OSDN Git Service

android-x86/art.git
9 years agoam 3e3364e2: Merge "Tidy logging code not using UNIMPLEMENTED."
Ian Rogers [Tue, 28 Oct 2014 04:51:55 +0000 (04:51 +0000)]
am 3e3364e2: Merge "Tidy logging code not using UNIMPLEMENTED."

* commit '3e3364e267117462894e0c3b7d9b413b25d28057':
  Tidy logging code not using UNIMPLEMENTED.

9 years agoam c57a66d5: Merge "ARM: Use hardfp calling convention between java to java call."
Ian Rogers [Tue, 28 Oct 2014 04:51:53 +0000 (04:51 +0000)]
am c57a66d5: Merge "ARM: Use hardfp calling convention between java to java call."

* commit 'c57a66d5aadfd682c63e34ec1b1f304260f8c333':
  ARM: Use hardfp calling convention between java to java call.

9 years agoam b62ff579: Merge "ART: Add div/rem zero check elimination flag"
Ian Rogers [Tue, 28 Oct 2014 04:51:50 +0000 (04:51 +0000)]
am b62ff579: Merge "ART: Add div/rem zero check elimination flag"

* commit 'b62ff579cd870b0bf213765b07d7b404d15ece7b':
  ART: Add div/rem zero check elimination flag

9 years agoam c6c9e5d4: Merge "ART: Prevent float inference back to SSA in-regs."
Ian Rogers [Tue, 28 Oct 2014 04:51:49 +0000 (04:51 +0000)]
am c6c9e5d4: Merge "ART: Prevent float inference back to SSA in-regs."

* commit 'c6c9e5d4511f2b23cab09f9d6f535f96fe33cb9b':
  ART: Prevent float inference back to SSA in-regs.

9 years agoam 98c271d5: Merge "Code cleanup."
Ian Rogers [Tue, 28 Oct 2014 04:51:48 +0000 (04:51 +0000)]
am 98c271d5: Merge "Code cleanup."

* commit '98c271d517bc4d25fc6879b4b8e35ea93885d9e2':
  Code cleanup.

9 years agoam 661cb3f6: Merge "Add more environment variable to control test granularity."
Nicolas Geoffray [Tue, 28 Oct 2014 04:51:26 +0000 (04:51 +0000)]
am 661cb3f6: Merge "Add more environment variable to control test granularity."

* commit '661cb3f65d3d6fd09f4823f4519890646260fbe6':
  Add more environment variable to control test granularity.

9 years agoam dc9667e5: Merge "Fix encoding of imul in x86_64 assembler."
Nicolas Geoffray [Tue, 28 Oct 2014 04:51:24 +0000 (04:51 +0000)]
am dc9667e5: Merge "Fix encoding of imul in x86_64 assembler."

* commit 'dc9667e50f6adf99f2d4cfea27b4abcbf42e6a3f':
  Fix encoding of imul in x86_64 assembler.

9 years agoam 54efee55: Merge "Make ObjectRegistry::InternalAdd GC safe"
Sebastien Hertz [Tue, 28 Oct 2014 04:51:23 +0000 (04:51 +0000)]
am 54efee55: Merge "Make ObjectRegistry::InternalAdd GC safe"

* commit '54efee55ccfc9eb411b9fde6742fe457016a391d':
  Make ObjectRegistry::InternalAdd GC safe

9 years agoam 64727aee: Merge "Revert "Revert "Implement long negate instruction in the optimizi...
Roland Levillain [Tue, 28 Oct 2014 04:51:17 +0000 (04:51 +0000)]
am 64727aee: Merge "Revert "Revert "Implement long negate instruction in the optimizing compiler."""

* commit '64727aeef8f4243f2058a19a43a937248a60dae1':
  Revert "Revert "Implement long negate instruction in the optimizing compiler.""

9 years agoam 8bde036e: Merge "Use test\'s `-d\' option when checking for a directory."
Roland Levillain [Tue, 28 Oct 2014 04:51:16 +0000 (04:51 +0000)]
am 8bde036e: Merge "Use test\'s `-d\' option when checking for a directory."

* commit '8bde036ebd74ce94477e65077bed6dea9c29616b':
  Use test's `-d' option when checking for a directory.

9 years agoam 32da24bb: Merge "Fix wrong unsigned to signed conversions."
Nicolas Geoffray [Tue, 28 Oct 2014 04:51:14 +0000 (04:51 +0000)]
am 32da24bb: Merge "Fix wrong unsigned to signed conversions."

* commit '32da24bb26885a9877adcef5b53f283d44b622d2':
  Fix wrong unsigned to signed conversions.

