OSDN Git Service

android-x86/art.git
7 years agoRemove the unused SMP instruction set feature option.
Serban Constantinescu [Fri, 23 Sep 2016 10:06:03 +0000 (11:06 +0100)]
Remove the unused SMP instruction set feature option.

The SMP option is currently not used in the ART codebase
(and it rots away).
Single CPU systems that run SMP code should be able to deal
with it nicely (otherwise the CPU has some serious problems).

Test: mma test-art-target && mma test-art-host

Change-Id: Ifaab42b29de1062866a0af7525de2f834a4a5a57

7 years agoMerge "ART: Make method handle runtime code callable from compiler."
Orion Hodson [Fri, 6 Jan 2017 09:04:47 +0000 (09:04 +0000)]
Merge "ART: Make method handle runtime code callable from compiler."

7 years agoMerge "Remove bogus DCHECK."
Treehugger Robot [Fri, 6 Jan 2017 09:04:18 +0000 (09:04 +0000)]
Merge "Remove bogus DCHECK."

7 years agoMerge "ARM: VIXL32: Fix crash in Exchange for stack slots."
Nicolas Geoffray [Fri, 6 Jan 2017 09:03:36 +0000 (09:03 +0000)]
Merge "ARM: VIXL32: Fix crash in Exchange for stack slots."

7 years agoMerge "Actually record arrays in classpath."
Nicolas Geoffray [Fri, 6 Jan 2017 09:00:15 +0000 (09:00 +0000)]
Merge "Actually record arrays in classpath."

7 years agoART: Make method handle runtime code callable from compiler.
Orion Hodson [Wed, 7 Dec 2016 11:35:37 +0000 (11:35 +0000)]
ART: Make method handle runtime code callable from compiler.

Most of this change is moving the existing method handles code, but it
also introduces a new header file, common_dex_operations.h, that has
some operations taken from interpreter_common.{h,cc} that are also used
by method handles (perform call, set field, get field).

Bug: 30550796
Test: m test-art-host
Change-Id: I2235e13770a5562950f2767f65a25ca273479150

7 years agoMerge "Prevent moving GC from running during redefinition."
Treehugger Robot [Fri, 6 Jan 2017 03:59:24 +0000 (03:59 +0000)]
Merge "Prevent moving GC from running during redefinition."

7 years agoPrevent moving GC from running during redefinition.
Alex Light [Fri, 6 Jan 2017 01:53:00 +0000 (17:53 -0800)]
Prevent moving GC from running during redefinition.

This could cause a deadlock during deoptimization.

Test: ART_USE_READ_BARRIER=true ART_READ_BARRIER_TYPE=TABLELOOKUP mma -j40 test-art-host
Change-Id: I57e4f1a50709bf4a1817227913e61f3ef434d04a

7 years agoMerge "Revert "Revert "Cleanup jvmti tests"""
Alex Light [Fri, 6 Jan 2017 00:46:17 +0000 (00:46 +0000)]
Merge "Revert "Revert "Cleanup jvmti tests"""

7 years agoMerge "Add some code info checking in GetCalleeSaveMethodCaller"
Mathieu Chartier [Thu, 5 Jan 2017 23:59:14 +0000 (23:59 +0000)]
Merge "Add some code info checking in GetCalleeSaveMethodCaller"

7 years agoRevert "Revert "Cleanup jvmti tests""
Alex Light [Thu, 5 Jan 2017 22:37:21 +0000 (14:37 -0800)]
Revert "Revert "Cleanup jvmti tests""

This reverts commit 2a9d198b9929560c9ab07691b3afbe275483be95.

Test: mma -j40 test-art-host
Test: ART_TEST_RUN_TEST_NDEBUG=true mma -j40 test-art-host

Change-Id: I14bfd02f74f93e4a6b853287fcaa69f1fb027256

7 years agoMerge "Reduce the number of fences needed for monitors"
Hans Boehm [Thu, 5 Jan 2017 22:15:30 +0000 (22:15 +0000)]
Merge "Reduce the number of fences needed for monitors"

7 years agoAdd some code info checking in GetCalleeSaveMethodCaller
Mathieu Chartier [Thu, 5 Jan 2017 18:17:55 +0000 (10:17 -0800)]
Add some code info checking in GetCalleeSaveMethodCaller

Will print info if the memory region is too small for the stack maps.

Test: test-art-host

Bug: 33924573

Change-Id: I03e0151724d3a0933f3bef004e0d16ba1c6134c4

7 years agoMerge "Tests on FP Math.abs()"
Aart Bik [Thu, 5 Jan 2017 20:34:32 +0000 (20:34 +0000)]
Merge "Tests on FP Math.abs()"

7 years agoMerge "Fix unquitable apps when debug enabled."
Treehugger Robot [Thu, 5 Jan 2017 19:13:38 +0000 (19:13 +0000)]
Merge "Fix unquitable apps when debug enabled."

