OSDN Git Service
buzbee [Fri, 18 Feb 2011 19:13:56 +0000 (11:13 -0800)]
[JIT] Fix profile trace dump
The debugging profile mode prints out a list of the top ten traces,
followed by recompilations. In some cases, it is possible that a trace
was requested, but did not finish compiling before the run ended. If
so, that could cause the dump to fail. This CL adds a check for null
codeAddress to detect those cases.
Change-Id: I415fd94d8fa9e270f75d5114fa5cc5d993bd6997
Elliott Hughes [Thu, 17 Feb 2011 00:12:03 +0000 (16:12 -0800)]
Merge "Ensure we always call inline natives." into dalvik-dev
Ben Cheng [Thu, 17 Feb 2011 00:02:33 +0000 (16:02 -0800)]
Merge "Add support to do suspend polling on backward branches in JIT'ed code." into dalvik-dev
jeffhao [Wed, 16 Feb 2011 23:55:26 +0000 (15:55 -0800)]
Merge "Fix for complex jsr nesting causing NullPointerException." into dalvik-dev
Ben Cheng [Thu, 3 Feb 2011 22:02:06 +0000 (14:02 -0800)]
Add support to do suspend polling on backward branches in JIT'ed code.
The polling is expensive for now as it is done through three
instructions: ld/ld/branch. As a result, a bunch of bonus stuff has
been worked on to mitigate the extra overhead:
- Cleaned up resource flags for memory disambiguation.
- Rewrote load/store elimination and scheduler routines to hide
the ld/ld latency for GC flag. Seperate the dependency checking into
memory disambiguation part and resource conflict part.
- Allowed code motion for Dalvik/constant/non-aliasing loads to be
hoisted above branches for null/range checks.
- Created extended basic blocks following goto instructions so that
longer instruction streams can be optimized as a whole.
Without the bonus stuff, the performance dropped about ~5-10% on some
benchmarks because of the lack of headroom to hide the polling latency
in tight loops. With the bonus stuff, the performance delta is between
+/-5% with polling code generated. With the bonus stuff but disabling
polling, the new bonus stuff provides consistent performance
improvements:
CaffeineMark 3.6%
Linpack 11.1%
Scimark 9.7%
Sieve 33.0%
Checkers 6.0%
As a result, GC polling is disabled by default but can be turned on
through the -Xjitsuspendpoll flag for experimental purposes.
Change-Id: Ia81fc85de3e2b70e6cc93bc37c2b845892003cdb
Elliott Hughes [Wed, 16 Feb 2011 22:30:45 +0000 (14:30 -0800)]
Ensure we always call inline natives.
Even though execute-inline is now a mandatory optimization, you can't be sure
the inline natives will be invoked that way. There's reflection and JNI, for
example, and there's the special case of String.equals that might be invoked
as Object.equals. This patch adds a regular native method corresponding to
each inline native, so that a corresponding libcore patch can drop its
implementations. (For example, despite the fact that we all believed last week
that the Java implementation of String.equals is never used, that turned out
not to be true: every HashMap lookup will have used it. This pair of patches
brings reality in line with our existing belief.)
Change-Id: I19e64c23bea83e91696206ca40ce4e3faf853040
Kenny Root [Wed, 16 Feb 2011 20:46:00 +0000 (12:46 -0800)]
am
ce0503dc: Clean up use of HAVE_ANDROID_OS
* commit '
ce0503dc677442f37b04277d888599d2cdb50ad6':
Clean up use of HAVE_ANDROID_OS
Jesse Wilson [Wed, 16 Feb 2011 19:29:32 +0000 (11:29 -0800)]
Merge "Emit compact ouput .dex files." into dalvik-dev
Jesse Wilson [Wed, 16 Feb 2011 02:28:53 +0000 (18:28 -0800)]
Emit compact ouput .dex files.
