always searched for two adjacent spill slots, even if
we only needed one. This small change fixes that.
Test: m test-art-host
Change-Id: I021d355e6602ffee687c8537a959232b1504dcf1
Vladimir Marko [Tue, 12 Jul 2016 15:02:31 +0000 (15:02 +0000)]
Merge "ARM64: Shorter fast-path for read barrier field load."
Vladimir Marko [Tue, 12 Jul 2016 14:14:26 +0000 (14:14 +0000)]
Merge "Two more patterns for instruction simplifier"
Vladimir Marko [Mon, 11 Jul 2016 18:30:56 +0000 (19:30 +0100)]
ARM64: Shorter fast-path for read barrier field load.
Reduces the aosp_flounder-userdebug 64-bit boot.oat by
~4.2MiB, i.e. ~3.5%, in the ART_USE_READ_BARRIER=true
configuration.
Test: Tested with ART_USE_READ_BARRIER=true on Nexus 9.
Bug:
29966877
Bug:
12687968
Change-Id: Ic9853ec82747a1702e4091fcf9cbf06674fcd5f6
Vladimir Marko [Tue, 12 Jul 2016 12:03:02 +0000 (12:03 +0000)]
Merge "ARM: Shorter fast-path for read barrier field load."
Maxim Kazantsev [Tue, 12 Jul 2016 09:55:33 +0000 (15:55 +0600)]
Two more patterns for instruction simplifier
This patch allows the instruction simplifier to cover
two more cases:
x = a + b and x = a - b
y = x - b y = x + b
In both cases y can be replaced with a, if the types
of arguments are integral.
Change-Id: I4735737f471e31057f1df76ddace5d4d1c79cde6
Test: New tests in 458-checker-instruction-simplification.
Signed-off-by: Max Kazantsev <maxim.kazantsev@intel.com>
Roland Levillain [Tue, 12 Jul 2016 11:07:50 +0000 (11:07 +0000)]
Merge "Rename kCall to kCallOnMainOnly"
Vladimir Marko [Mon, 11 Jul 2016 14:52:00 +0000 (15:52 +0100)]
ARM: Shorter fast-path for read barrier field load.
Reduces the aosp_hammerhead-userdebug boot.oat by 2.2MiB,
i.e. ~2.2%, in the ART_USE_READ_BARRIER=true configuration.
Test: Tested with ART_USE_READ_BARRIER=true on Nexus 5.
Bug:
29966877
Bug:
12687968
Change-Id: I4454150003e12a1aa7f0cf451627dc1ee9a495ae
Dmitriy Filchenko [Tue, 12 Jul 2016 00:41:28 +0000 (17:41 -0700)]
Update the get_process_name call.
BUG:
29824203
Test: test-art-host
Change-Id: Iba75b48b836b6ec249e6c77b9895ab3575f1027b
Oleksiy Vyalov [Mon, 11 Jul 2016 22:38:45 +0000 (22:38 +0000)]
Merge "Call prctl(PR_SET_PTRACER, PR_SET_PTRACER_ANY) if app is debuggable."
Oleksiy Vyalov [Tue, 21 Jun 2016 23:21:37 +0000 (16:21 -0700)]
Call prctl(PR_SET_PTRACER, PR_SET_PTRACER_ANY) if app is debuggable.
Enabled Yama security may prevent non-privileged debugger from
attaching to the app. However, if app is marked is debuggable
then it should acceptable to allow debugger with arbitrary pid
to connect the app.
Bug:
29549463
Change-Id: I50c79dbca272dc473e489cc61471628e2fdc4a03
(cherry picked from commit
e21145f6625a7faa939c1ca17fb95285aa245c16)
Aart Bik [Mon, 11 Jul 2016 19:33:07 +0000 (19:33 +0000)]
Merge "Fix clang static analyzer on potential memory leaks."