7 years agoMerge "Revert "Revert "Revert "Revert "Basic obsolete methods support"""""
Treehugger Robot [Thu, 5 Jan 2017 18:16:14 +0000 (18:16 +0000)]
Merge "Revert "Revert "Revert "Revert "Basic obsolete methods support"""""

7 years agoTests on FP Math.abs()
Aart Bik [Tue, 3 Jan 2017 17:53:00 +0000 (09:53 -0800)]
Tests on FP Math.abs()

Rationale:
As a "quality of implementation", rather than pure "spec compliance", we
require that Math.abs() clears the sign bit (but changes nothing else)
for all numbers, including NaN (signaling NaN may become quiet though).

These tests ensure that remains the case.

NOTE:
This is a better redo of reverted CL
https://android-review.googlesource.com/#/c/316351/

Test: test-art-host
Bug: 30758343

Change-Id: I2740f9aea19abd82335f1f000d92f1cdb469a1c3

7 years agoMerge "MIPS64: Implement table-based packed switch"
Treehugger Robot [Thu, 5 Jan 2017 17:37:56 +0000 (17:37 +0000)]
Merge "MIPS64: Implement table-based packed switch"

7 years agoMerge "Remove racy DCHECK on string dex cache."
Treehugger Robot [Thu, 5 Jan 2017 16:55:53 +0000 (16:55 +0000)]
Merge "Remove racy DCHECK on string dex cache."

7 years agoRemove racy DCHECK on string dex cache.
Nicolas Geoffray [Thu, 5 Jan 2017 15:27:03 +0000 (15:27 +0000)]
Remove racy DCHECK on string dex cache.

Test: test-art-host
Change-Id: I58940031d43d89f434d3c3239a218b99ebb1106b

7 years agoActually record arrays in classpath.
Nicolas Geoffray [Thu, 5 Jan 2017 15:23:19 +0000 (15:23 +0000)]
Actually record arrays in classpath.

With https://android-review.googlesource.com/#/c/316151/,
I was under the wrong impression arrays all have the same
access flags. They actually have the visibility of the inner
most component type.

Therefore, we still need to record visibility of array types whose
innermost component type is in the classpath.

Test: verifier_deps_test
Change-Id: If7b1004efb679e320330258f42ced83b8eedae87

7 years agoARM: VIXL32: Fix crash in Exchange for stack slots.
Artem Serov [Thu, 29 Dec 2016 16:21:49 +0000 (16:21 +0000)]
ARM: VIXL32: Fix crash in Exchange for stack slots.

In ParallelMoveResolverARMVIXL::Exchange(int mem1, int mem2)
scratch general purpose register was used without any spilling
(like in StoreToOffset) which led to lack of scratch register
for VLDR with big offset. Now it uses two scratch S-registers.

Test: ART_USE_VIXL_ARM_BACKEND=true m test-art-host
Test: ART_USE_VIXL_ARM_BACKEND=true m test-art-target
Change-Id: I0416a69e281d09a04dd1689efa5a8c1994c82638

7 years agoMerge "Refine receiver type when inlining based on inline caches."
Nicolas Geoffray [Thu, 5 Jan 2017 14:57:29 +0000 (14:57 +0000)]
Merge "Refine receiver type when inlining based on inline caches."

7 years agoMerge "MIPS32: java.lang.String.getChars"
Nicolas Geoffray [Thu, 5 Jan 2017 13:27:14 +0000 (13:27 +0000)]
Merge "MIPS32: java.lang.String.getChars"

7 years agoRefine receiver type when inlining based on inline caches.
Nicolas Geoffray [Wed, 4 Jan 2017 16:46:23 +0000 (16:46 +0000)]
Refine receiver type when inlining based on inline caches.

We know we're guarding for a specific type
(eg foo.getClass() == Foo.class), so we can propagate the
exact type when building the callee graph.

Test: ART_TEST_JIT=true test-art-host
Change-Id: I3e1858945afbb651cb8966b2b7d7755eccac09e1

7 years agoMerge "Revert "Disable 626-const-class-linking for JIT.""
Vladimir Marko [Thu, 5 Jan 2017 12:45:35 +0000 (12:45 +0000)]
Merge "Revert "Disable 626-const-class-linking for JIT.""

7 years agoMerge "Fix incorrect string hash value extension during cross-compilation."
Treehugger Robot [Thu, 5 Jan 2017 12:15:07 +0000 (12:15 +0000)]
Merge "Fix incorrect string hash value extension during cross-compilation."

7 years agoMerge "Limit Jack threads during tests"
Sebastien Hertz [Thu, 5 Jan 2017 10:28:56 +0000 (10:28 +0000)]
Merge "Limit Jack threads during tests"

7 years agoLimit Jack threads during tests
Sebastien Hertz [Wed, 4 Jan 2017 15:07:57 +0000 (16:07 +0100)]
Limit Jack threads during tests

In default mode, run Jack with 4 threads to help reduce memory
footprint and thread contention. These are the same settings than
the Android build.

