OSDN Git Service
David Brazdil [Fri, 26 Jun 2015 15:58:14 +0000 (16:58 +0100)]
ART: Bail out immediately on try/catch
Optimizing builds the graph even if it knows it will delegate to
Quick. This patch moves the decision at the beginning of the
TryCompile function to save a little bit of compile time.
Bug:
22115561
Change-Id: I18b5a686e10592080ea3c49d53a0dd81aa3b3010
Andreas Gampe [Fri, 26 Jun 2015 03:16:10 +0000 (03:16 +0000)]
Merge "ART: Fix kEverything compiler filter" into mnc-dev
Andreas Gampe [Tue, 19 May 2015 06:22:54 +0000 (23:22 -0700)]
ART: Only print stripped dex2oat command line
To curb logcat noise, strip many dex2oat parameters from the
logcat printout.
Bug:
20501758
(cherry picked from commit
046c706707ea0f16c804136e237ac7cbfdc897a1)
Change-Id: Ifc367f91f593916e0773af1ca950c798f129889f
buzbee [Thu, 25 Jun 2015 22:53:45 +0000 (15:53 -0700)]
ART: Fix kEverything compiler filter
Previously the kEverything filter failed to compile class
initializers. Now it will.
Bug:
19576257
(cherry picked from commit
c83329952b4a313e747c8835a73699e2cae5a6e2)
Change-Id: I189d2b5b379aee112c4e95f8d3e6c32abab6ed41
Andreas Gampe [Thu, 25 Jun 2015 17:26:40 +0000 (10:26 -0700)]
ART: Reset runtime_throw_failure flag
The flag is instruction-specific. It transports the info from Fail
to the main loop. It must be cleared after each iteration.
Introduce a second flag to store whether we saw such a failure at
all.
Update test expectations.
Bug:
22080519
(cherry picked from commit
d12e782bcee03ecb6dec41aa9673ef53b638dcea)
Change-Id: I32be914819946233babaa4cb7343844d97b61ba5
Andreas Gampe [Wed, 24 Jun 2015 23:25:33 +0000 (16:25 -0700)]
ART: Set methods to preverified in verify-none
Actually mark all methods to be verified, so that the interpreter
doesn't do access checks. Speedup is about 2x on a Nexus 6 when booting
in verify-none mode.
Bug:
22040916
(cherry picked from commit
1c284f2abdfe1b0ed87d5216021e69146f10b116)
Change-Id: I2d6c8df2a68eae6e358da2639d980cb8bfcc1560
Richard Uhler [Thu, 25 Jun 2015 15:21:12 +0000 (15:21 +0000)]
Merge "Do not try to compile resource-only dex files." into mnc-dev
Vladimir Marko [Tue, 23 Jun 2015 16:45:21 +0000 (17:45 +0100)]
Hard-fail get-/put-object to a non-reference field.
Bug:
21886894
(cherry picked from commit
414000ec4d728b5c85f8c6dee4f867fecde59b01)
Change-Id: Iafc32f0e45d26f3aaa2d521b98353e7cede16c6f
Nicolas Geoffray [Thu, 25 Jun 2015 09:01:47 +0000 (10:01 +0100)]
Only do some checks when compiling against the core image.
This will avoid false negatives when running dex2oatd on apks.
bug:
21865473
(cherry picked from commit
335005e2b3a179f26b7a8ae64ca60a1406b669bd)
Change-Id: Iac6dbe30c9d576077a0384b88696f79937d89471
Lazar Trsic [Wed, 24 Jun 2015 14:30:21 +0000 (16:30 +0200)]
[MIPS64] JNI Compiler: Sign-extend int function arguments
MIPS n64 ABI differs from arm64. Arguments smaller than the 8B stack
slot need to be sign-extended.
Use combination (lw,sd), instead of (lw,sw) for 4B values.
Change fixes software keyboard crash on mips64.
Bug:
21555893
(cherry picked from commit
f652d605753f1387e7797461b47116c5dcdf928d)
Change-Id: I7574d37f6039e9e8c9e0047254be71d28d4c829a
Andreas Gampe [Thu, 25 Jun 2015 04:12:06 +0000 (04:12 +0000)]
Merge "ART: Fix arm32 instrumentation exit stub" into mnc-dev
Mathieu Chartier [Thu, 25 Jun 2015 00:04:17 +0000 (17:04 -0700)]
Fix force copy
We now correctly pass the returned pointer back onto the release functions.
Bug:
22056708
Change-Id: I1a7300d3a4522a3c81b432ec742ae1c0bd00b51e
(cherry picked from commit
b735bd9c04aa291d0a1bdc2c0a094a1a75ad0596)
Andreas Gampe [Wed, 24 Jun 2015 06:28:59 +0000 (23:28 -0700)]
ART: Fix arm32 instrumentation exit stub
Properly store, pass and restore floating point result.
Bug:
22047833
Change-Id: If5d9efbe63334c9b6353f76ce6e4fe519602482a
Mathieu Chartier [Wed, 24 Jun 2015 20:25:25 +0000 (20:25 +0000)]
Merge "Fix imt conflict trampoline for mips64" into mnc-dev
Mathieu Chartier [Wed, 24 Jun 2015 20:11:06 +0000 (13:11 -0700)]
Fix imt conflict trampoline for mips64
Changed lwu to ld
Bug:
21815962
Change-Id: I137e871d9c00f83cd5efc9d3a87fa05e84dc5d23
Christopher Ferris [Wed, 24 Jun 2015 16:37:39 +0000 (09:37 -0700)]
Increase alt signal stack to 32K on host.
