OSDN Git Service

android-x86/dalvik.git
13 years agoThe jni-tips documentation has moved to frameworks/base.
Elliott Hughes [Thu, 14 Apr 2011 18:52:03 +0000 (11:52 -0700)]
The jni-tips documentation has moved to frameworks/base.

Change-Id: I77fbc3ed593c51f61de9590beaadaa52f08387f8

13 years agoFix debugger performance regression
buzbee [Wed, 13 Apr 2011 21:10:04 +0000 (14:10 -0700)]
Fix debugger performance regression

http://b/issue?id=4282910

The interpreter was reporting every instruction as a method entry point,
causing significant performance degradation in the debugger.

Change-Id: If6fb092d6ade7fc3d74be3c7f806bf2d6923a063

13 years agoMerge "Add missing JUMBO opcodes to volatile rewriting"
buzbee [Mon, 11 Apr 2011 21:20:45 +0000 (14:20 -0700)]
Merge "Add missing JUMBO opcodes to volatile rewriting"

13 years agoAdd missing JUMBO opcodes to volatile rewriting
buzbee [Mon, 11 Apr 2011 21:07:34 +0000 (14:07 -0700)]
Add missing JUMBO opcodes to volatile rewriting

The boolean/byte/char/short variants of iget/iput/sget/sput were not
being correctly converted to _VOLATILE opcodes for SMP builds.

Change-Id: Ic81d75f9c927325d54a98465c146b6d43ea204f4

13 years ago[JIT] Close code cache race window
buzbee [Sat, 9 Apr 2011 21:47:32 +0000 (14:47 -0700)]
[JIT] Close code cache race window

See http://b/issue?id=4271784 for details.

Three fixes:
   1.  Verify the code cache version hasn't changed between completion
       of the translation and registering it in JitTable
   2.  When code cache full detected during translating a trace, mark
       the "discard" flag on the work order.
   3.  [The actual cause of the bug] When doing a code cache flush,
       traverse the thread least and cancel any trace selections in
       progress.

Change-Id: Ifea70416d7d91637fb742fc8de11044a89358caa

13 years agoMerge "[JIT] Fix volatile test"
buzbee [Sat, 9 Apr 2011 01:01:17 +0000 (18:01 -0700)]
Merge "[JIT] Fix volatile test"

13 years ago[JIT] Fix volatile test
buzbee [Thu, 7 Apr 2011 22:22:06 +0000 (15:22 -0700)]
[JIT] Fix volatile test

On SMP systems, Dalvik opcodes referencing volatile fields will be
rewritten to their _VOLATILE variant.  On non-SMP systems, though,
this rewriting is not done.  The JIT, however, needs to know about
volatility for all systems in order to avoid performing unsafe
optimizations.  This change fixes the JIT's volatility test to be
either _VOLATILE opcode or the volatile flag in the field access bits
depending on SMP type.

Change-Id: I2edde58dc25f22cba88f62c5f1a2d125473309e6

13 years agoMerge "Add jniThrowExceptionFmt."
Elliott Hughes [Sat, 9 Apr 2011 00:20:54 +0000 (17:20 -0700)]
Merge "Add jniThrowExceptionFmt."

13 years agoAdd jniThrowExceptionFmt.
Elliott Hughes [Sat, 9 Apr 2011 00:15:16 +0000 (17:15 -0700)]
Add jniThrowExceptionFmt.

I still think you should be doing checks and throwing exceptions in Java,
and your native code should just crash if you lie to it, but this will
ease the translation of frameworks/base code over to using JNI more correctly.

Change-Id: I9f2512e2349452b82360b375911c814ab00db23b

13 years agoMerge "Fix method profiling [issue 4260697]"
buzbee [Fri, 8 Apr 2011 22:45:13 +0000 (15:45 -0700)]
Merge "Fix method profiling [issue 4260697]"

13 years agoRevert "[JIT] Fix volatile test" - breaks build
Joe Onorato [Fri, 8 Apr 2011 22:32:34 +0000 (15:32 -0700)]
Revert "[JIT] Fix volatile test" - breaks build

This reverts commit 11fb99d598ebe640719743a0d3bd7ed091e5be03.

13 years agoFix method profiling [issue 4260697]
buzbee [Fri, 8 Apr 2011 22:25:13 +0000 (15:25 -0700)]
Fix method profiling [issue 4260697]

In the new interpreter control mechanism, method tracing invoke
events are generated at the call site, whereas debug entry events
happen prior to the interpretation of the first instruction of a
method.  Because we were failing to check for active profiling
on entry to a new interpreter instance, we would miss the profile
event for the first method interpreted.  We subsequently (and correctly)
recorded the return event, and ended up with a mismatch causing traceview
to complain and die.

Change-Id: Ibd93dfcd16c2f1fc62f6a86626a902eaf5d4b160

13 years ago[JIT] Fix volatile test
buzbee [Thu, 7 Apr 2011 22:22:06 +0000 (15:22 -0700)]
[JIT] Fix volatile test

On SMP systems, Dalvik opcodes referencing volatile fields will be
rewritten to their _VOLATILE variant.  On non-SMP systems, though,
this rewriting is not done.  The JIT, however, needs to know about
volatility for all systems in order to avoid performing unsafe
optimizations.  This change fixes the JIT's volatility test to be
either _VOLATILE opcode or the volatile flag in the field access bits
depending on SMP type.

Change-Id: If485875c5abbf0147c3ef4b6d557faa89ed85426

13 years agoBug fix for rsubs being improperly generated.
jeffhao [Fri, 8 Apr 2011 19:14:27 +0000 (12:14 -0700)]
Bug fix for rsubs being improperly generated.

This will miss some opportunities to generate rsubs, and has been marked
by a TODO note where code needs to be added.