For the buildbot (where the Jack server is disabled), run Jack in
single-threaded mode to reduce the global memory consumption.

Bug: 31737153
Test: make -j32 test-art-host
Change-Id: I6a14efce92ea18dba008e6daa26cd6403e5eb503

7 years agoReduce the number of fences needed for monitors
Hans Boehm [Thu, 15 Dec 2016 21:12:59 +0000 (13:12 -0800)]
Reduce the number of fences needed for monitors

Add the necessary CasWeakAcquire primitives for LockWords.

Have MonitorEnter initially read the lockword using a
memory_order_relaxed operation. In the unlikely case we need more,
compensate with an explicit fence.

In the uncontended case, install the thin lock with Acquire,
rather than SequentiallyConsistent semantics.

Have MonitorExit use a Release instead of SequentiallyConsistent
CAS in the ReadBarrier case. Add TODO for the other case.

Together, these should usually eliminate 3 fences (or acq/rel)
per critical section.

Have Install() only use Release ordering.

Add TODO for inflation spinning, which looks to me like it could be
improved appreciably.

Drive-by fix:

GetMaxSpinsBeforeThinLockInflation spelling

Test: Build for several targets, boot, m art-test-host art-test-target

Change-Id: I2cab09723252065f6365e4234ee3249c69ece888

7 years agoMerge "ART: Remove dequicken map from VerifiedMethod"
Treehugger Robot [Wed, 4 Jan 2017 22:59:02 +0000 (22:59 +0000)]
Merge "ART: Remove dequicken map from VerifiedMethod"

7 years agoART: Remove dequicken map from VerifiedMethod
Andreas Gampe [Wed, 4 Jan 2017 20:04:05 +0000 (12:04 -0800)]
ART: Remove dequicken map from VerifiedMethod

Remove now unused VerifiedMethod's dequicken map. Follow-up to
c51c7ca7a779563d153c137c6bf01c3ea532a6c9.

Results on a large well-known app in speed compile with -j4:

Before:
 dex2oat took [...] native alloc=84MB [...] swap=208MB [...]
After:
 dex2oat took [...] native alloc=76MB [...] swap=208MB [...]

Bug: 34053922
Test: m clean-oat-host && m test-art-host
Change-Id: Ie95fd297299a0f5fac0c8c702773ee8372f662b6

7 years agoRemove bogus DCHECK.
Nicolas Geoffray [Wed, 4 Jan 2017 11:49:00 +0000 (11:49 +0000)]
Remove bogus DCHECK.

The code may be executed concurrently by multiple threads,
so the flag may already be set.

Test: test-art-host
bug:34074968
Change-Id: I162b90a853ce53e6b7063754bbfce5d68cffc82a

7 years agoMerge "Revert "Revert "Avoid scratch register exhaustion during ARM64 stack slot...
Roland Levillain [Wed, 4 Jan 2017 15:22:32 +0000 (15:22 +0000)]
Merge "Revert "Revert "Avoid scratch register exhaustion during ARM64 stack slot moves."""

7 years agoRevert "Revert "Avoid scratch register exhaustion during ARM64 stack slot moves.""
Roland Levillain [Wed, 4 Jan 2017 10:27:50 +0000 (10:27 +0000)]
Revert "Revert "Avoid scratch register exhaustion during ARM64 stack slot moves.""

This reverts commit 122ceb492f37e97d022cba2221a87368f1847f5f.

Test: m test-art-target on ARM64, with and without read barriers/heap poisoning.
Bug: 32545705
Change-Id: I4b447d762082eea8edfabeb070317d274e2f5bd0

7 years agoMerge "Support non-exact upper bound of HBoundType."
Nicolas Geoffray [Wed, 4 Jan 2017 12:38:29 +0000 (12:38 +0000)]
Merge "Support non-exact upper bound of HBoundType."

7 years agoMerge "Do not record dependencies on arrays."
Treehugger Robot [Wed, 4 Jan 2017 12:10:41 +0000 (12:10 +0000)]
Merge "Do not record dependencies on arrays."

7 years agoMerge "MIPS64: Drop redundant sign-extensions in integer conversions"
Treehugger Robot [Wed, 4 Jan 2017 11:47:50 +0000 (11:47 +0000)]
Merge "MIPS64: Drop redundant sign-extensions in integer conversions"

7 years agoMerge "MIPS32: Don't always do a null test in MarkGCCard"
Treehugger Robot [Wed, 4 Jan 2017 11:31:52 +0000 (11:31 +0000)]
Merge "MIPS32: Don't always do a null test in MarkGCCard"

7 years agoMerge "MIPS64: Improve integer comparison with constants"
Treehugger Robot [Wed, 4 Jan 2017 11:30:15 +0000 (11:30 +0000)]
Merge "MIPS64: Improve integer comparison with constants"

7 years agoMerge changes from topic 'VIXLUpdate'
Nicolas Geoffray [Wed, 4 Jan 2017 10:06:27 +0000 (10:06 +0000)]
Merge changes from topic 'VIXLUpdate'

* changes:
  ARM: VIXL32: Use a default code buffer capacity of 1Kb.
  ART: VIXL32: Fix assembler test after VIXL update.