9 years agoam 02567379: Merge "Fix JDWP Virtualmachine.Resume command"
Sebastien Hertz [Tue, 28 Oct 2014 04:50:40 +0000 (04:50 +0000)]
am 02567379: Merge "Fix JDWP Virtualmachine.Resume command"

* commit '02567379e7099682345fd59d45a23e17711c9ddd':
  Fix JDWP Virtualmachine.Resume command

9 years agoam a1e50cab: Merge "More use of WellKnownClasses for debugger"
Sebastien Hertz [Tue, 28 Oct 2014 04:50:39 +0000 (04:50 +0000)]
am a1e50cab: Merge "More use of WellKnownClasses for debugger"

* commit 'a1e50cab308af9f9a94e8bd370111eeff5634914':
  More use of WellKnownClasses for debugger

9 years agoam 6b8fbcb1: Merge "ART: Remove old LOG(INFO)"
Andreas Gampe [Tue, 28 Oct 2014 04:49:10 +0000 (04:49 +0000)]
am 6b8fbcb1: Merge "ART: Remove old LOG(INFO)"

* commit '6b8fbcb19f404b45ed01e9808f6d3254b13fc351':
  ART: Remove old LOG(INFO)

9 years agoam 754ff339: Merge "Use the passed-in \'name\' param to create a MemMap."
Mathieu Chartier [Tue, 28 Oct 2014 04:49:09 +0000 (04:49 +0000)]
am 754ff339: Merge "Use the passed-in \'name\' param to create a MemMap."

* commit '754ff339c776d69ac9caf7a783b3ff280d6bd08b':
  Use the passed-in 'name' param to create a MemMap.

9 years agoam d308804f: Merge "Make out-of-line mutex contention dependent on ART_USE_FUTEXES."
Ian Rogers [Tue, 28 Oct 2014 04:49:08 +0000 (04:49 +0000)]
am d308804f: Merge "Make out-of-line mutex contention dependent on ART_USE_FUTEXES."

* commit 'd308804f1af5d18324215b04d4c108759e527c4c':
  Make out-of-line mutex contention dependent on ART_USE_FUTEXES.

9 years agoam e8bd8ac7: Merge "ART: Use static_assert in down_cast"
Andreas Gampe [Tue, 28 Oct 2014 04:31:17 +0000 (04:31 +0000)]
am e8bd8ac7: Merge "ART: Use static_assert in down_cast"

* commit 'e8bd8ac7fb41545ef342c8120c2e83d2fef20a50':
  ART: Use static_assert in down_cast

9 years agoam 13fe5bbd: Merge "Add missing SHARED_LOCKS_REQUIRED attibute."
Chih-Hung Hsieh [Tue, 28 Oct 2014 04:31:16 +0000 (04:31 +0000)]
am 13fe5bbd: Merge "Add missing SHARED_LOCKS_REQUIRED attibute."

* commit '13fe5bbd636e4dd9b6d51e13932e6e55e354976d':
  Add missing SHARED_LOCKS_REQUIRED attibute.

9 years agoam b4e1a4d3: Merge "ART: Add basic tests for materialized conditions."
Nicolas Geoffray [Tue, 28 Oct 2014 04:31:15 +0000 (04:31 +0000)]
am b4e1a4d3: Merge "ART: Add basic tests for materialized conditions."

* commit 'b4e1a4d3156528d3e2b863f08c24793310a5b6bc':
  ART: Add basic tests for materialized conditions.

9 years agoam be8c0217: Merge "Fix concurrent start bytes race"
Mathieu Chartier [Tue, 28 Oct 2014 04:31:13 +0000 (04:31 +0000)]
am be8c0217: Merge "Fix concurrent start bytes race"

* commit 'be8c0217c858dcabc13cefecdeced8610e3e9d6f':
  Fix concurrent start bytes race

9 years agoam 55adc608: Merge "Revert "Implement long negate instruction in the optimizing compi...
Roland Levillain [Tue, 28 Oct 2014 04:31:12 +0000 (04:31 +0000)]
am 55adc608: Merge "Revert "Implement long negate instruction in the optimizing compiler.""

* commit '55adc608b04ddf4d126bfe92eaa444068ae78e8a':
  Revert "Implement long negate instruction in the optimizing compiler."

9 years agoam 46fdec13: Merge "Implement long negate instruction in the optimizing compiler."
Roland Levillain [Tue, 28 Oct 2014 04:31:03 +0000 (04:31 +0000)]
am 46fdec13: Merge "Implement long negate instruction in the optimizing compiler."

* commit '46fdec13b6dcaf932aa9fb1338f32df01aa0d959':
  Implement long negate instruction in the optimizing compiler.

9 years agoam e2b2cbf8: Merge "[optimizing compiler] Handle SUB_FLOAT/DOUBLE nodes"
Calin Juravle [Tue, 28 Oct 2014 04:31:02 +0000 (04:31 +0000)]
am e2b2cbf8: Merge "[optimizing compiler] Handle SUB_FLOAT/DOUBLE nodes"

* commit 'e2b2cbf8bffdf9ee3ece487fde9ac78652b4abaf':
  [optimizing compiler] Handle SUB_FLOAT/DOUBLE nodes

9 years agoam b1bcf57e: Merge "Use the shared smali test infrastructure for 800-smali."
Nicolas Geoffray [Tue, 28 Oct 2014 04:30:52 +0000 (04:30 +0000)]
am b1bcf57e: Merge "Use the shared smali test infrastructure for 800-smali."