Cherry-picked frmo dalvik-dev.

Change-Id: Iac272f16b78ab6d985b5aabef6bd360d36a5c7ce

13 years agoMerge "Scale the card count by bits per byte to derive a bit vector length."
Carl Shapiro [Thu, 7 Apr 2011 23:37:30 +0000 (16:37 -0700)]
Merge "Scale the card count by bits per byte to derive a bit vector length."

13 years agoScale the card count by bits per byte to derive a bit vector length.
Carl Shapiro [Thu, 7 Apr 2011 23:02:54 +0000 (16:02 -0700)]
Scale the card count by bits per byte to derive a bit vector length.

Change-Id: If3177d924193cd665f12758ce83f3be4cc5a66c8

13 years agoMerge "Improve JNI indirect ref diagnostics"
Andy McFadden [Thu, 7 Apr 2011 21:45:06 +0000 (14:45 -0700)]
Merge "Improve JNI indirect ref diagnostics"

13 years agoImprove JNI indirect ref diagnostics
Andy McFadden [Thu, 7 Apr 2011 17:36:56 +0000 (10:36 -0700)]
Improve JNI indirect ref diagnostics

Improved CheckJNI problem detection and failure reporting.  Also,
the indirect reference table data structure now responds to invalid
requests with a magic bogus object instead of NULL, so that failures
are immediate and obvious.

The extended checks in the indirect ref implementation are still on,
mostly so that bad code fails in an obvious way even when CheckJNI
is not enabled.

This change also includes a hack to allow weak globals to work.
Returning non-NULL for invalid refs turned a rarely-fails-weirdly
issue into an always-fails-loudly, which is good, but we need to
paper over it until the WGR fix is in place.  This will need to be
removed later.

Also, reduced some log levels.

Bug 4184721

Change-Id: Iaec75c71dbc85934366be2e717329b635d0fa49e

13 years agoUpdate the interpretation of --target-api.
Dan Bornstein [Thu, 7 Apr 2011 17:41:09 +0000 (10:41 -0700)]
Update the interpretation of --target-api.

Change-Id: I4ab158384d09470d84097d944ae1ddadeb2ad045

13 years agoFix x86 fast interpreter
buzbee [Tue, 5 Apr 2011 17:55:21 +0000 (10:55 -0700)]
Fix x86 fast interpreter

Was passing wrong value to dvmResolveField() in OP_IGET_WIDE_JUMBO

Change-Id: I7773fd0932d7d032fce30dc57b33ef6a8fbc7afb

13 years ago[JIT] Rely on the opcode only to determine volatile
buzbee [Mon, 4 Apr 2011 19:25:57 +0000 (12:25 -0700)]
[JIT] Rely on the opcode only to determine volatile

In the past, it was possible to have a volatile field with a
non-volatile opcode.  This is no longer the case, so this change
eliminates the volatile field flag check.

Change-Id: I1cface4e813144634b2f90732c76b0a16f08c304

13 years agoMerge "Do not emit the write barrier when storing to primitive volatile fields."
Bill Buzbee [Mon, 4 Apr 2011 19:10:09 +0000 (12:10 -0700)]
Merge "Do not emit the write barrier when storing to primitive volatile fields."

13 years agoMerge "Fix sizing of the modunion table on small heaps"
Brian Carlstrom [Mon, 4 Apr 2011 19:09:03 +0000 (12:09 -0700)]
Merge "Fix sizing of the modunion table on small heaps"

13 years agoDo not emit the write barrier when storing to primitive volatile fields.
Carl Shapiro [Mon, 4 Apr 2011 17:20:43 +0000 (10:20 -0700)]
Do not emit the write barrier when storing to primitive volatile fields.

Change-Id: I79cdac62baa40582bba160a04cbd4c8b2c9151a5

13 years agoFix sizing of the modunion table on small heaps
Brian Carlstrom [Mon, 4 Apr 2011 19:06:15 +0000 (12:06 -0700)]
Fix sizing of the modunion table on small heaps

Change-Id: I8696fb2f6d408049ec39609db34238edee171f79

13 years agoInclude strings.h for the ffs prototype.
Carl Shapiro [Mon, 4 Apr 2011 18:33:00 +0000 (11:33 -0700)]
Include strings.h for the ffs prototype.

Change-Id: I70a4dc5c8d9015cfc069611b7eba0860bb37a8d7

13 years agoUse x86 variant in dalvik for TARGET_ARCH_VARIANT=x86-atom
Brian Carlstrom [Sun, 3 Apr 2011 05:19:15 +0000 (22:19 -0700)]
Use x86 variant in dalvik for TARGET_ARCH_VARIANT=x86-atom

Change-Id: I1f7968a5bf8e24f3a3feb5f5a40ee2c21294a0c3

13 years agoMerge remote branch 'goog/dalvik-dev' into dalvik-dev-to-master
Brian Carlstrom [Fri, 1 Apr 2011 22:05:18 +0000 (15:05 -0700)]
Merge remote branch 'goog/dalvik-dev' into dalvik-dev-to-master

13 years agoChange the tread state before acquiring the code cache lock.
Ben Cheng [Fri, 1 Apr 2011 04:07:55 +0000 (21:07 -0700)]
Change the tread state before acquiring the code cache lock.

Fix a deadlock situation.
Bug: 4192964

Change-Id: I27f869d90d58f67e675a65444ebed6fdf2a5f518

13 years agoMerge "[JIT] Clear inCodeCacheFlag on extended singlestep" into dalvik-dev
buzbee [Thu, 31 Mar 2011 22:44:23 +0000 (15:44 -0700)]
Merge "[JIT] Clear inCodeCacheFlag on extended singlestep" into dalvik-dev

13 years ago[JIT] Clear inCodeCacheFlag on extended singlestep
buzbee [Thu, 31 Mar 2011 22:38:34 +0000 (15:38 -0700)]
[JIT] Clear inCodeCacheFlag on extended singlestep

The inJitCodeCache flag was not being reset if an extended
singlestep prevented a return to an existing translation.

