OSDN Git Service

android-x86/art.git
8 years agoMerge "ART: Fix wide stores in Optimizing"
David Brazdil [Mon, 23 Nov 2015 14:17:53 +0000 (14:17 +0000)]
Merge "ART: Fix wide stores in Optimizing"

8 years agoMerge "Optimizing: Avoid a PrettyMethod() call per compiled method."
Vladimir Marko [Mon, 23 Nov 2015 14:10:50 +0000 (14:10 +0000)]
Merge "Optimizing: Avoid a PrettyMethod() call per compiled method."

8 years agoOptimizing: Avoid a PrettyMethod() call per compiled method.
Vladimir Marko [Fri, 20 Nov 2015 13:02:22 +0000 (13:02 +0000)]
Optimizing: Avoid a PrettyMethod() call per compiled method.

PrettyMethod() is expensive so avoid it unless it's actually
needed.

Change-Id: I06ab29cc18e426b7dfb8c53b163757c88b53c27f

8 years agoART: Fix wide stores in Optimizing
David Brazdil [Thu, 19 Nov 2015 10:29:39 +0000 (10:29 +0000)]
ART: Fix wide stores in Optimizing

SsaBuilder::VisitStoreLocal did not take into account the following:
 (a) when storing a wide value, the high vreg must be invalidated,
 (b) when storing into the high vreg of a wide value, the low vreg
     must be invalidated.

Both situations cause overestimation of liveness but only (b) has
implications on correctness. CodeGenerator::EmitEnvironment will skip
the high vreg, causing deoptimizing and try/catch to load a wrong
value for that vreg.

In order to fix this bug, several changes had to be made to the
SsaBuilder:
 (1) phis need to be initialized with a type which matches its
     inputs' size,
 (2) eagerly created loop header phis may end up being undefined
     because of their corresponding vregs being invalidated inside
     the loop; these are marked dead during input setting,
 (3) the entire SSA-building algorithm should never revive an
     undefined loop header phi.

Bug: 25677992
Bug: https://code.google.com/p/android/issues/detail?id=194022

Change-Id: Id8a852e38c3f5ff1c2e608b1aafd6d5ac8311e32

8 years agoMerge "Do not change to the access check entrypoint when inlined."
Nicolas Geoffray [Mon, 23 Nov 2015 12:11:46 +0000 (12:11 +0000)]
Merge "Do not change to the access check entrypoint when inlined."

8 years agoDo not change to the access check entrypoint when inlined.
Nicolas Geoffray [Mon, 23 Nov 2015 12:04:37 +0000 (12:04 +0000)]
Do not change to the access check entrypoint when inlined.

The allocation entrypoint that deals with access checks does
not work with inlined methods.

Fixes 542-unresolved-access-check in jit mode.

Change-Id: I02290a8b2089fcf06e2216dabf8089920b529765

8 years agoMerge "Optimizing/ARM: Improve long shifts by 1."
Vladimir Marko [Mon, 23 Nov 2015 11:20:35 +0000 (11:20 +0000)]
Merge "Optimizing/ARM: Improve long shifts by 1."

8 years agoMerge "Optimizing/Thumb2: Improve load/store for large offsets."
Vladimir Marko [Mon, 23 Nov 2015 10:15:01 +0000 (10:15 +0000)]
Merge "Optimizing/Thumb2: Improve load/store for large offsets."

8 years agoMerge "Fix lint error."
Nicolas Geoffray [Mon, 23 Nov 2015 09:51:01 +0000 (09:51 +0000)]
Merge "Fix lint error."

8 years agoMerge "ART: Fix uninitialized variable"
David Brazdil [Mon, 23 Nov 2015 09:50:29 +0000 (09:50 +0000)]
Merge "ART: Fix uninitialized variable"

8 years agoFix lint error.
Nicolas Geoffray [Mon, 23 Nov 2015 09:50:18 +0000 (09:50 +0000)]
Fix lint error.

Change-Id: I29632dc7e49f7ec63040455fa40fcf87e9282e5e

8 years agoART: Fix uninitialized variable
David Brazdil [Mon, 23 Nov 2015 09:44:52 +0000 (09:44 +0000)]
ART: Fix uninitialized variable

Change-Id: I906de334b3c3cb1e36eff4944457f4598b7c174f

8 years agoMerge "Simplify boolean condition compared to 0"
David Brazdil [Mon, 23 Nov 2015 09:16:02 +0000 (09:16 +0000)]
Merge "Simplify boolean condition compared to 0"

8 years agoMerge "Revert "Add stats support for existing optimizations""
Nicolas Geoffray [Mon, 23 Nov 2015 08:59:29 +0000 (08:59 +0000)]
Merge "Revert "Add stats support for existing optimizations""

8 years agoRevert "Add stats support for existing optimizations"
Nicolas Geoffray [Mon, 23 Nov 2015 08:59:07 +0000 (08:59 +0000)]
Revert "Add stats support for existing optimizations"

Breaks the build. Please ensure your changes build.