7 years agoMerge "MIPS64: Implement branchless HCondition for floats"
Nicolas Geoffray [Wed, 4 Jan 2017 10:04:08 +0000 (10:04 +0000)]
Merge "MIPS64: Implement branchless HCondition for floats"

7 years agoMerge "Disable libcore test that requires IPv6"
Yi Kong [Wed, 4 Jan 2017 10:03:00 +0000 (10:03 +0000)]
Merge "Disable libcore test that requires IPv6"

7 years agoDisable libcore test that requires IPv6
Yi Kong [Wed, 4 Jan 2017 00:55:10 +0000 (00:55 +0000)]
Disable libcore test that requires IPv6

Test: N/A
Bug: 25178637
Bug: 34038136
Change-Id: I14c7451d269ee8f1163369d8c0d2209498466b66

7 years agoMerge "Clean up VerifiedMethod."
Treehugger Robot [Tue, 3 Jan 2017 22:33:12 +0000 (22:33 +0000)]
Merge "Clean up VerifiedMethod."

7 years agoClean up VerifiedMethod.
Nicolas Geoffray [Fri, 25 Nov 2016 15:46:48 +0000 (15:46 +0000)]
Clean up VerifiedMethod.

Remove unused methods, and devirt map.

Results on a large well-known app in speed compile with -j4:

Before:
 dex2oat took [...] native alloc=149MB [...] swap=208MB (218103808B)
After:
 dex2oat took [...] native alloc=84MB [...] swap=208MB (218103808B)

Bug: 34053922
Test: m clean-oat-host && m test-art-host
Change-Id: I0791c1e4ec7a98ae3c98321c582974db8e26617d

7 years agoMerge "Revert "Avoid scratch register exhaustion during ARM64 stack slot moves.""
Nicolas Geoffray [Tue, 3 Jan 2017 21:35:22 +0000 (21:35 +0000)]
Merge "Revert "Avoid scratch register exhaustion during ARM64 stack slot moves.""

7 years agoRevert "Avoid scratch register exhaustion during ARM64 stack slot moves."
Nicolas Geoffray [Tue, 3 Jan 2017 21:34:59 +0000 (21:34 +0000)]
Revert "Avoid scratch register exhaustion during ARM64 stack slot moves."

Checker test fails.
Bug: 32545705

This reverts commit 4ec76d28f0f808117272134347abf828eea80b91.

Change-Id: Ief14978596341399404c504ec4ca6b68c54fd63c

7 years agoMerge "MIPS64: Align register spills on 8-byte boundaries in slow paths"
Nicolas Geoffray [Tue, 3 Jan 2017 20:43:00 +0000 (20:43 +0000)]
Merge "MIPS64: Align register spills on 8-byte boundaries in slow paths"

7 years agoMIPS32: java.lang.String.getChars
Chris Larsen [Tue, 22 Nov 2016 00:04:53 +0000 (16:04 -0800)]
MIPS32: java.lang.String.getChars

Use memcpy(3) to copy characters under the assumption that memcpy()
has been hand optimized for best performance on the platform being
tested.

Test: run-test --optimizing 020-string
Test: run-test 020-string
Test: run-test --no-prebuild --optimizing 020-string
Test: run-test --no-prebuild 020-string
Test: run-test --optimizing 082-inline-execute
Test: run-test 082-inline-execute
Test: run-test --no-prebuild --optimizing 082-inline-execute
Test: run-test --no-prebuild 082-inline-execute
Test: mma -j2 ART_TEST_OPTIMIZING=true test-art-target-run-test
Test: booted MIPS32R2 emulator.

Note: Tested against both the MIPS32R2, and MIPS64R6 emulators.

Change-Id: I4192cf6244db120c8de5cc4932d4132acfc9740d

7 years agoMerge "Avoid scratch register exhaustion during ARM64 stack slot moves."
Roland Levillain [Tue, 3 Jan 2017 15:45:44 +0000 (15:45 +0000)]
Merge "Avoid scratch register exhaustion during ARM64 stack slot moves."

7 years agoAvoid scratch register exhaustion during ARM64 stack slot moves.
Roland Levillain [Mon, 7 Nov 2016 18:50:32 +0000 (18:50 +0000)]
Avoid scratch register exhaustion during ARM64 stack slot moves.

On ARM64, do not limit the selection of a scratch register
used in a move between two stack slots (or between two
double stack slots) to VIXL's floating-point register pool,
as it contains only one register (D31) and can be exhausted
in some rare cases. Instead, query both the core and the FP
register pools.

Add a regression run-test (626-checker-arm64-scratch-register).

