OSDN Git Service
Mathieu Chartier [Mon, 14 Mar 2016 20:14:30 +0000 (20:14 +0000)]
Merge "Fix cases where we miss instrumentation changes"
Mathieu Chartier [Sat, 12 Mar 2016 01:22:23 +0000 (17:22 -0800)]
Fix cases where we miss instrumentation changes
Moved allocation stack push to after we record the allocation since
it can cause thread suspension.
Added handling in entrypoint utils for thread suspension cases.
Keep the AllocRecordObjectMap around since we do not want to delete
it if there are any threads waiting on new_record_condition_. The
condition guards adding stack traces while the GC is running. If we
delete the map and there are still waiters that did not resume, it
caused a CHECK failure. This could happen in cases where one thread
disables allocation tracking while other threads are about to
resume from the condition.
Bug:
27506909
(cherry picked from commit
0f394021d6abfdf9ebea6c7f405ec936a812ea62)
Change-Id: I1dc51e9f000684b4032b57beab59d317ece26f06
Vladimir Marko [Mon, 14 Mar 2016 19:13:14 +0000 (19:13 +0000)]
Merge "Optimizing: Fix TypeConversion(And(x, const)) simplification."
Vladimir Marko [Mon, 14 Mar 2016 18:00:05 +0000 (18:00 +0000)]
Optimizing: Fix TypeConversion(And(x, const)) simplification.
Avoid introducing implicit conversions when simplifying the
expression TypeConversion(And(x, const)). Previously, when
we dropped the And, we could end up with a TypeConversion to
the same type which should be eliminated on subsequent pass
of the block's instructions; however, a subsequent dependent
TypeConversion in the same block would be processed earlier
and we would unexpectedly see its input as the conversion to
the same type, failing a DCHECK().
Bug:
27626509
Change-Id: I5874a9ceafbf635cf3391beea807ede8468ab5c3
Hiroshi Yamauchi [Mon, 14 Mar 2016 18:47:50 +0000 (18:47 +0000)]
Merge "Clean up the valgrind test targets."
Hiroshi Yamauchi [Mon, 14 Mar 2016 18:18:36 +0000 (11:18 -0700)]
Clean up the valgrind test targets.
Allow the 32-bit or the 64-bit valgrind tests to run separately or both.
Bug:
27156726
Change-Id: Ib439eb2f57d3302be152cb5af1a8960a41725f31
Vladimir Marko [Mon, 14 Mar 2016 17:42:19 +0000 (17:42 +0000)]
Merge "Optimizing: Do not re-record standby checks for dynamic BCE."
Nicolas Geoffray [Mon, 14 Mar 2016 15:54:54 +0000 (15:54 +0000)]
Merge "Revert "Enable compilation of secondary dex files""
Nicolas Geoffray [Mon, 14 Mar 2016 15:54:39 +0000 (15:54 +0000)]
Revert "Enable compilation of secondary dex files"
Reverting until we fix the buildbot failures on jdwp and libcore.
Bug:
26719109
This reverts commit
770b39eb564ec16c672ace83515535b055edb2b1.
Change-Id: Ifba366d2ddc9311737b6ea29c007475a40cc803f
Andreas Gampe [Mon, 14 Mar 2016 15:39:01 +0000 (15:39 +0000)]
Merge "ART: Check alignment of section offsets"
David Srbecky [Mon, 14 Mar 2016 14:19:40 +0000 (14:19 +0000)]
Merge "Add symbolizer option to generate debug information only."
David Srbecky [Mon, 14 Mar 2016 14:10:24 +0000 (14:10 +0000)]
Merge "Ignore empty maps when emitting DWARF variable locations."
Vladimir Marko [Mon, 14 Mar 2016 13:56:02 +0000 (13:56 +0000)]
Optimizing: Do not re-record standby checks for dynamic BCE.
Adding the checks to the standby vector invalidates the
vector storage used by range-based loop in the
BCEVisitor::Finish() as exposed by valgrind in image_test.