On host, when there is a crash, the code tries to do an unwind
from the signal handler. Under some conditions this can require
a stack larger than 16K, so increase this value to 32K.
The symptom of this is that you get an empty backtrace running
art tests when there is a crash.
Bug:
21949571
(cherry picked from commit
e34c71ce21dbc9fac3427df887e90e1311b6fa8e)
Change-Id: Ib1105a3470064d98843bd46f32d4fadb1d20e416
Jeff Hao [Wed, 24 Jun 2015 17:05:42 +0000 (17:05 +0000)]
Merge "Adding optimizing compiler test case for String.<init>." into mnc-dev
Nicolas Geoffray [Wed, 24 Jun 2015 14:53:03 +0000 (15:53 +0100)]
Use a flag from the verifier to know if we should compile.
Only used for the lack of bottom type in the aget-object case
for now. Could be used for more.
bug:
21865466
(cherry picked from commit
4824c27988c8eeb302791624bb3ce1d557b0db6c)
Change-Id: I2bb7fe1d4737bd92c1076b5193607d74d8761ee7
Nicolas Geoffray [Wed, 24 Jun 2015 13:57:44 +0000 (14:57 +0100)]
BCE: don't assume a bounds check always gets a HArrayLength.
Deoptimizations may change it to a HPhi.
bug:
22056703
(cherry picked from commit
8df886b9214802ad689316a1dedb00a6d102555c)
Change-Id: I8afcf88e3a12dbe4d87101e6a7cefb8b81e2bb96
Richard Uhler [Wed, 24 Jun 2015 15:44:19 +0000 (08:44 -0700)]
Do not try to compile resource-only dex files.
This changes behavior in the case where we are asked to load a dex
file that does not exist or has no classes.dex entry.
Previously we would run dex2oat, which would log an error message and
fail. Now we skip running dex2oat, we report the DexOptStatus as
kNoDexOptNeeded, and we do not try to fall back to the missing
original dex files.
Bug:
21722039
Change-Id: I90b1f4165138daac57bb9d7a354319005652c593
Nicolas Geoffray [Wed, 24 Jun 2015 13:25:09 +0000 (14:25 +0100)]
Fix codegen_test after DCE fix.
(cherry picked from commit
ea80942c39cb5e9bc3e38a388583d1646c1e2a3f)
bug:
22031382
Change-Id: If410045779adb9e7e611af22c988232b5bf9a902
Nicolas Geoffray [Wed, 24 Jun 2015 11:20:24 +0000 (12:20 +0100)]
Recompute dominator tree after DCE.
bug:
22031382
(cherry picked from commit
1f82ecc6a0c9f88d03d6d1a6d95eeb8707bd06c1)
Change-Id: I9a74edb185cb806045903dfe9695d9cc1a02e86b
Nicolas Geoffray [Wed, 24 Jun 2015 09:38:27 +0000 (10:38 +0100)]
BCE: Narrow instead of unconditionnaly overwrite the range.
bug:
21862741
(cherry picked from commit
a09ff9c11f07863ac57e6120a824f0d20dfaa284)
Change-Id: Ia8e903e09a7f9c2b8ef7cf3522f73f154534b81f
Nicolas Geoffray [Wed, 24 Jun 2015 11:41:20 +0000 (12:41 +0100)]
Fix another case of un-verified dead code.
bug:
22042796
https://code.google.com/p/android/issues/detail?id=178008
(cherry picked from commit
1efcc22cd1895c48adccbe49270d8e8583c2b12d)
Change-Id: I5c0d783e842da39cd3dcbb2f18ccf784e797a64f
Jeff Hao [Wed, 24 Jun 2015 00:34:04 +0000 (17:34 -0700)]
Adding optimizing compiler test case for String.<init>.
Bug:
21991156
Change-Id: Ibd0bd7d3b20680ac06ac08cb857595288c4e9b69
Andreas Gampe [Tue, 23 Jun 2015 23:15:18 +0000 (23:15 +0000)]
Andreas Gampe [Tue, 23 Jun 2015 21:18:17 +0000 (14:18 -0700)]
ART: Follow-up to
507cc6f83bf6379728f2dd20391f2ed5fbfe6371
Do not trigger DCHECK aborts for final and abstract classes when
they are used from other classes. Instead issue a soft error that
leads to a runtime failure.
Bug:
21873151
(cherry picked from commit
f23f33de8b3e9abd16720e0f48d43d63d41dee16)
Change-Id: Ie5251a176722811b3c74585222874313a53f7e02
Mathieu Chartier [Tue, 23 Jun 2015 22:54:48 +0000 (22:54 +0000)]
Merge "Fix field ordering for String in hprof" into mnc-dev
Mathieu Chartier [Tue, 23 Jun 2015 22:45:15 +0000 (15:45 -0700)]
Fix field ordering for String in hprof
We were inserting value field after the Object field instead of
the String fields.