* commit 'b1bcf57e9fcaa95447fc3b9dfd7d0068e6106cca':
  Use the shared smali test infrastructure for 800-smali.

9 years agoam 46de37a6: Merge "[optimizing compiler] Add float/double subtraction"
Calin Juravle [Tue, 28 Oct 2014 04:30:51 +0000 (04:30 +0000)]
am 46de37a6: Merge "[optimizing compiler] Add float/double subtraction"

* commit '46de37a6a314368cc5fc52743095e48a864ed9b5':
  [optimizing compiler] Add float/double subtraction

9 years agoam e833b110: Merge "ART: optimizing compiler: initial support for ARM64."
Nicolas Geoffray [Mon, 27 Oct 2014 09:41:55 +0000 (09:41 +0000)]
am e833b110: Merge "ART: optimizing compiler: initial support for ARM64."

* commit 'e833b110876fa8fe34a08f06d2521c74488180bb':
  ART: optimizing compiler: initial support for ARM64.

9 years agoam 37a71888: Merge "Remove obsolete TODOs from codegen tests"
Calin Juravle [Mon, 27 Oct 2014 09:41:54 +0000 (09:41 +0000)]
am 37a71888: Merge "Remove obsolete TODOs from codegen tests"

* commit '37a7188810e865a1ee0a7bdc2d01d62c1f1ea49e':
  Remove obsolete TODOs from codegen tests

9 years agoam 1e4dc259: Merge "[native bridge] Make sure we always unload the native bridge"
Calin Juravle [Mon, 27 Oct 2014 09:41:52 +0000 (09:41 +0000)]
am 1e4dc259: Merge "[native bridge] Make sure we always unload the native bridge"

* commit '1e4dc259b4242c1a03415b5b5f4aed7a23e53f79':
  [native bridge] Make sure we always unload the native bridge

9 years agoMerge "Tidy logging code not using UNIMPLEMENTED."
Ian Rogers [Sat, 25 Oct 2014 00:14:26 +0000 (00:14 +0000)]
Merge "Tidy logging code not using UNIMPLEMENTED."

9 years agoMerge "ARM: Use hardfp calling convention between java to java call."
Ian Rogers [Fri, 24 Oct 2014 23:28:08 +0000 (23:28 +0000)]
Merge "ARM: Use hardfp calling convention between java to java call."

9 years agoARM: Use hardfp calling convention between java to java call.
Zheng Xu [Thu, 23 Oct 2014 10:29:55 +0000 (18:29 +0800)]
ARM: Use hardfp calling convention between java to java call.

This patch default to use hardfp calling convention. Softfp can be enabled
by setting kArm32QuickCodeUseSoftFloat to true.

We get about -1 ~ +5% performance improvement with different benchmark
tests. Hopefully, we should be able to get more performance by address the left
TODOs, as some part of the code takes the original assumption which is not
optimal.

DONE:
1. Interpreter to quick code
2. Quick code to interpreter
3. Transition assembly and callee-saves
4. Trampoline(generic jni, resolution, invoke with access check and etc.)
5. Pass fp arg reg following aapcs(gpr and stack do not follow aapcs)
6. Quick helper assembly routines to handle ABI differences
7. Quick code method entry
8. Quick code method invocation
9. JNI compiler

TODO:
10. Rework ArgMap, FlushIn, GenDalvikArgs and affected common code.
11. Rework CallRuntimeHelperXXX().

Change-Id: I9965d8a007f4829f2560b63bcbbde271bdcf6ec2

9 years agoMerge "ART: Add div/rem zero check elimination flag"
Ian Rogers [Fri, 24 Oct 2014 22:26:23 +0000 (22:26 +0000)]
Merge "ART: Add div/rem zero check elimination flag"

9 years agoMerge "ART: Prevent float inference back to SSA in-regs."
Ian Rogers [Fri, 24 Oct 2014 21:57:36 +0000 (21:57 +0000)]
Merge "ART: Prevent float inference back to SSA in-regs."

9 years agoART: Prevent float inference back to SSA in-regs.
Stephen Kyle [Thu, 16 Oct 2014 14:02:42 +0000 (15:02 +0100)]
ART: Prevent float inference back to SSA in-regs.

.method public static getInt(I)I
    .registers 2
    const/4 v0, 0x0
    if-ne v0, v0, :after
    float-to-int v0, v0
    :exit
    add-int/2addr v0, v1
    return v0
    :after
    move v1, v0
    goto :exit
.end method

In this code sample, v1 is the single parameter to this method. In one
of the phi-nodes inserted between :exit and add-int/2addr, v1's two
incoming SSA regs are:
  - the initial def of v1 as a parameter
  - the v1 def'd at move v1, v0.
During type inference, because the 2nd def is a float (because of the
earlier float-to-int v0, v0) this will change the type of the 1st def to a
float as well, which is incorrect since the first parameter is known to be
non-float.

