This reverts commit 8b75e6bc35fb3f9c1e788dbd05084c0f4a60a0f3.

The FileCheck tests are not equivalent:

test/CodeGen/X86/tailcall-structret.ll:6:10: error: expected string not found in input
; CHECK: jmp init
         ^
<stdin>:1:2: note: scanning from here
 .section __TEXT,__text,regular,pure_instructions
 ^
<stdin>:13:2: note: possible intended match here
 jmp _init ## TAILCALL
 ^

llvm-svn: 175124

CoalescerPairs. Also, make it take a CoalescerPair directly like other methods
of RegisterCoalescer.

llvm-svn: 175123

of the copy is a subregister def. The current code assumes that it can do a full
def of the destination register, but it is not checking that the def operand is
read-undef. It also doesn't clear the subregister index of the destination in
the new instruction to reflect the full subregister def.

These issues were found running 'make check' with my next commit that enables
rematerialization in more cases.

llvm-svn: 175122

llvm-svn: 175121

Since functions with internal linkage don't have language linkage, it is valid
to overload them:

extern "C" {
       static int foo();
       static int foo(int);
}

So we mangle them.

llvm-svn: 175120

I added hasCLanguageLinkage while fixing some language linkage bugs some
time ago so that I wouldn't have to check all users of isExternC. It turned
out to be a much longer detour than expected, but this patch finally
merges the two again. The isExternC function now implements just the
standard notion of having C language linkage.

llvm-svn: 175119

Adding a new docs directory, with files auto-generated by sphinx-quickstart,
for user documentation for the various tools in the clang-tools-extra
repository.

Author: Jack Yang <jack.yang@intel.com>
Reviewers: gribozavr, silvas
llvm-svn: 175118

some cases where functions with no language linkage were being treated as having
C language linkage. In particular, don't warn in

extern "C" {
  static NonPod foo();
}

Since getLanguageLinkage checks the language linkage, the linkage computation
cannot use the language linkage. Break the loop by checking just the context
in the linkage computation.

llvm-svn: 175117

instantiation in order to permit devirtualization later in codegen, skip over
pure functions since those can't be devirtualization targets.

llvm-svn: 175116

llvm-svn: 175115

- test could be re-written at some point, but right now just trying to clean up known failures for the linux buildbot.

llvm-svn: 175114

llvm-svn: 175113

Some grammar fixes to 'Format String Checking', and reorder the text slightly to try to make the final code block actually get rendered.

llvm-svn: 175112

- PR-15260: lldb does not display correct value of 1-bit field
- PR-16261: lldb does not display size of (file/class) static array

llvm-svn: 175111

llvm-svn: 175110

llvm-svn: 175109

llvm-svn: 175108

llvm-svn: 175107

llvm-svn: 175106

llvm-svn: 175105

It's possible (e.g. after an LTO build) that an internal global may be used for
debugging purposes. If that's the case appending a '.b' to it makes it hard to
find that variable. Steal the name from the old GV before deleting it so that
they can find that variable again.

llvm-svn: 175104

Skip another two other tests asserting on Linux in RecordLayoutBuilder::updateExternalFieldOffset().

llvm-svn: 175103

llvm-svn: 175102

Parse objective C information as efficiently as possible and without taking dangerous runtime locks.

Reworked the way objective C information is parsed by:
1 - don't read all class names up front, this is about 500K of data with names
2 - add a 32 bit hash map that maps a hash of a name to the Class pointer (isa)
3 - Improved name lookups by using the new hash map
4 - split up reading the objc runtime info into dynamic and shared cache since the shared cache only needs to be read once.
5 - When reading all isa values, also get the 32 bit hash instead of the name
6 - Read names lazily now that we don't need all names up front
7 - Allow the hash maps to not be there and still have this function correctly

There is dead code in here with all of the various methods I tried. I want to check this in first to not lose any of it in case we need to revert to any of the extra code. I will promptly cleanup and commit again.

llvm-svn: 175101

in the course of property synthesis deterministic (ordered
by their type size), instead of having hashtable order
(as it is currently). // rdar://13192366

llvm-svn: 175100

inside an #if 0 block.

llvm-svn: 175098

llvm-svn: 175097

NOTE: This is a candidate for the Mesa stable branch.
llvm-svn: 175096

Skip test that causes lldb to assert due to inferior getting a SIGSEGV with si_code == SI_KERNEL (PR-15258)
- crash does not always reproduce

llvm-svn: 175095

- handle m_resume_state == eStateStopped || eStateSuspended in DoResume rather than asserting

llvm-svn: 175094

llvm-svn: 175093

Fixes PR15250!

llvm-svn: 175092

builtins.

<rdar://problem/13204027>

llvm-svn: 175091

llvm-svn: 175090

llvm-svn: 175089

llvm-svn: 175088

llvm-svn: 175087

instead of redefining separate instructions for them.

llvm-svn: 175086

llvm-svn: 175085

llvm-svn: 175084