Change-Id: I84f9e4053e3dc713a5ff994df9ecd2ab54e7fe56

13 years agoMerge "Generate code for loops formed with the new builder" into dalvik-dev
Ben Cheng [Thu, 31 Mar 2011 21:54:06 +0000 (14:54 -0700)]
Merge "Generate code for loops formed with the new builder" into dalvik-dev

13 years agoMerge "Delegate the queuing of cleared references to managed code." into dalvik-dev
Carl Shapiro [Thu, 31 Mar 2011 20:38:28 +0000 (13:38 -0700)]
Merge "Delegate the queuing of cleared references to managed code." into dalvik-dev

13 years agoGenerate code for loops formed with the new builder
Ben Cheng [Tue, 22 Mar 2011 21:09:09 +0000 (14:09 -0700)]
Generate code for loops formed with the new builder

Adapt the existing counted loop analysis and range/null check
elimination code to work with the new loop building heuristics.
Cleaned up the old ad-hoc loop builder.

Suspend polling is enabled by default for loops. The backward chaining
cell will be used in self-verification and profiling mode.

If the loop includes accesses to resolved fields/classes, abort code
generation for now and revert to the basic acyclic trace. Added
tests/090-loop-formation to make sure the JIT won't choke on such
instructions.

Change-Id: Idbc57df0a745be3b692f68c1acb6d4861c537f75

13 years agoMerge "Add missing JUMBO ops [fixed]" into dalvik-dev
buzbee [Thu, 31 Mar 2011 18:20:47 +0000 (11:20 -0700)]
Merge "Add missing JUMBO ops [fixed]" into dalvik-dev

13 years agoAdd missing JUMBO ops [fixed]
buzbee [Thu, 31 Mar 2011 18:18:28 +0000 (11:18 -0700)]
Add missing JUMBO ops [fixed]

Neglected to git-add the most recent version in Change I0c0b6d17.

Change-Id: I5f9e630f652edcf70ab893ade6f559056ed31f8f

13 years agoMerge "Improve diagnostics when non-String passed as a jstring." into dalvik-dev
Elliott Hughes [Thu, 31 Mar 2011 18:14:41 +0000 (11:14 -0700)]
Merge "Improve diagnostics when non-String passed as a jstring." into dalvik-dev

13 years agoImprove diagnostics when non-String passed as a jstring.
Elliott Hughes [Thu, 31 Mar 2011 18:06:36 +0000 (11:06 -0700)]
Improve diagnostics when non-String passed as a jstring.

For my particular bug, the cause was obvious by inspection of the calling
code, but this might be useful in harder cases.

Before:
    W(17868) JNI WARNING: jstring 0x3600001 points to non-string object (Check_GetStringUTFChars)  (dalvikvm)

Change-Id: I883a99936b8c2cd79c5a691f7c335245ff50b34f
After:
    W(19680) JNI WARNING: jstring 0x3600001 points to object of type Llibcore/io/Posix; (Check_GetStringUTFChars)  (dalvikvm)

13 years agoMerge "Add missing JUMBO ops to JIT" into dalvik-dev
buzbee [Thu, 31 Mar 2011 18:03:07 +0000 (11:03 -0700)]
Merge "Add missing JUMBO ops to JIT" into dalvik-dev

13 years agoAdd missing JUMBO ops to JIT
buzbee [Thu, 31 Mar 2011 17:55:04 +0000 (10:55 -0700)]
Add missing JUMBO ops to JIT

We were missing a dozen or so of the new jumbo ops in the JIT's
code generator.  Interestingly, our compiler error recovery mechanism was
good enough that we didn't notice (we die on an assert build,
but silently recover and continue on a production build).

Change-Id: I0c0b6d1704c47e81b39c7dcf7d1172dbdcd29856

13 years agoAdd missing jumbo opcodes to the self verification exclusion list
Ben Cheng [Thu, 31 Mar 2011 17:44:44 +0000 (10:44 -0700)]
Add missing jumbo opcodes to the self verification exclusion list

Change-Id: If38ce20b2ac30488118259e3b3bdfbc9f068e5c3

13 years agoDelegate the queuing of cleared references to managed code.
Carl Shapiro [Thu, 31 Mar 2011 02:35:34 +0000 (19:35 -0700)]
Delegate the queuing of cleared references to managed code.

Previously, the garbage collector would pass each cleared reference to
the heap worker thread for queuing.  The heap worker thread would then
perform a callback into managed code for each cleared reference which
assigned the reference to its reference queue.

With this change, the garbage collector instead links together all of
the cleared references and calls back into managed code exactly once
to hand off the references for processing.  This change makes the heap
worker thread and its data structures obsolete.

Change-Id: I28e02638f0877a7fd2ac96b9c3f5597a38541ebb

13 years agoMerge "Update a test to handle a different anonymous class name." into dalvik-dev
Jesse Wilson [Wed, 30 Mar 2011 23:40:25 +0000 (16:40 -0700)]
Merge "Update a test to handle a different anonymous class name." into dalvik-dev

13 years agoUpdate a test to handle a different anonymous class name.
Jesse Wilson [Wed, 30 Mar 2011 23:30:18 +0000 (16:30 -0700)]
Update a test to handle a different anonymous class name.

