| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
| |
A signed value comparison meant that on 64b systems comparisons
were false when pointers > 2GB were in use (as happens in long-running
tests). Fix this to be uint.
Implement a simple wrap-around in the MAP_32BIT emulation code.
Change-Id: I09870b4755f2dca676e42e701fbb6f6eb4bb95d0
|
| |
|
|
|
|
|
|
| |
Some compiled code (probably static methods) is still being invoked
with -Xint. Added an assert to detect this case.
Bug: 13250375
Change-Id: Iecfe8ef40c6c326962593db78e6e1d9f1c93842e
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Work-in-progress to allow arrays to fill usable size. Bug: 13028925.
Use C++11's override keyword on GCC >= 2.7 to ensure that we override GC and
allocator methods.
Move initial mirror::Class set up into a Functor so that all allocated objects
have non-zero sizes. Use this property to assert that all objects are never
larger than their usable size.
Other bits of GC related clean-up, missing initialization, missing use of
const, hot methods in .cc files, "unimplemented" functions that fail at
runtime in header files, reducing header file includes, move valgrind's space
into its own files, reduce number of array allocation routines.
Change-Id: Id5760041a2d7f94dcaf17ec760f6095ec75dadaa
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Introduce ClassLinker::FindArrayClass which performs an array class lookup
given the element/component class. This has a 16 element cache of recently
looked up arrays.
Pass the current thread to ClassLinker Find .. Class routines to avoid calls
to Thread::Current().
Avoid some uses of FindClass in the debugger where WellKnownClasses is a
faster and more compacting GC friendly alternative.
Change-Id: I60e231820b349543a7edb3ceb9cf1ce92db3c843
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
VerifyObject no longer resides in heap. You can now enable
VerifyObject for non-debug builds. VerifyStack is still slow, so it
is now guarded by its own flag.
Fixed the image writer to not use verification at places where
verification fails due to invalid reads.
Fixed RosAlloc to use SizeOf which doesn't call verify object.
Added a flag paremeter to some of the mirror getters / setters to
be able to selectively disable VerifyObject on certain calls.
Optimized the GC to not verify each object multiple times during
object scanning if verify object is enabled.
Added 3 verification options: verify reads, verify this, and verify
writes so that you can select how much verification you want for
mirror getters and setters.
Removed some useless DCHECKs which would slow debug builds without
providing any benefits.
TODO: RosAlloc verification doesn't currently work with verify
objects.
Bug: 12934910
Bug: 12879358
Change-Id: Ic61033104dfc334543f89b0fc0ad8cd4f4015d69
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Removes the class initialization blacklist and use transaction to detect and
revert class initialization attempting to invoke native method. This only
concerns class initialization happening at compilation time when generating an
image (like boot.art for the system).
In transactional mode, we log every object's field assignment and array update.
Therefore we're able to abort a transaction to restore values of fields and
array as they were before the transaction starts. We also log changes to the
intern string table so we can restore its state prior to transaction start.
Since transactional mode only happens at compilation time, we don't need to log
all these changes at runtime. In order to reduce the overhead of testing if
transactional mode is on/off, we templatize interfaces of mirror::Object and
mirror::Array, respectively responsible for setting a field and setting an
array element.
For various reasons, we skip some specific fields from transaction:
- Object's class and array's length must remain unchanged so garbage collector
can compute object's size.
- Immutable fields only set during class loading: list of fields, method,
dex caches, vtables, ... as all classes have been loaded and verified before a
transaction occurs.
- Object's monitor for performance reason.
Before generating the image, we browse the heap to collect objects that need to
be written into it. Since the heap may still holds references to unreachable
objects due to aborted transactions, we trigger one collection at the end of
the class preinitialization phase.
Since the transaction is held by the runtime and all compilation threads share
the same runtime, we need to ensure only one compilation thread has exclusive
access to the runtime. To workaround this issue, we force class initialization
phase to run with only one thread. Note this is only done when generating image
so application compilation is not impacted. This issue will be addressed in a
separate CL.
Bug: 9676614
Change-Id: I221910a9183a5ba6c2b99a277f5a5a68bc69b5f9
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Modify mirror objects so that references between them use an ObjectReference
value type rather than an Object* so that functionality to compress larger
references can be captured in the ObjectRefererence implementation.
ObjectReferences are 32bit and all other aspects of object layout remain as
they are currently.
