OSDN Git Service

android-x86/art.git
10 years agoMerge "Fixed ImageWriter timings not being written out."
Brian Carlstrom [Tue, 24 Jun 2014 19:16:20 +0000 (19:16 +0000)]
Merge "Fixed ImageWriter timings not being written out."

10 years agoFixed ImageWriter timings not being written out.
Alex Light [Tue, 24 Jun 2014 17:27:40 +0000 (10:27 -0700)]
Fixed ImageWriter timings not being written out.

During host runs of dex2oat the timings for ImageWriter would not be
written out.

Change-Id: I74ce8bae724d1e732a6a1efb1c82804473081693

10 years agoMerge "Revert "ART: Split out more cases of Load/StoreRef, volatile as parameter""
Andreas Gampe [Tue, 24 Jun 2014 18:44:29 +0000 (18:44 +0000)]
Merge "Revert "ART: Split out more cases of Load/StoreRef, volatile as parameter""

10 years agoRevert "ART: Split out more cases of Load/StoreRef, volatile as parameter"
Andreas Gampe [Tue, 24 Jun 2014 18:42:06 +0000 (18:42 +0000)]
Revert "ART: Split out more cases of Load/StoreRef, volatile as parameter"

This reverts commit 2689fbad6b5ec1ae8f8c8791a80c6fd3cf24144d.

Breaks the build.

Change-Id: I9faad4e9a83b32f5f38b2ef95d6f9a33345efa33

10 years agoMerge "ART: Split out more cases of Load/StoreRef, volatile as parameter"
Andreas Gampe [Tue, 24 Jun 2014 17:29:59 +0000 (17:29 +0000)]
Merge "ART: Split out more cases of Load/StoreRef, volatile as parameter"

10 years agoMerge "Add some more instruction support to optimizing compiler."
Dave Allison [Tue, 24 Jun 2014 16:08:12 +0000 (16:08 +0000)]
Merge "Add some more instruction support to optimizing compiler."

10 years agoAdd some more instruction support to optimizing compiler.
Dave Allison [Tue, 17 Jun 2014 03:44:29 +0000 (20:44 -0700)]
Add some more instruction support to optimizing compiler.

This adds a few more DEX instructions to the optimizing compiler's
builder (constants, moves, if_xx, etc).

Also:
* Changes the codegen for IF_XX instructions to use a condition
  rather than comparing a value against 0.
* Fixes some instructions in the ARM disassembler.
* Fixes PushList and PopList in the thumb2 assembler.
* Switches the assembler for the optimizing compiler to thumb2
  rather than ARM.

Change-Id: Iaafcd02243ccc5b03a054ef7a15285b84c06740f

10 years agoMerge "Use GCC on ARM64"
Sebastien Hertz [Tue, 24 Jun 2014 14:16:32 +0000 (14:16 +0000)]
Merge "Use GCC on ARM64"

10 years agoMerge "Log profile info only on debug builds."
Calin Juravle [Tue, 24 Jun 2014 10:57:35 +0000 (10:57 +0000)]
Merge "Log profile info only on debug builds."

10 years agoMerge "Always compile leafs during profile based compilation."
Calin Juravle [Tue, 24 Jun 2014 10:56:44 +0000 (10:56 +0000)]
Merge "Always compile leafs during profile based compilation."

10 years agoMerge "Use reader lock of bulk free lock when not freeing."
Mathieu Chartier [Tue, 24 Jun 2014 01:35:59 +0000 (01:35 +0000)]
Merge "Use reader lock of bulk free lock when not freeing."

10 years agoUse reader lock of bulk free lock when not freeing.
Mathieu Chartier [Tue, 24 Jun 2014 00:53:09 +0000 (17:53 -0700)]
Use reader lock of bulk free lock when not freeing.

Should help reduce contention observed in systrace.

Change-Id: Iadb81728d4ba797c3a68acea795b15d7f212e89b

10 years agoMerge "Fix implicit check option handling for non-cross-compiles"
Dave Allison [Mon, 23 Jun 2014 22:34:11 +0000 (22:34 +0000)]
Merge "Fix implicit check option handling for non-cross-compiles"

10 years agoFix implicit check option handling for non-cross-compiles
Dave Allison [Mon, 23 Jun 2014 21:46:53 +0000 (14:46 -0700)]
Fix implicit check option handling for non-cross-compiles

This fixes an issue where the compiler was not being told to
generate implicit checks but the runtime was expecting them.