Change-Id: Ie39f76b9aa7478c39e10165eba122e3aa2834461

13 years agoMerge "Fix interpreter debug attach" into dalvik-dev
buzbee [Wed, 30 Mar 2011 21:51:28 +0000 (14:51 -0700)]
Merge "Fix interpreter debug attach" into dalvik-dev

13 years agoFix interpreter debug attach
buzbee [Tue, 29 Mar 2011 17:26:07 +0000 (10:26 -0700)]
Fix interpreter debug attach

Fix a few miscellaneous bugs from the interpreter restructuring that were
causing a segfault on debugger attach.

Added a sanity checking routine for debugging.

Fixed a problem in which the JIT's threshold and on/off switch
wouldn't get initialized properly on thread creation.

Renamed dvmCompilerStateRefresh() to dvmCompilerUpdateGlobalState() to
better reflect its function.

Change-Id: I5b8af1ce2175e3c6f53cda19dd8e052a5f355587

13 years agoam 3a2e2a73: Merge "SamplingProfilerIntegration and SamplingProfiler improvements...
Brian Carlstrom [Wed, 30 Mar 2011 05:53:15 +0000 (22:53 -0700)]
am 3a2e2a73: Merge "SamplingProfilerIntegration and SamplingProfiler improvements (1/3)"

* commit '3a2e2a73b687ba68d879e49b465778c4d589cff1':
  SamplingProfilerIntegration and SamplingProfiler improvements (1/3)

13 years agoMerge "SamplingProfilerIntegration and SamplingProfiler improvements (1/3)"
Brian Carlstrom [Wed, 30 Mar 2011 05:35:12 +0000 (22:35 -0700)]
Merge "SamplingProfilerIntegration and SamplingProfiler improvements (1/3)"

13 years agoSamplingProfilerIntegration and SamplingProfiler improvements (1/3)
Brian Carlstrom [Mon, 28 Mar 2011 07:12:36 +0000 (00:12 -0700)]
SamplingProfilerIntegration and SamplingProfiler improvements (1/3)

Summary:

frameworks/base
- fix profiling to collect data beyond the first snapshot
- avoid many small files, accumulate data over process lifetime

libcore:
- add support for VM specific sampling, trying to cut down overhead
- added support for converting snapshot files to text format
- fixed race in profiler when stopping

dalvik
- added VMStack.setThreadStackTrace interface for filling a stack
  trace into an existing StackTraceElement[]

Details:

frameworks/base

    Changed snapshots from text to binary hprof format (bumping version to 3)
    Changed from one file per snapshot to one file per process lifetime.
    Restart profiling after snapshot.

core/java/com/android/internal/os/SamplingProfilerIntegration.java

    Add quick test in maybeSnapshot to avoid doing work when the
    SamplingProfilerIntegration is disabled. Make maybeSnapshot
    private. Remove unneeded memory allocation in handleLowMemory.

core/java/android/app/ActivityThread.java

libcore

    Added ThreadSampler interface. This allows VM specific thread
    sampling optimizations. The portable version continues to use
    Thread.getStackTrace().

dalvik/src/main/java/dalvik/system/profiler/ThreadSampler.java
dalvik/src/main/java/dalvik/system/profiler/PortableThreadSampler.java
dalvik/src/main/java/dalvik/system/profiler/SamplingProfiler.java

    Add VMStack.setThreadStackTrace and use in new DalvikThreadSampler
    to avoid allocating a full stack trace when only a limited depth
    is desired.

dalvik/src/main/java/dalvik/system/profiler/DalvikThreadSampler.java
dalvik/src/main/java/dalvik/system/VMStack.java

    Refactored BinaryHprof.readMagic out of BinaryHprofReader so it
    can be used by HprofBinaryToAscii converter to probing file
    types. Added magic number constant to be shared between readMagic
    and BinaryHprofWriter.

dalvik/src/main/java/dalvik/system/profiler/BinaryHprof.java
dalvik/src/main/java/dalvik/system/profiler/BinaryHprofReader.java
dalvik/src/main/java/dalvik/system/profiler/BinaryHprofWriter.java
dalvik/src/main/java/dalvik/system/profiler/HprofBinaryToAscii.java

    Removed unneeded HprofWriter interface. Changed to simpler static
    interface to write HprofData to binary and text formats.

dalvik/src/main/java/dalvik/system/profiler/HprofWriter.java
dalvik/src/main/java/dalvik/system/profiler/AsciiHprofWriter.java
dalvik/src/main/java/dalvik/system/profiler/BinaryHprofWriter.java
dalvik/src/test/java/dalvik/system/profiler/SamplingProfilerTest.java

    Added support for reading snapshot files created by
    SamplingProfilerIntegration by stripping the text header to allow
    easier conversion to the text format.

dalvik/src/main/java/dalvik/system/profiler/HprofBinaryToAscii.java

    Fixed race between Sampler and
    SamplingProfiler.stop. SamplingProfiler.stop previously simply
    called the Sampler's TimerTask.cancel method, but this does not
    wait for a currently running Sampler to finish. The TimerTask
    documentation says the only reliable way to do this is to have the
    run() cancel itself, so that is what is now done, with new code to
    ensure that SamplingProfiler.stop does not return until the
    Sampler has been terminated.

dalvik/src/main/java/dalvik/system/profiler/SamplingProfiler.java

dalvik

   Refactored VMStack_getThreadStackTrace to create helper getTraceBuf
   used to implement new VMStack_setThreadStackTrace. The new version
   interface fills an existing StackTraceElement[], avoid allocating
   unnecessary StackTraceElements.

vm/native/dalvik_system_VMStack.c

   Refactor dvmGetStackTraceRaw to create dvmSetStackTraceRaw which
   fills in an existing, potentially smaller, StackTraceElement[].

vm/Exception.c
vm/Exception.h

   Change stack depths to be size_t to avoid signed/unsigned comparison warnings.

vm/Ddm.c
vm/Exception.c
vm/Exception.h

Change-Id: Ie88a3f9e069f6984d790ee61c6f7ac5ed47d3388

13 years agoMerge "Improve diagnostics when RegisterNatives fails." into dalvik-dev
Elliott Hughes [Tue, 29 Mar 2011 21:59:02 +0000 (14:59 -0700)]
Merge "Improve diagnostics when RegisterNatives fails." into dalvik-dev

13 years agoImprove diagnostics when RegisterNatives fails.
Elliott Hughes [Tue, 29 Mar 2011 21:06:24 +0000 (14:06 -0700)]
Improve diagnostics when RegisterNatives fails.