Test: m test-art-target (on ARM64)
Bug: 32545705
Change-Id: I6203a4340e3c8b4f4879c07ed1be4c433c311c0f

7 years agoMerge "MIPS64: java.lang.String.getChars"
Treehugger Robot [Tue, 3 Jan 2017 11:13:50 +0000 (11:13 +0000)]
Merge "MIPS64: java.lang.String.getChars"

7 years agoMIPS64: Implement branchless HCondition for floats
Tijana Jakovljevic [Fri, 30 Dec 2016 08:23:01 +0000 (09:23 +0100)]
MIPS64: Implement branchless HCondition for floats

Test: mma test-art-target-run-test64 in QEMU

Change-Id: I595b5b7ddf9ebb19e872ed85f2e4098a835d9214

7 years agoMIPS64: Drop redundant sign-extensions in integer conversions
Goran Jakovljevic [Wed, 28 Dec 2016 15:21:48 +0000 (16:21 +0100)]
MIPS64: Drop redundant sign-extensions in integer conversions

Test: mma test-art-target-run-test64 in QEMU

Change-Id: I1dc7923d89528964771ddca288c5016d729cfb05

7 years agoMIPS64: Improve integer comparison with constants
Goran Jakovljevic [Wed, 28 Dec 2016 13:33:21 +0000 (14:33 +0100)]
MIPS64: Improve integer comparison with constants

Test: mma test-art-target-run-test64 in QEMU

Change-Id: Ib29855a3e268e0333602f47185c520c6cfb16b59

7 years agoMIPS32: Don't always do a null test in MarkGCCard
Goran Jakovljevic [Mon, 26 Dec 2016 13:21:43 +0000 (14:21 +0100)]
MIPS32: Don't always do a null test in MarkGCCard

Test: mma test-art-target-run-test on CI20

Change-Id: I38fac492bb76b06b9cffc9be16944fb22ed2db70

7 years agoMerge "Revert "Tests on FP Math.abs()""
Treehugger Robot [Fri, 23 Dec 2016 00:44:54 +0000 (00:44 +0000)]
Merge "Revert "Tests on FP Math.abs()""

7 years agoMIPS64: Align register spills on 8-byte boundaries in slow paths
Alexey Frunze [Thu, 22 Dec 2016 21:54:23 +0000 (13:54 -0800)]
MIPS64: Align register spills on 8-byte boundaries in slow paths

64-bit loads/stores would otherwise be split into pairs of
32-bit ones.

Test: booted MIPS64 (with 2nd arch MIPS32R6) in QEMU
Test: test-art-target-run-test-optimizing (MIPS64R6) in QEMU

Change-Id: I4846d11b52b71507dfd5ca2e27b3f2a5befcc58e

7 years agoRevert "Tests on FP Math.abs()"
Aart Bik [Thu, 22 Dec 2016 22:54:04 +0000 (22:54 +0000)]
Revert "Tests on FP Math.abs()"

This reverts commit b67877704edf8a32e71ce825d7ab4703be39f8ce.

On  art-relocate-no-patchoat (linux)
we saw

#################### info
# Tests on floating-point Math.abs.
#################### diffs
--- expected.txt 2016-12-22 21:33:16.796152568 +0000
+++ output.txt 2016-12-22 21:44:16.685738052 +0000
@@ -1 +1,3 @@
-passed
+Exception in thread "main" java.lang.Error: Expected: 0x7f800001, found: 0x7fc00001
+ at Main.expectEquals32(Main.java:140)
+ at Main.main(Main.java:85)
####################

Change-Id: If8d1bce9d4d20668661af82bbd08940787cd8252

7 years agoMerge "Tests on FP Math.abs()"
Treehugger Robot [Thu, 22 Dec 2016 21:00:40 +0000 (21:00 +0000)]
Merge "Tests on FP Math.abs()"

7 years agoARM: VIXL32: Use a default code buffer capacity of 1Kb.
Scott Wakeling [Tue, 20 Dec 2016 09:43:32 +0000 (09:43 +0000)]
ARM: VIXL32: Use a default code buffer capacity of 1Kb.

Test: mma test-art-host && mma test-art-target

Change-Id: I238c40e775338b63cbdb8868c09d0555b0c3b077

7 years agoART: VIXL32: Fix assembler test after VIXL update.
Artem Serov [Wed, 21 Dec 2016 18:42:21 +0000 (18:42 +0000)]
ART: VIXL32: Fix assembler test after VIXL update.

Veneer pool is emitted 4 bytes later, so the expected output for
the test has been adjusted.