This reverts commit 06241b1b07fb031b7d2cf55f4b78d3444d07cc2d.

Change-Id: I68b18f99a9882719bf6654d3313531a7965b8483

8 years agoMerge "Explicitly add HLoadClass/HClinitCheck for HNewInstance."
Nicolas Geoffray [Mon, 23 Nov 2015 08:51:44 +0000 (08:51 +0000)]
Merge "Explicitly add HLoadClass/HClinitCheck for HNewInstance."

8 years agoMerge "Add stats support for existing optimizations"
Nicolas Geoffray [Mon, 23 Nov 2015 08:51:21 +0000 (08:51 +0000)]
Merge "Add stats support for existing optimizations"

8 years agoAdd stats support for existing optimizations
Jean-Philippe Halimi [Thu, 3 Sep 2015 15:28:38 +0000 (17:28 +0200)]
Add stats support for existing optimizations

This patch adds support for the --dump-stats facility with existing
optimizations.

Change-Id: I68751b119a030952a11057cb651a3c63e87e73ea
Signed-off-by: Jean-Philippe Halimi <jean-philippe.halimi@intel.com>
8 years agoMerge "Revert "Dynamic BCE (based on induction range analysis)""
Aart Bik [Sat, 21 Nov 2015 05:40:12 +0000 (05:40 +0000)]
Merge "Revert "Dynamic BCE (based on induction range analysis)""

8 years agoRevert "Dynamic BCE (based on induction range analysis)"
Aart Bik [Sat, 21 Nov 2015 05:21:52 +0000 (05:21 +0000)]
Revert "Dynamic BCE (based on induction range analysis)"

This reverts commit 0b5849be045c5683d4a6b6b6c306abadba5f0fcc.

Change-Id: Id33f5da42bbdfb1aff7e2281417c8a7aa492df05
Rationale: so close :-( but bullhead-userdebug (linux) build in git_mnc-dr-dev-plus-aosp reported a breakage with a type inconsistency (long vs int in probably the codegen of dynamic bce); no time to investigate and fix this fully before my trip, so rolling back for now

8 years agoMerge "Dynamic BCE (based on induction range analysis)"
Aart Bik [Sat, 21 Nov 2015 01:03:49 +0000 (01:03 +0000)]
Merge "Dynamic BCE (based on induction range analysis)"

8 years agoDynamic BCE (based on induction range analysis)
Aart Bik [Mon, 19 Oct 2015 21:59:26 +0000 (14:59 -0700)]
Dynamic BCE (based on induction range analysis)

Rationale: A rewritten dynamic BCE that uses induction variable analysis
           to generate the run-time tests before a loop in order to
           eliminate bounds-checks from its body. This CL removes now
           obsoleted induction related code inside the BCE module.
           Also, the dynamic test generation is placed more strategically,
           since we missed a few cases where static analysis does better.

Most significant performance improvements (after filtering noise) is about:
  Linpack +20%
  LU    > +10%

Change-Id: I4e7b8bab0288beff6f98a14856e3536103d32742

8 years agoMerge "lambda: Add support for invoke-interface for boxed innate lambdas"
Igor Murashkin [Sat, 21 Nov 2015 00:55:35 +0000 (00:55 +0000)]
Merge "lambda: Add support for invoke-interface for boxed innate lambdas"

8 years agolambda: Add support for invoke-interface for boxed innate lambdas
Igor Murashkin [Fri, 23 Oct 2015 00:37:50 +0000 (17:37 -0700)]
lambda: Add support for invoke-interface for boxed innate lambdas

Lambda closures created with the 'create-lambda' instruction
(termed "innate lambdas") can be turned into an object with 'box-lambda'.

This CL enables support for those kinds of lambdas to work with
'invoke-interface' by generating a proxy class for the lambda.

Note: MIPS32/64 support not included.

Bug: 24618608
Bug: 25107649
Change-Id: Ic8f1bb66ebeaed4097e758a50becf1cff6ccaefb

8 years agoMerge "Fix null pointer in processing of enum annotations."
Jeff Hao [Fri, 20 Nov 2015 23:07:59 +0000 (23:07 +0000)]
Merge "Fix null pointer in processing of enum annotations."

8 years agoFix null pointer in processing of enum annotations.
Jeff Hao [Fri, 20 Nov 2015 22:56:09 +0000 (14:56 -0800)]
Fix null pointer in processing of enum annotations.