Change-Id: I69ca23b53a542db7e7a18d819795e795bf0822c0
http://b/
3447216
Kenny Root [Wed, 16 Feb 2011 18:15:52 +0000 (10:15 -0800)]
Clean up use of HAVE_ANDROID_OS
HAVE_ANDROID_OS was defined as "1" for targets, but never defined as "0"
for non-targets. Changing them to #ifdef should be safe and matches all
the other uses of HAVE_ANDROID_OS throughout the system.
Change-Id: I077aee39b08ce82327a571d9aca0cae422a0ebb5
Andy McFadden [Wed, 16 Feb 2011 15:54:43 +0000 (07:54 -0800)]
Merge "Barrier after construction of finalizable object" into dalvik-dev
Andy McFadden [Wed, 16 Feb 2011 15:54:17 +0000 (07:54 -0800)]
Merge "Fix some asm .size directives" into dalvik-dev
Andy McFadden [Tue, 15 Feb 2011 23:46:36 +0000 (15:46 -0800)]
Fix some asm .size directives
We were missing a .size directive for dvmPlatformInvoke, and the
directive for the mterp handlers wasn't being handled right. Threw
in a bonus directive for the entry point and the "assist debugger"
stuff that wraps method calls.
Bug
3456786
Change-Id: Ideee64a496e54eb09008410e9e9eba652b59f403
Carl Shapiro [Tue, 15 Feb 2011 23:37:02 +0000 (15:37 -0800)]
Merge "Use memchr to scan the card table." into dalvik-dev
Carl Shapiro [Tue, 15 Feb 2011 23:19:13 +0000 (15:19 -0800)]
Use memchr to scan the card table.
Change-Id: I8066b2818d360d3f383d1b90bbbea0b2983d4519
Andy McFadden [Tue, 15 Feb 2011 15:48:05 +0000 (07:48 -0800)]
Barrier after construction of finalizable object
If thread 1 creates an object, and thread 2 executes the finalizer,
then thread 2 must be guaranteed to see a fully-constructed object.
To this end, we need to emit a store/store barrier before returning
from the object's constructor.
We already do this for objects with final fields, so this is a
minor tweak.
While working on this I noticed that Enum overrides finalize()
(for the explicit purpose of making sure that no subclass can do
so), which the VM was taking as a signal that all enumerated types
are finalizable. In practice this doesn't matter, since the only
instances are the enum elements themselves, and we'd only GC them
if the enum class itself went away. However, setting it correctly
means less work for dexopt and has no measurable impact on class
loading time.
Bug
3403518
Change-Id: Ifa890b5e7ef1cda2a554ba54f25c4148272c3537
jeffhao [Tue, 15 Feb 2011 22:41:45 +0000 (14:41 -0800)]
Fix for complex jsr nesting causing NullPointerException.
Happens when one jsr calls another that ultimates throws an exception.
The subroutine inliner assumes the return block of the first jsr is
reachable when it's not, causing access to a null field. Fixed by
checking the field for null before accessing it.
Change-Id: Id1fb376c9f14ffebc77cdbd253a713eb6d949c1f
Jesse Wilson [Tue, 15 Feb 2011 22:21:05 +0000 (14:21 -0800)]
Deduplicate type lists on merged .dex files.
Change-Id: Ic089fecebb66544dfdb422421cce47a6f203a31a
http://b/
3447216
Jesse Wilson [Tue, 15 Feb 2011 19:10:13 +0000 (11:10 -0800)]
Merge "Fix some bugs in the new dx instruction code." into dalvik-dev
Jesse Wilson [Tue, 15 Feb 2011 03:01:34 +0000 (19:01 -0800)]
Fix some bugs in the new dx instruction code.
http://b/
3447216
Change-Id: I9a77541b994f184da0e389840e5cac728ad6c072
buzbee [Tue, 15 Feb 2011 18:40:49 +0000 (10:40 -0800)]
Merge "Remove spurious code from bad merge/pilot error" into dalvik-dev
buzbee [Tue, 15 Feb 2011 18:29:38 +0000 (10:29 -0800)]
Remove spurious code from bad merge/pilot error
Looks like I spuriously duplicated two lines of assembly code in
footer.S around a call to method profiling. Removed.