This fix checks during phi-node type-inference if an SSA reg that is the
initial def of a parameter vreg is about to be set as float when it was
not previously, and skips the inference if so.

In this case, when using a hard-float ABI, having the in-reg v1 set as
float causes FlushIns() to read the argument to the method from an FP reg,
when the argument will be passed in a core reg by any caller.

Also included is a smali test for this bug: compare difference between
./run-test --64 800
./run-test --64 --interpreter 800
when the vreg_analysis patch has not been applied.
(Requires 64-bit because 32-bit ARM currently does not use hard-float.)

getInt(I)I should return its argument, but it returns an incorrect
value.

Change-Id: I1d4b5be6a931fe853279e89dd820820f29823da1
Signed-off-by: Stephen Kyle <stephen.kyle@arm.com>
9 years agoTidy logging code not using UNIMPLEMENTED.
Ian Rogers [Fri, 24 Oct 2014 21:20:06 +0000 (14:20 -0700)]
Tidy logging code not using UNIMPLEMENTED.

Change-Id: I7a79c1671a6ff8b2040887133b3e0925ef9a3cfe

9 years agoMerge "Code cleanup."
Ian Rogers [Fri, 24 Oct 2014 20:43:21 +0000 (20:43 +0000)]
Merge "Code cleanup."

9 years agoam 1122c7fc: Merge "Implement int bit-wise not operation in the optimizing compiler."
Roland Levillain [Fri, 24 Oct 2014 20:18:17 +0000 (20:18 +0000)]
am 1122c7fc: Merge "Implement int bit-wise not operation in the optimizing compiler."

* commit '1122c7fc277130814dd1d47b5e8be3ac35b1a5fc':
  Implement int bit-wise not operation in the optimizing compiler.

9 years agoam fd9e9561: Merge "C++11 related clean-up of DISALLOW_.."
Ian Rogers [Fri, 24 Oct 2014 20:18:14 +0000 (20:18 +0000)]
am fd9e9561: Merge "C++11 related clean-up of DISALLOW_.."

* commit 'fd9e95619d94608687af201bb296b807716cc7c3':
  C++11 related clean-up of DISALLOW_..

9 years agoam 79f4ec3f: Merge "Fix arm64 bug introduced by refactoring."
Ian Rogers [Fri, 24 Oct 2014 20:18:13 +0000 (20:18 +0000)]
am 79f4ec3f: Merge "Fix arm64 bug introduced by refactoring."

* commit '79f4ec3fb3bccc4436328eef132c7c6e7db9ef32':
  Fix arm64 bug introduced by refactoring.

9 years agoMerge "Add more environment variable to control test granularity."
Nicolas Geoffray [Fri, 24 Oct 2014 15:53:15 +0000 (15:53 +0000)]
Merge "Add more environment variable to control test granularity."

9 years agoAdd more environment variable to control test granularity.
Nicolas Geoffray [Fri, 24 Oct 2014 12:49:08 +0000 (13:49 +0100)]
Add more environment variable to control test granularity.

Change-Id: I7d96b6841db097c51262aa26dba8f331ca7891ff

9 years agoMerge "Fix encoding of imul in x86_64 assembler."
Nicolas Geoffray [Fri, 24 Oct 2014 15:13:16 +0000 (15:13 +0000)]
Merge "Fix encoding of imul in x86_64 assembler."

9 years agoFix encoding of imul in x86_64 assembler.
Nicolas Geoffray [Fri, 24 Oct 2014 14:43:49 +0000 (15:43 +0100)]
Fix encoding of imul in x86_64 assembler.

Change-Id: I5b97f5698ed8ec9d0759d0e1eba8be29119c16c5

9 years agoMerge "Make ObjectRegistry::InternalAdd GC safe"
Sebastien Hertz [Fri, 24 Oct 2014 14:22:05 +0000 (14:22 +0000)]
Merge "Make ObjectRegistry::InternalAdd GC safe"

9 years agoam aea6888b: Merge "Tidy up logging."
Ian Rogers [Fri, 24 Oct 2014 13:01:50 +0000 (13:01 +0000)]
am aea6888b: Merge "Tidy up logging."

* commit 'aea6888b056be21adf762e066c7f33b8939b8a06':
  Tidy up logging.

9 years agoMerge "Revert "Revert "Implement long negate instruction in the optimizing compiler."""
Roland Levillain [Fri, 24 Oct 2014 10:48:38 +0000 (10:48 +0000)]
Merge "Revert "Revert "Implement long negate instruction in the optimizing compiler."""

9 years agoMerge "Use test's `-d' option when checking for a directory."
Roland Levillain [Fri, 24 Oct 2014 10:45:47 +0000 (10:45 +0000)]
Merge "Use test's `-d' option when checking for a directory."