Bug:
27597089
Change-Id: Ib0f0e8cdfdb7211a64a26836e085cb99fb2ce8b8
Calin Juravle [Mon, 14 Mar 2016 11:38:24 +0000 (11:38 +0000)]
Merge "Enable compilation of secondary dex files"
Roland Levillain [Mon, 14 Mar 2016 10:18:29 +0000 (10:18 +0000)]
Merge "Revert "Re-enable test that was causing TimeoutExceptions on ARM64.""
David Srbecky [Sat, 12 Mar 2016 21:10:04 +0000 (21:10 +0000)]
Ignore empty maps when emitting DWARF variable locations.
This is rewrite of https://android-review.googlesource.com/#/c/202115
The aim in both cases is to avoid gaps in generated locations.
(which helps to keep the size of the generated DWARF down).
However, the previous CL was a bit too eager with extending of
variable scope and reporting locations. We might have reported
variable in scope when in fact, it was not.
This CL implements simpler solution by just filtering stackmaps
without dex register maps at first opportunity. This should
ensure that locations for breakpoint locations are completely
accurate as originally intended.
Change-Id: I98378716c0ef5ef46b12181502904621eb6ecf2f
Roland Levillain [Sat, 12 Mar 2016 13:42:33 +0000 (13:42 +0000)]
Revert "Re-enable test that was causing TimeoutExceptions on ARM64."
The test is failing again on the ARMv8 concurrent collector
configuration
(https://build.chromium.org/p/client.art/builders/armv8-concurrent-collector/builds/1833/steps/test%20libcore/logs/stdio).
This reverts commit
9e9411cc6edf5fa8be0cabd11203fac09e4bac13.
Bug:
26155567
Change-Id: I81b543a48cbbc1eca99efcc430965222ca6edfec
Bill Buzbee [Fri, 11 Mar 2016 23:05:59 +0000 (23:05 +0000)]
Merge "ART: mterp arm/arm64 cleanup"
buzbee [Fri, 11 Mar 2016 17:51:11 +0000 (09:51 -0800)]
ART: mterp arm/arm64 cleanup
Assembly code cleanup in response to comments from already-submitted
CL: https://android-review.googlesource.com/#/c/188977/
Change-Id: I0ea85c5759a08cb50ef3e97dc5cf79b3ba041640
Stephen Hines [Fri, 11 Mar 2016 21:19:48 +0000 (21:19 +0000)]
Merge "Enable clang for ARM builds."
Hiroshi Yamauchi [Fri, 11 Mar 2016 21:18:21 +0000 (21:18 +0000)]
Merge "Assembly region TLAB allocation fast path for arm."
Hiroshi Yamauchi [Wed, 24 Feb 2016 20:51:18 +0000 (12:51 -0800)]
Assembly region TLAB allocation fast path for arm.
This is for the CC collector.
Share the common fast path code with the tlab fast path code.
Speedup (on N5):
BinaryTrees: 2291 -> 902 ms (-60%)
MemAllocTest: 2137 -> 1845 ms (-14%)
Bug:
9986565
Bug:
12687968
Change-Id: Ica63094ec2f85eaa4fd04d202a20090399275d85
Christopher Ferris [Fri, 11 Mar 2016 20:14:18 +0000 (20:14 +0000)]
Merge "ART: Allow unwinding unattached threads"
Vladimir Marko [Fri, 11 Mar 2016 19:48:08 +0000 (19:48 +0000)]
Merge "Integrate BitwiseNegated into shared framework."
Roland Levillain [Fri, 11 Mar 2016 19:28:44 +0000 (19:28 +0000)]
Merge "Typo in comment in art::mirror::Object::IdentityHashCode."
David Srbecky [Fri, 11 Mar 2016 19:25:45 +0000 (19:25 +0000)]
Merge "Fix DCHECK in OatQuickMethodHeader::GetFrameSizeInBytes()"
David Srbecky [Fri, 11 Mar 2016 18:48:55 +0000 (18:48 +0000)]
Fix DCHECK in OatQuickMethodHeader::GetFrameSizeInBytes()
We should check whether the frame is divisible by kStackAlignment.