Bug: 15747876
Change-Id: I65e7475bac245c44d5094eb666d67bc1af327ab1

10 years agoMerge "Move trampolines to thumb2 instead of ARM"
Dave Allison [Tue, 24 Jun 2014 01:56:51 +0000 (01:56 +0000)]
Merge "Move trampolines to thumb2 instead of ARM"

10 years agoART: Split out more cases of Load/StoreRef, volatile as parameter
Andreas Gampe [Mon, 23 Jun 2014 20:23:04 +0000 (13:23 -0700)]
ART: Split out more cases of Load/StoreRef, volatile as parameter

Splits out more cases of ref registers being loaded or stored. For
code clarity, adds volatile as a flag parameter instead of a separate
method.

On ARM64, continue cleanup. Add flags to print/fatal on size mismatches.

Change-Id: I30ed88433a6b4ff5399aefffe44c14a5e6f4ca4e

10 years agoMove trampolines to thumb2 instead of ARM
Dave Allison [Mon, 23 Jun 2014 20:19:59 +0000 (13:19 -0700)]
Move trampolines to thumb2 instead of ARM

Currently the entrypoint trampolines are compiled using the ARM32 assembler.
This changes that to use the thumb2 assembler for consistency with the
other compiled code.

Bug: 15455279
Change-Id: I6bacdf359e2ae4c0967fb5cd9dc694af11f802ec

10 years agoLog profile info only on debug builds.
Calin Juravle [Mon, 23 Jun 2014 14:22:29 +0000 (15:22 +0100)]
Log profile info only on debug builds.

Change-Id: I80cce951bf4cce988c9ab8373be3e335860eadf3

10 years agoAlways compile leafs during profile based compilation.
Calin Juravle [Mon, 23 Jun 2014 13:40:31 +0000 (14:40 +0100)]
Always compile leafs during profile based compilation.

Bug:12877748
Change-Id: Idbee9eaa87c419454728de195bdbdd9b0e6b7b50

10 years agoMerge "x86_64: Correct fix for cmp-long"
Bill Buzbee [Mon, 23 Jun 2014 20:16:10 +0000 (20:16 +0000)]
Merge "x86_64: Correct fix for cmp-long"

10 years agox86_64: Correct fix for cmp-long
Serguei Katkov [Mon, 23 Jun 2014 06:23:38 +0000 (13:23 +0700)]
x86_64: Correct fix for cmp-long

We cannot rely on the sign of the sub instruction because
LONG_MAX - LONG_MIN = -1 and the sign will indicate that
LONG_MAX < KONG_MIN and it is incorrect.

The fix also contains small improvement for load wide constant.

Change-Id: I74df70d7c198cebff5cad8c1d5614c1d29b79a1b
Signed-off-by: Serguei Katkov <serguei.i.katkov@intel.com>
10 years agoMerge "Mac doesn't define MAP_ANONYMOUS"
Ian Rogers [Mon, 23 Jun 2014 21:09:35 +0000 (21:09 +0000)]
Merge "Mac doesn't define MAP_ANONYMOUS"

10 years agoMac doesn't define MAP_ANONYMOUS
Ian Rogers [Mon, 23 Jun 2014 21:07:03 +0000 (14:07 -0700)]
Mac doesn't define MAP_ANONYMOUS

Fix mac build.

Change-Id: I607e8fff227259fbf27e468193a705c8c3fc1b59

10 years agoMerge "Revert "Revert "ART: Fix StoreValue to use RefDisp when necessary."""
Andreas Gampe [Mon, 23 Jun 2014 16:04:54 +0000 (16:04 +0000)]
Merge "Revert "Revert "ART: Fix StoreValue to use RefDisp when necessary."""

10 years agoRevert "Revert "ART: Fix StoreValue to use RefDisp when necessary.""
Andreas Gampe [Mon, 23 Jun 2014 15:39:00 +0000 (15:39 +0000)]
Revert "Revert "ART: Fix StoreValue to use RefDisp when necessary.""

This reverts commit e98297bafe84e2ac15eb73739f27826fcaae1203.

No matter if holding a reference or not, the store needs to be tagged
as updating a Dalvik register.

Change-Id: Icb80e1b6f82103cff33ee971faaaa02f9ad4a683

10 years agoMerge "Limit the size of files created by run-test to 2MB."
Ian Rogers [Mon, 23 Jun 2014 15:11:25 +0000 (15:11 +0000)]
Merge "Limit the size of files created by run-test to 2MB."

