Fixed the redo.py script to emit correct arch and compiler options after dotest.py was switched over to use argparse.

llvm-svn: 165507

- Teach it about dadd[i] instructions and move pseudo-instruction
- Make it parse the register names correctly (for N32 / N64)

llvm-svn: 165506

Add in the first step of the multiple pointer support. This adds in support to the data layout for specifying a per address space pointer size.
The next step is to update the optimizers to allow them to optimize the different address spaces with this information.

llvm-svn: 165505

Determine supported archs for compiler-rt libraries on Linux by trying to compile a simple executable

llvm-svn: 165504

Fixup for r165097: build 32-bit ASan compiler-rt library on 64-bit Linux only if just-built clang can build simple 32-bit executables

llvm-svn: 165503

write out the macro history for that macro. Similarly, we need to cope
with reading a macro definition that has been #undef'd.

Take advantage of this new ability so that global code-completion
results can refer to #undef'd macros, rather than losing them
entirely. For multiply defined/#undef'd macros, we will still get the
wrong result, but it's better than getting no result.

llvm-svn: 165502

Clearing a SmallPtrSet is still expensive, split it out from OverloadCandidateSet::clear and don't do it on destruction.

llvm-svn: 165501

llvm-svn: 165498

llvm-svn: 165497

llvm-svn: 165496

llvm-svn: 165495

llvm-svn: 165494

llvm-svn: 165493

llvm-svn: 165492

DeadArgumentElimination pass can replace one LLVM function with another,
invalidating a pointer stored in debug info metadata entry for this function.
To fix this, we collect debug info descriptors for functions before
running a DeadArgumentElimination pass and "patch" pointers in metadata nodes
if we replace a function.

llvm-svn: 165490

opaque layer is responsible for knowing where that specific attribute is stored.

llvm-svn: 165489

We use the enums to query whether an Attributes object has that attribute. The
opaque layer is responsible for knowing where that specific attribute is stored.

llvm-svn: 165488

Patch by: Laszlo Nagy

llvm-svn: 165486

llvm-svn: 165485

llvm-svn: 165484

Separate AVXCC and SSECC printing for cmpps/pd/ss/sd and add masking before the switch statement. This keeps the unreachable default case from being hit if the instruction was created with an intrinsic with too large of an immediate.

llvm-svn: 165483

lit -jN works on cygwin in most cases, but still sometimes I can see stalls with iterative run on the buildbot.

llvm-svn: 165482

llvm-svn: 165481

In parseMSRMaskOperand, add an explicit check for the operand being an identifier instead of just having an assert.

llvm-svn: 165480

Remove some dead code from ARMAsmPrinter. Add virtual and LLVM_OVERRIDE to the other methods. Mark some of the helper methods as private.

llvm-svn: 165479

Remove some unused code in the X86AsmPrinter. Add LLVM_OVERRIDE and virtual keywords to the remaining interface methods.

llvm-svn: 165478

  * Add relocation reference support for Hexagon and PowerPC.
    - Hexagon has many relocation types, this patch covers just
      a subset, more will be added.
    - PowerPC support was added so that we can verify mixed endian
      environments.
  * Add some basic test cases
  * Update ldd-core to accept -endian flag

llvm-svn: 165477

Thanks to Benjamin for the raw test case. This one took about 50 times
longer to reduce than to fix. =/

llvm-svn: 165476

llvm-svn: 165475

llvm-svn: 165474

ASTContext so that it can be widely available.

llvm-svn: 165473

canonical method; avoid storing them again for an out-of-line definition.

llvm-svn: 165472

whether we try to call an external program to load symbols unconditionally,
or if we check the user's preferences before calling it.

ProcessMachCore now sets CanJIT to false - we can't execute code in a core file.

DynamicLoaderDarwinKernel::OSKextLoadedKextSummary::LoadImageUsingMemoryModule changed
to load the kernel from an on-disk file if at all possible.
Don't load the kext binaries out of memory from the remote systems - their linkedit doesn't
seem to be in a good state and we'll error out down in SymbolVendorMacOSX if we try to use
the in-memory images.
Call Symbols::DownloadObjectAndSymbolFile to get the kext/kernel binary -- the external
program may be able to give us a file path on the local filesystem instead of reading
the binary / dSYM over a network drive every time.  Fall back to calling 
Target::GetSharedModule() like before if DownloadObjectAndSymbolFile fails.

llvm-svn: 165471

llvm-svn: 165470

llvm-svn: 165469

llvm-svn: 165468

llvm-svn: 165467

No functionality change intended.

llvm-svn: 165466

llvm-svn: 165465

llvm-svn: 165464