Bug:
22043800
Change-Id: Ibaddbd7299562949e32eceb625feac325539e3ce
Jeff Hao [Tue, 23 Jun 2015 03:20:30 +0000 (20:20 -0700)]
Make dex file verifier check that indexes are increasing.
Note that this still allows duplicates.
Bug:
21868011
Change-Id: Ib0e91c816eccfaf3b0dac340c05ffd5dc4df17cc
Nicolas Geoffray [Mon, 22 Jun 2015 22:12:45 +0000 (23:12 +0100)]
Do not overwrite an input register in shift operations.
'second_reg' is an input register that can survive the instruction.
Instead use the output register as a temporary result.
bug:
21667432
(cherry picked from commit
a4f3581da73b83484a30ab499c4f8ad43b378dab)
Change-Id: Ic1f399964911b8a9fc57352130c92b2a0a1b8e0d
Mathieu Chartier [Tue, 23 Jun 2015 00:14:43 +0000 (00:14 +0000)]
Merge "Fix another miranda method moving GC bug" into mnc-dev
Mathieu Chartier [Sun, 14 Jun 2015 21:14:37 +0000 (14:14 -0700)]
New experimental GC stress mode
Tries to do a GC for every unique call stack (up to 16 frames).
The goal is to catch moving GC bugs and lock violations without being
rediculously slow. Some tests fail on 64 bits, 32 bit host doesn't
work. N5 is booting.
Added runtime -Xgc options: gcstress and nogcstress.
Bug:
21664466
(cherry picked from commit
310008008c90fea246efd00cb99ee7ded97c5209)
Change-Id: Icb8e420f2048e8ee83bcca7937563166a2638f5c
Jeff Hao [Mon, 22 Jun 2015 22:26:54 +0000 (22:26 +0000)]
Merge "Change GC to allow strings to use large object space." into mnc-dev
Jeff Hao [Mon, 22 Jun 2015 22:14:49 +0000 (15:14 -0700)]
Change GC to allow strings to use large object space.
Bug:
21209641
Change-Id: Ib007a30e575e391e2b5bddb816a9d2e82502c7d5
Andreas Gampe [Mon, 22 Jun 2015 21:36:40 +0000 (21:36 +0000)]
Merge "ART: Allow checking RETURN_VOID_NO_BARRIER on unresolved classes" into mnc-dev
Jeff Hao [Mon, 22 Jun 2015 21:29:54 +0000 (14:29 -0700)]
Make large string use large object space.
Bug:
21209641
Change-Id: I731a920c02f29f0321f2077093516aaa0f8d782f
Andreas Gampe [Mon, 22 Jun 2015 18:35:46 +0000 (11:35 -0700)]
ART: Allow checking RETURN_VOID_NO_BARRIER on unresolved classes
We must check that all fields are non-final, even for unresolved
classes. Fall back to the dex file for this information.
Bug:
22005084
Bug: https://code.google.com/p/android/issues/detail?id=177694
(cherry picked from commit
68df32091cc79468c81956633e279523b48ea8bf)
Change-Id: If2126d1be5dccadbbedcdccabde7398688d6e5ca
Mathieu Chartier [Sat, 20 Jun 2015 03:24:45 +0000 (20:24 -0700)]
Fix another miranda method moving GC bug
Need to copy miranda methods over before we allocate the new vtable
or else we may have stale miranda gc roots.
Bug:
21664466
Change-Id: Ib3e415bb9e7df7abfa18c98fe01f790fa39622dc
Vladimir Marko [Mon, 22 Jun 2015 13:35:42 +0000 (14:35 +0100)]
ART: Don't do pre-linked calls to the interpreter bridge.
Bug:
21897404
(cherry picked from commit
c08ab29796eeaa3b5863fbd987cd96fe7538d40c)
Change-Id: I66531379fb4468a8ddee063e664bf6593abbe7d5
Andreas Gampe [Sat, 20 Jun 2015 05:58:47 +0000 (22:58 -0700)]
ART: Disallow classes that are abstract and final
Make the verifier fail such classes.
Bug:
21873151
(cherry picked from commit
507cc6f83bf6379728f2dd20391f2ed5fbfe6371)
Change-Id: I217f3d71f44bccdcee7ca830e092c807928bed39
Nicolas Geoffray [Mon, 22 Jun 2015 14:15:07 +0000 (14:15 +0000)]
Revert "Revert "Add support for inlining already sharpened interface calls.""
bug:
21867144
This reverts commit
ee39360775066e6b9920348c86ea98bf01d0facf.
Change-Id: I69bf85ce7c352b1a22f36f0f2a136f95b4e73bd6
Nicolas Geoffray [Mon, 22 Jun 2015 14:11:58 +0000 (15:11 +0100)]
Revert "Add support for inlining already sharpened interface calls. DO NOT MERGE ANYWHERE"
This reverts commit
c362782609ca6fbc355ccc9ea9110fe671db9671.
Reverting in order to make a commit that will merge.
bug:
21867144
Nicolas Geoffray [Mon, 22 Jun 2015 10:06:43 +0000 (11:06 +0100)]
Revert "Revert "Use IsAssignableFrom instead of IsSubclass for robustness.""
Don't use IsAssignableFrom, but check beforehand if the referrer is an
interface. Otherwise, we are being too aggressive on removing clinit
checks on interfaces (being a subclass doesn't imply the interface
has been initialized).
bug:
21870666
This reverts commit
463580ca5a1e75e27ad0207537ffc6252091326a.