10 years agoLimit the size of files created by run-test to 2MB.
Ian Rogers [Sun, 22 Jun 2014 05:58:05 +0000 (22:58 -0700)]
Limit the size of files created by run-test to 2MB.

In MemMap::MapAnonymous on the host, avoid creating ashmem regions that are
backed by files, when the size of the file will be greater than the ulimit.

Change-Id: I7cbf2ac59bf5869ed85850bea8d71898f3b7e1c7

10 years agoMerge "X86 Dis: Add missing mov byte; Add size suffixes"
Ian Rogers [Sun, 22 Jun 2014 05:31:52 +0000 (05:31 +0000)]
Merge "X86 Dis: Add missing mov byte; Add size suffixes"

10 years agoX86 Dis: Add missing mov byte; Add size suffixes
Mark Mendell [Thu, 19 Jun 2014 01:02:29 +0000 (21:02 -0400)]
X86 Dis: Add missing mov byte; Add size suffixes

Yet another instruction not disassembled properly.
Add 'b', 'w', 'q' to opcodes to diffferentiate between various versions
and make it more understandable.

Change-Id: Ib794aac660bc8bc4900bfa49eab5aed682996adc
Signed-off-by: Mark Mendell <mark.p.mendell@intel.com>
10 years agoMerge "Revert "ART: Fix StoreValue to use RefDisp when necessary.""
Ian Rogers [Mon, 23 Jun 2014 14:54:11 +0000 (14:54 +0000)]
Merge "Revert "ART: Fix StoreValue to use RefDisp when necessary.""

10 years agoRevert "ART: Fix StoreValue to use RefDisp when necessary."
Ian Rogers [Sun, 22 Jun 2014 07:47:53 +0000 (07:47 +0000)]
Revert "ART: Fix StoreValue to use RefDisp when necessary."

This reverts commit 596bea8f6331fe8ad6eb456afa1f7dfe1b92e09c.

Change-Id: I1942bfc5127e8a631c5b24d107d304c7cd994e78

10 years agoMerge "Use command line android root in dex2oat to find image."
Ian Rogers [Sat, 21 Jun 2014 23:15:24 +0000 (23:15 +0000)]
Merge "Use command line android root in dex2oat to find image."

10 years agoMerge "x86_64: Enable some optimizations"
Bill Buzbee [Sat, 21 Jun 2014 14:59:51 +0000 (14:59 +0000)]
Merge "x86_64: Enable some optimizations"

10 years agoMerge "x86_64: Fix fp-to-core conversion"
Bill Buzbee [Sat, 21 Jun 2014 14:59:44 +0000 (14:59 +0000)]
Merge "x86_64: Fix fp-to-core conversion"

10 years agoMerge "x86_64: Fix comp-long"
Bill Buzbee [Sat, 21 Jun 2014 14:59:34 +0000 (14:59 +0000)]
Merge "x86_64: Fix comp-long"

10 years agox86_64: Fix comp-long
Serguei Katkov [Fri, 20 Jun 2014 11:17:16 +0000 (18:17 +0700)]
x86_64: Fix comp-long

If result and source are the same VR then implementation does
a wrong thing clearing dest in the beginning.

Change-Id: I66d97a828aa38121676561858119ccd44aaa4d50
Signed-off-by: Serguei Katkov <serguei.i.katkov@intel.com>
10 years agoART: Fix StoreValue to use RefDisp when necessary.
Andreas Gampe [Fri, 20 Jun 2014 19:48:12 +0000 (12:48 -0700)]
ART: Fix StoreValue to use RefDisp when necessary.

Split from https://android-review.googlesource.com/#/c/98605

Change-Id: Ib3a03428ddbec659591b838404bbfa1553fe18e9

10 years agoMerge "More verbose logs for JDWP events"
Sebastien Hertz [Mon, 23 Jun 2014 08:02:59 +0000 (08:02 +0000)]
Merge "More verbose logs for JDWP events"

10 years agoMore verbose logs for JDWP events
Sebastien Hertz [Fri, 11 Apr 2014 14:01:17 +0000 (16:01 +0200)]
More verbose logs for JDWP events

Logs more information about the JDWP event being reported with VLOG(jdwp). This
helps mainly when debugging JDWP with -verbose:jdwp option.

Change-Id: I49495bdab03dbcbdc0b6c00a8df718bfedbad91d

10 years agox86_64: Fix fp-to-core conversion
Serguei Katkov [Fri, 20 Jun 2014 09:45:52 +0000 (16:45 +0700)]
x86_64: Fix fp-to-core conversion

Long max value cannot be represented with double precision
and check fp against max_long does not guard invocation of
conversion instruction. As a result conversion ends up with
min long instead of max long.

