This required disabling a PowerPC optimization that did the following:
input:
x = BUILD_VECTOR <i32 16, i32 16, i32 16, i32 16>
lowered to:
tmp = BUILD_VECTOR <i32 8, i32 8, i32 8, i32 8>
x = ADD tmp, tmp

The add now gets folded immediately and we're back at the BUILD_VECTOR we
started from. I don't see a way to fix this currently so I left it disabled
for now.

Fix some trivially foldable X86 tests too.

llvm-svn: 174325

llvm-svn: 174324

llvm-svn: 174322

llvm-svn: 174321

We found that findAll has been implemented incorrectly multiple times
by various people using the matchers. To prevent further wasted
development effort, it makes sense to add it as convenience matcher
implemented as eachOf(m, forEachDescendant(m)).

This patch also updates the docs with the new matchers.

llvm-svn: 174320

llvm-svn: 174319

llvm-svn: 174318

definition

Current practice is not to use 'inline' in:

  class Foo {
  public:
    inline void bar() {
      // ...
    }
  };

llvm-svn: 174317

llvm-svn: 174316

eachOf gives closure on the forEach and forEachDescendant matchers.
Before, it was impossible to implement a findAll matcher, as matching
the node or any of its descendants was not expressible (since anyOf
only triggers the first match).

llvm-svn: 174315

This is a follow up to r174309 to actually make it work.

llvm-svn: 174314

Makefile.config.

This is implied at the bottom of the help text of configure (besides
CC/CXX/LDFLAGS, already passed to Makefile.config).

For backward compatibility, the values of CFLAGS and CXXFLAGS defaults
to empty, overriding the default values provided by autoconf (for
example, '-g -O2' when CC=gcc').

$(CPP) is not used by our makefiles. Therefore, the value of CPP is
not passed to Makefile.config, despite beeing mentioned by 'configure
--help'.

llvm-svn: 174313

llvm-svn: 174312

llvm-svn: 174311

If there are string literals on either side of a '<<', chances are
high that they represent logically separate concepts. Otherwise,
the author could just have just a single literal (possible split
over multiple lines).

So, we can now nicely format things like:
cout << "somepacket = {\n"
     << "  val a = " << ValueA << "\n"
     << "  val b = " << ValueB << "\n"
     << "}";

llvm-svn: 174310

We can now (even in non-bin-packing modes) format:
someFunction(1, /* comment 1 */
             2, /* comment 2 */
             3, /* comment 3 */
             aaa);

llvm-svn: 174309

Two minor changes:
* Slight penalty for breaking at "," as opposed to ";".
* Don't apply bin-packing rules to for-loops.

Before:
for (int aaaaaa = aaaaaaaaaa; aaaaaa < bbbbbbbb; ++aaaaaa,
         ++ccccccccccccccc) {}

After:
for (int aaaaaa = aaaaaaaaaa; aaaaaa < bbbbbbbb;
     ++aaaaaa, ++ccccccccccccccc) {}

llvm-svn: 174308

This combines several changes:
* Calculation token type (e.g. for * and &) in the AnnotatingParser.
* Calculate the scope binding strength in the AnnotatingParser.
* Let <> and [] scopes bind stronger than () and {} scopes.
* Add minimal debugging output.

llvm-svn: 174307

This change lets us bootstrap LLVM/Clang under ASan and MSan. It contains
fixes for 2 issues:

- X86JIT reads return address from stack, which MSan does not know is
  initialized.
- bugpoint tests run binaries with RLIMIT_AS. This does not work with certain
  Sanitizers.

We are no longer including config.h in Compiler.h with this change.

llvm-svn: 174306

llvm-svn: 174305

llvm-svn: 174304

clang/Analysis: Fix r174245, a valgrind error in AnalysisDeclContext::getBody(bool &IsAutosynthesized), to initialize IsAutosynthesized explicitly.

llvm-svn: 174303

Removed reference to LLVM as a project (since in LangRef it is used solely as a reference to the IR). Thanks silvas!

llvm-svn: 174301

Swift has a renaming dependency if we load into D subregisters. We don't have a
way of distinguishing between insertelement operations of values from loads and
other values. Therefore, we are pessimistic for now (The performance problem
showed up in example 14 of gcc-loops).

radar://13096933

llvm-svn: 174300

Added support to the cmake build to turn off uninitialized use warnings
for gcc. This cleans the build up somewhat.

Used logic simpler than found in autoconf by making use of the fact that
although gcc won't complain about unsupported -Wno-* flags it *will*
complain about unsupported -W flags.

Reviewers: gribozavr, doug.gregor, chandlerc
llvm-svn: 174299

llvm-svn: 174298

llvm-svn: 174297

This is an improvement of r173630, that handles the following case:

  struct VirualDestrClass
  {
    VirualDestrClass(int arg);
    virtual ~VirualDestrClass();
  };

  struct ConstrWithCleanupsClass
  {
    ConstrWithCleanupsClass(const VirualDestrClass& cplx = VirualDestrClass(42));
  };

  ConstrWithCleanupsClass cwcNoArg;

That was printed as:

  ConstrWithCleanupsClass cwcNoArg();

llvm-svn: 174296

Added instance variable/initializers/getter/setters for new keyword externally initialized to GlobalVariable. No *TRUE* functionality change.

I am going to add in the actual test cases with the actual functionality
changes in a later patch because I want to include some test cases.

To be clear when I say no *TRUE* functionality change I mean that this
patch (like it says in the title) only contains getters/setters and sets
up a default initial value of the instance variable to false so that
this patch does not affect any other uses of Global Variable.h.

llvm-svn: 174295

llvm-svn: 174294

This can yield dramatic speedups of dynamic_cast for simple inheritance trees,
at least with libsupc++. Neither libcxxabi nor libcxxrt make use of this
hint currently, it was never implemented because clang didn't support it.

There was some concern about the number of class hierarchy walks this change
introduces. If it turns out to be an issue we can add caching either at the cast
pair level or even deeper, but we also do a lot of walks in Sema so this
codepath is probably fairly optimized already.

llvm-svn: 174293

llvm-svn: 174292

We can now format:
SomeArrayOfSomeType a = { { { 1, 2, 3 } }, { { 1, 2, 3 } },
                          { { 111111111111111111111111111111,
                              222222222222222222222222222222,
                              333333333333333333333333333333 } },
                          { { 1, 2, 3 } }, { { 1, 2, 3 } } };

Before, we did strange things there.

llvm-svn: 174291

llvm-svn: 174290

This allows the optimizer to CSE dynamic_casts.

llvm-svn: 174289

llvm-svn: 174288

llvm-svn: 174287

This removes quite a few 'cxstring::' qualifications where they are obvious.

llvm-svn: 174286

llvm-svn: 174285

llvm-svn: 174284