It broke the self host and clang-x86_64-darwin10-RA.

llvm-svn: 139259

llvm-svn: 139258

Need branch relocation support to distinguish this encoding from the
16-bit Thumb1 encoding w/o the explicit .w suffix. That comes later, though.

llvm-svn: 139257

llvm-svn: 139256

llvm-svn: 139255

The immediate offset of the non-writeback i8 form (encoding T4) allows
negative offsets only. The positive offset form of the encoding is the
LDRT instruction. Immediate offsets in the range [0,255] use encoding T3
instead.

llvm-svn: 139254

ctor-initializer, remember to call the Sema action to generate default
ctor-initializers. What a delightful little miscompile. Fixes PR10578
/ <rdar://problem/9877267>.

llvm-svn: 139253

llvm-svn: 139252

llvm-svn: 139251

Second of a three-patch series aiming to fix MSR/MRS on Cortex-M. This adds predicate checking to the Disassembler.

llvm-svn: 139250

llvm-svn: 139249

llvm-svn: 139248

In some cases such as interpreters using indirectbr, the CFG can be very
complicated, and live range splitting may be forced to insert a large
number of phi-defs.  When that happens, traceSiblingValue can spend a
lot of time zipping around in the CFG looking for defs and reloads.

This patch causes more information to be cached in SibValues, and the
cached values are used to terminate searches early.  This speeds up
spilling by 20x in one interpreter test case.  For more typical code,
this is just a 10% speedup of spilling.

llvm-svn: 139247

(MachThreadList::EnableHardwareWatchpoint()) where the watchpoint is not associated
with a thread and the current thread, if set, is returned, otherwise we return the
first thread.

Plus minor change to RNBRemote::HandlePacket_z() to use the existing macros to check
the validity of break_id/watch_id.

llvm-svn: 139246

Fix atomic load and store on x86 to pass -verify-machineinstrs (and possibly fix some subtle bugs involving passes which check mayStore()).

This isn't exactly ideal, but it is good enough for the moment.

llvm-svn: 139245

llvm-svn: 139244

There is no 16-bit wide encoding, so the .w suffix isn't needed (indeed, isn't
documented as allowed). Also add the missing '!' token on the _UPD
variant.

llvm-svn: 139243

Choose 32-bit vs. 16-bit encoding when there's no .w suffix in post-processing
as match classes are insufficient to handle the context-sensitiveness of
the writeback operand's legality for the 16-bit encodings.

llvm-svn: 139242

llvm-svn: 139240

llvm-svn: 139238

Refactor instprinter and mcdisassembler to take a SubtargetInfo. Add -mattr= handling to llvm-mc. Reviewed by Owen Anderson.

llvm-svn: 139237

duplicate tests are eliminated (for example if the two functions both have
a catch clause catching the same type, ensure the redundant one is removed).
Note that it would probably be safe to say that eh.typeid.for is 'const',
but since two calls to it with the same argument can give different results
(but only if the calls are in different functions), it seems more correct to
mark it only 'pure'; this doesn't get in the way of the optimization.

llvm-svn: 139236

in GC mode. // rdar://10073896

llvm-svn: 139235

Handle explicit 'ia' suffix via a MnemonicAlias (pre-existing).

llvm-svn: 139234

(not assert) early.

llvm-svn: 139233

llvm-svn: 139232

llvm-svn: 139230

llvm-svn: 139229

function. This is really the beginning of the second phase of
refactorings here. The end goal is to have (roughly) three interfaces:

1) Base class to format a single diagnostic suitable for display on the
   console.
2) Extension of the base class which also displays a caret diagnostic
   suitable for display on the console.
3) An adaptor that implements the DiagnosticClient by delegating to #1
   and/or #2 as appropriate.

Once we have these, things like libclang's formatDiagnostic can use #1
and #2 to provide really well formatted (and consistently formatted!)
textual formatting of diagnostics.

Getting there is going to be quite a bit of shuffling. I'm basically
sketching out where the interface boundaries can be drawn for #1 and #2
within the existing classes. That lets me shuffle with a minimum of fuss
and delta. Once that's done, and any of the related interfaces that need
to change are updated, I'll hoist these into separate headers and
re-implement libclang in terms of their interfaces. Long WIP, but
comments at each step welcome. =D

llvm-svn: 139228

CaretDiagnostic. It's completely generic, with nothing to do with the
diagnostic client or info APIs.

llvm-svn: 139227

llvm-svn: 139226

llvm-svn: 139225

llvm-svn: 139224

llvm-svn: 139223

are returning a pointer to pointer.

llvm-svn: 139222

Relax the MemOperands on atomics a bit.  Fixes -verify-machineinstrs failures for atomic laod/store on ARM.

(The fix for the related failures on x86 is going to be nastier because we actually need Acquire memoperands attached to the atomic load instrs, etc.)

llvm-svn: 139221

name.

llvm-svn: 139220

ExprResult LHS, RHS,
Expr *LHSExpr, *RHSExpr
QualType LHSType, RHSType

Functions changed:
DiagnoseLogicalAndInLogicalOrLHS()
DiagnoseBinOpPrecedence()
ActOnBinOp()
BuildBinOp()

llvm-svn: 139219

ExprResult LHS, RHS,
Expr *LHSExpr, *RHSExpr
QualType LHSType, RHSType

Functions changed:
CreateBuiltinBinOp()
DiagnoseBitwisePrecedence()

llvm-svn: 139218

a stack array of a magical size with an assert() that we never
overflowed it. That seems incredibly risky. We also have a very nice API
for bundling up a vector we expect to usually have a small size without
loss of functionality or security if the size is excessive.

The fallout is to remove the last pointer+size parameter pair that are
traced through the recursive caret diagnostic emission.

llvm-svn: 139217