The patch changes the guard check to not allow conversion
instruction for max long.

Change-Id: Ied761051ec27cf6c833040c25a2c61ab9fcea414
Signed-off-by: Serguei Katkov <serguei.i.katkov@intel.com>
10 years agoUse GCC on ARM64
Sebastien Hertz [Fri, 20 Jun 2014 10:22:11 +0000 (12:22 +0200)]
Use GCC on ARM64

This automatically enables the computed-goto interpreter as default.

Bug: 15763315
Bug: 15405621
Change-Id: I4f6ed5f7cad96333bc8dc1029e9b90ba1b6916df

10 years agoMerge "ART: Method verifier must check invoke-virtuals for non-zero args"
Andreas Gampe [Sat, 21 Jun 2014 01:38:46 +0000 (01:38 +0000)]
Merge "ART: Method verifier must check invoke-virtuals for non-zero args"

10 years agoART: Method verifier must check invoke-virtuals for non-zero args
Andreas Gampe [Fri, 20 Jun 2014 01:13:29 +0000 (18:13 -0700)]
ART: Method verifier must check invoke-virtuals for non-zero args

Compiler checks rely on all instructions, not just reachable ones,
so add two new verifier flags.

Bug: 15755602
Change-Id: Ia9c2146cf82d94ce4d69fb6f7be6450137bb84bd

10 years agoMerge "Shared single GC iteration accounting for all GCs."
Mathieu Chartier [Fri, 20 Jun 2014 20:01:28 +0000 (20:01 +0000)]
Merge "Shared single GC iteration accounting for all GCs."

10 years agoShared single GC iteration accounting for all GCs.
Mathieu Chartier [Sun, 15 Jun 2014 22:15:43 +0000 (15:15 -0700)]
Shared single GC iteration accounting for all GCs.

Previously, each garbage collector had data that was only used
during collection. Since only one collector can be running at any
given time, we can make this data be shared between all collectors.
This reduces memory usage since we don't need to have redundant
information for each GC types. Also reduced how much code is required
to sweep spaces.

Bug: 9969166
Change-Id: I31caf0ee4d572f75e0c66863fe7db12c08ae08e7

10 years agoMerge "ART: Do not try to pretty-print method in dex file"
Andreas Gampe [Fri, 20 Jun 2014 05:58:42 +0000 (05:58 +0000)]
Merge "ART: Do not try to pretty-print method in dex file"

10 years agoART: Do not try to pretty-print method in dex file
Andreas Gampe [Fri, 20 Jun 2014 03:24:22 +0000 (20:24 -0700)]
ART: Do not try to pretty-print method in dex file

The code is called in verification when it is not clear yet whether
structures are valid. Simplify warning message.

Bug: 15754233
Change-Id: I85858a71143f41444ce9f00536d901c268afe547

10 years agoMerge "ART: Method verifier needs to check 2-reg values"
Andreas Gampe [Fri, 20 Jun 2014 03:39:07 +0000 (03:39 +0000)]
Merge "ART: Method verifier needs to check 2-reg values"

10 years agoMerge "ART: ReadFully needs to fail when EOF"
Andreas Gampe [Fri, 20 Jun 2014 19:53:56 +0000 (19:53 +0000)]
Merge "ART: ReadFully needs to fail when EOF"

10 years agoART: ReadFully needs to fail when EOF
Andreas Gampe [Fri, 20 Jun 2014 17:45:30 +0000 (10:45 -0700)]
ART: ReadFully needs to fail when EOF

Besides an explicit error, we should stop trying when we unexpectedly
reach the end of the file.

Change-Id: I16d601e1b5bcbc39cb8c4dd7fd3dbb7d69016579

10 years agoMerge "ART: Fix StoreValue to use RefDisp when necessary."
Andreas Gampe [Fri, 20 Jun 2014 19:53:22 +0000 (19:53 +0000)]
Merge "ART: Fix StoreValue to use RefDisp when necessary."