This assumes that people spell the method name correctly, but make mistakes
in the signature (which seems to be true in practice).

Before:
    W( 7757) ERROR: Unable to find decl for native Llibcore/io/Posix;.sendfile:(Ljava/io/FileDescriptor;Ljava/io/FileDescriptor;Llibcore/util/MutableInt;J)J  (dalvikvm)
    E( 7757) RegisterNatives failed for 'libcore/io/Posix', aborting  (JNIHelp)

After:
    E(23966) ERROR: couldn't find native method  (dalvikvm)
    E(23966) Requested: Llibcore/io/Posix;.sendfile:(Ljava/io/FileDescriptor;Ljava/io/FileDescriptor;Llibcore/util/MutableLong;J)Z  (dalvikvm)
    E(23966) Candidate: Llibcore/io/Posix;.sendfile:(Ljava/io/FileDescriptor;Ljava/io/FileDescriptor;Llibcore/util/MutableLong;J)J  (dalvikvm)
    E(23966) RegisterNatives failed for 'libcore/io/Posix', aborting  (JNIHelp)

Change-Id: I434aee679bd7c2760404b7790b996f7f86ed9b44

13 years agoInit classes used by DDMS
Andy McFadden [Tue, 29 Mar 2011 16:58:27 +0000 (09:58 -0700)]
Init classes used by DDMS

Not all call paths ensure class initialization.

Change-Id: Ic12ac5c59f3ffa2d7c04e84ea37264f65a7630be

13 years agoMerge "Add safepoint callback for gc" into dalvik-dev
buzbee [Tue, 29 Mar 2011 01:12:30 +0000 (18:12 -0700)]
Merge "Add safepoint callback for gc" into dalvik-dev

13 years agoAdd safepoint callback for gc
buzbee [Thu, 24 Mar 2011 23:41:03 +0000 (16:41 -0700)]
Add safepoint callback for gc

This change adds a safepoint callback registration mechanism.
On a per-thread (or all-thread) basis, you pass in a function to be
called at the next safe point by the target thread.  That if that
function returns 0, it will be automatically disarmed.  If not,
the callback will stay in effect and the function will be called
on all subsequent safe points.

GC is the expected customer of this feature.

Change-Id: Icd3b93128b1fd547e142d047a12df7ae8ee646e3

13 years agoRemove unused security manager code.
Jesse Wilson [Mon, 28 Mar 2011 23:05:04 +0000 (16:05 -0700)]
Remove unused security manager code.

Change-Id: Iecc602b8235eb3feeac31e198bdaf61c65dd9fbf
http://b/4087759

13 years agoMerge "Support reflective lookup of single fields." into dalvik-dev
Jesse Wilson [Mon, 28 Mar 2011 22:57:58 +0000 (15:57 -0700)]
Merge "Support reflective lookup of single fields." into dalvik-dev

13 years agoMerge "Prep DDMS stuff in InitRefs" into dalvik-dev
Andy McFadden [Mon, 28 Mar 2011 22:48:27 +0000 (15:48 -0700)]
Merge "Prep DDMS stuff in InitRefs" into dalvik-dev

13 years agoPrep DDMS stuff in InitRefs
Andy McFadden [Mon, 28 Mar 2011 22:03:21 +0000 (15:03 -0700)]
Prep DDMS stuff in InitRefs

Find some methods and field offsets.

Also removed a duplicate entry.

Change-Id: I4c23af6cfe6e7d4f5e8767c5a9a30d3b441ad425

13 years agoMerge "Fix a couple of asserts" into dalvik-dev
Andy McFadden [Mon, 28 Mar 2011 22:32:34 +0000 (15:32 -0700)]
Merge "Fix a couple of asserts" into dalvik-dev

13 years agoFix a couple of asserts
Andy McFadden [Mon, 28 Mar 2011 22:23:23 +0000 (15:23 -0700)]
Fix a couple of asserts

Polarity was reversed.

Change-Id: I15af89267a703210ae5d08a9cca529b2f0cd48c6

13 years agoresolved conflicts for merge of ac815100 to dalvik-dev
buzbee [Mon, 28 Mar 2011 21:50:04 +0000 (14:50 -0700)]
resolved conflicts for merge of ac815100 to dalvik-dev

Change-Id: I724cce041f14f91aec9ec5ca6cc535ede6d94f70

13 years agoSupport reflective lookup of single fields.
Jesse Wilson [Mon, 28 Mar 2011 21:26:34 +0000 (14:26 -0700)]
Support reflective lookup of single fields.

Change-Id: Ie7058fa9de8212b0f52d5d9b0ec7d46242a851fc
http://b/4087759

13 years agoFix x86-atom build
buzbee [Mon, 28 Mar 2011 20:51:38 +0000 (13:51 -0700)]
Fix x86-atom build

Fix an apparent merge problem in the x86-atom build.  The
handler table still had a reference to the unused 0xF1 opcode
rather than the new OP_RETURN_VOID_BARRIER.