Change-Id: I5936a6e90b38c09775b4c3a8212641e639c4181c
Aart Bik [Fri, 11 Mar 2016 18:43:28 +0000 (18:43 +0000)]
Merge "Tests for round() method."
Hiroshi Yamauchi [Fri, 11 Mar 2016 18:24:22 +0000 (18:24 +0000)]
Merge "Suppress a valgrind libc issue."
Hiroshi Yamauchi [Thu, 10 Mar 2016 21:49:14 +0000 (13:49 -0800)]
Suppress a valgrind libc issue.
To prevent our valgrind tests from failing.
Bug:
27596582
Bug:
27156726
Change-Id: Ic55256bd09fdffcf2a0c8fa4fa7951bc648b86a8
David Srbecky [Fri, 11 Mar 2016 18:19:11 +0000 (18:19 +0000)]
Merge "Revert "Revert "Allow duplicated methods in different DWARF line tables."""
David Srbecky [Fri, 11 Mar 2016 14:35:45 +0000 (14:35 +0000)]
Revert "Revert "Allow duplicated methods in different DWARF line tables.""
This reverts commit
8862fac4a0b97d827d2808146d2d79b8d799b998.
Change-Id: I4d5629df4580b7ac08a5cb04924c56eecad3ad25
David Srbecky [Fri, 11 Mar 2016 18:16:35 +0000 (18:16 +0000)]
Merge "DWARF: Compensate for compiler's off-by-one instruction."
David Srbecky [Fri, 11 Mar 2016 17:41:52 +0000 (17:41 +0000)]
Merge "Symbolize native debug information for AOTed methods in boot.oat."
David Srbecky [Mon, 7 Mar 2016 20:47:29 +0000 (20:47 +0000)]
Symbolize native debug information for AOTed methods in boot.oat.
The oatdump symbolizer will produce all DWARF information now,
not just the symtab. This allows us to recover almost all native
debug information after the fact, even for non-debuggable oat files.
The main drawback is that non-debuggable oat files do not have
enough stackmaps to provide reasonable line stepping experience.
We might also be missing inlined methods from backtraces.
Change-Id: I73ff6f43a3a0e5a0907af353f74f04e9b0129590
David Srbecky [Fri, 11 Mar 2016 17:35:48 +0000 (17:35 +0000)]
Merge "Add trampoline symbols to native debug info."
David Srbecky [Fri, 11 Mar 2016 17:11:44 +0000 (17:11 +0000)]
Add trampoline symbols to native debug info.
The debugger needs them to unwind through the trampolines and to
understand what is happening in the call stack.
Change-Id: Ia554058c3796788adcd7336d620a7734eb366905
David Srbecky [Fri, 11 Mar 2016 17:33:22 +0000 (17:33 +0000)]
Merge "Minor optimizations of debug::WriteCFISection."
David Srbecky [Fri, 11 Mar 2016 16:54:22 +0000 (16:54 +0000)]
Minor optimizations of debug::WriteCFISection.
The method might be passed method infos without CFI in some cases.
Use the sorting phase as a chance to filter them out.
This makes sure we do not allocate memory and sort methods if there
is in fact no work to do.
Also change the sort to stable - for the sake of determinism.
Change-Id: I97d57d77e8b709d0d49d6971f66b955efcbb57b0
Roland Levillain [Fri, 11 Mar 2016 16:27:27 +0000 (16:27 +0000)]
Typo in comment in art::mirror::Object::IdentityHashCode.
Change-Id: I5380d1c34deda32c668e48ed5f80e96b29f5b0b5
Roland Levillain [Fri, 11 Mar 2016 16:47:02 +0000 (16:47 +0000)]
Merge "MIPS: Don't use $t8 for calling entrypoints"
Nicolas Geoffray [Fri, 11 Mar 2016 16:37:23 +0000 (16:37 +0000)]
Merge "Revert "Revert "Revert "Revert "Change condition to opposite if lhs is constant"""""
David Srbecky [Fri, 11 Mar 2016 14:25:00 +0000 (14:25 +0000)]
DWARF: Compensate for compiler's off-by-one instruction.