Expand fields in objects holding pointers so they can hold 64bit pointers. Its
expected the size of these will come down by improving where we hold compiler
meta-data.
Stub out x86_64 architecture specific runtime implementation.
Modify OutputStream so that reads and writes are of unsigned quantities.
Make the use of portable or quick code more explicit.
Templatize AtomicInteger to support more than just int32_t as a type.
Add missing, and fix issues relating to, missing annotalysis information on the
mutator lock.
Refactor and share implementations for array copy between System and uses
elsewhere in the runtime.
Fix numerous 64bit build issues.
Change-Id: I1a5694c251a42c9eff71084dfdd4b51fff716822
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Adds GetWithoutChecks and SetWithoutChecks methods in PrimitiveArray and use
them in the interpreter. Updates Get and Set methods to rely on them and adds
some DCHECK to control exception flow.
Renames IsValidIndex into CheckIsValidIndex to reflect it can throw an
exception. It's also more consistent with ObjectArray::CheckIsAssignable.
Make ThrowArrayIndexOutOfBoundsException private in Array since it's only used
by Array::CheckIsValidIndex.
Updates DoFilledNewArray to use SetWithoutChecks rather than Set.
Change-Id: I2fd314d77a67cf969843d499b86d04ca7b7a43e6
|
| |\ |
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
To comply with a moving collector, we used to load invoke's receiver (for non
static methods only) from the caller shadow frame after resolving the method
itself (in case the garbage collector is triggered inside) but before passing
invoke arguments, including loading receiver from the caller's shadow frame
into the callee's shadow frame. Therefore, we used to load the receiver 3 times
in the fast path but only twice in the slow path.
The slow path is rarely used (only in method requiring extra runtime checks) so
we now move this extra reload to the slow path. Therefore an invoke using the
fast path loads the receiver twice while the slow path loads it 3 times.
I don't expect much improvement here. The main reason is to keep extra code in
the slow path.
Change-Id: I10e96b10de4b8c2992e276bd564bc3e2f191779c
|
| |/
|
|
|
|
|
|
|
|
|
|
| |
Also, make jni_internal_test execute via the interpreter rather than compile
methods. Add tests for passing negative array sizes to JNI routines new
functions. Re-enable the tests NewStringNullCharsNonzeroLength and
NewDirectBuffer_GetDirectBufferAddress_GetDirectBufferCapacity. Test and
explicitly fail if the initial value argument to NewObjectArray isn't
assignable to that type of array.
Use unchecked ObjectArray::Set with NewObjectArray with an initial value.
Change-Id: If3491cb5f974b42cf70c1b850819265f9963ee48
|
| |
|
|
| |
Change-Id: I87f452e338bd4ff0587e3fc7b0bec3f08a1e7fe6
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Only the Class#forName(String) call was handled during the compilation
stage, but VMClassLoader#loadClass and VMClassLoader#findLoadedClass
gets called through various other channels. Add these so we cover most
of the use cases without going too far up the call stack.
This was encountered during compilation of code that did:
MyClass.class.getClassLoader().loadClass(...)
during an attempt to avoid class initialization. Note that the existing
Class#forName handling doesn't appear to ensure the class is initialized
like would normally happen when Class#forName is called in the running
VM. This is a correctness issue, but is left for another change to
correct.
Change-Id: I14628a1d1d5a5dd1a885cbd24455a73bb0531489
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Adds support for switching entrypoints during runtime. Enables
addition of new allocators with out requiring significant copy
paste. Slight speedup on ritzperf probably due to more inlining.
TODO: Ensuring that the entire allocation path is inlined so
that the switch statement in the allocation code is optimized
out.
Rosalloc measurements:
4583
4453
4439
4434
4751
After change:
4184
4287
4131
4335
4097
Change-Id: I1352a3cbcdf6dae93921582726324d91312df5c9
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The compacting collector is currently similar to semispace. It works by
copying objects back and forth between two bump pointer spaces. There
are types of objects which are "non-movable" due to current runtime
limitations. These are Classes, Methods, and Fields.
Bump pointer spaces are a new type of continuous alloc space which have
no lock in the allocation code path. When you allocate from these it uses
atomic operations to increase an index. Traversing the objects in the bump
pointer space relies on Object::SizeOf matching the allocated size exactly.
