llvm-svn: 197507

The branch offset for a R_MIPS_PC16 relocation is indeed a 16-bit signed
immediate.

llvm-svn: 197506

Trying to fix buildbots after r197503 (test passes locally).

<rdar://problem/15627766>

llvm-svn: 197505

llvm-svn: 197504

This reverts commit r197481, recommiting r197469 with an extra fix.

The vastart_save_xmm_regs pseudo-instruction expands to a test and a
branch, so it modifies EFLAGS.  Mark it so, or else the scheduler might
place it in the middle of another test+branch.

This fixes a bug exposed by r192750, which changed the initial scheduler
to source-order as part of enabling the MI Scheduler for X86.

This re-commit changes the VASTART_SAVE_XMM_REGS custom inserter not to
try to save %flags, and adds a test that catches the bad behavior of
r197469.

<rdar://problem/15627766>

llvm-svn: 197503

llvm-svn: 197502

This will be tested from clang.

llvm-svn: 197501

This makes it identical to the string llvm produces.

llvm-svn: 197500

These will later be used as tests for the dumper part of http://llvm-reviews.chandlerc.com/D2232

llvm-svn: 197499

llvm-svn: 197498

1) Introduce TryConsumeToken() to handle the common test-and-consume pattern.
   This brings about readability improvements in the parser and optimizes to avoid
   redundant checks in the common case.

2) Eliminate the ConsumeCodeCompletionTok special case from ConsumeToken(). This
   was used by only one caller which has been switched over to the more
   appropriate ConsumeCodeCompletionToken() function.

llvm-svn: 197497

Now that we emit diagnostics for keyword-as-identifier hacks (-Wkeyword-compat)
we can go ahead and simplify some of the old revertible keyword support.

This commit adds a TryIdentKeywordUpgrade() function to mirror the recently
added TryKeywordIdentFallback() and uses it to replace the hard-coded list of
REVERTIBLE_TYPE_TRAITs.

llvm-svn: 197496

llvm-svn: 197495

Formatting this:
   void f() { // 1 space initial indent.
      int i;
  #define A   \
      int i;  \
     int j;
        int k; // Format this line.
   }
   void f() {
       #define A 1 // Format this line.
   }

Before:
   void f() { // 1 space initial indent.
      int i;
  #define A   \
      int i;  \
     int j;
     int k;  // Format this line.
   }
   void f() {
   #define A 1  // Format this line.
   }

After:
   void f() { // 1 space initial indent.
      int i;
  #define A   \
      int i;  \
     int j;
      int k;  // Format this line.
   }
   void f() {
  #define A 1  // Format this line.
   }

llvm-svn: 197494

llvm-svn: 197493

http://llvm.org/bugs/show_bug.cgi?id=18045

Short issue description:
For X86 machines with sse < sse4.1 we got failures for some
particular load/store vector sequences:

$ clang-trunk -m32 -O2 test-case.c
fatal error: error in backend: Cannot select: 0x4200920: v4i32,ch = load 0x41d6ab0, 0x4205850,
      0x41dcb10<LD16[getelementptr inbounds ([4 x i32]* @e, i32 0, i32 0)](align=4)> [ORD=82]
      [ID=58]
  0x4205850: i32 = X86ISD::Wrapper 0x41d5490 [ORD=26] [ID=43]
    0x41d5490: i32 = TargetGlobalAddress<[4 x i32]* @e> 0 [ORD=26] [ID=23]
  0x41dcb10: i32 = undef [ID=2]

The reason is that EltsFromConsecutiveLoads could emit such load instruction
both before and after legalize stage. Though this instruction is not legal for
machines with SSSE3 and lower.

The fix: In EltsFromConsecutiveLoads, if we have passed legalize stage, we
check whether nodes it emits are legal. 