The compiler generates stackmaps with PC value which is *after*
the instruction rather PC of the instruction itself. This causes
trouble when generating native line number mapping since the
branch instruction itself will not be mapped, and consequently
we get incorrect line number in backtraces. Add code to compensate
for this. See the in-code comment for more details.
Change-Id: I72c992e6d08a767f314290d562421b251ae60732
Goran Jakovljevic [Fri, 11 Mar 2016 14:22:18 +0000 (15:22 +0100)]
MIPS: Don't use $t8 for calling entrypoints
When jumping to entrypoints, $t9 register must be used. Value of $gp
is calculated based on value from $t9 and it should contain the
address of the first instruction of the entrypoint.
This CL enables MIPS32 and MIPS64 to boot.
Change-Id: Idf0f7f479cceafabb88b1b087da190dbda86e1c3
Nicolas Geoffray [Fri, 11 Mar 2016 16:15:32 +0000 (16:15 +0000)]
Merge "Fix invariant in reference type propagation."
David Srbecky [Thu, 10 Mar 2016 15:32:37 +0000 (15:32 +0000)]
Add symbolizer option to generate debug information only.
There is no need to copy the .rodata and .text sections since the
consumers can get their data form the original ELF file. We mark
those section as SHT_NOBITS instead, which is the standard way to
create ELF file with separate debug information only.
The new option is called --only-keep-debug and it essentially has
the same effect as --only-keep-debug option for "strip" command.
Change-Id: I4f7488c04bbad86288beeda11c3bf295890e740d
David Srbecky [Fri, 11 Mar 2016 14:35:33 +0000 (14:35 +0000)]
Merge "Revert "Allow duplicated methods in different DWARF line tables.""
David Srbecky [Fri, 11 Mar 2016 14:34:47 +0000 (14:34 +0000)]
Revert "Allow duplicated methods in different DWARF line tables."
This reverts commit
3c5462364daee17e651299f1c370f965f34baed8.
Change-Id: Icc63c47cd53abeff2a470ae84715012a1ef988ac
David Srbecky [Fri, 11 Mar 2016 14:28:00 +0000 (14:28 +0000)]
Merge "Allow duplicated methods in different DWARF line tables."
David Srbecky [Fri, 11 Mar 2016 14:25:56 +0000 (14:25 +0000)]
Merge "Allow generation of native debug info for multiple methods."
David Srbecky [Fri, 11 Mar 2016 14:25:40 +0000 (14:25 +0000)]
Merge "Refactor MethodDebugInfo (input of DWARF writer)."
David Srbecky [Fri, 11 Mar 2016 11:49:20 +0000 (11:49 +0000)]
Allow duplicated methods in different DWARF line tables.
This makes the compilation units more self-contained.
If method is mentioned in the compilation unit, we can
also find its line table there. Otherwise, we would have
to search through all of them.
Change-Id: I0cdfb9006e796e41e123fc1f4fecd15312570068
Nicolas Geoffray [Fri, 11 Mar 2016 13:35:51 +0000 (13:35 +0000)]
Fix invariant in reference type propagation.
Also some cleanups.
Change-Id: I7f0ec7d06b4bab10dbfa230c757447d311658f93
Artem Serov [Tue, 9 Feb 2016 17:15:29 +0000 (17:15 +0000)]
Integrate BitwiseNegated into shared framework.
Share implementation between arm and arm64.
Change-Id: I0dd12e772cb23b4c181fd0b1e2a447470b1d8702
Anton Shamin [Mon, 15 Feb 2016 06:48:36 +0000 (12:48 +0600)]
Revert "Revert "Revert "Revert "Change condition to opposite if lhs is constant""""
This reverts commit
d4aee949b3dd976295201b5310f13aa2df40afa1.