Change-Id: I0b10656e15eba0d16af8784ae9ca09c33b5096c0
Andy McFadden [Mon, 14 Feb 2011 22:44:35 +0000 (14:44 -0800)]
resolved conflicts for merge of
6e2af6d2 to dalvik-dev
Change-Id: Ib5ec0912c0c5d7041b794351fcca10c41c4294ec
Andy McFadden [Mon, 14 Feb 2011 22:04:42 +0000 (14:04 -0800)]
Fix "all stubs" interpreter
The "all stubs" interpreter is the recommended starting point for
anyone creating a "fast" interpreter on a new CPU. This cleans
up some minor bit rot in the code and documentation.
Bug
3452689
Change-Id: I6bc3b0b5da11955d842d5fc5f16cb1a2209d4a89
Andy McFadden [Mon, 14 Feb 2011 21:02:19 +0000 (13:02 -0800)]
Merge "Remove gclog.py" into dalvik-dev
Andy McFadden [Mon, 14 Feb 2011 20:29:38 +0000 (12:29 -0800)]
Remove gclog.py
The GC event log entries are no longer generated, so we no longer
need this.
Change-Id: I247bbc44d628ac5e54dfe031d09eb14edbff8d4d
Ben Cheng [Mon, 14 Feb 2011 18:02:04 +0000 (10:02 -0800)]
Merge "Add scheduling barrier at the end of IT blocks for FP comparisons." into dalvik-dev
jeffhao [Fri, 11 Feb 2011 23:43:07 +0000 (15:43 -0800)]
Fix to allow SCCP to correctly propagate division and remainer ops.
Also updated expected values for the const collector test.
Change-Id: Iedcf17d776c60cb6174f7a7c9f75be84d2c38020
Ben Cheng [Sat, 12 Feb 2011 01:01:08 +0000 (17:01 -0800)]
Add scheduling barrier at the end of IT blocks for FP comparisons.
Bug:
3448446
Change-Id: I98e2bbc4886443ba3c27c2963d7540fcee5790bb
Andy McFadden [Sat, 12 Feb 2011 00:00:54 +0000 (16:00 -0800)]
Merge "Rename invoke-direct-empty to invoke-object-init" into dalvik-dev
Andy McFadden [Fri, 11 Feb 2011 23:26:10 +0000 (15:26 -0800)]
Rename invoke-direct-empty to invoke-object-init
The invoke-direct-empty instruction was introduced to remove the
overhead of calling the empty Object constructor. We now need it
to do some extra work on behalf of object construction, so it's
appropriate to change the instruction name to match the role it
fills rather than the more general role it was hoped to fill.
No functional changes.
Bug
3342343
Change-Id: I65dd6a2c00c99581c9a19b16fe193b70642c8fbb
Carl Shapiro [Fri, 11 Feb 2011 00:46:55 +0000 (16:46 -0800)]
Fast fail impossible allocations.
If the size_t type cannot be used to represent any component of an
array size the allocation is be rejected. With this change, overflow
is checked for in the element size computation and total array size
computation. More preconditions are checked by asserts. Misleading
comments have been removed.
Change-Id: I067869c3404b9da591939555c6f9904f52ca7bd7
Carl Shapiro [Fri, 11 Feb 2011 20:06:33 +0000 (12:06 -0800)]
Ensure a dalvik-cache directory exists before running dalvik.
The dalvik tests require a dalvik-cache directory otherwise dalvikvm
will fail with an unhelpful error message. On the device it is safe
to assume a dalvik-cache directory exists but this is not a safe
assumption on the host. With this change the script tries to create a
dalvik-cache directory on the host if it does not exist.