9 years agoMerge "Fix wrong unsigned to signed conversions."
Nicolas Geoffray [Fri, 24 Oct 2014 10:44:50 +0000 (10:44 +0000)]
Merge "Fix wrong unsigned to signed conversions."

9 years agoRevert "Revert "Implement long negate instruction in the optimizing compiler.""
Roland Levillain [Thu, 23 Oct 2014 17:12:09 +0000 (18:12 +0100)]
Revert "Revert "Implement long negate instruction in the optimizing compiler.""

This reverts commit 30ca3d847fe72cfa33e1b2473100ea2d8bea4517.

Change-Id: I188ca8d460d55d3a9966bcf31e0588575afa77d2

9 years agoMerge "Fix JDWP Virtualmachine.Resume command"
Sebastien Hertz [Fri, 24 Oct 2014 07:49:38 +0000 (07:49 +0000)]
Merge "Fix JDWP Virtualmachine.Resume command"

9 years agoMerge "More use of WellKnownClasses for debugger"
Sebastien Hertz [Fri, 24 Oct 2014 07:49:17 +0000 (07:49 +0000)]
Merge "More use of WellKnownClasses for debugger"

9 years agoMerge "ART: Remove old LOG(INFO)"
Andreas Gampe [Fri, 24 Oct 2014 01:09:58 +0000 (01:09 +0000)]
Merge "ART: Remove old LOG(INFO)"

9 years agoMerge "Use the passed-in 'name' param to create a MemMap."
Mathieu Chartier [Fri, 24 Oct 2014 01:03:32 +0000 (01:03 +0000)]
Merge "Use the passed-in 'name' param to create a MemMap."

9 years agoART: Remove old LOG(INFO)
Andreas Gampe [Fri, 24 Oct 2014 00:59:19 +0000 (17:59 -0700)]
ART: Remove old LOG(INFO)

Clean up output, it's annoying.

Change-Id: Ifc1972dd3e988f36f781dc989d23cdb93e6f828e

9 years agoMerge "Make out-of-line mutex contention dependent on ART_USE_FUTEXES."
Ian Rogers [Fri, 24 Oct 2014 00:55:27 +0000 (00:55 +0000)]
Merge "Make out-of-line mutex contention dependent on ART_USE_FUTEXES."

9 years agoMake out-of-line mutex contention dependent on ART_USE_FUTEXES.
Ian Rogers [Fri, 24 Oct 2014 00:48:20 +0000 (17:48 -0700)]
Make out-of-line mutex contention dependent on ART_USE_FUTEXES.

Fix Mac build. Also fix Linux compilation if ART_USE_FUTEXES is disabled.

Change-Id: I51cb1d70b5548ea6121ff7567b9546bad0894e01

9 years agoam b08f4dcf: Merge "X86 Long Min/Max: Avoid calling SRegToVReg with -1"
Vladimir Marko [Thu, 23 Oct 2014 22:30:12 +0000 (22:30 +0000)]
am b08f4dcf: Merge "X86 Long Min/Max: Avoid calling SRegToVReg with -1"

* commit 'b08f4dcf90215ed49e0b796ab3e609bd605be8ba':
  X86 Long Min/Max: Avoid calling SRegToVReg with -1

9 years agoMerge "ART: Use static_assert in down_cast"
Andreas Gampe [Thu, 23 Oct 2014 18:32:06 +0000 (18:32 +0000)]
Merge "ART: Use static_assert in down_cast"

9 years agoART: Use static_assert in down_cast
Andreas Gampe [Thu, 23 Oct 2014 18:24:08 +0000 (11:24 -0700)]
ART: Use static_assert in down_cast

Use C++11 to write an actual compile-time assert.

Change-Id: I36bd94adbf6c732e103720308e1e6bf11065f474

9 years agoMerge "Add missing SHARED_LOCKS_REQUIRED attibute."
Chih-Hung Hsieh [Thu, 23 Oct 2014 18:21:02 +0000 (18:21 +0000)]
Merge "Add missing SHARED_LOCKS_REQUIRED attibute."

9 years agoAdd missing SHARED_LOCKS_REQUIRED attibute.
Chih-Hung Hsieh [Thu, 23 Oct 2014 18:07:52 +0000 (11:07 -0700)]
Add missing SHARED_LOCKS_REQUIRED attibute.

BUG: 18093707
Change-Id: I66e5c7650c67bb308472ff0da7faed6dff4ead53

9 years agoFix wrong unsigned to signed conversions.
Nicolas Geoffray [Thu, 23 Oct 2014 17:32:13 +0000 (18:32 +0100)]
Fix wrong unsigned to signed conversions.

The HIntConstant node takes an int32_t, so we have to keep things signed.

Change-Id: Ib3fa50e87f99118d320cbb381f619d5be9287530

9 years agoMerge "ART: Add basic tests for materialized conditions."
Nicolas Geoffray [Thu, 23 Oct 2014 17:28:50 +0000 (17:28 +0000)]
Merge "ART: Add basic tests for materialized conditions."