Change-Id: I505b8c9863c310a3a708f580b00d425b750c9541
Nicolas Geoffray [Fri, 11 Mar 2016 11:33:13 +0000 (11:33 +0000)]
Merge "Fix thread race when fetching the ProfilingInfo object."
Nicolas Geoffray [Fri, 11 Mar 2016 09:57:57 +0000 (09:57 +0000)]
Fix thread race when fetching the ProfilingInfo object.
Problem is:
1) Compiler fetches the ProfilingInfo of A, it's null.
2) Mutator creates the ProfilingInfo.
3) Compiler notifies it's not using A anymore, calls
ProfilingInfo::DecrementInlineUse -> Crash as we expected
ProfilingInfo::IncrementUse to be called before.
Also update some namings to better reflect what is going on.
Change-Id: I55ea4c5d81988131467095e18a0d13a8be9d0ef7
Andreas Gampe [Tue, 1 Mar 2016 22:38:37 +0000 (14:38 -0800)]
ART: Allow unwinding unattached threads
Partial revert of commit
ed8b723c5f3989d2593ec21c65c96d6d8bf25579.
Make it (constexpr) configurable whether we allow unwinding native
stacks of unattached threads.
Bug:
27449879
Change-Id: I307a80c9a0166f33fa8a41e492d7eb0d77d0e44f
Christopher Ferris [Thu, 10 Mar 2016 22:12:34 +0000 (22:12 +0000)]
Merge "Log when why an Unwind failed."
Mathieu Chartier [Thu, 10 Mar 2016 20:07:55 +0000 (20:07 +0000)]
Merge "Allocate interrupted exception before re-acquiring lock"
Andreas Gampe [Thu, 10 Mar 2016 20:06:27 +0000 (20:06 +0000)]
Merge "ART: Release all resource on MonitorPool destruction"
Andreas Gampe [Thu, 10 Mar 2016 16:33:45 +0000 (08:33 -0800)]
ART: Release all resource on MonitorPool destruction
To be valgrind-clean, we need to release the current metadata and
all chunks on destruction.
Bug:
27156726
Change-Id: Ia51cea139a6e9669975b6ac045f5223cd68f1f6a
Mathieu Chartier [Thu, 10 Mar 2016 18:49:35 +0000 (10:49 -0800)]
Allocate interrupted exception before re-acquiring lock
Allocating the monitor exception after acquiring the lock can cause
a deadlock in the following scenario:
Reference queue daemon interrupted from wait() on
ReferenceQueue.class. Tries to allocate exception and starts a GC
that blocks on WaitForGcToComplete.
Some other thread is already doing a GC, and tries to enqueue the
cleared refereneces. This deadlocks trying to lock Reference.class
in ReferenceQueue.add.
Bug:
27508829
Change-Id: Icbc18b6b8d1e906c3f7413810d6cefdda06eb921
Christopher Ferris [Thu, 10 Mar 2016 00:03:00 +0000 (16:03 -0800)]
Log when why an Unwind failed.
Bug:
27449879
Change-Id: If7ae5f0991da6fd64dd45e22d69eecef1388a28e
Aart Bik [Thu, 10 Mar 2016 18:59:10 +0000 (10:59 -0800)]
Tests for round() method.
Rationale:
While working on some ideas on implementing round(),
I wrote these tests. We might as well use them to
ensure any future work does not break (test also
provides an easy place to add future "difficult" values).
bug=
26327751
Change-Id: I911a3e1e5fe45147bc1b7214225c9aec4e979965
Vladimir Marko [Thu, 10 Mar 2016 18:48:37 +0000 (18:48 +0000)]
Merge "ART: Do not use vixld - workaround to fix dex2oatds."
Aart Bik [Thu, 10 Mar 2016 18:39:18 +0000 (18:39 +0000)]
Merge "Avoid generating dead code on frame enter/exit. This includes stack operations and, on x86, call/pop to read PC."