10 years agoMerge "Quick compiler: enable Arm64 compilation"
buzbee [Fri, 20 Jun 2014 19:45:44 +0000 (19:45 +0000)]
Merge "Quick compiler: enable Arm64 compilation"

10 years agoQuick compiler: enable Arm64 compilation
buzbee [Thu, 19 Jun 2014 23:08:48 +0000 (16:08 -0700)]
Quick compiler: enable Arm64 compilation

Change-Id: I19058defaff1c60b4b6b4d1ad82e0a42d50506a6

10 years agoMerge "ART: Fix register overlap checks in CopyArgumentRegs"
Bill Buzbee [Fri, 20 Jun 2014 19:42:30 +0000 (19:42 +0000)]
Merge "ART: Fix register overlap checks in CopyArgumentRegs"

10 years agoART: Fix register overlap checks in CopyArgumentRegs
Andreas Gampe [Fri, 20 Jun 2014 18:34:17 +0000 (11:34 -0700)]
ART: Fix register overlap checks in CopyArgumentRegs

This is a make-shift CL. Expect this to be replaced by a reworked
implementation.

Change-Id: Ia74697d1436efd2971bc4c791fabed66d2e9d72d

10 years agoMerge "Add more read barriers for JNI roots."
Hiroshi Yamauchi [Fri, 20 Jun 2014 19:30:53 +0000 (19:30 +0000)]
Merge "Add more read barriers for JNI roots."

10 years agoAdd more read barriers for JNI roots.
Hiroshi Yamauchi [Wed, 18 Jun 2014 20:47:35 +0000 (13:47 -0700)]
Add more read barriers for JNI roots.

To make it possible to concurrently scan the JNI global roots (that
is, the roots visited by JavaVMExt::VisitRoots()), add read barriers
to the indirect reference table and the reference table.

Also, add read barriers to the jmethodID/jfieldID decode routines
(ScopedObjectAccessAlreadyRunnable::DecodeField/DecodeMethod) so that
we can concurrently handle (encoded) fields and methods.

Bug: 12687968
Change-Id: I3df4e4e622a572ff0ea8d44b2dc70a4d6b3ba058

10 years agoART: Method verifier needs to check 2-reg values
Andreas Gampe [Fri, 20 Jun 2014 00:29:48 +0000 (17:29 -0700)]
ART: Method verifier needs to check 2-reg values

The method verifier did not check against the second register
for a J or D parameter from a method signature.

The register line had a wrong DCHECK that did not catch this even
in debug mode.

Bug: 15751498
Change-Id: Ic6af08bf4704b3ab0f308dd9f0da28691a4cb024

10 years agoMerge "Quick compiler: remove Arm64 "EXPERIMENTAL" filter"
buzbee [Thu, 19 Jun 2014 23:05:02 +0000 (23:05 +0000)]
Merge "Quick compiler: remove Arm64 "EXPERIMENTAL" filter"

10 years agoQuick compiler: remove Arm64 "EXPERIMENTAL" filter
buzbee [Thu, 19 Jun 2014 22:04:12 +0000 (15:04 -0700)]
Quick compiler: remove Arm64 "EXPERIMENTAL" filter

Support for the compilation of all dex opcodes should now exist.
We'll go ahead and leave the filter code in for a new weeks, as it
may be handy if we need a quick workaround.

Change-Id: I8f1857136d78347b7d2337db2c793ad41ad45cb9

10 years agoMerge "x86_64: Several fixes required for enabling promotion"
Bill Buzbee [Thu, 19 Jun 2014 21:44:43 +0000 (21:44 +0000)]
Merge "x86_64: Several fixes required for enabling promotion"

10 years agoMerge "x86_64: Fix neg_double"
Bill Buzbee [Thu, 19 Jun 2014 21:17:44 +0000 (21:17 +0000)]
Merge "x86_64: Fix neg_double"

10 years agoMerge "ART: Log information when skipping method during compilation"
Andreas Gampe [Thu, 19 Jun 2014 20:43:58 +0000 (20:43 +0000)]
Merge "ART: Log information when skipping method during compilation"

10 years agoART: Log information when skipping method during compilation
Andreas Gampe [Thu, 19 Jun 2014 18:34:06 +0000 (11:34 -0700)]
ART: Log information when skipping method during compilation

Add a reason for skipping a method to the log.

Change-Id: I3b31ee64cce6b531b25397f646cdfee650704ad6

10 years agoMerge "x86_64: Enable core.oat/boot.oat compilation"
Bill Buzbee [Thu, 19 Jun 2014 20:18:35 +0000 (20:18 +0000)]
Merge "x86_64: Enable core.oat/boot.oat compilation"

10 years agox86_64: Fix neg_double
Alexei Zavjalov [Wed, 18 Jun 2014 10:18:36 +0000 (17:18 +0700)]
x86_64: Fix neg_double

In a case, when src and dest regs are the same, previous
implementation of the neg_double bytecode returns an
incorrect result.

This implementation uses shifts and xor and works for both
cases.