Aart Bik [Sat, 9 Jul 2016 01:32:45 +0000 (18:32 -0700)]
Fix clang static analyzer on potential memory leaks.
Rationale:
First was not really a leak, but the combination of stack/heap
allocated data structures with a single re-allocation was a bit
complex for static analysis. Second was a true leak.
BUG=
30040701
Change-Id: I10c2095f3061bd2ad49f24f7c25fcd7ca0f95111
Serban Constantinescu [Thu, 7 Jul 2016 17:03:19 +0000 (18:03 +0100)]
Rename kCall to kCallOnMainOnly
This patch renames kCall to kCallOnMainOnly in preparation for
the next patch in this series which will be adding kCallOnMainAndSlowPath.
Note: With this patch there will be places where we use kCallOnMainOnly
even though we call on the slow path too. The next patch in this series
will fix that.
Test: ART host tests.
Change-Id: Iabfdb0901990d163be5d780f3bdd2fab6fa17b32
Roland Levillain [Mon, 11 Jul 2016 09:24:43 +0000 (09:24 +0000)]
Merge "Have LOCAL_ASFLAGS honor debug/non-debug configuration."
Treehugger Robot [Sat, 9 Jul 2016 02:27:50 +0000 (02:27 +0000)]
Merge "Fix unsafe weak reference usage in test 141"
Mathieu Chartier [Thu, 30 Jun 2016 22:20:26 +0000 (15:20 -0700)]
Fix unsafe weak reference usage in test 141
There were patterns like:
WeakRef<Class> klass = ...
klass.get().<method>
These are unsafe since background GC could unload and null out the
weak reference. We now pass the weak reference to another function.
Deleted bad and unsafe test.
Test: test-art-host
Bug:
29889416
Change-Id: I6a8daf678262757e54e5817ec63b7c33f4b156ab
Jeff Hao [Fri, 8 Jul 2016 20:06:58 +0000 (20:06 +0000)]
Merge "Revert "Revert "Optimize IMT"""
Chris Larsen [Fri, 15 Apr 2016 17:00:56 +0000 (10:00 -0700)]
MIPS64: Highest/Lowest Bit Intrinsic Support
- int java.lang.Integer.highestOneBit(int)
- long java.lang.Long.highestOneBit(long)
- int java.lang.Integer.lowestOneBit(int)
- long java.lang.Long.lowestOneBit(long)
Change-Id: I0db7858542130d759ee7d17ecb1eadb7ec3dc04c
Treehugger Robot [Fri, 8 Jul 2016 17:31:52 +0000 (17:31 +0000)]
Merge "Fix missing patchoat dependency for dex2oat test."
Aart Bik [Fri, 8 Jul 2016 17:26:30 +0000 (17:26 +0000)]
Merge "MIPS64: java.lang.Math.round"
Roland Levillain [Fri, 8 Jul 2016 16:28:23 +0000 (16:28 +0000)]
Merge "Math Round Intrinsic Implementations For Java8."
Richard Uhler [Fri, 8 Jul 2016 16:22:33 +0000 (09:22 -0700)]
Fix missing patchoat dependency for dex2oat test.
Change-Id: I7d00e676b895da805a1e476214cb4c92d7581e2b
Test: rm -r out && m test-art-host-gtest-dex2oat_test
Test: rm -r out && m test-art-host-gtest-oat_file_assistant_test
Vladimir Marko [Fri, 8 Jul 2016 15:43:25 +0000 (15:43 +0000)]
Merge "ARM: Remove unnecessary VMOV from float/double-to-int."
Roland Levillain [Fri, 8 Jul 2016 15:15:17 +0000 (15:15 +0000)]
Merge "Fix the definition of THREAD_CURRENT_IBASE_OFFSET."
Treehugger Robot [Fri, 8 Jul 2016 14:36:43 +0000 (14:36 +0000)]
Merge "ARM: Use 64-bit literals for LoadDImmediate()."