Vladimir Marko [Thu, 10 Mar 2016 17:25:50 +0000 (17:25 +0000)]
ART: Do not use vixld - workaround to fix dex2oatds.
This is a quick workaround for ODR violations caused
by linking libvixl.a compiled without VIXL_DEBUG with
the libartd-compiler.a compiled with VIXL_DEBUG.
Bug:
27588884
Change-Id: Ib1af165f177f125f03cdd99777dff4c2912f6405
Aart Bik [Thu, 3 Mar 2016 00:06:13 +0000 (16:06 -0800)]
Avoid generating dead code on frame enter/exit.
This includes stack operations and, on x86, call/pop to read PC.
bug=
26997690
Rationale:
(1) If method is fully intrinsified, and makes no calls in slow
path or uses special input, no need to require current method.
(2) Invoke instructions with HasPcRelativeDexCache() generate code
that reads the PC (call/pop) on x86. However, if the invoke is
an intrinsic that is later replaced with actual code, this PC
reading code may be dead.
Example X86 (before/after):
0x0000108c: 83EC0C sub esp, 12
0x0000108f: 890424 mov [esp], eax <-- not needed
0x00001092:
E800000000 call +0 (0x00001097)
0x00001097: 58 pop eax <-- dead code to read PC
0x00001098:
F30FB8C1 popcnt eax, ecx
0x0000109c:
F30FB8DA popcnt ebx, edx
0x000010a0: 03D8 add ebx, eax
0x000010a2: 89D8 mov eax, ebx
0x000010a4: 83C40C add esp, 12 <-- not needed
0x000010a7: C3 ret
0x0000103c:
F30FB8C1 popcnt eax, ecx
0x00001040:
F30FB8DA popcnt ebx, edx
0x00001044: 03D8 add ebx, eax
0x00001046: 89D8 mov eax, ebx
0x00001048: C3 ret
Example ARM64 (before/after):
0x0000103c:
f81e0fe0 str x0, [sp, #-32]!
0x00001040:
f9000ffe str lr, [sp, #24]
0x00001044:
dac01020 clz x0, x1
0x00001048:
f9400ffe ldr lr, [sp, #24]
0x0000104c:
910083ff add sp, sp, #0x20 (32)
0x00001050:
d65f03c0 ret
0x0000103c:
dac01020 clz x0, x1
0x00001040:
d65f03c0 ret
Change-Id: I8377db80c9a901a08fff4624927cf4a6e585da0c
Vladimir Marko [Thu, 10 Mar 2016 17:11:44 +0000 (17:11 +0000)]
Merge "Add Mips abiflags section to the oatfile."
Roland Levillain [Thu, 10 Mar 2016 17:09:37 +0000 (17:09 +0000)]
Merge "MIPS32: java.lang.Thread, and java.lang.String intrinsics:"
Roland Levillain [Thu, 10 Mar 2016 15:28:34 +0000 (15:28 +0000)]
Merge "Disable some image_test cases with concurrent collector."
Nicolas Geoffray [Thu, 10 Mar 2016 15:25:02 +0000 (15:25 +0000)]
Merge "Tweak JIT hotness threshold."
Roland Levillain [Thu, 10 Mar 2016 15:16:31 +0000 (15:16 +0000)]
Disable some image_test cases with concurrent collector.
The following image_test cases sometimes fail on ART Builbot's
x86 concurrent collector configuration:
- ImageTest.WriteReadUncompressed
- ImageTest.WriteReadLZ4
- ImageTest.WriteReadLZ4HC
Disable them to make the build turn green again, while we
investigate the failures.
Bug:
27578460
Change-Id: I46126e4690e6300e3bfa771c1b6a560272ecb1da
Nicolas Geoffray [Tue, 8 Mar 2016 16:50:21 +0000 (16:50 +0000)]
Tweak JIT hotness threshold.
bug:
23128949
bug:
27398183
Change-Id: Iffde6ba064e54546827cb8fc2a670baedf2e2409
Nicolas Geoffray [Thu, 10 Mar 2016 14:23:35 +0000 (14:23 +0000)]
Merge "MIPS32: Fix test 082-inline-execute"
Roland Levillain [Thu, 10 Mar 2016 14:21:40 +0000 (14:21 +0000)]
Merge "Fix and enable java.lang.StringFactory intrinsics."