Bug: 25802263
Change-Id: Ib20b7049fd3824a5eb3e396d34ef32574771d074

8 years agoExplicitly add HLoadClass/HClinitCheck for HNewInstance.
Nicolas Geoffray [Thu, 19 Nov 2015 13:29:02 +0000 (13:29 +0000)]
Explicitly add HLoadClass/HClinitCheck for HNewInstance.

bug:25735083
bug:25173758

Change-Id: Ie81cfa4fa9c47cc025edb291cdedd7af209a03db

8 years agoMerge "Pass DexPathList.Element array to openDexFileNative"
Mathieu Chartier [Fri, 20 Nov 2015 21:52:05 +0000 (21:52 +0000)]
Merge "Pass DexPathList.Element array to openDexFileNative"

8 years agoMerge "MIPS32: Miscellaneous bit manipulations routines:"
Andreas Gampe [Fri, 20 Nov 2015 21:27:59 +0000 (21:27 +0000)]
Merge "MIPS32: Miscellaneous bit manipulations routines:"

8 years agoMerge "Kill stale dalvikvm processes at the end of the jdwp test."
Hiroshi Yamauchi [Fri, 20 Nov 2015 18:55:24 +0000 (18:55 +0000)]
Merge "Kill stale dalvikvm processes at the end of the jdwp test."

8 years agoMerge "Fix --jvm on run-tests 960 and 961"
Alex Light [Fri, 20 Nov 2015 18:51:59 +0000 (18:51 +0000)]
Merge "Fix --jvm on run-tests 960 and 961"

8 years agoKill stale dalvikvm processes at the end of the jdwp test.
Hiroshi Yamauchi [Fri, 20 Nov 2015 00:24:31 +0000 (16:24 -0800)]
Kill stale dalvikvm processes at the end of the jdwp test.

Bug: 25800335
Change-Id: I938437d6e1ecd362b30a1dfc4db7a25538cf3211

8 years agoFix --jvm on run-tests 960 and 961
Alex Light [Fri, 20 Nov 2015 18:37:46 +0000 (10:37 -0800)]
Fix --jvm on run-tests 960 and 961

Change-Id: I2c20e96a27c92b15a201abeca6bc3461a0912703

8 years agoPass DexPathList.Element array to openDexFileNative
Mathieu Chartier [Fri, 20 Nov 2015 18:29:42 +0000 (10:29 -0800)]
Pass DexPathList.Element array to openDexFileNative

App images will use this to check for conflicts. It is required to
pass down since the class loader won't have the element array until
after all of the elements in the dex path list are loaded.

Bug: 22858531
Change-Id: I4f3d85b5e0ad542298fc8458bafe6504c263dc41

8 years agoMerge "Use arc4random when available to select delta for image relocation."
Alex Light [Fri, 20 Nov 2015 17:48:47 +0000 (17:48 +0000)]
Merge "Use arc4random when available to select delta for image relocation."

8 years agoMerge "Encode function signatures properly in DWARF."
David Srbecky [Fri, 20 Nov 2015 17:37:16 +0000 (17:37 +0000)]
Merge "Encode function signatures properly in DWARF."

8 years agoSimplify boolean condition compared to 0
Mark Mendell [Tue, 17 Nov 2015 16:16:56 +0000 (11:16 -0500)]
Simplify boolean condition compared to 0

CaffeineMarkRR Logic has some boolean flipping which can be helped by
some simplification.

Simplify non-FP (A COND_OP B) != 0 to A OPPOSITE_COND_OP B.
This is better than the original code, which would use a HBooleanNot
after the condition.

Also simplify non-FP (A COND_OP B) == 1 to A OPPOSITE_COND_OP B.

Move GetOppositeCondition to nodes.h/nodes.cc to share with Boolean
Simplification, renaming it to InsertOppositeCondition, as it inserts
the new HInstruction (unless it is a constant).

Change-Id: I34ded7758836e375de0d6fdba9239d2d451928d0
Signed-off-by: Mark Mendell <mark.p.mendell@intel.com>
8 years agoEncode function signatures properly in DWARF.
David Srbecky [Mon, 9 Nov 2015 18:05:48 +0000 (18:05 +0000)]
Encode function signatures properly in DWARF.

The signatures were previously stored as plain string.
The proper way in DWARF is to store them as structured tree of tags.
(for example, DW_TAG_subprogram containing DW_TAG_formal_parameter)

Note that this makes the debug sections smaller since DWARF
signatures are actually more efficient than just plain strings.

Change-Id: I6afbce28340570666d8674d07c0e324aad561dd5

8 years agoOptimizing/ARM: Improve long shifts by 1.
Vladimir Marko [Fri, 20 Nov 2015 15:08:11 +0000 (15:08 +0000)]
Optimizing/ARM: Improve long shifts by 1.

Implement long
    Shl(x,1) as LSLS+ADC,
    Shr(x,1) as ASR+RRX and
    UShr(x,1) as LSR+RRX.

