GitLab

Also fixed some lines that were too long, and a few other minor
details.

Change-Id: I6efba5fb6e03eb5d0a300fddb2a75bf8e2f175cb

Use method signatures, field types and types embedded in dex
insns for type inference. Perform the type inference in two
phases, first a simple pass that records all types implied
by individual insns, and then an iterative pass to propagate
those types further via phi, move, if-cc and aget/aput insns.

Bug: 19419671
Change-Id: Id38579d48a44fc5eadd13780afb6d370093056f9

The ELF specification defines several types which differ between
32-bit ELF and 64-bit ELF.  We used to template all ELF-related
methods on all of those types which was very verbose.

This CL wraps all the types as typedefs in ElfTypes32 and ElfTypes64.
One of those wrappers is then used as the template parameter.

Change-Id: I65247c2c79d92a7c4799e988cf3e4a1b10eb4788

A couple of instructions don't pass the 'Address' to EmitRex64.  This
will cause the incorrect register number to be assembled if the register
is >= 8.

This may cause bad code to be generated in some cases.

Change-Id: I2907ae8b7629ee95d542e3fab429318994a78938
Signed-off-by: Mark Mendell <mark.p.mendell@intel.com>

This happens pretty often and pollutes the logs.

Change-Id: I074783d3cf3519a5186d2dd81c821d97071302e7

HBasicBlock::InsertPhiAfter would not update the last_instruction
pointer when adding at the end of the list. This could cause problems
when iterating over phis backwards. Fortunately, we don't do that
anywhere in the existing code.

Change-Id: I4487265bf2cf3d3819623fafd7ce7c359bac190e

This will allow Checker to differentiate between them.

Change-Id: I972d452a64f1c19d370567b9ef3d620f9b835a9a

Its use in intrinsics has been bogus. In all other instances
it's been used under the assumption that the inferred type
matches the return type of associated calls. However, if the
type inference identifies a type mismatch, the assumption
doesn't hold and there isn't necessarily a valid value that
the function could reasonably return.

Bug: 19918641
Change-Id: I050934e6f9eb00427d0b888ee29ae9eeb509bb3f

On docs this doubles the amount of instructions removed.

When taking the decision to move a null check to the user we did not
verify if the next instruction checks the same object.

Change-Id: I2f4533a4bb18aa4b0b6d5e419f37dcccd60354d2

Change-Id: If3452b616df48ecec0d85ac34d8af9253c3ea925

This is a bogus assertion. We need to handle the case
though to free the existing tlsPtr_.long_jump_context.

Bug: 20449310
Change-Id: Ie1e49dd1fec9defab69f081d1f129aa8dc1dc6bf

Merge "Class.forName(..., ..., null) fixes - Add test for Class.forName(..., ..., null) - Simplify VMStack.getClosestUserClassLoader based on new behavior of Class.forName(..., ..., null)"

- Add test for Class.forName(..., ..., null)
- Simplify VMStack.getClosestUserClassLoader based on new behavior of Class.forName(..., ..., null)

Change-Id: I6bc470e20fa177e8a3debe55c90a84eef7ef518e

Moves functionality to ART from libcore. Precursor to moving
ArtMethods to native. Mostly performance improvements.

N5 perf before (irrelevant results removed):
            Class_getConstructor  962.87 ===========
         Class_getDeclaredMethod 2394.37 ============================
                 Class_getMethod 2509.20 ==============================
               Class_newInstance 1999.81 =======================
                  Method_invokeI 1439.02 =================
          Method_invokePreBoxedI 1415.82 ================
            Method_invokeStaticI 1456.24 =================
    Method_invokeStaticPreBoxedI 1427.32 =================
            Method_invokeStaticV  814.47 =========
                  Method_invokeV  816.56 =========
After:
                       benchmark      ns linear runtime
            Class_getConstructor 1302.04 ================
         Class_getDeclaredMethod 1459.01 ==================
                 Class_getMethod 1560.40 ===================
               Class_newInstance 2029.94 =========================
                  Method_invokeI 1312.89 ================
          Method_invokePreBoxedI 1255.01 ===============
            Method_invokeStaticI 1289.13 ===============
    Method_invokeStaticPreBoxedI 1196.52 ==============
            Method_invokeStaticV  790.82 =========
                  Method_invokeV  791.73 =========

Performance improvements are more than just fixing regressions introduced
in: http://android-review.googlesource.com/#/c/146069/

Bug: 19264997

Change-Id: Ife79c469fdb09f30e3aefcfc3e0ce5ed32303fce

Add a dex2oat option for compiled-methods, a more granular filter
than compiled-classes. Add compiler-driver support for it.

Refactor dex2oat to reuse file reading.

Add a test to oat_test.

Change-Id: I78d0d040bce7738b4bb7aabe7768b5788d2587ac

Add code that tries to provoke dex-cache collisions.

Change-Id: I58c7905597a501f3731efe4c6e1e2f7772e3b312

Allow constant and memory addresses to more X86_64 instructions.

Add memory formats to X86_64 instructions to match.

Fix a bug in cmpq(CpuRegister, const Address&).

Allow mov <addr>,immediate (instruction 0xC7) to be a valid faulting
instruction.

Change-Id: I5b8a409444426633920cd08e09f687a7afc88a39
Signed-off-by: Mark Mendell <mark.p.mendell@intel.com>

Fixes issue 20125400 - ART: Need .eh_frame_hdr and PT_GNU_EH_FRAME
for libunwind.

.eh_frame_hdr serves two purposes. Firstly, it can optionally contain
binary search table for fast eh_frame lookup. This is important for
C++ exception handling, but we do not need it so we omit it.
Secondly, it contains a relative .eh_frame pointer which makes it
easier for run-time code to locate the .eh_frame section.
libunwind seems to rely on this relative pointer.

Bug: 20125400
Change-Id: I7c1e3f68d914f70781404c508395831a3296a7da

The repo manifest on the build server includes prebuilts
only for Arm and host.

Change-Id: Ic9fc762f3e9a4421cc2293b6e95caf196551cbad

This reverts commit 69dcf642.

Change-Id: I1e01c17a3aa8b001ad5659260802e5f5073bf14e