Change-Id: I137d9a90298ec225d80435d35558da8abb69cd01
Signed-off-by: Alexei Zavjalov <alexei.zavjalov@intel.com>
10 years agoMerge "ART: Implicit checks in the compiler are independent from Runtime"
Andreas Gampe [Thu, 19 Jun 2014 06:56:51 +0000 (06:56 +0000)]
Merge "ART: Implicit checks in the compiler are independent from Runtime"

10 years agoART: Implicit checks in the compiler are independent from Runtime
Andreas Gampe [Tue, 17 Jun 2014 23:36:07 +0000 (16:36 -0700)]
ART: Implicit checks in the compiler are independent from Runtime

When cross-compiling, those flags are independent. This is an
initial CL that helps bypass fatal failures when cross-compiling,
as not all architectures support (and have turned on) implicit
checks.

The actual transport for the target architecture when it is
different from the runtime needs to be implemented in a follow-up
CL.

Bug: 15703710
Change-Id: Idc881a9a4abfd38643b862a491a5af9b8841f693

10 years agoMerge "ART: Reserve 8B for float literals on ARM64"
Andreas Gampe [Thu, 19 Jun 2014 06:42:33 +0000 (06:42 +0000)]
Merge "ART: Reserve 8B for float literals on ARM64"

10 years agoART: Reserve 8B for float literals on ARM64
Andreas Gampe [Thu, 19 Jun 2014 06:19:07 +0000 (23:19 -0700)]
ART: Reserve 8B for float literals on ARM64

This is a cludge to make sure literal pools will be 8B aligned.
A better approach would be post-processing - that way we could
pack the floats.

Change-Id: Ia9c6f53e0f6e37b4be79e1efe7f14feb49f2052b

10 years agoMerge "ART: Change rrr add and sub for ARM64"
Andreas Gampe [Thu, 19 Jun 2014 02:39:37 +0000 (02:39 +0000)]
Merge "ART: Change rrr add and sub for ARM64"

10 years agoART: Change rrr add and sub for ARM64
Andreas Gampe [Thu, 19 Jun 2014 00:45:32 +0000 (17:45 -0700)]
ART: Change rrr add and sub for ARM64

OpRegRegImm will fall back to loading a constant into a register
and then doing the operation with three registers. That is, for
example, the case when we allocate large stack frames. However,
the currently chosen operations are add/sub shifted, which does
*not* allow to specify SP (x31 will be interpreted as xzr). Switch
to add/sub extended. There won't be a practical difference, as we
do not call with anything other than 0 shift.

Change-Id: I2b78df9f044d2963e3e890777c855b339952f9f4

10 years agoMerge "Assign a lower Locklevel for Mutexes used in QuasiAtomic operations. This...
Ian Rogers [Wed, 18 Jun 2014 23:24:40 +0000 (23:24 +0000)]
Merge "Assign a lower Locklevel for Mutexes used in QuasiAtomic operations. This fixes the CompilerDriverTest on MIPS. The test was failing when Transaction::Abort() acquired intern_table_lock_ and log_lock_ and subsequently performed QuasiAtomic operations."

10 years agoMerge "ART: Target-dependent stack overflow, less check elision"
Andreas Gampe [Thu, 19 Jun 2014 02:39:21 +0000 (02:39 +0000)]
Merge "ART: Target-dependent stack overflow, less check elision"

10 years agoART: Target-dependent stack overflow, less check elision
Andreas Gampe [Thu, 19 Jun 2014 00:01:15 +0000 (17:01 -0700)]
ART: Target-dependent stack overflow, less check elision

Refactor the separate stack overflow reserved sizes from thread.h
into instruction_set.h and make sure they're used in the compiler.

Refactor the decision on when to elide stack overflow checks:
especially with large interpreter stack frames, it is not a good
idea to elide checks when the frame size is even close to the
reserved size. Currently enforce checks when the frame size is
>= 2KB, but make sure that frame sizes 1KB and below will elide
the checks (number from experience).

Bug: 15728765
Change-Id: I016bfd3d8218170cbccbd123ed5e2203db167c06

10 years agoMerge "Ensure classes are initialized rather than initializing."
Ian Rogers [Wed, 18 Jun 2014 23:56:39 +0000 (23:56 +0000)]
Merge "Ensure classes are initialized rather than initializing."

10 years agoEnsure classes are initialized rather than initializing.
Ian Rogers [Wed, 18 Jun 2014 23:07:20 +0000 (16:07 -0700)]
Ensure classes are initialized rather than initializing.