9 years agoMerge "Fix concurrent start bytes race"
Mathieu Chartier [Thu, 23 Oct 2014 17:11:44 +0000 (17:11 +0000)]
Merge "Fix concurrent start bytes race"

9 years agoMerge "Revert "Implement long negate instruction in the optimizing compiler.""
Roland Levillain [Thu, 23 Oct 2014 17:11:21 +0000 (17:11 +0000)]
Merge "Revert "Implement long negate instruction in the optimizing compiler.""

9 years agoRevert "Implement long negate instruction in the optimizing compiler."
Roland Levillain [Thu, 23 Oct 2014 17:07:44 +0000 (18:07 +0100)]
Revert "Implement long negate instruction in the optimizing compiler."

This reverts commit 66ce173a40eff4392e9949ede169ccf3108be2db.

9 years agoFix concurrent start bytes race
Mathieu Chartier [Thu, 23 Oct 2014 00:18:34 +0000 (17:18 -0700)]
Fix concurrent start bytes race

Previously, we set concurrent start bytes to max int when we
requested a concurrent GC, but there was a race if another thread
was doing another GC and had already completed GrowForUtilization
but had not yet finished the GC. This meant that the thread doing
the GC would update the concurrent start bytes properly, but the
allocating thread would re-update it to max int. Then when the
concurrent GC thread woke up, it would call WaitForGcToComplete
and see that there was a collector running and avoid doing the
concurrent GC, leaving the concurrent start bytes set to max int.

This meant that there would be no more concurrent GC until either
the next explicit GC or the next GC for alloc.

The fix is to only set concurrent start bytes to max int inside of
the CollectGarbageInternal code such that there isn't any way for
two threads to race.

Bug: 17942071

(cherry picked from commit 0133ec454d8dd3fa5ffe35649b5704aa18f15a49)

Change-Id: Ibc23dd2f937a2b5fda51ed064634c034dbb6e668

9 years agoART: Add basic tests for materialized conditions.
Alexandre Rames [Wed, 1 Oct 2014 11:55:56 +0000 (12:55 +0100)]
ART: Add basic tests for materialized conditions.

Change-Id: I4acef30cc6a48b5fe07d55db6b9cf0d093b326ee

9 years agoMerge "Implement long negate instruction in the optimizing compiler."
Roland Levillain [Thu, 23 Oct 2014 16:14:56 +0000 (16:14 +0000)]
Merge "Implement long negate instruction in the optimizing compiler."

9 years agoImplement long negate instruction in the optimizing compiler.
Roland Levillain [Thu, 23 Oct 2014 15:38:33 +0000 (16:38 +0100)]
Implement long negate instruction in the optimizing compiler.

- Add support for the neg-long (long integer two's
  complement negate) instruction in the optimizing compiler.
- Add a 64-bit NEG instruction (negq) to the x86-64
  assembler.
- Generate ARM, x86 and x86-64 code for integer HNeg nodes.
- Put neg-related tests into test/415-optimizing-arith-neg.

Change-Id: I1fbe9611e134408a6b8745d1df20ab6ffa5e50f2

9 years agoMerge "[optimizing compiler] Handle SUB_FLOAT/DOUBLE nodes"
Calin Juravle [Thu, 23 Oct 2014 16:04:25 +0000 (16:04 +0000)]
Merge "[optimizing compiler] Handle SUB_FLOAT/DOUBLE nodes"

9 years ago[optimizing compiler] Handle SUB_FLOAT/DOUBLE nodes
Calin Juravle [Thu, 23 Oct 2014 16:01:13 +0000 (17:01 +0100)]
[optimizing compiler] Handle SUB_FLOAT/DOUBLE nodes

Also add 414-optimizing-arith-sub to
TEST_ART_BROKEN_OPTIMIZING_ARM64_RUN_TESTS.

Change-Id: Ia1b7ff7857bc0e488b9b8ed8a36efb1a3a9bad36

9 years agoUse test's `-d' option when checking for a directory.
Roland Levillain [Thu, 23 Oct 2014 15:40:59 +0000 (16:40 +0100)]
Use test's `-d' option when checking for a directory.

Change-Id: I2d46ba01ec13daa7aa0b78e447eccbc4f1527a19

9 years agoMerge "Use the shared smali test infrastructure for 800-smali."
Nicolas Geoffray [Thu, 23 Oct 2014 15:34:12 +0000 (15:34 +0000)]
Merge "Use the shared smali test infrastructure for 800-smali."

9 years agoMerge "[optimizing compiler] Add float/double subtraction"
Calin Juravle [Thu, 23 Oct 2014 15:25:59 +0000 (15:25 +0000)]
Merge "[optimizing compiler] Add float/double subtraction"

9 years ago[optimizing compiler] Add float/double subtraction
Calin Juravle [Thu, 23 Oct 2014 14:38:15 +0000 (15:38 +0100)]
[optimizing compiler] Add float/double subtraction

- for arm, x86, x86_64
- add tests
- a bit of clean up

Change-Id: I3761b0d908aca3e3c5d60da481fafb423ff7c9b9

9 years agoMake ObjectRegistry::InternalAdd GC safe
Sebastien Hertz [Thu, 23 Oct 2014 13:39:33 +0000 (15:39 +0200)]
Make ObjectRegistry::InternalAdd GC safe

Because a call to IdentityHashCode may cause GC, the object pointer
may become invalid (if the object has been moved) on next uses. We
now access the object through a Handle to be GC safe.