Goran Jakovljevic [Tue, 8 Mar 2016 14:26:51 +0000 (15:26 +0100)]
MIPS32: Fix test 082-inline-execute
This is fixing the test when run with --debuggable flag.
Change-Id: Ifd11c28a77f1c6f6fe5951c5c27309187b29d3ee
Nicolas Geoffray [Thu, 10 Mar 2016 10:47:57 +0000 (10:47 +0000)]
Merge "Clear inline caches at each full GC."
David Srbecky [Wed, 9 Mar 2016 11:44:44 +0000 (11:44 +0000)]
Allow generation of native debug info for multiple methods.
Slight tweak in the API which I will need in the near future.
Change-Id: I45954ae16710bc941a9a06a3a17c70798315ca53
David Srbecky [Mon, 7 Mar 2016 17:33:57 +0000 (17:33 +0000)]
Refactor MethodDebugInfo (input of DWARF writer).
Do not pass CompiledMethod pointer through since it is only available
during AOT compile but not during JIT compile or at runtime. Creating
mock CompiledMethod just pass data is proving increasingly tricky, so
copy the fields that we need to MethodDebugInfo instead.
Change-Id: I820297b41e769fcac488c0ff2d2ea0492bb13ed8
Andreas Gampe [Thu, 10 Mar 2016 01:44:24 +0000 (01:44 +0000)]
Merge "ART: Fix assumption in class profile collection"
Mathieu Chartier [Thu, 10 Mar 2016 00:51:40 +0000 (00:51 +0000)]
Merge "Use PwriteFully and write image header last"
Mathieu Chartier [Wed, 9 Mar 2016 19:14:50 +0000 (11:14 -0800)]
Use PwriteFully and write image header last
Prevent corrupted images if dex2oat gets killed or if the image
writer is interrupted during writing.
Bug:
22858531
Bug:
27561308
(cherry picked from commit
dba5a70977da0a28cec2bfc8261d52a177738477)
Change-Id: If4f2c43bcc3cf918b5d2780f1709225b5a4ce116
Mathieu Chartier [Thu, 10 Mar 2016 00:34:23 +0000 (00:34 +0000)]
Merge "Use per character suffix for multi image"
Andreas Gampe [Wed, 9 Mar 2016 23:52:21 +0000 (15:52 -0800)]
ART: Fix assumption in class profile collection
The dex cache may contain erroneous classes. Filter, instead of
DCHECK.
Bug:
27500691
(cherry picked from commit
4309f6129795ade42591b872e44a33cec95f4e79)
Change-Id: Ic99bca3a7e54a5e7893c801ec5ac92cbf690cd67
Mathieu Chartier [Wed, 9 Mar 2016 00:50:20 +0000 (16:50 -0800)]
Use per character suffix for multi image
E.g. if you pull the oat files from the device, you get oat files
with the following names:
system@framework@boot.art
But the name stored in the image header is:
boot.art
So we need to append system@framework@ as a file name prefix to each
image file to get the actual one.
Required for adding oatdump support for app images.
Bug:
22858531
Bug:
27408512
(cherry picked from commit
05752398db7f25d7892b62cb39615eee468f1db8)
Change-Id: Ia6fb34b137c5e285818d39c3b9794cd4ce6c3219
Andreas Gampe [Wed, 9 Mar 2016 17:30:50 +0000 (17:30 +0000)]
Merge "ART: Enable native stack dumping on timeout"
Andreas Gampe [Wed, 9 Mar 2016 17:30:18 +0000 (17:30 +0000)]
Merge "ART: Use optimizing compiler in image_test"
Nicolas Geoffray [Mon, 7 Mar 2016 14:29:04 +0000 (14:29 +0000)]
Clear inline caches at each full GC.