Roland Levillain [Fri, 8 Jul 2016 13:21:44 +0000 (14:21 +0100)]
Fix the definition of THREAD_CURRENT_IBASE_OFFSET.
The difference between between
art::Thread::tlsPtr_.mterp_current_ibase and
art::Thread::tlsPtr_.thread_local_objects is sizeof(size_t),
not sizeof(void*) (even if they have the same value on all
supported platforms/compilers in practice).
Test: ART host tests.
Change-Id: I70d196732499bd0cd073983de564547d08ee59ef
xueliang.zhong [Fri, 27 May 2016 17:56:13 +0000 (18:56 +0100)]
Math Round Intrinsic Implementations For Java8.
According to Java 8 API definition Math.round() performs:
Return the closest long or int to the argument,
with ties rounding to positive infinity.
Before this patch, the Math.round intrinsics were disabled,
because they followed the old definition of Math.round: floor(input+0.5).
This patch reimplement Math.round intrinsic to follow the new definition.
Change-Id: I453c01a9857a3d461b230e4d4e98c2d7bdf7a3c7
Vladimir Marko [Thu, 7 Jul 2016 14:37:02 +0000 (15:37 +0100)]
ARM: Use 64-bit literals for LoadDImmediate().
And rewrite the medium-range long/fp literal to use
MOVW+ADD+LDRD/VLDR because the old instruction sequence
was broken if the "ADD ip, pc" was not 4-byte aligned.
Test: assembler_thumb2_test has been updated. Standard
ART test suite has been run on host and Nexus 5.
Change-Id: I37c6a62aa6e77c6a9701b5a1fb4db2e666c1eae9
Nicolas Geoffray [Fri, 8 Jul 2016 12:51:01 +0000 (12:51 +0000)]
Merge "Implement System.arraycopy intrinsic on x86."
Roland Levillain [Fri, 8 Jul 2016 11:19:03 +0000 (11:19 +0000)]
Merge "Mark ReadBarrierJni as a direct entry point on MIPS32."
Richard Uhler [Tue, 19 Apr 2016 19:41:04 +0000 (12:41 -0700)]
Use "" to indicate the oat filename could not be computed.
Instead of having an extra boolean flag.
Test: oat file assistant tests.
Change-Id: I91249ba862522e691c6b67fd28a2eac58dcde07e
Richard Uhler [Mon, 18 Apr 2016 17:40:50 +0000 (10:40 -0700)]
Compute and cache oat file status in OatFileAssistant.
Instead of computing and caching IsOutOfDate + IsUpToDate separately.
Test: oat file assistant tests.
Change-Id: Ic93812d99c547f54a1c29315d9a3572180cee8d1
Richard Uhler [Fri, 15 Apr 2016 22:18:56 +0000 (15:18 -0700)]
profile_changed should not effect GetBestOatFile.
This change moves the check for whether a profile changed from
GetBestOatFile to GetDexOptStatus, because profile_changed should not
effect what oat files are loaded.
Test: OatFileAssistantTest.ProfileOatUpToDate
Change-Id: Iafd12677f20d2844809337d1d83b688f17461cc0
Richard Uhler [Fri, 15 Apr 2016 22:51:21 +0000 (15:51 -0700)]
Make OpenImageSpace static and ArtFileName internal.
Test: m build-art-host
Test: m build-art-host-tests
Change-Id: I40595d158e63cc547eba988b6c19df00ff589b05
Bill Buzbee [Thu, 7 Jul 2016 21:53:29 +0000 (21:53 +0000)]
Merge "Revert "Revert "ART: Add Mterp export pc poison testing mode"""
Treehugger Robot [Thu, 7 Jul 2016 17:53:56 +0000 (17:53 +0000)]
Merge "Add missing dependency required for ahat-test."
Nicolas Geoffray [Wed, 6 Jul 2016 11:09:12 +0000 (12:09 +0100)]
Implement System.arraycopy intrinsic on x86.