(cherry picked from commit
b783b408112d1797da646f576a40f94bcb5162f3)
Change-Id: Ida03f453c9b0d4dda87a5696098f9ffbd69a3aa3
Nicolas Geoffray [Sat, 20 Jun 2015 21:44:47 +0000 (22:44 +0100)]
Use compiling class instead of outer class for access checks.
bug:
21869678
(cherry picked from commit
afd06414598e011693137ba044e38756609b2179)
Change-Id: I982ccdf46e3f4d0cc1901439e0bc2dcfa0fc661f
Nicolas Geoffray [Fri, 19 Jun 2015 12:32:41 +0000 (13:32 +0100)]
Use compiling class instead of referrer for access checks.
bug:
21869670
(cherry picked from commit
30451743493bc2f589e96d68ab98f557461f97f9)
Change-Id: Ia18986797d166470801e7a9a690c054d2dc57952
Nicolas Geoffray [Fri, 19 Jun 2015 15:17:05 +0000 (16:17 +0100)]
Be careful with predecessor/successor index.
When we simplify the CFG, we must preserve things that were already
simplified. For example, the index in the predecessor list or
successor list of a block must be preserved for ensuring the
first block is a loop pre header.
bug:
21867463
(cherry picked from commit
8b20f88b0a8d1b374dd5eaae289d19734c77b8f8)
Change-Id: I2581b5a50942290da96cd9ec876f6f2573e0a6c4
Nicolas Geoffray [Mon, 22 Jun 2015 11:25:02 +0000 (11:25 +0000)]
Merge "Don't check the return type in the builder." into mnc-dev
Nicolas Geoffray [Thu, 18 Jun 2015 17:04:12 +0000 (18:04 +0100)]
Don't check the return type in the builder.
The verifier will not complain about dead code. For the compiler,
it is benign.
bug:
21863767
(cherry picked from commit
69505f86a1a653bf86c1f2b423b3e0605137d5d1)
Change-Id: I2bc7888b39c91862c6dade130eccdf92f19b3f6f
Nicolas Geoffray [Mon, 22 Jun 2015 11:07:17 +0000 (11:07 +0000)]
Merge "Do not expect 0 or 1 only when comparing a boolean." into mnc-dev
Nicolas Geoffray [Fri, 19 Jun 2015 10:42:07 +0000 (11:42 +0100)]
Do not expect 0 or 1 only when comparing a boolean.
bug:
21866529
(cherry picked from commit
3c4ab80c102ff1bfc0e74d4abddbf5454bf4008d)
Change-Id: Ibdc0d4a9730bfc6e7307282276f084dae5ac55c1
Nicolas Geoffray [Fri, 19 Jun 2015 09:35:42 +0000 (10:35 +0100)]
Verifier: check an aput, even if we know it will fail.
bug:
21867457
(cherry picked from commit
66389fbeb41c981648bb2e1e7e2f43089f095591)
Change-Id: Id01edeb871f6bc59518eb341e5b76f021cfdfe06
Nicolas Geoffray [Thu, 18 Jun 2015 14:46:47 +0000 (15:46 +0100)]
Run a simplification pass before code generation.
The code generators assume things that only the instruction
simplier ensures. So it has to be run last in case previous
optimiziations broke those assumptions.
bug:
21865464
(cherry picked from commit
b2bdfce7f805b00668a2521b1c939a0aafb2be49)
Change-Id: Ibf4384a911e400eb7586bbf6b4edd6351034cbd8
Nicolas Geoffray [Mon, 22 Jun 2015 09:25:56 +0000 (09:25 +0000)]
Merge "Bailout from compilation if an invoke is malformed." into mnc-dev
Nicolas Geoffray [Thu, 18 Jun 2015 10:11:27 +0000 (11:11 +0100)]
Bailout from compilation if an invoke is malformed.
Because the verifier does not check trivially dead instructions,
the compilers must prepare for bogus instructions. This change
fixes the case the arguments for an invoke do not match the
formal parameters.
bug:
21865459
(cherry picked from commit
2e33525bd4eb892246b4c244c6d4ebf6c6d07501)
Change-Id: I392f86eafefde28263fe35a31f17b398ff8dfc24
Nicolas Geoffray [Mon, 22 Jun 2015 09:16:04 +0000 (09:16 +0000)]
Merge "Fix wrong DCHECK in bounds check elimination." into mnc-dev
Nicolas Geoffray [Sat, 20 Jun 2015 22:49:01 +0000 (23:49 +0100)]
Fix wrong DCHECK in bounds check elimination.
The lower range of an array length instruction can
be changed by other instructions than HBoundsCheck,
like HNewArray.
bug:
21862741
(cherry picked from commit
8d82a0c2b2b12f259ccb357d3b1e699c68ad0400)
Change-Id: I1bb1a4f4c6673509dd3fb5184c32992bed876250
Nicolas Geoffray [Wed, 17 Jun 2015 22:09:05 +0000 (23:09 +0100)]
Fix a bug in optimizing when the null constant has been DCE.
If it has been DCE, we should create a new one, instead of
using the old one.