Also remove unused methods.

Bug: 18098424
Change-Id: I38fb55c3a6be62c4d98d4c94272a9cfeba327598

9 years agoMerge "ART: optimizing compiler: initial support for ARM64."
Nicolas Geoffray [Thu, 23 Oct 2014 12:48:02 +0000 (12:48 +0000)]
Merge "ART: optimizing compiler: initial support for ARM64."

9 years agoART: optimizing compiler: initial support for ARM64.
Alexandre Rames [Thu, 23 Oct 2014 09:03:10 +0000 (10:03 +0100)]
ART: optimizing compiler: initial support for ARM64.

The ARM64 port uses VIXL for code generation, to which it defers work
like label binding and branch resolving, register type coherency
checking, and immediate values handling.

Change-Id: I0a44508c0c991f472a63e67b3469cdd878fe1a68
Signed-off-by: Serban Constantinescu <serban.constantinescu@arm.com>
Signed-off-by: Alexandre Rames <alexandre.rames@arm.com>
9 years agoMerge "Remove obsolete TODOs from codegen tests"
Calin Juravle [Thu, 23 Oct 2014 11:48:28 +0000 (11:48 +0000)]
Merge "Remove obsolete TODOs from codegen tests"

9 years agoRemove obsolete TODOs from codegen tests
Calin Juravle [Thu, 23 Oct 2014 11:32:11 +0000 (12:32 +0100)]
Remove obsolete TODOs from codegen tests

The features are already exercised by the art test 411-optimizing-arith.

Change-Id: Id008931e0ed8206ced11ecc85a80a7e4aef3e68e

9 years agoMerge "[native bridge] Make sure we always unload the native bridge"
Calin Juravle [Thu, 23 Oct 2014 10:58:45 +0000 (10:58 +0000)]
Merge "[native bridge] Make sure we always unload the native bridge"

9 years ago[native bridge] Make sure we always unload the native bridge
Calin Juravle [Wed, 22 Oct 2014 20:02:23 +0000 (21:02 +0100)]
[native bridge] Make sure we always unload the native bridge

libnativebridge may allocate some resources during loading and
intialization and de-allocate them when unloading. This makes sure that
we don't leak anything.

Bug: 18097480
Change-Id: I901f2d3c2ab1efb2875388f99e8c5c111ce82d5a

9 years agoMore use of WellKnownClasses for debugger
Sebastien Hertz [Mon, 13 Oct 2014 09:27:13 +0000 (11:27 +0200)]
More use of WellKnownClasses for debugger

Uses WellKnownClasses cached fields for ThreadGroupReference-related
JDWP commands.
Adds fields of java.lang.ThreadGroup and java.util.ArrayList used by
the debugger.

Change-Id: I54f7ef7e44cb61bec95657a69177bba1fbb7dc75

9 years agoUse the shared smali test infrastructure for 800-smali.
Nicolas Geoffray [Thu, 23 Oct 2014 09:29:33 +0000 (10:29 +0100)]
Use the shared smali test infrastructure for 800-smali.

No need for a special build file now.

Change-Id: I7f6813553548b48fb5f54640506d1a361d9db3c1

9 years agoMerge "Implement int bit-wise not operation in the optimizing compiler."
Roland Levillain [Thu, 23 Oct 2014 09:17:41 +0000 (09:17 +0000)]
Merge "Implement int bit-wise not operation in the optimizing compiler."

9 years agoImplement int bit-wise not operation in the optimizing compiler.
Roland Levillain [Wed, 22 Oct 2014 17:06:21 +0000 (18:06 +0100)]
Implement int bit-wise not operation in the optimizing compiler.

- Add support for the not-int (integer one's complement
  negate) instruction in the optimizing compiler.
- Extend the HNot control-flow graph node type and make it
  inherit from HUnaryOperation.
- Generate ARM, x86 and x86-64 code for integer HNeg nodes.
- Exercise these additions in the codegen_test gtest, as there
  is not direct way to assess the support of not-int from a
  Java source.  Indeed, compiling a Java expression such as
  `~a' using javac and then dx generates an xor-int/lit8 Dex
  instruction instead of the expected not-int Dex instruction.
  This is probably because the Java bytecode has an `ixor'
  instruction, but there's not instruction directly
  corresponding to a bit-wise not operation.

Change-Id: I223aed75c4dac5785e04d99da0d22e8d699aee2b

9 years agoMerge "C++11 related clean-up of DISALLOW_.."
Ian Rogers [Thu, 23 Oct 2014 06:06:55 +0000 (06:06 +0000)]
Merge "C++11 related clean-up of DISALLOW_.."