Remove the simplification substituting Shl(x,1) with
ADD(x,x) as it interferes with some other optimizations
instead of helping them. And since it didn't help 64-bit
architectures anyway, codegen is the correct place for it.
This is now implemented for ARM and x86, so only mips32 can
be improved.

Change-Id: Idd14f23292198b2260189e1497ca5411b21743b3

8 years agoMerge "Opt compiler: More strength reduction for multiplications."
David Brazdil [Fri, 20 Nov 2015 15:43:58 +0000 (15:43 +0000)]
Merge "Opt compiler: More strength reduction for multiplications."

8 years agoOpt compiler: More strength reduction for multiplications.
Alexandre Rames [Fri, 20 Nov 2015 15:02:45 +0000 (15:02 +0000)]
Opt compiler: More strength reduction for multiplications.

We transform code looking like

   MUL dst, src, (2^n + 1)

into

   SHL tmp, src, n
   ADD dst, src, tmp

and code looking like

   MUL dst, src, (2^n - 1)

into

   SHL tmp, src, n
   SUB dst, tmp, src

Change-Id: Ia620ab68758caa70a01530b88cd65dd0444376d7

8 years agoOptimizing/Thumb2: Improve load/store for large offsets.
Vladimir Marko [Thu, 19 Nov 2015 21:13:52 +0000 (21:13 +0000)]
Optimizing/Thumb2: Improve load/store for large offsets.

This reduces the boot.oat size on Nexus 5 by 568KiB (0.8%).

Also change 32-bit ADD/SUB immediate to use the recommended
encoding T3 when both T3 and T4 are available.

Change-Id: I174382bda2b22da70560b947f5536acf8c1814a9

8 years agoMerge "Clean up the special input in HInvokeStaticOrDirect."
Vladimir Marko [Fri, 20 Nov 2015 10:53:56 +0000 (10:53 +0000)]
Merge "Clean up the special input in HInvokeStaticOrDirect."

8 years agoUse arc4random when available to select delta for image relocation.
Alex Light [Thu, 19 Nov 2015 19:03:10 +0000 (11:03 -0800)]
Use arc4random when available to select delta for image relocation.

Bug: 25776162

Change-Id: I5a5ca5d55c231508b61d4960049ab42e19c017e2

8 years agoMerge "Revert "Use arc4random when available to select delta for image relocation.""
Alex Light [Fri, 20 Nov 2015 01:24:59 +0000 (01:24 +0000)]
Merge "Revert "Use arc4random when available to select delta for image relocation.""

8 years agoRevert "Use arc4random when available to select delta for image relocation."
Alex Light [Fri, 20 Nov 2015 01:24:47 +0000 (01:24 +0000)]
Revert "Use arc4random when available to select delta for image relocation."

This reverts commit 7ecbd49c6c78e6c633883aa6766675df8abaa7dd.

Change-Id: Ifb37e23584722b31cb2369bcc9b91da6146d2cf6

8 years agoMerge "Use arc4random when available to select delta for image relocation."
Alex Light [Fri, 20 Nov 2015 01:17:20 +0000 (01:17 +0000)]
Merge "Use arc4random when available to select delta for image relocation."

8 years agoUse arc4random when available to select delta for image relocation.
Alex Light [Thu, 19 Nov 2015 19:03:10 +0000 (11:03 -0800)]
Use arc4random when available to select delta for image relocation.

Bug: 25776162

Change-Id: I1c4cc992977b5d6fe782ff819903a802a40391ee

8 years agoMerge "Remove thread suspension assertion for SuspendAllForDebugger"
Mathieu Chartier [Fri, 20 Nov 2015 00:41:58 +0000 (00:41 +0000)]
Merge "Remove thread suspension assertion for SuspendAllForDebugger"

8 years agoMerge "Create parent class loader for dex2oat"
Mathieu Chartier [Fri, 20 Nov 2015 00:35:00 +0000 (00:35 +0000)]
Merge "Create parent class loader for dex2oat"

8 years agoCreate parent class loader for dex2oat
Mathieu Chartier [Fri, 20 Nov 2015 00:05:58 +0000 (16:05 -0800)]
Create parent class loader for dex2oat

This means we also put the class loader in the app image so that it
can be used for verifying the dex file order.

Bug: 22858531
Change-Id: I30761b59421c8a24cffd62b469134b25d2929e2e

8 years agoMerge "Relax CAS in some uses of Object::AtomicSetReadBarrierPointer."
Hiroshi Yamauchi [Thu, 19 Nov 2015 22:51:15 +0000 (22:51 +0000)]
Merge "Relax CAS in some uses of Object::AtomicSetReadBarrierPointer."