P.S.: If you get failure in time from 12:00 and till 22:00 (UTC-8),
perhaps I'll slow with response, so you better reject this commit. Thanks!

llvm-svn: 197492

If 'symbolize' flag is not set, we still want to transform virtual address
to module+offset pair in the call to Symbolizer::SymbolizeCode().
See https://code.google.com/p/address-sanitizer/issues/detail?id=251 for
more details.

llvm-svn: 197491

llvm-svn: 197490

Add an interface for telling LSan that a region of memory is to be treated as a
source of live pointers. Useful for code which stores pointers in mapped memory.

llvm-svn: 197489

llvm-svn: 197488

(Yes, the irony is not lost on me :-) ).

llvm-svn: 197486

Instead, mark the module as unavailable so that clang errors as soon as
someone tries to build this module.

This works towards the long-term goal of not stat'ing the header files at all
while reading the module map and instead read them only when the module is
being built (there is a corresponding FIXME in parseHeaderDecl()).  However, it
seems non-trivial to get there and this unblock us and moves us into the right
direction.

Also changed the implementation to reuse the same DiagnosticsEngine.

llvm-svn: 197485

This is covered by msan_test depending on glibc version and/or resolver setup.

llvm-svn: 197484

when using structured exception handling (SEH) on Windows 64.

http://llvm-reviews.chandlerc.com/D2378

Patch by Jonathan Liu!

llvm-svn: 197483

Added implementation of "truncate" from integer type (i64/i32/i16/i8) to i1.

llvm-svn: 197482

This reverts commit r197469.

The sanitizer and dragonegg buildbots are failing, I think because of
this change.  Reverting until I figure out why.

llvm-svn: 197481

Executable files do not use a string table, so section names longer than 8
characters are not permitted. Long section names should just be truncated.

llvm-svn: 197470

The vastart_save_xmm_regs pseudo-instruction expands to a test and a
branch, so it modifies EFLAGS.  Mark it so, or else the scheduler might
place it in the middle of another test+branch.

This fixes a bug exposed by r192750, which turned on the MI Scheduler
for X86.

<rdar://problem/15627766>

llvm-svn: 197469

With llvm-config.exe --bindir --libdir --build-mode, on Visual Studio 2010,

In build tree:
  (OBJ_ROOT)/bin/MinSizeRel
  (OBJ_ROOT)/lib/MinSizeRel
  MinSizeRel

In installed tree:
  (INSTALL_PREFIX)/bin
  (INSTALL_PREFIX)/lib
  MinSizeRel

This is enhancements since r196283.

llvm-svn: 197467

llvm-svn: 197466

Allow MachineCSE to coalesce trivial subregister copies the same way that it coalesces normal copies.

Without this, MachineCSE is powerless to handle redundant operations with truncated source operands.

This required fixing the 2-addr pass to handle tied subregisters. It isn't clear what combinations of subregisters can legally be tied, but the simple case of truncated source operands is now safely handled:

     %vreg11<def> = COPY %vreg1:sub_32bit; GR32:%vreg11 GR64:%vreg1
     %vreg12<def> = COPY %vreg2:sub_32bit; GR32:%vreg12 GR64:%vreg2
     %vreg13<def,tied1> = ADD32rr %vreg11<tied0>, %vreg12<kill>, %EFLAGS<imp-def>

Test case: cse-add-with-overflow.ll.

This exposed an existing bug in
PPCInstrInfo::commuteInstruction. Thanks to Rafael for the test case:
PowerPC/crash.ll.

llvm-svn: 197465

llvm-svn: 197464

llvm-svn: 197462

llvm-svn: 197461

[CMake][Standalone] Fix llvm-tblgen's directory to use LLVM_TOOLS_BINARY_DIR instead of LLVM_BINARY_DIR.

LLVM_BINARY_DIR points OBJROOT.

FIXME: llvm-config should show *real* bindir with --bindir.
llvm-svn: 197460

Standalone scripts shall be rewritten.

llvm-svn: 197459

Re-reading the comment I updated in previous commit, it's better to make
it more explicit and avoid ambiguity more effectively.

llvm-svn: 197458

llvm-svn: 197457

Make debugging ISel a bit easier by printing out a dump of the generated
code at the end.

llvm-svn: 197456

Tidy up built-in record creation to reduce code duplication.

Continuation of r197336.

llvm-svn: 197452