GitLab

This change uses libcorkscrew to show native stacks for threads in kNative or,
unlike dalvikvm, kVmWait --- working on the runtime directly I've found it
somewhat useful to be able to see _which_ internal resource we're waiting on.
We can always take that back out (or make it oatexecd-only) if it turns out to
be too noisy/confusing for app developers.

This change also lets us rank mutexes and enforce -- in oatexecd -- that you
take locks in a specific order.

Both of these helped me test the third novelty: removing the heap locking from
ScopedThreadListLock. I've manually inspected all the callers and added a
ScopedHeapLock where I think one is necessary. In manual testing, this makes
jdb a lot less prone to locking us up. There still seems to be a problem with
the JDWP VirtualMachine.Resume command, but I'll look at that separately. This
is a big enough and potentially disruptive enough change already.

Change-Id: Iad974358919d0e00674662dc8a69cc65878cfb5c

Change the Arm & Mips instruction templaces to record instruction
size in bytes rather than half-words.  Also includes a few Mips
changes to get us in compilable state.

Change-Id: I5a4f6cbd0cb0569805d9dfbd341c244152e59ac7

Updating the MIPS target to use the now common codegen routines.
Still much to do, but the general structure is sufficient to allow
work to begin on the other target.

Change-Id: I0d288fdfb59c8e76fad73185fdd56b345e87b604

MIPS ICS devices today are mips32r2. Tested LLVM on MIPS devices without
synci and rdhwr instructions to make sure LLVM/mips32r2 is working.

Change-Id: If1db7fe268e9f3b7d27eedbacd1d55620d57612a

Change-Id: Id5be7b5484623bb40687df400b5b4370dbb7cdbe

Change-Id: Ib13fc8c0c8f39288a5c6e994928c0f4787a16a4c

Change-Id: Ied2808806dabc03e115402e8a453ce57c1318693

Materialize the bitcode in the end of CompileClass,
so that the code generation time won't be calculated
in method compilation time.

Change-Id: I1e2bd0b23bcb5c55bd966a31b1df0b8b34922f94

Change-Id: Ib30784651bc30b841db1e4248bef061933b1fae9

Change-Id: I14ce622686d44e975e2760e2f1c1e0e64e83125f

Change-Id: I40a77c816231663bf19c635af69f48f5e10363fc

Cleanup some code to follow the coding convention,
such as changing "T const*" to "const T*".

Change-Id: I1a96f17a86b7bfafea2474834d6a1eaaa5a834f6

Change-Id: I56b946a6dd2410ef7835068c0a104cb582fe4a10

Change-Id: I81f79997e339a0f8d947daddf44b1f18b4c276f2

Change-Id: I0b897b7dddc927c8f8dbdc73f3598304a435819e

The passed tests include the 12 originally ARM-only tests. Now it works
on x86 as well if defined(ART_USE_LLVM_COMPILER).

Change-Id: I03092637fa4f0979ca77e0cac06e5d31a867e465

I added a new part to the test, but only ran it in --dev mode, which doesn't
check the expectations, which I forgot to update. Also make naming and comment
improvements.

Change-Id: Ie2cac9f3b0837eda4a821d972f72c96e0316337a

also add --host-prefix defaulting to oatdump matching oatexec

Change-Id: I0e82f78510be18dcb09dbb1ec14931fe46173a68

Continuing to move target-specific code from the Arm
code generator into the independent realm.  This will be
done in multiple small steps.

In this CL, the focus is on unifying the LIR data structure and
various enums that don't really need to be target specific. Also
creates two new shared source files: GenCommon.cc (to hold
top-level code generation functions) and GenInvoke.cc (which
is likely to be shared only by the Arm and Mips targets).

Also added is a makefile hack to build for Mips (which we'll
eventually remove when the compiler support multiple targets
via the command line) and various minor cleanups.

Overall, this CL moves more than 3,000 lines of code from
target dependent to target independent.

Change-Id: I431ca4ae728100ed7d0e9d83a966a3f789f731b1

Change-Id: I8faf7427944324c9bac12573d217cde2a2e658f1

Merge "Fix exception delivery for exceptions caught by a static method called on behalf of <clinit>." into dalvik-dev

We'd already fixed ToDexPC, but needed to fix ToNativePC in the same way.

Change-Id: If8594beec710a39529b15499faeeedd41c9a42f2

Merge "Add a test for correct delivery of exceptions caught by methods called from <clinit>." into dalvik-dev

Change-Id: Ie770240962959b92753e4b55acc0f81f1afe2b68

(The unused code was used for heap trimming in dalvikvm.)

Change-Id: I9087726629f442a54952cc0495daaa6066430141

Change-Id: Ibb897cd2b59284b1b6915d3f6f53a110ca72275e

Change-Id: Ia9a52cced70b8c44d1354e3342ad27f212c8d993

Generic failures detected at compile time will not be fixed at runtime,
so classes with generic failures are now added to a rejected classes
set and not compiled at all. This prevents the compiler from
segfaulting on certain occasions when registers and indexes are out of
range, etc. There is one remaining segfault in the vm-tests now.

Change-Id: Id67c12fd13f3e993a6a16a8625620daa0ea496cb

I used the wrong bound for finding the end address, and just dumping
plain bytes is far more useful. These bytes can be filtered through
"llvm-mc -disassemble -triple=thumbv7-unknown-linux" to get mostly-readable
disassembly (it doesn't seem to cope with the single-register PUSH our
methods start with, but gets in sync).

Change-Id: Ic7e07f1330b352110a4517c1939ec7c11ae65e0a

Change-Id: Ia987e1392dc1814dcb29e2a26c8a5d7b12d86f21

Change-Id: Iafe3ff9f528c6a1573b7358b16566af745ff3c9f

Extract method-related information, such as class_loader,
class_linker, dex_file, dex_cache, code_item, method_idx,
and access_flags from art::CompilationUnit, so that we
can use them in 2 different code generators.

Change-Id: I20631cc73b6f01e9646a983156f3fcb066d732db

Create a helper class to either keep in a short local array or allocate
the arg array. This benefits JNI methods with a small number of
arguments by avoiding an allocation.

Change-Id: I88a9ff6b7ff88b16021813650e9425697e934353

Change-Id: Ib34abfb5c1d2542ff3c3ae9417cd65d174bb6834