8 years agoRelax CAS in some uses of Object::AtomicSetReadBarrierPointer.
Hiroshi Yamauchi [Wed, 18 Nov 2015 01:52:15 +0000 (17:52 -0800)]
Relax CAS in some uses of Object::AtomicSetReadBarrierPointer.

Aside from the call to change the object from gray to black, the CAS in
AtomicSetReadBarrierPointer doesn't need to be a release CAS.

CC collector Ritz EAAC GC time: 28.4 -> 26.5s (-6.7%) on N5.

Bug: 12687968

Change-Id: Ic162717981b98e7f1e762fd6b27bd0c146187ca4

8 years agoMerge "Fix cmdline parser test."
Nicolas Geoffray [Thu, 19 Nov 2015 18:18:19 +0000 (18:18 +0000)]
Merge "Fix cmdline parser test."

8 years agoFix cmdline parser test.
Nicolas Geoffray [Thu, 19 Nov 2015 18:17:41 +0000 (18:17 +0000)]
Fix cmdline parser test.

Change-Id: I6f03552d8d48e875f80a69455397146b132ed623

8 years agoMerge "Allow NullConstant to be untyped in GraphVisualiser."
Calin Juravle [Thu, 19 Nov 2015 17:46:50 +0000 (17:46 +0000)]
Merge "Allow NullConstant to be untyped in GraphVisualiser."

8 years agoAllow NullConstant to be untyped in GraphVisualiser.
Mark Mendell [Mon, 16 Nov 2015 17:21:53 +0000 (12:21 -0500)]
Allow NullConstant to be untyped in GraphVisualiser.

The NullConstant may be added to the graph during other passes that
happen between ReferenceTypePropagation and Inliner (e.g.
InstructionSimplifier). If the inliner doesn't run or doesn't inline
anything, the NullConstant remains untyped.

The infrastructure to properly type NullConstants everywhere is to
complex to add for the benefits

Bug: 25786318

Change-Id: I904a3e605b57f8cac9936e82f19a4994c7b1a82a
Signed-off-by: Mark Mendell <mark.p.mendell@intel.com>
8 years agoMerge "MIPS64: Add java.lang.String.equals intrinsic."
Andreas Gampe [Thu, 19 Nov 2015 16:12:57 +0000 (16:12 +0000)]
Merge "MIPS64: Add java.lang.String.equals intrinsic."

8 years agoClean up the special input in HInvokeStaticOrDirect.
Vladimir Marko [Thu, 19 Nov 2015 15:48:33 +0000 (15:48 +0000)]
Clean up the special input in HInvokeStaticOrDirect.

Change-Id: I4042aefbdac1a8c236d00e2e7145349a64f6486b

8 years agoMerge "Rename options for shorter names."
Nicolas Geoffray [Thu, 19 Nov 2015 15:48:01 +0000 (15:48 +0000)]
Merge "Rename options for shorter names."

8 years agoRename options for shorter names.
Nicolas Geoffray [Thu, 19 Nov 2015 14:25:43 +0000 (14:25 +0000)]
Rename options for shorter names.

Change-Id: I2ea98ead4cd99d6a51cbfa149b0aab5e036ffb96

8 years agoMerge "ARM read barrier support for concurrent GC in Optimizing."
Roland Levillain [Thu, 19 Nov 2015 14:01:37 +0000 (14:01 +0000)]
Merge "ARM read barrier support for concurrent GC in Optimizing."

8 years agoMerge "Fix ClinitCheck pruning."
Vladimir Marko [Thu, 19 Nov 2015 13:12:09 +0000 (13:12 +0000)]
Merge "Fix ClinitCheck pruning."

8 years agoFix ClinitCheck pruning.
Vladimir Marko [Fri, 13 Nov 2015 14:47:00 +0000 (14:47 +0000)]
Fix ClinitCheck pruning.

Make sure we merge the ClinitCheck only with LoadClass and
HInvokeStaticOrDirect that is a part of the very same dex
instruction. This fixes incorrect stack traces from class
initializers (wrong dex pcs).

Rewrite the pruning to do all the ClinitCheck merging when
we see the ClinitCheck, instead of merging ClinitCheck into
LoadClass and then LoadClass into HInvokeStaticOrDirect.
When we later see an HInvokeStaticOrDirect with an explicit
check (i.e. not merged), we know that some other instruction
is doing the check and the invoke doesn't need to, so we
mark it as not requiring the check at all. (Previously it
would have been marked as having an implicit check.)