Change-Id: Ie0c69ff8dcd68b6b04a763f06861edd308406f69
Elliott Hughes [Thu, 10 Feb 2011 23:47:58 +0000 (15:47 -0800)]
Merge "Fix an out-of-date comment." into dalvik-dev
Andy McFadden [Thu, 10 Feb 2011 23:47:57 +0000 (15:47 -0800)]
Merge "Tweak the 64-bit load/store code" into dalvik-dev
Andy McFadden [Thu, 10 Feb 2011 23:32:16 +0000 (15:32 -0800)]
Tweak the 64-bit load/store code
The old version would either move data through a union, call memcpy(),
or just cast the pointer to a wider type and dereference it. The
latter turns out to be an aliasing violation, and it recently started
causing test failures on the "host" build.
We now use memcpy() for x86 and unions for ARM.
Bug
3431820
Change-Id: I302a7f49f7ae88ac96b8f7fef3d9260ac64d631b
Elliott Hughes [Thu, 10 Feb 2011 23:25:04 +0000 (15:25 -0800)]
Fix an out-of-date comment.
Change-Id: Ie3a93604533e9dab59e892506e2c0258d046837b
Elliott Hughes [Thu, 10 Feb 2011 22:21:44 +0000 (14:21 -0800)]
Merge "Move more system propery handling into managed code." into dalvik-dev
Elliott Hughes [Thu, 10 Feb 2011 20:03:34 +0000 (12:03 -0800)]
Move more system propery handling into managed code.
Bug:
3413364
Change-Id: I044ed1b11e919bb1c589c626b613faf85157fb64
jeffhao [Thu, 10 Feb 2011 21:02:51 +0000 (13:02 -0800)]
Merge "Reverting SCCP change to division because the system doesn't start." into dalvik-dev
jeffhao [Thu, 10 Feb 2011 20:46:14 +0000 (12:46 -0800)]
Reverting SCCP change to division because the system doesn't start.
Temporarily reverting change to allow the system to boot while the
root cause is determined.
Change-Id: Ife1fb6fc99fead9eb72fff8415969fc4955ff7cd
jeffhao [Thu, 10 Feb 2011 17:52:00 +0000 (09:52 -0800)]
Merge "Added support for integer division to SCCP." into dalvik-dev
jeffhao [Thu, 10 Feb 2011 02:29:00 +0000 (18:29 -0800)]
Added support for integer division to SCCP.
Division generates a result that must be fetched with a
move-result-pseudo instruction. SCCP originally had no way of tying the
division with the result it generates. This change should allow proper
constant propagation when division of integers is involved.
Change-Id: Ib7c5d2dd26eea3ab6545b613a540f0161a8e1642
Elliott Hughes [Thu, 10 Feb 2011 01:29:30 +0000 (17:29 -0800)]
Merge "Move more system property handling into managed code." into dalvik-dev
Elliott Hughes [Thu, 10 Feb 2011 00:52:36 +0000 (16:52 -0800)]
Move more system property handling into managed code.
Bug:
3413364
Change-Id: I0f22ea9ed5327e4fef9d4395025173a839ea9ce2
Carl Shapiro [Thu, 10 Feb 2011 00:41:57 +0000 (16:41 -0800)]
Merge "Split addNewHeap into separate definitions." into dalvik-dev
Carl Shapiro [Thu, 10 Feb 2011 00:36:12 +0000 (16:36 -0800)]
Split addNewHeap into separate definitions.
Change-Id: I47c4edb8dddf1e8c256ff2def7045fa236b5b1fd
Carl Shapiro [Wed, 9 Feb 2011 23:35:43 +0000 (15:35 -0800)]
Merge "Remove the unused argument to dvmCollectGarbage." into dalvik-dev
jeffhao [Wed, 9 Feb 2011 22:44:40 +0000 (14:44 -0800)]
Merge "SCCP propagates more constants and prunes unexecuted paths from branches." into dalvik-dev
jeffhao [Tue, 8 Feb 2011 01:48:13 +0000 (17:48 -0800)]
SCCP propagates more constants and prunes unexecuted paths from branches.