Also remove wrong comments when doing the raw copying.
test:run-test, 537-checker-arraycopy, 610-arraycopy
Change-Id: I2495bc03cde8ccad03c93f7722dd29bf85138525
Bill Buzbee [Thu, 7 Jul 2016 14:42:43 +0000 (14:42 +0000)]
Revert "Revert "ART: Add Mterp export pc poison testing mode""
This reverts commit
d7af522c0f03cd926b13bbef7c21a8a504969f36.
Fixes mips assembly typo: "mov" -> "move"
TEST: Previously tested arm/arm64/x86/x86_64 via test-art-host,
test-art-target plus injected error. Neglected to test mips
builds. Now those tested as well.
Change-Id: If2f9fb06973f8c1220a7d7699d79e6b9f1ce3197
Treehugger Robot [Thu, 7 Jul 2016 13:38:44 +0000 (13:38 +0000)]
Merge "Fix System.arraycopy when doing same array copying."
Vladimir Marko [Thu, 7 Jul 2016 11:07:44 +0000 (12:07 +0100)]
ARM: Remove unnecessary VMOV from float/double-to-int.
Test: Run standard ART test suite on Nexus 5.
Change-Id: I780fd0cca68f89401d2a114e1022bed498d02979
Nicolas Geoffray [Thu, 7 Jul 2016 11:07:42 +0000 (12:07 +0100)]
Fix System.arraycopy when doing same array copying.
At compile time, if constant source < constant destination, and we don't
know if the arrays are the same, then we must emit code that checks
if the two arrays are the same. If so, we jump to the slow path.
test:610-arraycopy
Change-Id: Ida67993d472b0ba4056d9c21c68f6e5239421f7d
Artem Udovichenko [Thu, 30 Jun 2016 09:18:25 +0000 (09:18 +0000)]
Revert "Revert "Optimize IMT""
This reverts commit
88f288e3564d79d87c0cd8bb831ec5a791ba4861.
Change-Id: I49605d53692cbec1e2622e23ff2893fc51ed4115
Nicolas Geoffray [Thu, 7 Jul 2016 08:28:47 +0000 (08:28 +0000)]
Merge "Revert "Revert "Inline and optimize interface calls."""
Bill Buzbee [Wed, 6 Jul 2016 23:31:30 +0000 (23:31 +0000)]
Merge "Revert "ART: Add Mterp export pc poison testing mode""
Bill Buzbee [Wed, 6 Jul 2016 23:28:26 +0000 (23:28 +0000)]
Revert "ART: Add Mterp export pc poison testing mode"
This reverts commit
e664cfef4d373a037dcf18637bea6b41e0624302.
Change-Id: I521dfccd6c829946c93290163bf817b8a2c368e9
Richard Uhler [Fri, 1 Jul 2016 20:10:56 +0000 (13:10 -0700)]
Add missing dependency required for ahat-test.
Change-Id: I8f7afc97a82fadcdd2f18032900f2b8d8b2059bf
Test: rm -r out/ && m ahat-test
Test: rm -r out/ && m build-art-host
Test: rm -r out/ && m build-art-host-tests
Test: rm -r out/ && m build-art-target
Test: rm -r out/ && m build-art-target-tests
Bill Buzbee [Wed, 6 Jul 2016 21:06:06 +0000 (21:06 +0000)]
Merge "ART: Add Mterp export pc poison testing mode"
Richard Uhler [Wed, 6 Jul 2016 20:02:20 +0000 (20:02 +0000)]
Merge "Revert "Add missing dependency required for ahat-test.""
Richard Uhler [Wed, 6 Jul 2016 19:59:33 +0000 (19:59 +0000)]
Revert "Add missing dependency required for ahat-test."
This reverts commit
13f6d39da8ba6450dc861c4bfd09d21f3b403d4b.