Remove the restriction on merging with inlined invoke static
as this is not necessary anymore. This was a workaround for
    X.test():
       invoke-static C.foo() [1]
    C.foo():
       invoke-static C.bar() [2]
After inlining and GVN we have
    X.test():
       LoadClass C (from [1])
       ClinitCheck C (from [1], to be merged to LoadClass)
       InvokeStaticOrDirect C.bar() (from [2])
and the LoadClass must not be merged into the invoke as this
would cause the resolution trampoline to see an inlined
frame from the not-yet-loaded class C during the stack walk
and try to load the class. However, we're not allowed to
load new classes at that point, so an attempt to do so leads
to an assertion failure. With this CL, LoadClass is not
merged when it comes from a different instruction, so we can
guarantee that all inlined frames seen by the stack walk in
the resolution trampoline belong to already loaded classes.

Change-Id: I2b8da8d4f295355dce17141f0fab2dace126684d

8 years agoMerge "Increase code cache after 1 full collection."
Nicolas Geoffray [Thu, 19 Nov 2015 11:40:15 +0000 (11:40 +0000)]
Merge "Increase code cache after 1 full collection."

8 years agoIncrease code cache after 1 full collection.
Nicolas Geoffray [Wed, 18 Nov 2015 11:15:22 +0000 (11:15 +0000)]
Increase code cache after 1 full collection.

Also add a max capacity option.

Change-Id: Icd442b72e9be0c6b091b588b4c4473c69b7cde10

8 years agoMerge "Add jack as a build target for the buildbot."
Nicolas Geoffray [Thu, 19 Nov 2015 11:29:36 +0000 (11:29 +0000)]
Merge "Add jack as a build target for the buildbot."

8 years agoAdd jack as a build target for the buildbot.
Nicolas Geoffray [Thu, 19 Nov 2015 11:02:50 +0000 (11:02 +0000)]
Add jack as a build target for the buildbot.

This avoids dist targets to try on killing the jack server.

Change-Id: Icb913f3ba70ef4c5abd46ebac56cb283e17f0f12

8 years agoMerge "ART: Change Init{From,Without}Image to return bool"
Andreas Gampe [Thu, 19 Nov 2015 06:27:53 +0000 (06:27 +0000)]
Merge "ART: Change Init{From,Without}Image to return bool"

8 years agoART: Change Init{From,Without}Image to return bool
Andreas Gampe [Tue, 17 Nov 2015 19:52:46 +0000 (11:52 -0800)]
ART: Change Init{From,Without}Image to return bool

Rewrite some CHECKs to return false. For a common failure (missing)
image, this improves the abort (as it's not a runtime abort with
lots of stack traces anymore).

Change-Id: I717b1db74950267ced0ad3bafa1aed1693680062

8 years agoMerge "Add immune spaces abstraction"
Mathieu Chartier [Thu, 19 Nov 2015 01:55:31 +0000 (01:55 +0000)]
Merge "Add immune spaces abstraction"

8 years agoAdd immune spaces abstraction
Mathieu Chartier [Tue, 17 Nov 2015 00:05:55 +0000 (16:05 -0800)]
Add immune spaces abstraction

ImmuneSpaces is a set of spaces which are not reclaimable by the GC in
the current collection. This set of spaces does not have requirements
about space adjacency like the old ImmuneRegion. ImmuneSpaces generates
the largest immune region for the GC. Since there is no requirement on
adjacency, it is possible to have multiple non-adjacent applicaton
image files.

For image spaces, we also look at the oat code which is normally after
the application image. In this case, we add the code as part of the
immune region. This is required to have both the boot image and the
zygote space be in the same immune region (for performance reasons).

Bug: 22858531

Change-Id: I5103b31c0e39ad63c594f5557fc848a3b288b43e

8 years agoMerge "Fix an assert in lse."
Mingyao Yang [Wed, 18 Nov 2015 23:33:40 +0000 (23:33 +0000)]
Merge "Fix an assert in lse."

8 years agoFix an assert in lse.
Mingyao Yang [Wed, 18 Nov 2015 23:20:37 +0000 (15:20 -0800)]
Fix an assert in lse.

Wrong assert. Just deleted it.

Change-Id: I6f051609d87b2c6da081ee63f783ce99462f8a09

8 years agoMerge "Revert "Revert "Enable store elimination for singleton objects."""
Mingyao Yang [Wed, 18 Nov 2015 21:29:47 +0000 (21:29 +0000)]
Merge "Revert "Revert "Enable store elimination for singleton objects."""

8 years agoRevert "Revert "Enable store elimination for singleton objects.""
Mingyao Yang [Mon, 2 Nov 2015 18:56:59 +0000 (10:56 -0800)]
Revert "Revert "Enable store elimination for singleton objects.""

This reverts commit 55d02cf056f993aeafebd54e7b7c68c7a48507c9, and
makes the following change:
Currently we leverage loop side effects to decide whether heap values are
killed by the loop. Stores need to be kept if heap values may be killed
by loops and the corresponding loads cannot be eliminated. Similar thing
need to be done for each predecessor when we merge predecessor heap values.
To do that, the HInstanceFieldSet instruction itself is put in the heap
value array instead of the value of the store instruction. The store
instruction may be added to possibly_removed_stores_ first, but can later
be removed from possibly_removed_stores_ when it's found out that the store
needs to be kept due to merging/loop side effects.