This fixes occasional failures of 141-class-unload.
Also fix a bug where clearing inline caches also cleared the dex
pc associated with it.
bug:
26846185
bug:
23128949
Change-Id: I77bf1dee229d7764c3cc21440829c7fba7b37001
Calin Juravle [Wed, 9 Mar 2016 17:17:03 +0000 (17:17 +0000)]
Enable compilation of secondary dex files
DexClassLoader should be allowed to compile the secondary dex files if
needed. Adding interpret-only flag when using JIT prevented that.
Note that primary apks are loaded with a PathClassLoader and which
should not attempt to compile the dex files.
Bug:
26719109
Change-Id: I29323bc48160c3c70fea4244e502e43a4b0a678c
Andreas Gampe [Fri, 19 Feb 2016 00:53:41 +0000 (16:53 -0800)]
ART: Use optimizing compiler in image_test
Expose some compiler options so we can run the image test with
the optimizing compiler without running into a stack overflow.
Also allow a variable amount of threads in CreateCompilerDriver.
Use 16 as a middle ground on the host to speed up the otherwise
now slowed-down test.
Bug:
27240085
Bug:
27552475
Change-Id: I8db5055d32ae722c8f430903244faa9166cc4886
Aart Bik [Wed, 9 Mar 2016 17:13:34 +0000 (17:13 +0000)]
Merge "Implement the 1.8 unsafe memory fences directly in HIR."
Andreas Gampe [Wed, 9 Mar 2016 16:23:41 +0000 (16:23 +0000)]
Merge "ART: Fix missing include"
Vladimir Marko [Wed, 9 Mar 2016 16:16:49 +0000 (16:16 +0000)]
Merge "ART: Write bit fields together in ComputeSpecialAccessorInfo()."
Roland Levillain [Wed, 9 Mar 2016 16:14:00 +0000 (16:14 +0000)]
Fix and enable java.lang.StringFactory intrinsics.
The following intrinsics were not considered by the
intrinsics recognizer:
- StringNewStringFromBytes
- StringNewStringFromChars
- StringNewStringFromString
This CL enables them and add tests for them.
This CL also:
- Fixes the locations of the ARM64 & MIPS64
StringNewStringFromString intrinsics.
- Fixes the definitions of the FOUR_ARG_DOWNCALL macros on
ARM and x86, which are used to implement the
art_quick_alloc_string_from_bytes* runtime entry points.
- Fixes PC info (stack maps) recording in the
StringNewStringFromBytes, StringNewStringFromChars and
StringNewStringFromString ARM, ARM64 & MIPS64 intrinsics.
Bug:
27425743
Change-Id: I38c00d3f0b2e6b64f7d3fe9146743493bef9e45c
Nicolas Geoffray [Wed, 9 Mar 2016 15:46:58 +0000 (15:46 +0000)]
Merge "Stop oat loading spam on host."
Nicolas Geoffray [Wed, 9 Mar 2016 15:36:23 +0000 (15:36 +0000)]
Stop oat loading spam on host.
Change-Id: I336669f49bfbfc76bb5e9d3b4f1738709e070b27
David Srbecky [Wed, 9 Mar 2016 15:05:50 +0000 (15:05 +0000)]
Merge "Mark DWARF lines in non-debuggable methods as "not a statement"."
Vladimir Marko [Wed, 9 Mar 2016 13:45:39 +0000 (13:45 +0000)]
ART: Write bit fields together in ComputeSpecialAccessorInfo().
Avoid function calls between storing individual bit fields
to allow the compiler (gcc/clang) to merge those writes
together. Valgrind then marks the memory as "defined" while
individual bit field writes would leave it "undefined" and
later trigger the valgrind error:
Conditional jump or move depends on uninitialised value(s)
on DCHECK()s using the bit fields.
Bug:
27552451
Change-Id: If6de5cbe231f99da0f974a0fc9a36c14e3dc071e