Change-Id: Iaa6329c15f7891fd793846a2d19979679aed9af2

13 years agoMerge "Yet more InitRefs-ing" into dalvik-dev
Andy McFadden [Mon, 28 Mar 2011 17:05:39 +0000 (10:05 -0700)]
Merge "Yet more InitRefs-ing" into dalvik-dev

13 years agoYet more InitRefs-ing
Andy McFadden [Fri, 25 Mar 2011 14:37:31 +0000 (07:37 -0700)]
Yet more InitRefs-ing

Fix up some early initialization stuff.

Also, stop pretending that the top of non-main-thread stacks is a
Runnable.

Change-Id: I3e89c6320351c160cfa0bea3cf6bbfd26a5b7c12

13 years agoMerge "Try a bit harder to look like gdb(1)." into dalvik-dev
Elliott Hughes [Fri, 25 Mar 2011 21:19:15 +0000 (14:19 -0700)]
Merge "Try a bit harder to look like gdb(1)." into dalvik-dev

13 years agoMerge "Native code to fast path annotation lookups." into dalvik-dev
Jesse Wilson [Fri, 25 Mar 2011 20:56:40 +0000 (13:56 -0700)]
Merge "Native code to fast path annotation lookups." into dalvik-dev

13 years agoTry a bit harder to look like gdb(1).
Elliott Hughes [Fri, 25 Mar 2011 20:53:43 +0000 (13:53 -0700)]
Try a bit harder to look like gdb(1).

Before:

    W(27206) #9  ...
    W(27206) #10  ...

After:

    W(30066) #9  ...
    W(30066) #10 ...

Bug: http://b/4176271
Change-Id: If75b4f0bc9553bce0747a818718f5a8873c30565

13 years agoMerge "Automatically dump stack traces in dvmAbort on the host." into dalvik-dev
Elliott Hughes [Fri, 25 Mar 2011 20:38:13 +0000 (13:38 -0700)]
Merge "Automatically dump stack traces in dvmAbort on the host." into dalvik-dev

13 years agoAutomatically dump stack traces in dvmAbort on the host.
Elliott Hughes [Fri, 25 Mar 2011 18:58:52 +0000 (11:58 -0700)]
Automatically dump stack traces in dvmAbort on the host.

Before:

    E(25569) VM aborting  (dalvikvm)

After:

    E( 9806) VM aborting  (dalvikvm)
    W( 9806) Obtained 7 stack frames.  (dalvikvm)
    W( 9806) #0  /usr/local/google/home/enh/dd4/out/host/linux-x86/bin/../lib/libdvm.so(dvmPrintNativeBackTrace+0x3f) [0xf7678f2f]  (dalvikvm)
    W( 9806) #1  /usr/local/google/home/enh/dd4/out/host/linux-x86/bin/../lib/libdvm.so(dvmAbort+0x52) [0xf7679072]  (dalvikvm)
    W( 9806) #2  /usr/local/google/home/enh/dd4/out/host/linux-x86/bin/../lib/libdvm.so(dvmStartup+0x67f) [0xf767a45f]  (dalvikvm)
    W( 9806) #3  /usr/local/google/home/enh/dd4/out/host/linux-x86/bin/../lib/libdvm.so(JNI_CreateJavaVM+0x1ee) [0xf768152e]  (dalvikvm)
    W( 9806) #4  /usr/local/google/home/enh/dd4/out/host/linux-x86/bin/dalvikvm() [0x8048ae3]  (dalvikvm)
    W( 9806) #5  /lib32/libc.so.6(__libc_start_main+0xe6) [0xf6ed5bd6]  (dalvikvm)
    W( 9806) #6  /usr/local/google/home/enh/dd4/out/host/linux-x86/bin/dalvikvm() [0x8048761]  (dalvikvm)

This makes continuous build crashes more directly useful, and saves having to
switch to the device/break out gdb when doing simple debugging.

Bug: 4176271
Change-Id: I49b40fd56e6aa89d1ddc7c7f299314761456fc28

13 years agoNative code to fast path annotation lookups.
Jesse Wilson [Fri, 25 Mar 2011 17:44:15 +0000 (10:44 -0700)]
Native code to fast path annotation lookups.

This also removes duplication between Method and Constructor.

Change-Id: I19cdcfdfe5cc3ce28a9be401a1bc58a5a6e8ec81
http://code.google.com/p/android/issues/detail?id=7811

13 years agoMerge "Some more InitRefs-ing" into dalvik-dev
Andy McFadden [Fri, 25 Mar 2011 14:35:48 +0000 (07:35 -0700)]
Merge "Some more InitRefs-ing" into dalvik-dev

13 years agoOptimize annotations.
Jesse Wilson [Thu, 24 Mar 2011 23:43:50 +0000 (16:43 -0700)]
Optimize annotations.

See libcore change I5912b526dfa87a61a95fbe7b7efc48d0dca3ad38

Change-Id: I89ed0be331eea3d9c0d02bebeaa719b686d532ca
http://code.google.com/p/android/issues/detail?id=7811

13 years agoSome more InitRefs-ing
Andy McFadden [Thu, 24 Mar 2011 20:45:14 +0000 (13:45 -0700)]
Some more InitRefs-ing

Uncaught exception handling and FileDescriptor.descriptor.

Change-Id: Ia7429417a47e9a3e3d327b246c3ad33c6c77dfa0

13 years agoMerge "Various dx fixes to make it preserve locals and debug info." into dalvik-dev
jeffhao [Thu, 24 Mar 2011 21:29:49 +0000 (14:29 -0700)]
Merge "Various dx fixes to make it preserve locals and debug info." into dalvik-dev

13 years agoVarious dx fixes to make it preserve locals and debug info.
jeffhao [Thu, 24 Mar 2011 21:01:25 +0000 (14:01 -0700)]
Various dx fixes to make it preserve locals and debug info.