Test: test-art-host
Test: test-art-target
Change-Id: I3d656224fd4151904b8096486adecb6ef1eafea6

7 years agoMIPS64: Implement table-based packed switch
Alexey Frunze [Wed, 21 Dec 2016 01:24:59 +0000 (17:24 -0800)]
MIPS64: Implement table-based packed switch

Test: booted MIPS64 (with 2nd arch MIPS32R6) in QEMU
Test: test-art-target-run-test-optimizing (MIPS64R6) in QEMU
Test: test-art-host-gtest

Change-Id: I333dca43fca57ae7e6021bb84585487c889417c3

7 years agoRevert "Revert "Revert "Revert "Basic obsolete methods support""""
Alex Light [Wed, 21 Dec 2016 16:20:29 +0000 (08:20 -0800)]
Revert "Revert "Revert "Revert "Basic obsolete methods support""""

A GetDeclaringClass()->GetDexCache() got inserted during the
merge/review process meaning that we would try to access incorrect
dex-cache in obsolete methods in some situations.

Also when using tracing we would loop forever (or at least until an
OOM error) in test 916 due to tracing forcing InterpretOnly mode
meaning methods would never be jitted.

Bug: 32369913
Bug: 33630159

Test: ART_TEST_TRACE=true \
      ART_TEST_JIT=true   \
      ART_TEST_INTERPRETER=true mma -j40 test-art-host

This reverts commit f6abcda293b115a9d7d8a26376ea2dcf2d1dc510.

Change-Id: I0773bfcba52e3cd51a83be815c6a50c189558f48

7 years agoMerge "Refactored compilation on host to own method. Added dx option."
Aart Bik [Wed, 21 Dec 2016 16:59:12 +0000 (16:59 +0000)]
Merge "Refactored compilation on host to own method. Added dx option."

7 years agoDo not record dependencies on arrays.
Nicolas Geoffray [Wed, 21 Dec 2016 15:06:46 +0000 (15:06 +0000)]
Do not record dependencies on arrays.

We do not record dependencies on arrays with component types in
the compiled DEX files, as the only thing that might change is their
access flags. If we were to change these flags in a breaking way, we would
need to enforce full verification again anyways.

Saves on doing type resolution during vdex update.

Test: test-art-host
Change-Id: Ib85e172a5cb99c2999eed8b6b37390aa6b5cac37

7 years agoSupport non-exact upper bound of HBoundType.
Nicolas Geoffray [Tue, 20 Dec 2016 09:25:42 +0000 (09:25 +0000)]
Support non-exact upper bound of HBoundType.

It used to not affect the analysis, but in order to recognize:
if (foo.getClass() == Foo.class)

We now need to differentiate between an exact and a non-exact
upper bound. Until now the "exact" flag on upper bounds was
there for convenience:
1) re-use of the same data structure
2) we mostly cared about HLoadClass as upper bounds.

Test: test-art-host
Test: master build with prebuilts
Test: apk scanner
Change-Id: I3a1f5e4d489a715a4545becee4c8c80f7d152fc0

7 years agoMerge "Update vdex in place if input_vdex_fd == output_vdex_fd"
Nicolas Geoffray [Wed, 21 Dec 2016 14:12:41 +0000 (14:12 +0000)]
Merge "Update vdex in place if input_vdex_fd == output_vdex_fd"

7 years agoMerge "Point optimizations for vdex."
Nicolas Geoffray [Wed, 21 Dec 2016 13:30:48 +0000 (13:30 +0000)]
Merge "Point optimizations for vdex."

7 years agoMerge "Optimize OatWriter when we don't compile any method."
Nicolas Geoffray [Wed, 21 Dec 2016 12:30:10 +0000 (12:30 +0000)]
Merge "Optimize OatWriter when we don't compile any method."

7 years agoMerge "Remove soon to be obsolete call kinds for direct calls."
Nicolas Geoffray [Wed, 21 Dec 2016 12:17:42 +0000 (12:17 +0000)]
Merge "Remove soon to be obsolete call kinds for direct calls."

7 years agoPoint optimizations for vdex.
Nicolas Geoffray [Wed, 21 Dec 2016 10:29:43 +0000 (10:29 +0000)]
Point optimizations for vdex.

- Do not record assignability due to not optimized CHECKCAST.
- Do not record that j.l.Object must not be assignable to other types.
- Chase the super class boundary to avoid recording a dependency
  on a local class. This avoids doing type resolution of that class when
  verifying the VerifierDeps.

Test: test-art-host
bug: 30937355
Change-Id: Ibcee205451f23958c759ddcca7f88fe9003d37a9

7 years agoOptimize OatWriter when we don't compile any method.
Nicolas Geoffray [Tue, 20 Dec 2016 21:15:00 +0000 (21:15 +0000)]
Optimize OatWriter when we don't compile any method.

- Don't write any quickening info.
- Don't visit methods.

Saves ~20% of compilation times.

Test: test-art-host

Change-Id: Ib18fd06c0ca42308e1d81401de0ee3e6297de0ce

7 years agoUpdate vdex in place if input_vdex_fd == output_vdex_fd
Nicolas Geoffray [Tue, 20 Dec 2016 13:17:09 +0000 (13:17 +0000)]
Update vdex in place if input_vdex_fd == output_vdex_fd

Test: test-art-host
bug: 30937355
Change-Id: Ib8180d67996faec518d9092725b5de00d4dba9f6

7 years agoMerge "Disable 080-oom-fragmentation."
Nicolas Geoffray [Wed, 21 Dec 2016 10:26:25 +0000 (10:26 +0000)]
Merge "Disable 080-oom-fragmentation."