9 years agoC++11 related clean-up of DISALLOW_..
Ian Rogers [Thu, 23 Oct 2014 05:06:39 +0000 (22:06 -0700)]
C++11 related clean-up of DISALLOW_..

Move DISALLOW_COPY_AND_ASSIGN to delete functions. By no having declarations
with no definitions this prompts better warning messages so deal with these
by correcting the code.
Add a DISALLOW_ALLOCATION and use for ValueObject and mirror::Object.
Make X86 assembly operand types ValueObjects to fix compilation errors.
Tidy the use of iostream and ostream.
Avoid making cutils a dependency via mutex-inl.h for tests that link against
libart. Push tracing dependencies into appropriate files and mutex.cc.
x86 32-bit host symbols size is increased for libarttest, avoid copying this
in run-test 115 by using symlinks and remove this test's higher than normal
ulimit.
Fix the RunningOnValgrind test in RosAllocSpace to not use GetHeap as it
returns NULL when the heap is under construction by Runtime.

Change-Id: Ia246f7ac0c11f73072b30d70566a196e9b78472b

9 years agoMerge "Fix arm64 bug introduced by refactoring."
Ian Rogers [Thu, 23 Oct 2014 05:03:25 +0000 (05:03 +0000)]
Merge "Fix arm64 bug introduced by refactoring."

9 years agoUse the passed-in 'name' param to create a MemMap.
Kyungmin Lee [Thu, 23 Oct 2014 00:32:05 +0000 (09:32 +0900)]
Use the passed-in 'name' param to create a MemMap.

Change-Id: I15c33b23751f4c32cb40c9514b602845feb47f9f
Signed-off-by: Kyungmin Lee <snailee@gmail.com>
9 years agoMerge "Tidy up logging."
Ian Rogers [Wed, 22 Oct 2014 19:10:23 +0000 (19:10 +0000)]
Merge "Tidy up logging."

9 years agoTidy up logging.
Ian Rogers [Wed, 22 Oct 2014 06:31:19 +0000 (23:31 -0700)]
Tidy up logging.

Move gVerboseMethods to CompilerOptions. Now "--verbose-methods=" option to
dex2oat rather than runtime argument "-verbose-methods:".
Move ToStr and Dumpable out of logging.h, move LogMessageData into logging.cc
except for a forward declaration.
Remove ConstDumpable as Dump methods are all const (and make this so if not
currently true).
Make LogSeverity an enum and improve compile time assertions and type checking.
Remove log_severity.h that's only used in logging.h.
With system headers gone from logging.h, go add to .cc files missing system
header includes.
Also, make operator new in ValueObject private for compile time instantiation
checking.

Change-Id: I3228f614500ccc9b14b49c72b9821c8b0db3d641

9 years agoMerge commit 'ad6f0d28f69352c19cf7a1a954e33027dd3a3529' into HEAD
Bill Yi [Wed, 22 Oct 2014 15:45:57 +0000 (08:45 -0700)]
Merge commit 'ad6f0d28f69352c19cf7a1a954e33027dd3a3529' into HEAD

9 years agoMerge "X86 Long Min/Max: Avoid calling SRegToVReg with -1"
Vladimir Marko [Wed, 22 Oct 2014 15:22:43 +0000 (15:22 +0000)]
Merge "X86 Long Min/Max: Avoid calling SRegToVReg with -1"

9 years agoMerge "Fix make file comment"
Calin Juravle [Wed, 22 Oct 2014 15:00:43 +0000 (15:00 +0000)]
Merge "Fix make file comment"

9 years agoFix make file comment
Calin Juravle [Wed, 22 Oct 2014 14:54:34 +0000 (15:54 +0100)]
Fix make file comment

Change-Id: I2b2261aea00b934411979dbca1e00d5a75030407

9 years agoMerge "Minor fix in codegen tests."
Calin Juravle [Wed, 22 Oct 2014 14:57:06 +0000 (14:57 +0000)]
Merge "Minor fix in codegen tests."

9 years agoMinor fix in codegen tests.
Calin Juravle [Wed, 22 Oct 2014 14:54:12 +0000 (15:54 +0100)]
Minor fix in codegen tests.

Change-Id: I9b843536353d4f820b969895d5f75ee9b679aff0

9 years agoFix arm64 bug introduced by refactoring.
Stephen Kyle [Wed, 22 Oct 2014 14:23:46 +0000 (15:23 +0100)]
Fix arm64 bug introduced by refactoring.

- Commit 1d8cdbc5202378 made changes to the parameters for
artSet64InstanceFromCode, but did not change THREE_ARG_DOWNCALL
to THREE_ARG_REF_DOWNCALL for arm64.

Change-Id: If256b0bd32dbd1c95f61e0afcb35975bb09a8b97
Signed-off-by: Stephen Kyle <stephen.kyle@arm.com>