The SCCP pass now more aggressively propagates constant values through
the code. Combined with changes to the LiteralOpUpgrader, instructions
with known constant results will be replaced with a simple const
instruction.
In addition, the SCCP pass can now find branches with constant conditions
and remove the branch path that never ends up being executed. Because of
the way finally blocks end up being replicated, this tends to prune away
error handling when no exception occurs, and hard code error handling
when an exception happens.
Change-Id: I6f3330151ec387c8a1e7ce098ff6cdb8d0ce5606
Carl Shapiro [Wed, 9 Feb 2011 22:20:14 +0000 (14:20 -0800)]
Remove the unused argument to dvmCollectGarbage.
Change-Id: Id7bf60b99b33cb8f058f9a500dc63cfe01b8e3c2
Carl Shapiro [Wed, 9 Feb 2011 03:54:55 +0000 (19:54 -0800)]
Remove the private CLZ implementation in the GC.
This was a hold over from a time when the compiler did not properly
support the CLZ intrinsic.
Change-Id: I224481868c8d6fd3e0382d9cde3a2ffd02cfd39c
Carl Shapiro [Wed, 9 Feb 2011 21:30:40 +0000 (13:30 -0800)]
Merge "Move the duplicated alignment macros to a common location." into dalvik-dev
Carl Shapiro [Wed, 9 Feb 2011 21:30:24 +0000 (13:30 -0800)]
Merge "Replace dvmValidateObject with ordinary null pointer checks." into dalvik-dev
Carl Shapiro [Wed, 9 Feb 2011 21:28:42 +0000 (13:28 -0800)]
Merge "Let tryMalloc field over-sized allocation errors." into dalvik-dev
Andy McFadden [Wed, 9 Feb 2011 21:21:26 +0000 (13:21 -0800)]
Merge "Add huge-array OOM test" into dalvik-dev
Andy McFadden [Wed, 9 Feb 2011 21:12:36 +0000 (13:12 -0800)]
Add huge-array OOM test
Added a regression test for allocation of 4GB+ arrays.
Change-Id: Ibac82bbba9ef25b0850386a35e0f5b8554abbaa3
Ben Cheng [Wed, 9 Feb 2011 18:04:12 +0000 (10:04 -0800)]
Merge "Misc goodies in the JIT in preparation for more aggressive code motion." into dalvik-dev
Andy McFadden [Wed, 9 Feb 2011 16:24:13 +0000 (08:24 -0800)]
am
6f562a3a: Merge "Make Release(JNI_COMMIT) faster"
* commit '
6f562a3aee2b866d1f8a8a5773d81e0593069b3e':
Make Release(JNI_COMMIT) faster
Andy McFadden [Wed, 9 Feb 2011 16:24:07 +0000 (08:24 -0800)]
am
81abf28f: Remove JNI_FORCE_C
* commit '
81abf28f00ba97fdf6b085d1d91a5e73f9a1e799':
Remove JNI_FORCE_C
Andy McFadden [Wed, 9 Feb 2011 16:10:49 +0000 (08:10 -0800)]
Merge "Make Release(JNI_COMMIT) faster"
Carl Shapiro [Wed, 9 Feb 2011 04:41:28 +0000 (20:41 -0800)]
Let tryMalloc field over-sized allocation errors.
Change-Id: I691b8eeca74b7e018f86753b00b7f117d5872916
Carl Shapiro [Wed, 9 Feb 2011 04:19:17 +0000 (20:19 -0800)]
Replace dvmValidateObject with ordinary null pointer checks.