A class can be being initialized on a different thread, in that case other
threads should block trying to access the class. The initializing state shows
the class is being initialized but not that its safe for other threads to
access. Change occurances of IsInitializing to IsInitialized primarily in
slow-path code.
Bug: 15347354

Change-Id: Ib586d0a385be6086a890dfbf8868d76f16767fac

10 years agoMerge "ART: Start implementation of OpRegRegRegExtend for ARM64"
Andreas Gampe [Thu, 19 Jun 2014 17:23:37 +0000 (17:23 +0000)]
Merge "ART: Start implementation of OpRegRegRegExtend for ARM64"

10 years agoART: Start implementation of OpRegRegRegExtend for ARM64
Andreas Gampe [Thu, 19 Jun 2014 08:10:07 +0000 (01:10 -0700)]
ART: Start implementation of OpRegRegRegExtend for ARM64

We need a sign-extending add for packed-switch and sparse-switch,
as the 32b values are signed offsets. This starts an implementation
that is sufficient for the use cases.

Change-Id: Ib5bae24b902077346a97d5e9e061533f9cdfcdb0

10 years agox86_64: Enable some optimizations
Dmitry Petrochenko [Fri, 13 Jun 2014 07:51:01 +0000 (14:51 +0700)]
x86_64: Enable some optimizations

Except kLoadStoreElimination and kPromoteRegs.

Change-Id: Iaebf6652641fe32c9e9b21469e4f898af2d67250
Signed-off-by: Dmitry Petrochenko <dmitry.petrochenko@intel.com>
10 years agox86_64: Enable core.oat/boot.oat compilation
Dmitry Petrochenko [Fri, 13 Jun 2014 08:15:32 +0000 (15:15 +0700)]
x86_64: Enable core.oat/boot.oat compilation

This patch enabled compilation of image.

Change-Id: Idacce4a3898964d3a05762925dba236d7254a254
Signed-off-by: Dmitry Petrochenko <dmitry.petrochenko@intel.com>
10 years agox86_64: Several fixes required for enabling promotion
Dmitry Petrochenko [Wed, 18 Jun 2014 12:11:41 +0000 (19:11 +0700)]
x86_64: Several fixes required for enabling promotion

This patch includes x86_64 fixes which solve the issues
reproducible with promotion optimization enabled.

Change-Id: I3a3b0d290380d639705cd58f38b8d62b1eb930d3
Signed-off-by: Serguei Katkov <serguei.i.katkov@intel.com>
Signed-off-by: Dmitry Petrochenko <dmitry.petrochenko@intel.com>
10 years agoMerge "Fix systrace logging, total paused time, and bytes saved message."
Mathieu Chartier [Tue, 17 Jun 2014 23:00:01 +0000 (23:00 +0000)]
Merge "Fix systrace logging, total paused time, and bytes saved message."

10 years agoFix systrace logging, total paused time, and bytes saved message.
Mathieu Chartier [Tue, 17 Jun 2014 22:04:40 +0000 (15:04 -0700)]
Fix systrace logging, total paused time, and bytes saved message.

Moved the GC top level systrace logging to be inside of Collector::Run.
This prevents cases where we forgot to call it such as background
compaction. Fixed a unit error regarding total pause time. Fixed
negative bytes saved to use the word "expanded".

Bug: 15702709

Change-Id: Ic2991ecad2daa000d0aee9d559b8bc77d8c160aa

10 years agoMerge "Add x86 inlined abs method for float/double"
Vladimir Marko [Wed, 18 Jun 2014 11:14:22 +0000 (11:14 +0000)]
Merge "Add x86 inlined abs method for float/double"

10 years agoAdd x86 inlined abs method for float/double
Yixin Shou [Wed, 5 Mar 2014 11:07:48 +0000 (06:07 -0500)]
Add x86 inlined abs method for float/double

Add the optimized implementation of inlined abs method for
float/double for X86 side.

Change-Id: I2f367542f321d88a976129f9f7156fd3c2965c8a
Signed-off-by: Yixin Shou <yixin.shou@intel.com>
10 years agoMerge "Fix art_quick_instrumentation_exit for ARM64"
Sebastien Hertz [Thu, 19 Jun 2014 14:55:02 +0000 (14:55 +0000)]
Merge "Fix art_quick_instrumentation_exit for ARM64"

10 years agoFix art_quick_instrumentation_exit for ARM64
Sebastien Hertz [Thu, 19 Jun 2014 09:51:41 +0000 (11:51 +0200)]
Fix art_quick_instrumentation_exit for ARM64

Stores the integer result in x0 into the stack. Also avoids clobbering integer
result in x0 by setting it to xSELF after copying it to x2 when calling the
artInstrumentationMethodExitFromCode function.