Runtime changes:
JNI::GetArrayElements returns copies objects if you attempt to get the
backing data of a movable array. For GetArrayElementsCritical, we return
direct backing storage for any types of arrays, but temporarily disable
the GC until the critical region is completed.
Added a new runtime call called VisitObjects, this is used in place of
the old pattern which was flushing the allocation stack and walking
the bitmaps.
Changed image writer to be compaction safe and use object monitor word
for forwarding addresses.
Added a bunch of added SIRTs to ClassLinker, MethodLinker, etc..
TODO: Enable switching allocators, compacting on background, etc..
Bug: 8981901
Change-Id: I3c886fd322a6eef2b99388d19a765042ec26ab99
|
| |
|
|
|
|
| |
These showed up in compaction work.
Change-Id: Iac8eb0a1395c25aabba9f2e0ff6b01fc6180bdca
|
| |
|
|
|
|
|
|
|
|
|
| |
All template functions using thread analysis must be explicitly instantiated
with the thread analysis attributes to enable thread analysis. We use macros
to do this since there are many variants of a same function depending the
number of template arguments.
Also add documentation of these functions.
Change-Id: I3c79acc2f0a6a8dfb5c42924439145292dd68812
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This CL improves the initialization of the new shadow frame (callee) from the
current shadow frame (caller) following invoke format (non-range or range).
This also removes the use of Instruction::GetArgs for non-range invoke when we
can go the fast route (no use of shorty). This avoids decoding arguments into
the stack from the instruction and then read these arguments from the stack. We
now do both at once. Note the slow path (doing runtime checks) is not modified
since it should be rarely used.
Performance improved by 3% in average on our benchmarks. Good news it is more
important on benchmark doing lots of invokes:
- Dhrystone: +11%
- DeltaBlue: +7%
- CaffeineMark Method: +16%
Bug: 10668955
Change-Id: I5c230777b2388a78dca7b3a88189e007c1c2bb7a
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Some cleanup in invocation stuff:
- Get the number of invoke arguments from instruction (vA) rather than get it
from its code item. This benefits to native invoke since we no longer need to
parse the method's shorty. Also pass the low 16 bits of instructions to avoid
fetching it twice when reading vA.
- Remove "is_static" tests by taking advantage of invoke type template argument
rather than testing method's access flags.
- Ensure Instruction::GetArgs is inlined.
- Check exception when initializing method's class when transitioning from
interpreter to compiled code (artInterpreterToCompiledCodeBridge).
- Move UnstartedRuntimeInvoke function to interpreter_common.cc and make it
static as it's only used by DoInvoke and DoInvokeVirtualQuick functions.
- Avoid duplicating code in ShadowFrame::Create.
Performance remains the same according to benchmarks. Hopefully, this should be
addressed in next CLs, especially by improving new shadow frame initialization.
Bug: 10668955
Change-Id: I514b8f098d0ef3e35921ceb770383aac1a9c7902
|
| |
|
|
|
|
|
|
|
|
| |
The verifier should not treat use of unresolved references as a reason to reject
the entire class. Instead, the verifier treats the instruction as a throw. If
that class is run, the interpreter with extra checks will throw an exception.
Bug: 10457426
Change-Id: I3799da843a7ffb3519bbf6dc13a6276519d9cb95
|
|
|
This CL adds a new implementation of the interpreter using computed goto table.
In order to keep the switch-based implementation, it reorders things as the
following:
- Keep interpreter entrypoints into interpreter.h/.cc files.
- Move common interpreter parts to interpreter_common.h/.cc files.
- Separate both implementations to their own modules.
The interpreter implementation can be selected by changing the value of the
kInterpreterImplKind global variable (see interpreter.cc file). The default one
remains the switch-based implementation.
Also updates the exception handling (FindNextInstructionFollowingException) and
SPARSE_SWITCH switch handling (DoSparseSwitch) routines to share code between both
implementations.
Finally, adds a PACKED_SWITCH handling routine (DoPackedSwitch) so we are
consistent with SPARSE_SWITCH handling.
The computed goto implementation use two handlers table: one for normal
instruction handling and one for instrumentation handling. The current handlers
table to be used is updated on backward branch depending on whether there is
listener to DEX pc change.
Bug: 10602809
Change-Id: Ibb53bcc68be75c473fe5440835e78fc9a74381b3
|