Change-Id: I8eddca7fac83e014d591bdd06ce2df634672243f
Carl Shapiro [Wed, 9 Feb 2011 04:10:56 +0000 (20:10 -0800)]
Move the duplicated alignment macros to a common location.
Change-Id: I11cb1c3034a3a4e7d8e8c0793f5c85fa623155b2
Carl Shapiro [Wed, 9 Feb 2011 02:16:51 +0000 (18:16 -0800)]
Merge "Relocate the always in-lined functions from HeapBitmap.h." into dalvik-dev
Carl Shapiro [Wed, 9 Feb 2011 01:43:43 +0000 (17:43 -0800)]
Relocate the always in-lined functions from HeapBitmap.h.
The three essential bitmap access methods have been moved to the new
HeapBitmapInlines.h header file. All other bitmap methods have been
moved into HeapBitmap.c. As an added bonus, source indexing and cross
reference tools should correctly locate all heap bitmap definitions.
Change-Id: I35703403f4f79d121c9c3f411b7d7110d92739d0
Ben Cheng [Wed, 9 Feb 2011 01:09:25 +0000 (17:09 -0800)]
Misc goodies in the JIT in preparation for more aggressive code motion.
- Set up resource masks correctly for Thumb push/pop when LR/PC are involved.
- Preserve LR around simulated heap references under self-verification mode.
- Compact a few simple flags in ArmLIR into bit fields.
- Minor performance tuning in TEMPLATE_MEM_OP_DECODE
Change-Id: Id73edac837c5bb37dfd21f372d6fa21c238cf42a
Andy McFadden [Wed, 9 Feb 2011 00:24:34 +0000 (16:24 -0800)]
Make Release(JNI_COMMIT) faster
The Release<primitive>ArrayElements and ReleasePrimitiveArrayCritical
calls take a "mode" argument. If you pass in JNI_COMMIT, the storage
isn't actually released; instead, the buffer contents are flushed to
the backing storage. If the VM has handed you a pointer to the raw
data rather than a copy, the call is a no-op.
In its present incarnation, it's a no-op that changes the thread state,
adding overhead and the possibility of suspension. This change moves
the test for JNI_COMMIT outside the enter/exit macros.
Bug
3430203
Change-Id: I8426647ee02b2336c5c90146028fc207cfbacc3a
Andy McFadden [Wed, 9 Feb 2011 00:12:33 +0000 (16:12 -0800)]
Remove JNI_FORCE_C
This hasn't been useful in a really long time.
Change-Id: Ieea01f5d051bc8fc671f7414af3bed626ece75f1
Andy McFadden [Wed, 9 Feb 2011 00:05:45 +0000 (16:05 -0800)]
am
1fbcc3ad: Merge "Added JNI hack to support JNI hack"
* commit '
1fbcc3ad1344f6997f34db0ab8296b0867c84b71':
Added JNI hack to support JNI hack
Andy McFadden [Wed, 9 Feb 2011 00:03:05 +0000 (16:03 -0800)]
Merge "Added JNI hack to support JNI hack"
Andy McFadden [Tue, 8 Feb 2011 20:07:32 +0000 (12:07 -0800)]
Added JNI hack to support JNI hack
Some of our graphics code wants to get a pointer, release it, and then
continue to use it. The "forcecopy" mode of CheckJNI was designed
to find any code that does this. It succeeded.
To support the behavior, we provide a JNI helper function that does
the dirty work. It passes a magic value into the Get and Release calls
that causes "forcecopy" to skip the copy. When forcecopy is not
enabled, the values are simply ignored.
To avoid any possibility of the function getting published in the NDK,
the function is not described in JNIHelp.h.