Bug: 15690361
Change-Id: I9fb873cf32f194779f299607dda8a7316a99c528

10 years agoMerge "Add an option to enable collecting dex pc in ART profiler"
Calin Juravle [Thu, 19 Jun 2014 12:09:33 +0000 (12:09 +0000)]
Merge "Add an option to enable collecting dex pc in ART profiler"

10 years agoAdd an option to enable collecting dex pc in ART profiler
Wei Jin [Mon, 9 Jun 2014 23:19:15 +0000 (16:19 -0700)]
Add an option to enable collecting dex pc in ART profiler

This CL allows the ART profiler to collect both method signature and
dex pc. The type of the profile data is controlled by the option
"-Xprofile-type:{method,dexpc}". To avoid conflicting with the
original profile data based on method signatures, the new profile
data are stored in files with extension ".pc".

Change-Id: I8afb2541d386bff77c5b07fc9367d817a79f58e1

10 years agoMerge "Add mark compact collector."
Mathieu Chartier [Thu, 19 Jun 2014 20:10:51 +0000 (20:10 +0000)]
Merge "Add mark compact collector."

10 years agoAdd mark compact collector.
Mathieu Chartier [Tue, 10 Jun 2014 18:22:31 +0000 (11:22 -0700)]
Add mark compact collector.

The mark compact collector is a 4 phase collection, doing a normal
full mark_sweep, calculating forwarding addresses of objects in the
from space, updating references of objects in the from space, and
moving the objects in the from space.

Support is diabled by default since it needs to have non movable
classes and field arrays. Performance numbers is around 50% as fast.

The main advantage that this has over semispace is that the worst
case memory usage is 50% since we only need one space isntead of two.

TODO: Make field arrays and classes movable. This causes complication
since Object::VisitReferences relies on these, so if we update the
fields of an object but another future object uses this object to
figure out what fields are reference fields it doesn't work.

Bug: 14059466

Change-Id: I661ed3b71ad4dde124ef80312c95696b4a5665a1

10 years agoMerge "Fix GetMethodID to find an interface method from a super-interface"
Brian Carlstrom [Wed, 18 Jun 2014 16:44:50 +0000 (16:44 +0000)]
Merge "Fix GetMethodID to find an interface method from a super-interface"

10 years agoFix GetMethodID to find an interface method from a super-interface
Brian Carlstrom [Wed, 18 Jun 2014 15:34:01 +0000 (08:34 -0700)]
Fix GetMethodID to find an interface method from a super-interface

Bug: 15651032
Change-Id: I5df113e9489a1615e901f03bed92ed2a1a42bd67

10 years agoMerge "Add missing parens to Makefile rule."
Nicolas Geoffray [Wed, 18 Jun 2014 15:08:40 +0000 (15:08 +0000)]
Merge "Add missing parens to Makefile rule."

10 years agoAdd missing parens to Makefile rule.
Nicolas Geoffray [Wed, 18 Jun 2014 15:07:49 +0000 (16:07 +0100)]
Add missing parens to Makefile rule.

Change-Id: I5bff24a696a07c734d05d0ee42f3cf6cd57c0904

10 years agoMerge "Build rules for running run-test with optimizing compiler."
Nicolas Geoffray [Wed, 18 Jun 2014 14:37:12 +0000 (14:37 +0000)]
Merge "Build rules for running run-test with optimizing compiler."

10 years agoMerge "Enable background compaction by default"
Mathieu Chartier [Tue, 17 Jun 2014 22:28:15 +0000 (22:28 +0000)]
Merge "Enable background compaction by default"

10 years agoMerge "AArch64: Fix OpCmpMemImmBranch."
Bill Buzbee [Wed, 18 Jun 2014 13:17:33 +0000 (13:17 +0000)]
Merge "AArch64: Fix OpCmpMemImmBranch."

10 years agoAArch64: Fix OpCmpMemImmBranch.
Zheng Xu [Tue, 17 Jun 2014 10:17:31 +0000 (18:17 +0800)]
AArch64: Fix OpCmpMemImmBranch.

The temp register can be 64-bit in some cases(ArgReg or RefReg).
Always compare 32-bit value no matter what the temp register is.

Change-Id: Ib237dd081da0b5900b8c2418df1621d3245cb03d