Also move the first DCE to a place where it could actually
be useful.
bug:
21870788
(cherry picked from commit
18e6873c469b48aaed22148451523479eece98e3)
Change-Id: I3b3ab2dafe8ce5fb60868fd1a6ef0eeefe666e0c
Nicolas Geoffray [Sun, 21 Jun 2015 17:57:02 +0000 (18:57 +0100)]
Fix String Change baseline compiler errors.
The String Change adds multiple move results if the uninitialized string
is in multiple registers. This adds StoreLocals on the same instruction,
which isn't allowed. Now, a LoadLocal is added for each extra move
needed.
bug:
21902634
(cherry picked from commit
aa919207d2fb63af11d72d3b7cdbc435769565af)
Change-Id: I057d14cdac437d06eec20caaddd430c304e58196
David Srbecky [Sat, 20 Jun 2015 04:01:22 +0000 (05:01 +0100)]
Use signed encoding when using relative CFI addresses.
This is required for gdb to work.
libunwind works with either encoding.
(cherry picked from commit
17065880693d1b15ffeb60b9955a2d092839977f)
Bug:
21924613
Change-Id: I4e4f1cf9c65d48fa885a5993eeeed0253a3f2579
Andreas Gampe [Sat, 20 Jun 2015 03:37:46 +0000 (20:37 -0700)]
ART: Check for expected args for instance methods
Don't assume that the receiver is expected. This was writing out
of bounds (or triggering a DCHECK).
Bug:
21817284
Bug:
21872240
Bug: https://code.google.com/p/android/issues/detail?id=176571
Bug: https://code.google.com/p/android/issues/detail?id=176572
(cherry picked from commit
ef0b1a1dbc99fe7c92f9598cbfc164763c1b66d7)
Change-Id: I7cad2c1dfc39feb0b1b4660deeb43afdcb844ba3
Taiju Tsuiki [Mon, 13 Apr 2015 05:21:00 +0000 (14:21 +0900)]
ART: Check var-arg count earlier in method verifier
Check the count in vararg instructions before filling the temp
array. Avoids a DCHECK.
Bug:
21869663
Bug:
20170976
(cherry picked from commit
29498a23bcfe47a7134552aacad5524ecb484a49)
Change-Id: I1ff93502ab84c5967ad54d1b50dba517da791637
Mingyao Yang [Fri, 19 Jun 2015 23:45:43 +0000 (23:45 +0000)]
Merge "ART: Fix BCE lint issue" into mnc-dev
Andreas Gampe [Thu, 11 Jun 2015 01:33:26 +0000 (18:33 -0700)]
ART: Fix BCE lint issue
Bug:
21034044
(cherry picked from commit
45d68f138a31a3ff9b45cda313f0ba27f1431f26)
Change-Id: I7f382a3124955eff5c0b96ca39ec67fb658fa3d0
Mathieu Chartier [Fri, 19 Jun 2015 22:19:22 +0000 (22:19 +0000)]
Merge "Fix GC map generation edge case" into mnc-dev
Mathieu Chartier [Fri, 19 Jun 2015 20:55:27 +0000 (13:55 -0700)]
Fix GC map generation edge case
Previously we checked the number of bits, but then rounded up to a
byte boundary. If ref_bitmap_bits was 65535, it would pass the check
but not actually fit in the 13 bits for ref_bitmap_bytes since we
rounded up to 8192 bytes.
Regression test infeasible due to smali limitations.
Bug:
21888453
Change-Id: If4b769431f27ecca14fe7852d017bcde01b3e146
Nicolas Geoffray [Wed, 17 Jun 2015 09:17:49 +0000 (10:17 +0100)]
Remove bogus DCHECK in BCE.
When creating a phi for the array length when we add HDeoptimization
nodes, we might update accesses in inner loops to use that phi instead
of the array length. The BCE phase was not expecting this case.
Bug:
21034044
(cherry picked from commit
3cde6227678cf62e06bca264671d1e957456ac3d)
Change-Id: I639f4ea6f5889726142041a42736183f162c7437
Mingyao Yang [Tue, 19 May 2015 23:01:59 +0000 (16:01 -0700)]
Fix premature deoptimization if the loop body isn't entered.
Add a test between initial_ and end_ to see if the loop body is entered.
If the loop body isn't entered at all, we jump to the loop header. Loop header is
still executed and is going to test the condition again and loop body won't be
entered. This makes sure no deoptimization is triggered if the loop body isn't
even entered.
Bug:
21034044
(cherry picked from commit
3584bce5b1f45e5741d3a6ca24884a36320ecb6b)
Change-Id: I2b6de1f22fbc4568ca419f76382ebd87806d9694
David Srbecky [Fri, 19 Jun 2015 18:25:56 +0000 (18:25 +0000)]
Merge "ART stack unwinding fixes for libunwind/gdb/lldb." into mnc-dev
Jeff Hao [Wed, 17 Jun 2015 23:04:26 +0000 (16:04 -0700)]
Fix case where block has no predecessor for StringChange.
Removes part that checks for throwing half of instruction. It's no
longer necessary. Also adds regression test.
Bug:
21902684
Change-Id: Ic600165e6b3719de3d83a73b8a1fa64473668fc8
Vladimir Marko [Fri, 19 Jun 2015 13:56:38 +0000 (14:56 +0100)]
Quick: Handle total high/low register overlap on arm/mips.