7 years agoMerge "Revert "Revert "Revert "Basic obsolete methods support""""
Nicolas Geoffray [Wed, 21 Dec 2016 09:26:36 +0000 (09:26 +0000)]
Merge "Revert "Revert "Revert "Basic obsolete methods support""""

7 years agoRevert "Revert "Revert "Basic obsolete methods support"""
Nicolas Geoffray [Wed, 21 Dec 2016 09:26:18 +0000 (09:26 +0000)]
Revert "Revert "Revert "Basic obsolete methods support"""

Fails in tracing mode

Bug: 32369913
Bug: 33630159

This reverts commit ce77fc0e7f60a15354bb20c356537cbf8b53b722.

Change-Id: I1bdcf6ad467f2e31f9c5d0c3c987b90a4f5efc69

7 years agoMerge "Revert "Create test for field access in obsolete methods.""
Nicolas Geoffray [Wed, 21 Dec 2016 09:25:42 +0000 (09:25 +0000)]
Merge "Revert "Create test for field access in obsolete methods.""

7 years agoRevert "Create test for field access in obsolete methods."
Nicolas Geoffray [Wed, 21 Dec 2016 09:25:19 +0000 (09:25 +0000)]
Revert "Create test for field access in obsolete methods."

Fails on ndebug.

This reverts commit 322214405bf43684dbbde2fce549b2dad53fb04a.

Change-Id: I6a6c6c4bfb230eef1b7c4cb35f7e8051561d9e78

7 years agoDisable 080-oom-fragmentation.
Nicolas Geoffray [Wed, 21 Dec 2016 08:58:27 +0000 (08:58 +0000)]
Disable 080-oom-fragmentation.

bug:33795328
Test: test-art-host
Change-Id: I1aa87d8564c930f5ad5f8d5c4ac75f432cb6e7e7

7 years agoMerge "Revert "Cleanup jvmti tests""
Nicolas Geoffray [Wed, 21 Dec 2016 08:56:10 +0000 (08:56 +0000)]
Merge "Revert "Cleanup jvmti tests""

7 years agoRevert "Cleanup jvmti tests"
Nicolas Geoffray [Wed, 21 Dec 2016 08:55:49 +0000 (08:55 +0000)]
Revert "Cleanup jvmti tests"

Breaks ndebug.

This reverts commit ca5b9dccced35a2dbf9626d92d3a44886aee097b.

Change-Id: I1f45fa43f554e9796f397b2f9f111689c3845692

7 years agoRefactored compilation on host to own method. Added dx option.
Aart Bik [Wed, 21 Dec 2016 03:06:37 +0000 (19:06 -0800)]
Refactored compilation on host to own method. Added dx option.

Test: jfuzzing itself
Bug: 31267855
Change-Id: I9224075ea17b57c8c36b74fa4b9e401e26d3c96e

7 years agoMerge "Temporarily disable tests 914, 915, 917, & 918 on JIT."
Alex Light [Wed, 21 Dec 2016 01:57:50 +0000 (01:57 +0000)]
Merge "Temporarily disable tests 914, 915, 917, & 918 on JIT."

7 years agoTemporarily disable tests 914, 915, 917, & 918 on JIT.
Alex Light [Wed, 21 Dec 2016 00:15:45 +0000 (16:15 -0800)]
Temporarily disable tests 914, 915, 917, & 918 on JIT.

Since these tests are very sensitive to the positioning of
runtime methods due to bugs we will temporarily disable them.

Bug: 33630159
Bug: 33616143
Test: ART_TEST_JIT=true mma -j40 test-art-host
Change-Id: I38795029eed2ea8416d24a259f3a7925d95684c6

7 years agoRevert "Disable 626-const-class-linking for JIT."
Vladimir Marko [Tue, 20 Dec 2016 21:26:13 +0000 (21:26 +0000)]
Revert "Disable 626-const-class-linking for JIT."

The failures were presumably fixed by
    https://android-review.googlesource.com/314764

Test: m ART_TEST_JIT=true test-art-host-run-test-626-const-class-linking
Bug: 33567581

This reverts commit 0392920cb9f3dc76cbc9d2339b9ca00ba8a8269c.

Change-Id: I13ab17269f456580d57ca090ff44e8d087d770d4

7 years agoTests on FP Math.abs()
Aart Bik [Mon, 19 Dec 2016 21:57:31 +0000 (13:57 -0800)]
Tests on FP Math.abs()

Rationale:
As a "quality of implementation" rather than pure "spec compliance"
we require that Math.abs() clears the sign bit (but changes nothing
else) for all numbers, including NaN.

These tests ensure that remains the case.