Breaks 'build-art-target'.
https://build.chromium.org/p/client.art/builders/mips64-emulator-debug/builds/5651/steps/build%20target/logs/stdio
Change-Id: I1466f92308955604ecfcac65deff86cb581d6fdc
Treehugger Robot [Wed, 6 Jul 2016 18:21:34 +0000 (18:21 +0000)]
Merge "Add missing dependency required for ahat-test."
buzbee [Thu, 23 Jun 2016 20:35:39 +0000 (13:35 -0700)]
ART: Add Mterp export pc poison testing mode
One of the optimizations of the Mterp assembly interpreter is
that it attempts to limit the number of times it updates the saved
dex_pc_ value stored in the shadow frame. Instead, it maintains
a pointer to the current dex instruction in a dedicated register.
However, whenever execution flow exits the immediate mterp
environment and some other part of the runtime needs to know the
current dex pc (for example - exception throws), mterp must
save - or "export" the dex_pc back into it's home location in
the shadow frame.
Forgetting to do this is a typical bug when developing mterp
instruction handlers. And, the failures show up in odd and
sometimes hard to track down ways. This CL adds a testing mode
that poisons the saved dex_pc prior to each instruction
interpretation. Thus, if any individual handler neglects to
properly export the current dex pc and then goes on to use it
we'll segfault out with an identifiable fault address: 0xdead00ff.
TEST: test-art-target (Nexus 5x, Nexus 6p), test-art-host. Verified
expected failure by artificially neglecting an EXPORT_PC.
Bug:
26572899
Change-Id: I704fbb8ece693b94341aed0ffbb51a6acfb89697
Nicolas Geoffray [Wed, 6 Jul 2016 14:18:23 +0000 (14:18 +0000)]
Revert "Revert "Inline and optimize interface calls.""
This reverts commit
19dc255bf94a4229de8627a2079ee6f0e9005e2d.
Change-Id: Ifc3a92280878d4db1b460d486137497b3456beae
Nicolas Geoffray [Wed, 6 Jul 2016 14:02:13 +0000 (14:02 +0000)]
Merge "Revert "Inline and optimize interface calls.""
Nicolas Geoffray [Wed, 6 Jul 2016 14:01:50 +0000 (14:01 +0000)]
Revert "Inline and optimize interface calls."
Went ahead too quickly.
This reverts commit
d9faceaa8da92f4a56c2907de949081bd42faf79.
Change-Id: I1610deaf89b38037cf2786d135a59c48b10ced0c
Treehugger Robot [Wed, 6 Jul 2016 13:37:49 +0000 (13:37 +0000)]
Merge "Inline and optimize interface calls."
Roland Levillain [Wed, 6 Jul 2016 13:20:52 +0000 (13:20 +0000)]
Merge "Fix stack alignment in x86 read barrier entry points."
Roland Levillain [Wed, 6 Jul 2016 12:45:52 +0000 (13:45 +0100)]
Mark ReadBarrierJni as a direct entry point on MIPS32.
The ReadBarrierJni MIPS32 entry point is a direct reference
to the native (C++) method art::ReadBarrierJni.
Test: Build ART for MIPS32.
Bug:
12687968
Change-Id: I6475855f4696690a681f69931618a055be65a00a
Nicolas Geoffray [Wed, 6 Jul 2016 09:19:23 +0000 (10:19 +0100)]
Inline and optimize interface calls.
- Support for inlining methods with invoke-interface (the
previous performance limitation is now fixed with the new
ImtConflictTable).
- Turn non inlineable invoke-interface into invoke-virtual to
avoid any potential execution of the conflict trampoline.
test:609-checker-inline-interface
Change-Id: Ibe33bc945deaada8fb78541e6e0bf034ebb52420
Vladimir Marko [Wed, 6 Jul 2016 12:12:28 +0000 (12:12 +0000)]
Merge "Expect spurious wake-up from futex(2)/FUTEX_WAIT call."