OpRegCopyWide() in arm and mips backends didn't handle the
total register overlap when the registers holding the source
and destination pairs are the same but in reverse order.
Bug:
21897012
(cherry picked from commit
8958f7f8702327e713264d0538ab5dec586f3738)
Change-Id: I20afce6cc3213e7f7b3edaef91f3ec29c469f877
Vladimir Marko [Fri, 19 Jun 2015 11:58:22 +0000 (12:58 +0100)]
Quick: Fix optimizations for empty if blocks.
If a block ending with if-eqz or if-nez has the same "taken"
and "fallthrough", we cannot assume that the value has been
checked against zero in one of the succesors. This affects
the null check elimination pass as well as GVN. Refactor all
those checks to a single function in BasicBlock and check
that the "taken" and "falthrough" are different when needed.
Bug:
21614284
(cherry picked from commit
f11c420c448baffac6a70ac0884d481ab347e257)
Change-Id: I062e0042de3470ce8680b586487b9c7acbd206bc
David Brazdil [Thu, 18 Jun 2015 11:40:01 +0000 (12:40 +0100)]
ART: Remove old DCHECK that trips Baseline
Codegen verified that the entry block always falls through to the next
block. While this is the case with Optimizing, it doesn't hold for
Baseline but it doesn't need to since codegen handles it fine.
Bug:
21913514
Change-Id: I751ef227e6cf103af3e7fc35fca4b01c663385a1
(cherry picked from commit
015c7e63604c038e866d7af3850c557403cddc8b)
Jeff Hao [Fri, 19 Jun 2015 04:19:08 +0000 (04:19 +0000)]
Merge "Fix StringChange for optimizing compiler." into mnc-dev
Jeff Hao [Fri, 19 Jun 2015 04:16:08 +0000 (21:16 -0700)]
Fix StringChange for optimizing compiler.
Uses optimizing compiler more and fixes x86_64 invoke codegen.
Bug:
21902634
Change-Id: Ia2a87d013c4746b107014a04a22a0a37269cfdb2
David Srbecky [Thu, 23 Apr 2015 01:57:06 +0000 (18:57 -0700)]
ART stack unwinding fixes for libunwind/gdb/lldb.
dex2oat can already generate unwinding and symbol information which
allows tools to create backtrace of mixed native and Java code.
This is a cherry pick from aosp/master which fixes several issues.
Most notably:
* It enables generation of ELF-64 on 64-bit systems (in dex2oat, C
compilers already produce ELF-64). Libunwind requires ELF-64 on
64-bit systems for backtraces to work.
* It enables loading of ELF files with dlopen. This is required for
libunwind to be able to generate backtrace of current process (i.e.
the process requesting backtrace of itself).
* It adds unit test to test the above (32 vs 64 bit, in-proces vs
out-of-process, application code vs framework code).
* Some other fixes or clean-ups which should not be of much
significance but which are easier to include to make the
important CLs cherry-pick cleanly.
This is squash of the following commits from aosp/master:
7381010 ART: CFI Test
e1bbed2 ART: Blacklist CFI test for non-compiled run-tests
aab9f73 ART: Blacklist CFI test for JIT
4437219 ART: Blacklist CFI test for Heap Poisoning
a3a49fe Switch to using ELF-64 for 64-bit architectures.
297ed22 Write 64-bit address in DWARF if we are on 64-bit architecture.
24981a1 Set correct size of PT_PHDR ELF segment.
1a146bf Link .dynamic to .dynstr
67a0653 Make some parts of ELF more (pointer) aligned.
f50fa82 Enable 64-bit CFI tests.
49e1fab Use dlopen to load oat files.
5dedb80 Add more logging output for dlopen.
aa03870 Find the dlopened file using address rather than file path.
82e73dc Release dummy MemMaps corresponding to dlopen.
5c40961 Test that we can unwind framework code.
020c543 Add more log output to the CFI test.
88da3b0 ART: Fix CFI test wrt/ PIC
a70e5b9 CFI test: kill the other process in native code.
ad5fa8c Support generation of CFI in .debug_frame format.
90688ae Fix build - large frame size of ElfWriterQuick<ElfTypes>::Write.
97dabb7 Fix build breakage in dwarf_test.
388d286 Generate just single ARM mapping symbol.
f898087 Split .oat_patches to multiple sections.
491a7fe Fix build - large frame size of ElfWriterQuick<ElfTypes>::Write (again).
8363c77 Add --generate-debug-info flag and remove the other two flags.
461d72a Generate debug info for core.oat files.
Bug:
21924613
Change-Id: I3f944a08dd2ed1df4d8a807da4fee423fdd35eb7
Mathieu Chartier [Thu, 18 Jun 2015 23:48:52 +0000 (16:48 -0700)]
Fix moving GC bug in DoFilledNewArray
Previously we read from componentClass after allocating the array.
Bug:
21783443
(cherry picked from commit
52ea33b10370d60d4ce877aec529626537b7813b)
Change-Id: I5283982edab479434e27416509e1436b4176fe01
Mathieu Chartier [Thu, 18 Jun 2015 17:17:26 +0000 (10:17 -0700)]
Fix moving GC bugs in proxy stub for X86/X86_64
Needed to restore the refs.