**NOTE**
This CL depends on the pending libcore change
(https://android-review.googlesource.com/#/c/316368/).

Test: test-art-host
Bug: 30758343

Change-Id: I01044a73992b2e89d97060ed8bbb76abea0933b9

7 years agoMerge "Fix transfer over rejected induction."
Aart Bik [Tue, 20 Dec 2016 20:46:20 +0000 (20:46 +0000)]
Merge "Fix transfer over rejected induction."

7 years agoFix transfer over rejected induction.
Aart Bik [Tue, 20 Dec 2016 19:13:03 +0000 (11:13 -0800)]
Fix transfer over rejected induction.

Rationale:
With the more precise rejection of narrowing
linear induction, parent rules should be
prepared to reject failed transfers. Also
added a bit more comments to clarify rules.
With regression tests.

Bug: 33774618
Test: test-art-host
Change-Id: I4a206e51d4359ab383379914dd4697fc81903547

7 years agoMerge "Handle ART_USE_READ_BARRIER as a product variable."
Hiroshi Yamauchi [Tue, 20 Dec 2016 18:56:19 +0000 (18:56 +0000)]
Merge "Handle ART_USE_READ_BARRIER as a product variable."

7 years agoMerge "Create test for field access in obsolete methods."
Alex Light [Tue, 20 Dec 2016 18:45:54 +0000 (18:45 +0000)]
Merge "Create test for field access in obsolete methods."

7 years agoMerge changes Ie74f2dbe,I092d2e95,I5a75f5c4
Treehugger Robot [Tue, 20 Dec 2016 17:49:02 +0000 (17:49 +0000)]
Merge changes Ie74f2dbe,I092d2e95,I5a75f5c4

* changes:
  Fix 616-cha assumptions.
  Fix JIT crash due to unverified dead code
  Use interpret-only instead of verify-at-runtime when testing JIT

7 years agoCreate test for field access in obsolete methods.
Alex Light [Tue, 20 Dec 2016 16:55:35 +0000 (08:55 -0800)]
Create test for field access in obsolete methods.

Test: ./test/run-test --host 918
Change-Id: Idf27a0a713e12876e35de5c24bf5d12984881927

7 years agoMerge changes I8cc87bac,If1a92e47
Alex Light [Tue, 20 Dec 2016 16:42:09 +0000 (16:42 +0000)]
Merge changes I8cc87bac,If1a92e47

* changes:
  Cleanup jvmti tests
  Revert "Revert "Basic obsolete methods support""

7 years agoFix 616-cha assumptions.
Calin Juravle [Mon, 19 Dec 2016 15:38:07 +0000 (15:38 +0000)]
Fix 616-cha assumptions.

616-cha relies that some classes are not loaded at startup. Disable the
app image to prevent that.

Test: ART_TEST_JIT=true m test-art-host-run-test-616-cha
Change-Id: Ie74f2dbeae0f898d42051f248492126615ffdad6

7 years agoFix JIT crash due to unverified dead code
Calin Juravle [Tue, 20 Dec 2016 14:36:59 +0000 (14:36 +0000)]
Fix JIT crash due to unverified dead code

The JIT compiler assumes that it only gets completely verified code. To
work around potential unverified dead code it uses kAccDontBotherCompile
flag set during runtime verification. However, if a class is verified
during a prior dex2oat the flag is not persisted and JIT happily things
that everything is ok.

The simplest fix is to mark classes with potential unverified dex code
as verify at runtime. We only do this for apps and assume that
everything in the boot image is well formed.

Test: m test-art-host
Bug: 31000839
Change-Id: I092d2e9553cd1c577036d78e8563a7a39d6cb7b9

7 years agoUse interpret-only instead of verify-at-runtime when testing JIT
Calin Juravle [Tue, 20 Dec 2016 14:00:48 +0000 (14:00 +0000)]
Use interpret-only instead of verify-at-runtime when testing JIT

That's because we run most of our products with interpret-only and not
verify-at-runtime.

Test: ART_TEST_JIT=true m test-art-host
Change-Id: I5a75f5c47a9e811db3ee7c3475a03fb648dba80d

7 years agoFix incorrect string hash value extension during cross-compilation.
Alexey Grebenkin [Fri, 2 Dec 2016 14:44:54 +0000 (17:44 +0300)]
Fix incorrect string hash value extension during cross-compilation.

Previouly, having a 32-bit Android device and a 64-bit host to compile
boot.oat could lead to an interning table be fulfilled using one hash
function and be worked with using another hash function (which is caused
by sign extension).
Target case is any string with a negative .GetHashCode().

Test: test-art-host-gtest-intern_table_test

Change-Id: I3f417e1ac990ef681f0651160292130e9b3632f0

7 years agoMerge "ARM: VIXL32: Use 16-bit encoding for B(cond,...) where possible."
Nicolas Geoffray [Tue, 20 Dec 2016 13:52:48 +0000 (13:52 +0000)]
Merge "ARM: VIXL32: Use 16-bit encoding for B(cond,...) where possible."