The SCCP pass checks if a register has associated local info before
attepting to prune branches or replace values with constants. Also, the
register allocator reserves a register for each local in a method. This
could be later improved by only reserving registers for the lengths of
their lifetimes as specified by the local info table.

Change-Id: I654ac014a9aac530f1db0db7d5cdef7535ff49bf

13 years agoInterpreter fix - limit suspend checks
buzbee [Thu, 24 Mar 2011 19:56:41 +0000 (12:56 -0700)]
Interpreter fix - limit suspend checks

Register maps are generated only for a subset of instructions (to
limit memory consumption).  The previous interpreter restructuring
was doing suspend checks at locations that had no register maps.
This CL limits suspend checks to instructions that have register
maps, and also expands that set to include foward as well as
backwards branches.

Change-Id: Ia2b2e1096efe524fdb2a5fd4be171a4216b6a3b4

13 years agoMerge "Dump the current code stream when the assembler fails on size issues." into...
Ben Cheng [Thu, 24 Mar 2011 19:08:12 +0000 (12:08 -0700)]
Merge "Dump the current code stream when the assembler fails on size issues." into dalvik-dev

13 years agoDump the current code stream when the assembler fails on size issues.
Ben Cheng [Wed, 23 Mar 2011 21:37:40 +0000 (14:37 -0700)]
Dump the current code stream when the assembler fails on size issues.

Since the assembler is very robust and will recover from such problems,
adding the verbose/noisy mode will make it easier to detect overly
aggressive optimizations that don't actually work.

Example:

D/dalvikvm( 2348): Assembler abort #1 on 1
D/dalvikvm( 2348): kThumbBCond@16: delta=260
                 :

Instruction at 0x16 is a conditional branch:
D/dalvikvm( 2348): 0x16 (0016): beq     0x0000001a (L0xb6c0c)
                 :

Label at L0xb6c0c is a PC reconstruction cell:
D/dalvikvm( 2348): L0xb6c0c:
D/dalvikvm( 2348): -------- reconstruct dalvik PC : 0x401854d6 @ +0x002b
D/dalvikvm( 2348): 0x11e (011e): ldr     r0, [r15pc, #0]
D/dalvikvm( 2348): 0x122 (0122): b       0x00000126 (L0xb685c)

where 0x11e - 0x16 - 4 = 260

Change-Id: Icbc3dae581949f5976722e24e38f04ec882c7d79

13 years agoMerge "Various JDWP updates" into dalvik-dev
Andy McFadden [Thu, 24 Mar 2011 18:28:50 +0000 (11:28 -0700)]
Merge "Various JDWP updates" into dalvik-dev

13 years agoVarious JDWP updates
Andy McFadden [Wed, 23 Mar 2011 23:15:36 +0000 (16:15 -0700)]
Various JDWP updates

More tweaks to JDWP support:

- Mask off access flags not defined by the Java programming language
  from fields and methods.
- Use InitRefs to look up ThreadGroup fields.
- Use memcpy() for some 64-bit accesses.

Change-Id: I4ff9f28ad7e6f1519a2befa557ddb5cf41e0f4cf

13 years agoam 4624f350: am 17aff091: am b1101e84: am 486ac3c0: Removing inaccurate and unmaintai...
Dan Bornstein [Thu, 24 Mar 2011 01:24:12 +0000 (18:24 -0700)]
am 4624f350: am 17aff091: am b1101e84: am 486ac3c0: Removing inaccurate and unmaintained docs.

* commit '4624f350313daaae52068c5d1d12b88859d43e54':
  Removing inaccurate and unmaintained docs.

13 years agoam 17aff091: am b1101e84: am 486ac3c0: Removing inaccurate and unmaintained docs.
Dan Bornstein [Thu, 24 Mar 2011 00:58:34 +0000 (17:58 -0700)]
am 17aff091: am b1101e84: am 486ac3c0: Removing inaccurate and unmaintained docs.

* commit '17aff091bad641b81e2609cfce69739986597d40':
  Removing inaccurate and unmaintained docs.

13 years agoam b1101e84: am 486ac3c0: Removing inaccurate and unmaintained docs.
Dan Bornstein [Thu, 24 Mar 2011 00:56:38 +0000 (17:56 -0700)]
am b1101e84: am 486ac3c0: Removing inaccurate and unmaintained docs.

* commit 'b1101e847312da8ee98a26e25058f93ba381f1ac':
  Removing inaccurate and unmaintained docs.

13 years agoam 486ac3c0: Removing inaccurate and unmaintained docs.
Dan Bornstein [Thu, 24 Mar 2011 00:53:04 +0000 (17:53 -0700)]
am 486ac3c0: Removing inaccurate and unmaintained docs.

* commit '486ac3c0a90b05be5af4b82801ef325e83241057':
  Removing inaccurate and unmaintained docs.

13 years agoMerge "Use PrintWriter rather than PrintStream in dex analysis tools." into dalvik-dev
Jesse Wilson [Wed, 23 Mar 2011 23:46:04 +0000 (16:46 -0700)]
Merge "Use PrintWriter rather than PrintStream in dex analysis tools." into dalvik-dev

13 years agoUse PrintWriter rather than PrintStream in dex analysis tools.
Jesse Wilson [Wed, 23 Mar 2011 23:21:01 +0000 (16:21 -0700)]
Use PrintWriter rather than PrintStream in dex analysis tools.

Writers make it easier to create strings, useful when these
tools are used as building blocks for other tools.

Change-Id: I8eac49719b130551df216f42eaafb942203da0d0

13 years agoRemoving inaccurate and unmaintained docs.
Dan Bornstein [Wed, 23 Mar 2011 22:26:58 +0000 (15:26 -0700)]
Removing inaccurate and unmaintained docs.