Change-Id: I4f7bb1960f7b47240873e00ff1adac46fc102a02

8 years agoMIPS32: Miscellaneous bit manipulations routines:
Chris Larsen [Wed, 28 Oct 2015 17:08:56 +0000 (10:08 -0700)]
MIPS32: Miscellaneous bit manipulations routines:

- short java.lang.Short.reverseBytes(short)
- int java.lang.Integer.reverseBytes(int)
- long java.lang.Long.reverseBytes(long)
- float java.lang.Float.intBitsToFloat(int)
- double java.lang.Double.longBitsToDouble(long)
- int java.lang.Float.floatToRawIntBits(float)
- long java.lang.Double.doubleToRawLongBits(double)

Change-Id: Id9803349d465c28756820e90e2cbe633f3f40a44

8 years agoMerge "Remove incorrect iterator increment"
Mathieu Chartier [Wed, 18 Nov 2015 20:06:43 +0000 (20:06 +0000)]
Merge "Remove incorrect iterator increment"

8 years agoRemove incorrect iterator increment
Mathieu Chartier [Wed, 18 Nov 2015 18:24:43 +0000 (10:24 -0800)]
Remove incorrect iterator increment

Erase already goes to the next element.

Bug: 25738102
Change-Id: Iea31a7fb27ce9f810acf019a7b59dda7704b8f26

8 years agoMerge "Opt compiler: Arm64 packed-switch jump tables."
Vladimir Marko [Wed, 18 Nov 2015 17:09:04 +0000 (17:09 +0000)]
Merge "Opt compiler: Arm64 packed-switch jump tables."

8 years agoOpt compiler: Arm64 packed-switch jump tables.
Zheng Xu [Wed, 18 Nov 2015 09:46:25 +0000 (17:46 +0800)]
Opt compiler: Arm64 packed-switch jump tables.

In this patch, we set a rough threshold and only generate jump table
with limited number of HIRs in the graph. This is because current VIXL
can only handle Adr with label in the range of +/-1Mb.

Change-Id: I42bff2095ec26caeacc5efc90afebe34e229b518

8 years agoMerge "Fix bogus DCHECK and rename DidForkFromZygote."
Nicolas Geoffray [Wed, 18 Nov 2015 10:56:51 +0000 (10:56 +0000)]
Merge "Fix bogus DCHECK and rename DidForkFromZygote."

8 years agoFix bogus DCHECK and rename DidForkFromZygote.
Nicolas Geoffray [Tue, 17 Nov 2015 13:50:21 +0000 (13:50 +0000)]
Fix bogus DCHECK and rename DidForkFromZygote.

The method can also be called for non-zygote forked processes.

(cherry picked from commit 9d157e48e0abc48f08fd98e4c5e918d029c1b84c)

Change-Id: Id4f57b77c1a7f802f80e066afcf30afecc198bfc

8 years agoMerge "Remove DEBUG_JIT from Zygote flags."
Nicolas Geoffray [Wed, 18 Nov 2015 10:47:50 +0000 (10:47 +0000)]
Merge "Remove DEBUG_JIT from Zygote flags."

8 years agoMerge "Implement common super type in reference type propagation."
Calin Juravle [Wed, 18 Nov 2015 10:41:41 +0000 (10:41 +0000)]
Merge "Implement common super type in reference type propagation."

8 years agoMerge "Support deoptimization only to set vreg"
Sebastien Hertz [Wed, 18 Nov 2015 10:29:22 +0000 (10:29 +0000)]
Merge "Support deoptimization only to set vreg"

8 years agoMerge "Fix StackVisitor::GetVReg for reference in shadow frame"
Sebastien Hertz [Wed, 18 Nov 2015 10:17:46 +0000 (10:17 +0000)]
Merge "Fix StackVisitor::GetVReg for reference in shadow frame"

8 years agoRemove DEBUG_JIT from Zygote flags.
Nicolas Geoffray [Thu, 5 Nov 2015 11:32:24 +0000 (11:32 +0000)]
Remove DEBUG_JIT from Zygote flags.

The flag is being obsolete by the move to JIT.

(cherry picked from commit 0f042e04efba887557e40f981bd9c41b121c4652)

Change-Id: I32c3183e791690964c00fab02a6ee7bbacf6b665

8 years agoMerge "RosAlloc fast path inline in assembly for x86 32bit."
Hiroshi Yamauchi [Tue, 17 Nov 2015 21:53:34 +0000 (21:53 +0000)]
Merge "RosAlloc fast path inline in assembly for x86 32bit."