Bug
3409356
Change-Id: Ibd20d12ba6d3d3236ebf5760f7ccaa8c557e3774
Kenny Root [Tue, 8 Feb 2011 20:07:50 +0000 (12:07 -0800)]
am
e3f9c573: Free some strings in error paths
* commit '
e3f9c573d0edb186f4583fea1dbcdf9c12c0c36c':
Free some strings in error paths
Carl Shapiro [Tue, 8 Feb 2011 19:45:18 +0000 (11:45 -0800)]
Merge "Remove the deprecated gcSoftReferences method." into dalvik-dev
Kenny Root [Wed, 2 Feb 2011 23:07:00 +0000 (15:07 -0800)]
Free some strings in error paths
Change-Id: I54629ca65c30960f930ce6886e0915e360e20237
Carl Shapiro [Tue, 8 Feb 2011 04:38:38 +0000 (20:38 -0800)]
Remove the deprecated gcSoftReferences method.
This code was used by the zygote to force a garbage collection that
never preserves SoftReference referents. With this change, the
SoftReference processor defaults to never preserve when running in
zygote mode.
Change-Id: I90edadd05b732badc401baead6986d37c71a5afc
Andy McFadden [Mon, 7 Feb 2011 23:56:12 +0000 (15:56 -0800)]
am
85019f27: Avoid double-free on error.
* commit '
85019f27b4140f86450948f3486a64743d1f8671':
Avoid double-free on error.
Andy McFadden [Mon, 7 Feb 2011 23:49:13 +0000 (15:49 -0800)]
Avoid double-free on error.
In certain circumstances on a -user build we could end up trying to
free a buffer twice.
Bug
3431477
Change-Id: I1cce1da0e32aabc85d5c5c69346224c6db49b8ee
jeffhao [Mon, 7 Feb 2011 21:58:20 +0000 (13:58 -0800)]
Updated expected outputs for some dx tests.
Recent changes in code generation caused the generated output to no
longer match the expected results of these tests. They've now been
updated properly.
Change-Id: I223075e2b1ff297d37022d77de74304b084ca900
Dan Bornstein [Mon, 7 Feb 2011 20:22:18 +0000 (12:22 -0800)]
Fix expected output.
The expected output of these tests changed due to incidental (to the
tests) changes in dx, in particular renaming the payload opcodes and
simplifying some error output.
Change-Id: Ia16cd484202a86b9ddab1ae4794a384b2ec53d9a
Dan Bornstein [Mon, 7 Feb 2011 20:06:54 +0000 (12:06 -0800)]
Fix dex processing when no output file is specified.
This case is mainly exercised in test code.
Change-Id: Ifbcbec3f001cae925090796ce8ed69f6a986384c
Dan Bornstein [Mon, 7 Feb 2011 19:31:24 +0000 (11:31 -0800)]
Merge "Move DexException to dx.util." into dalvik-dev
Brian Carlstrom [Sun, 6 Feb 2011 01:10:56 +0000 (17:10 -0800)]
Merge commit '
e502b66edd734e8265bc2bd8f1c2addb64b63e73' into honeycomb-release-to-dalvik-dev
The Android Automerger [Sat, 5 Feb 2011 06:25:58 +0000 (22:25 -0800)]
merge in honeycomb-release history after reset to honeycomb
The Android Automerger [Sat, 5 Feb 2011 05:55:43 +0000 (21:55 -0800)]
merge in honeycomb-release history after reset to honeycomb
Dan Bornstein [Sat, 5 Feb 2011 01:59:14 +0000 (17:59 -0800)]
Move DexException to dx.util.
Change-Id: I86bf83bc3adcc55eb8bbeac2fc28822cb5f2d2b9
Carl Shapiro [Sat, 5 Feb 2011 01:36:49 +0000 (17:36 -0800)]
Remove unused and racey debugging code.
Change-Id: I335a9304b7ffdcba8062cd33ce1a0d721f345a73
The Android Automerger [Sat, 5 Feb 2011 01:35:37 +0000 (17:35 -0800)]
merge in honeycomb-release history after reset to honeycomb
Carl Shapiro [Sat, 5 Feb 2011 01:29:40 +0000 (17:29 -0800)]
Remove and outdated comment and logging macro.