(cherry picked from commit
9346ff0cfad6344d0bf4eaa69362dbe1987ac054)
Bug:
21907554
Change-Id: I562906dff07dcaa78dfb39646ba9ab35a5f56c6c
Sebastien Hertz [Tue, 9 Jun 2015 12:09:14 +0000 (14:09 +0200)]
Follow up on CL 151605
- Fixes return type of StackedShadowFrameRecord::GetType
- Makes StackedShadowFrameType an enum class (scoped enum)
- Moves DeoptimizationReturnValueRecord and StackedShadowFrameRecord
to thread.cc file and use forward declaration in thread.h header
- Fixes tools/generate-operator-out.py for scoped enum classes.
Bug:
20845490
(cherry picked from commit
f795869da0a1fa006fdcdacd8afb6149a63fc1a7)
Change-Id: I6b67e288b1db563699161e58ec2e2330d42dd8f5
Mingyao Yang [Mon, 18 May 2015 19:12:50 +0000 (12:12 -0700)]
Fix nested deoptimization.
Handle nested deoptimization cases. Create a stacked shadow frame
records to keep track of deoptimization shadow frames. Shadow frames
under construction can be tracked in the same stack.
Bug:
20845490
(cherry picked from commit
1f2d3ba6af52cf6f566deb38b7e07735c9a08fb6)
Change-Id: I768285792c29e7c3cfcd21e7a2600802506024d8
Vladimir Marko [Thu, 18 Jun 2015 16:40:00 +0000 (17:40 +0100)]
Quick: Don't expect move-exception in every catch block.
The dalvik bytecode doesn't require a move-exception in
a catch handler that ignores the exception.
Bug:
21873167
(cherry picked from commit
2d1a0a408fd148f7b2a2d670e6942ec3d920f875)
Change-Id: Ia00e5791fabdd1b88e9a63b741600ed63d918703
Mathieu Chartier [Wed, 17 Jun 2015 23:11:12 +0000 (16:11 -0700)]
Fix some java_lang_Class related moving GC bugs
There was some missing handles around mirror::Class*.
(cherry picked from commit
05b7226787f1470ad93f6f632fed60f70bc8631e
Bug:
21898408
Change-Id: Icb754074dfb469473101d20d6873a5bc3274abc5
David Brazdil [Thu, 18 Jun 2015 16:13:16 +0000 (16:13 +0000)]
Merge "ART: Allow PackedSwitch instructions with zero targets" into mnc-dev
David Brazdil [Wed, 17 Jun 2015 17:20:52 +0000 (18:20 +0100)]
ART: Allow PackedSwitch instructions with zero targets
Optimizing and the interpreter wrongly assumed that a PackedSwitch
always has at least one target. This patch removes the corresponding
DCHECKs and adds a regression test case.
This is a resubmission of CL I32b7033ed38de6f1d1a6ee5d5bf12f3a47c9b37e
Bug:
21863783
Change-Id: I04e6e124bdd16591ba27c79490e6ce183c36b691
(cherry picked from commit
2ef645ba50544b879a82ea30e606f18c9af98917)
Vladimir Marko [Wed, 17 Jun 2015 21:00:09 +0000 (22:00 +0100)]
Quick: Fix marking unused registers in DCE.
Bug:
21702651
(cherry picked from commit
511d408909b7fe9be3c95e032cc7f426f4a01e6e)
Change-Id: I059e0b841543334cd08a27431e0ea853b42e4f35
Vladimir Marko [Mon, 15 Jun 2015 17:52:54 +0000 (18:52 +0100)]
ART: Fix reporting initialized classes by CompilerDriver.
Fix a bug where the CompilerDriver was erroneously reporting
classes as initialized during AOT compilation when they were
not guaranteed to be initialized at runtime.
This fix prevents the Quick compiler from inlining calls to
static methods in classes that are not guaranteed to be
initialized, so that the runtime performs the initialization
required for correctness.
Bug:
21847756
(cherry picked from commit
07785bb98dc8bbe192970e0f4c2cafd338a8dc68)
Change-Id: I60c7361cb6e8f51be20a3cbfcae19f3240bdfbed
Nicolas Geoffray [Thu, 18 Jun 2015 09:05:21 +0000 (09:05 +0000)]
Merge "Add support for inlining already sharpened interface calls. DO NOT MERGE ANYWHERE" into mnc-dev
Nicolas Geoffray [Tue, 9 Jun 2015 14:43:38 +0000 (15:43 +0100)]
Add support for inlining already sharpened interface calls. DO NOT MERGE ANYWHERE
bug:
21867144
Test had to change a bit because it was using some features from
aosp not present in mnc-dev.
(cherry picked from commit
3507105caea0d209c66a95d5ec5d739949105c5d)
Change-Id: If12ea9a5c64e86b5bd3a5fd4a943bff60fe49762
Nicolas Geoffray [Mon, 15 Jun 2015 08:01:08 +0000 (09:01 +0100)]
Fix reference comparison after redundant phi elimination.