8 years agoMerge "Refactor some patching logic"
Mathieu Chartier [Tue, 17 Nov 2015 20:10:24 +0000 (20:10 +0000)]
Merge "Refactor some patching logic"

8 years agoARM read barrier support for concurrent GC in Optimizing.
Roland Levillain [Tue, 17 Nov 2015 19:35:12 +0000 (19:35 +0000)]
ARM read barrier support for concurrent GC in Optimizing.

This first implementation uses slow paths to instrument heap
reference loads and GC root loads for the concurrent copying
collector, respectively calling the artReadBarrierSlow and
artReadBarrierForRootSlow runtime entry points.

Notes:
- This implementation does not instrument HInvokeVirtual
  nor HInvokeInterface instructions (for class reference
  loads), as the corresponding read barriers are not stricly
  required with the current concurrent copying collector.
- Intrinsics which may eventually call (on slow path) are
  disabled when read barriers are enabled, as the current
  slow path infrastructure does not support this case.
- When read barriers are enabled, the code generated for a
  HArraySet instruction always go into the array set slow
  path for object arrays (delegating the operation to the
  runtime), as we are lacking a mechanism to keep a
  temporary register live accross a runtime call (needed for
  the instrumentation of type checking code, which requires
  two successive read barriers).

Bug: 12687968
Change-Id: I92e8db414d029f952c07f3d3a98069e46dfdbc2a

8 years agoImplement common super type in reference type propagation.
Calin Juravle [Wed, 11 Nov 2015 16:58:31 +0000 (16:58 +0000)]
Implement common super type in reference type propagation.

Currently only if both types are classes.

Change-Id: I06e98211ead56875a42bd17f099e319b107a50d4

8 years agoMerge "Minor fixes and cleaning of arm64 static and direct calls code."
Vladimir Marko [Tue, 17 Nov 2015 17:40:57 +0000 (17:40 +0000)]
Merge "Minor fixes and cleaning of arm64 static and direct calls code."

8 years agoMerge "ART: add kryo to a53 #835769 & #843419 erratum exception list"
Vladimir Marko [Tue, 17 Nov 2015 17:40:34 +0000 (17:40 +0000)]
Merge "ART: add kryo to a53 #835769 & #843419 erratum exception list"

8 years agoMerge "ART: Refactor GenerateTestAndBranch"
David Brazdil [Tue, 17 Nov 2015 16:46:49 +0000 (16:46 +0000)]
Merge "ART: Refactor GenerateTestAndBranch"

8 years agoMerge "Change roots view to "rooted"."
Richard Uhler [Tue, 17 Nov 2015 16:21:12 +0000 (16:21 +0000)]
Merge "Change roots view to "rooted"."

8 years agoART: Refactor GenerateTestAndBranch
David Brazdil [Thu, 12 Nov 2015 18:37:00 +0000 (18:37 +0000)]
ART: Refactor GenerateTestAndBranch

Each code generator implements a method for generating condition
evaluation and branching to arbitrary labels. This patch refactors
it for better clarity but also to generate fewer jumps when the true
branch is the fallthrough successor.

This is preliminary work for implementing HSelect.

Change-Id: Iaa545a5ecbacb761c5aa241fa69140cf6eb5952f

8 years agoSupport deoptimization only to set vreg
Mingyao Yang [Fri, 31 Jul 2015 23:40:24 +0000 (16:40 -0700)]
Support deoptimization only to set vreg

This CL removes the old API that was used to update vreg. We now rely
exclusively on deoptimization to change the value of a vreg (mainly
from the debugger). This allows to have only one mechanism working
with both Quick and Optimizing compilers. It also remove run-test
455-set-vreg which was the only user of the old API.

This CL also renames StackVisitor::SetVReg[Pair]FromDebugger to
StackVisitor::SetVReg[Pair] to reflect that change.

Bug: 25428216
Change-Id: Ib09510185280dbd0f6d02c52549ae9671c187e32

8 years agoFix StackVisitor::GetVReg for reference in shadow frame
Sebastien Hertz [Tue, 17 Nov 2015 09:35:39 +0000 (10:35 +0100)]
Fix StackVisitor::GetVReg for reference in shadow frame

Also fixes 457-regs run-test that used to pass an incorrect
vreg kind.

Bug: 25428216
Change-Id: Ic31b51ca0f72350db6325f23087b06e93055d750

8 years agoRefactor some patching logic
Mathieu Chartier [Fri, 13 Nov 2015 18:42:08 +0000 (10:42 -0800)]
Refactor some patching logic

Reduce duplication since app images in-place relocation will require
this code also.

Bug: 22858531
Change-Id: Ibb901b67267e27ef3bc2a0baff77189d4dcd018a