Change-Id: I39fe34f69276d2e3d26d5d296cda4e7240f4d845
Andy McFadden [Sat, 5 Feb 2011 01:08:41 +0000 (17:08 -0800)]
resolved conflicts for merge of
d6e32b81 to dalvik-dev
Change-Id: Ib038cc0e68651c63dbcf9a0df2a5607ab2c2d6c7
Andy McFadden [Thu, 3 Feb 2011 00:55:41 +0000 (16:55 -0800)]
Ensure word atomicity in System.arraycopy
The VM needs to ensure that updates to 16-bit and 32-bit fields
and array elements happen atomically. System.arraycopy was calling
memmmove(), which in bionic happens to copy a byte at a time.
The new plan is to have bionic provide a platform-optimized memmove
variant that makes guarantees about word sizes. If it is not present
(either because it hasn't been written for the current platform,
or we're not linking against bionic), we will use a trivial copy loop.
We assume that the memmove-by-word implementation does whatever
it needs to for best performance. For example, if the function
determines that memcpy() is word-safe and will be faster, it can
call that. The VM no longer makes assumptions about the internal
workings of libc functions.
I also rearranged the code to avoid making indirect calls, reading
function addresses from memory, and using integer multiplication
instructions when a simple shift will do. (gcc now generates the
whole thing as one function.)
Also, added some primitive array copy tests to 011-array-copy.
Bug
3398352
Change-Id: I4677ee1c87998697a93e61f39a48b3f391e8e11b
The Android Automerger [Fri, 4 Feb 2011 19:49:07 +0000 (11:49 -0800)]
merge in honeycomb-release history after reset to honeycomb
Dan Bornstein [Fri, 4 Feb 2011 18:44:08 +0000 (10:44 -0800)]
Move the instruction representation code.
It now gets its own subpackage, reducing the clutter in the base .io
package. I was on the fence about CodeReader and ended up just leaving
it be, but maybe that one wants to move too.
Change-Id: I87c574ce2086025bcb9583627a545160084ba88e
The Android Automerger [Fri, 4 Feb 2011 18:26:04 +0000 (10:26 -0800)]
merge in honeycomb-release history after reset to honeycomb
Dan Bornstein [Fri, 4 Feb 2011 17:45:23 +0000 (09:45 -0800)]
Merge "Clean up targets and data payloads." into dalvik-dev
The Android Automerger [Fri, 4 Feb 2011 15:19:01 +0000 (07:19 -0800)]
merge in honeycomb-release history after reset to honeycomb
Brian Carlstrom [Fri, 4 Feb 2011 05:47:30 +0000 (21:47 -0800)]
Merge commit '
e6cad708e93dab280335d3f69917048c34cfc640' into honeycomb-release-to-dalvik-dev
Dan Bornstein [Fri, 4 Feb 2011 01:05:28 +0000 (17:05 -0800)]
Clean up targets and data payloads.
In particular:
* Fix target calculation to be properly relative when encoding.
* Add data payload variants of DecodedInstruction, and en/decode using them.
* Remove the uninterpreted data member from the base DecodedInstruction.
This change introduces the concept of an "address map" that is built
up while reading or writing code, which is used when en/decoding
switch data payload instructions, so that they can decode targets into
absolute addresses and encode targets into relative addresses. The
deal is that the relative addresses encoded in these payload
instructions are relative to the address of the dependant switch
instruction and not the address of the payload instruction.
Change-Id: I43cb014f1af75a31bbf0b757e4854a9dc865eab7
The Android Automerger [Thu, 3 Feb 2011 23:26:36 +0000 (15:26 -0800)]
merge in honeycomb-release history after reset to honeycomb
Dan Bornstein [Thu, 3 Feb 2011 23:18:24 +0000 (15:18 -0800)]
Merge "Make some "argument shape" be represented by subclassing." into dalvik-dev