Otherwise, the graph could remain mistyped.
bug:
21776173
(cherry picked from commit
51d400d4ebd41b9fb4d67ac3179f8fb66a090fdd)
Change-Id: Iff36dfa4e79b14a9dd85c37e0fbb9e1080dd0364
Mathieu Chartier [Wed, 17 Jun 2015 21:14:36 +0000 (14:14 -0700)]
Add missing EndAssertNoThreadSuspension
Bug:
21872507
Change-Id: I20f9920e4a12511ce498c252bca955d82a4f4102
Oleksiy Vyalov [Sun, 14 Jun 2015 22:17:26 +0000 (15:17 -0700)]
Set default visibility on art_sigsegv_fault.
libart.so is built with flag -fvisibility=protected
which complicates setting breakpoints on art_sigsegv_fault.
According to the GCC manual "Protected visibility indicates
that the symbol will be placed in the dynamic symbol table,
but that references within the defining module will bind
to the local symbol.". This implies that a breakpoint on
a protected symbol will not be hit.
Without this change gdbclient with ART debugging enabled
doesn't break at art_sigsegv_fault as expected.
Bonus build fix included (commit
c37769b6) :
Fix lint error.
Blank line at the start of a code block.
(cherry picked from commit
f775bab533cd385c00ee30b56bdcb6f962fc5239)
(cherry picked from commit
c37769b63faf7b120a6ed80f5cfb92667f5090c4)
bug:
21893559
Change-Id: Ie388f027f717a7bd2351a0163a38e50d3acbba8f
Mathieu Chartier [Sat, 13 Jun 2015 00:18:37 +0000 (00:18 +0000)]
Merge "Change mips64 lwu to ld for loading caller ArtMethod*" into mnc-dev
Mathieu Chartier [Fri, 12 Jun 2015 23:05:54 +0000 (16:05 -0700)]
Change mips64 lwu to ld for loading caller ArtMethod*
Bug:
21815962
Change-Id: Id589f4a49349d817d3dc9128109048292fe0ed33
Sebastien Hertz [Thu, 11 Jun 2015 16:42:58 +0000 (18:42 +0200)]
JDWP: update handling of 8-bit and 16-bit fields
We must use special getter/setter for types smaller than int:
boolean, byte, char and short. However, JDWP still requires to treat:
- int and float as 32-bit values
- long and double as 64-bit values
Bug:
21746739
(cherry picked from commit
05c26b30843dd8283c703356e75d51726e8b8e3a)
Change-Id: I2d667291974e191f2c9b10311ce02435b902dd0c
Sebastien Hertz [Mon, 1 Jun 2015 15:33:12 +0000 (17:33 +0200)]
JDWP: asynchronous invoke command handling
The JDWP thread used to wait for the result of a method invocation
running in an event thread. But doing that prevents the JDWP thread
from processing incoming commands from the debugger if the event
thread gets suspended by a debug event occurring in another thread.
In Android Studio (or another IDE), this leads to the debugger being
blocked (with the famous message "Waiting until last debugger command
completes" of Android Studio / IntelliJ) because it is actually
waiting for the reply of its latest command while the JDWP thread
cannot process it.
This CL changes the way invoke commands (ClassType.InvokeCommand,
ClassType.NewInstance and ObjectReference.InvokeCommand) are handled
in the ART runtime.
The JDWP thread no longer waits for the event thread to complete the
method invocation. It now simply waits for the next JDWP command to
process. This means it does not send any reply for invoke commands,
except if the information given by the debugger is wrong. In this
case, it still sends a reply with the appropriate error code.
The event thread is now responsible for sending the reply (containing
the result and the exception object of the invoked method) before
going back to the suspended state.
In other words, we add special handling for invoke commands so they
are handled asynchronously while other commands remained handled
synchronously. In the future, we may want to handle all commands
asynchronously (using a queue of reply/event for instance) to remove
the special handling code this CL is adding.
Now the JDWP thread can process commands while a thread is invoking
a method, it is possible for the debugger to detach (by sending a
VirtualMachine.Dispose command) before the invocation completes. In
that situation, we must not suspend threads again (including the
event thread that executed the method) because they would all remain
suspended forever.
Also minor cleanup of the use of JDWP constants and update comments.
Bug:
21515842
Bug:
18899981
(cherry picked from commit
cbc5064ff05179b97b416f00ca579c55e38cd7d9)
Change-Id: I8d31006043468913ee8453212e6d16e11fcfe4ea
Tim Murray [Fri, 12 Jun 2015 00:21:24 +0000 (00:21 +0000)]
Merge "Add cortex-a53 as a valid instruction set variant for arm64." into mnc-dev
Tim Murray [Fri, 12 Jun 2015 00:09:18 +0000 (17:09 -0700)]
Add cortex-a53 as a valid instruction set variant for arm64.
Change-Id: I720e2cc4acd94e367dae4a21c13aa90a3e30b7cc
Mathieu Chartier [Thu, 11 Jun 2015 23:21:42 +0000 (16:21 -0700)]
Fix compaction bug in Class_getDeclaredMethodsUnchecked
Added handle to fix the bug.
Bug:
21638351
Change-Id: I1c3abea33aa825d3a28c1fc5cb415508686ad93e
Dmitriy Ivanov [Thu, 11 Jun 2015 21:54:53 +0000 (21:54 +0000)]
Merge "Let classloader provide correct LD_LIBRARY_PATH" into mnc-dev
Mathieu Chartier [Thu, 11 Jun 2015 21:45:53 +0000 (21:45 +0000)]
Merge "Move image intern table into image" into mnc-dev