Change-Id: Ifef37063e844e4eb9ccc8111a1f1263fe26bd517

13 years agoMerge "Kill off code, and leave a note." into dalvik-dev
Dan Bornstein [Wed, 23 Mar 2011 19:49:26 +0000 (12:49 -0700)]
Merge "Kill off code, and leave a note." into dalvik-dev

13 years agoKill off code, and leave a note.
Dan Bornstein [Wed, 23 Mar 2011 19:00:20 +0000 (12:00 -0700)]
Kill off code, and leave a note.

Based on discussion yesterday, the code in Copying.c was already
superfluous. I'm leaving dvmForEachRequiredReference(), except with a
note, just in case it ends up being needed. If by the time I'm done
with the next round of VM changes, it turns out not to be needed, then
I'll just go ahead and axe it.

Change-Id: Ib7cf06ba75851c7ffd98108c4c73cddee9259ddf

13 years agoInterpreter restructuring
buzbee [Wed, 2 Mar 2011 23:43:48 +0000 (15:43 -0800)]
Interpreter restructuring

This is a restructuring of the Dalvik ARM and x86 interpreters:

  o Combine the old portstd and portdbg interpreters into a single
    portable interpreter.
  o Add debug/profiling support to the fast (mterp) interpreters.
  o Delete old mechansim of switching between interpreters.  Now, once
    you choose an interpreter at startup, you stick with it.
  o Allow JIT to co-exist with profiling & debugging (necessary for
    first-class support of debugging with the JIT active).
  o Adds single-step capability to the fast assembly interpreters without
    slowing them down (and, in fact, measurably improves their performance).
  o Remove old "polling for safe point" mechanism.  Breakouts now achieved
    via modifying base of interpreter handler table.
  o Simplify interpeter control mechanism.
  o Allow thread-granularity control for profiling & debugging

The primary motivation behind this change was to improve the responsiveness
of debugging and profiling and to make it easier to add new debugging and
profiling capabilities in the future.  Instead of always bailing out to the
slow debug portable interpreter, we can now stay in the fast interpreter.

A nice side effect of the change is that the fast interpreters
got a healthy speed boost because we were able to replace the
polling safepoint check that involved a dozen or so instructions
with a single table-base reload.  When combined with the two earlier CLs
related to this restructuring, we show a 5.6% performance improvement
using libdvm_interp.so on the Checkers benchmark relative to Honeycomb.

Change-Id: I8d37e866b3618def4e582fc73f1cf69ffe428f3c

13 years agoMerge "Fix some exception issues" into dalvik-dev
Andy McFadden [Wed, 23 Mar 2011 14:45:33 +0000 (07:45 -0700)]
Merge "Fix some exception issues" into dalvik-dev

13 years agoMerge "Get rid of some inconsistent formatting." into dalvik-dev
Carl Shapiro [Wed, 23 Mar 2011 02:44:12 +0000 (19:44 -0700)]
Merge "Get rid of some inconsistent formatting." into dalvik-dev

13 years agoGet rid of some inconsistent formatting.
Carl Shapiro [Wed, 23 Mar 2011 01:19:05 +0000 (18:19 -0700)]
Get rid of some inconsistent formatting.

Change-Id: I5d5130702e17bb95c835b9ccc3ec0aa86eda9dea

13 years agoLeft a note for future editors.
Dan Bornstein [Tue, 22 Mar 2011 23:49:41 +0000 (16:49 -0700)]
Left a note for future editors.

Change-Id: Ic96f58baabdf0d1a694b51efc991de5caef66a38

13 years agoPull the globals iteration into InitRefs.
Dan Bornstein [Tue, 22 Mar 2011 21:47:44 +0000 (14:47 -0700)]
Pull the globals iteration into InitRefs.

This gives the new GC a way to iterate over the global object refs
defined in Globals.h, without having to have detailed knowledge of the
list of such refs. However, this change doesn't actually modify the GC
code in question, since it is not currently in a working state anyway.

InitRefs.c probably wants a slightly different name. Suggestions welcome.

Bonus: Cleaned up a vaguely-related clause in Class.c.

Change-Id: I406987c56ddf310c61d877574071c46f3717fc72

13 years agoFix some exception issues
Andy McFadden [Tue, 22 Mar 2011 23:17:09 +0000 (16:17 -0700)]
Fix some exception issues

The function that obtained an exception's message string was making a
bad assumption.  This has been corrected.

Also, in the invoke-object-init functions, we now test for a pending
exception after the call to dvmSetFinalizable().

Also, make the test for pending exception at the end of VM startup
an error rather than an assert.

Bug 4121213

Change-Id: I6912e5c79d63e8dda1a1dc2e788c7a8edcf487aa

13 years agoMerge "Fix dalvik test 089" into dalvik-dev
Andy McFadden [Tue, 22 Mar 2011 22:10:10 +0000 (15:10 -0700)]
Merge "Fix dalvik test 089" into dalvik-dev

13 years agoFix dalvik test 089
Andy McFadden [Tue, 22 Mar 2011 22:04:52 +0000 (15:04 -0700)]
Fix dalvik test 089

The local variable used to hold the field reference was a u2.

Also, removed the "don't include debug info" options from the build
script.

Change-Id: I14b0fd4ae5af8dcd52f4b22e6b8fe9f68fd7bc89

13 years agoRemove more dead SecurityManager cruft.
Elliott Hughes [Tue, 22 Mar 2011 20:34:40 +0000 (13:34 -0700)]
Remove more dead SecurityManager cruft.

(Requires corresponding libcore change.)

Change-Id: I86aac9dda6708173fd2b5a1c660ef20ea11fca03