This patch adjust the r171506 to make all DWARF enconding pc-relative
for PPC64. It also adds the R_PPC64_REL32 relocation handling in MCJIT
(since the eh_frame will not generate PIC-relative relocation) and also
adds the emission of stubs created by the TTypeEncoding.

llvm-svn: 171979

Fixes pr14861.

llvm-svn: 171978

llvm-svn: 171977

subclass of TargetMachine which "forwards" all operations to an
existing internal TargetMachine member variable. In the usage context the
specific-machine class derived from TargetMachine is not visible,
only a reference to the generic base class TargetMachine. Although
getSubtargetImpl() is public in specific-machine classes derived from
TargetMachine, the TargetMachine class unfortunately has
getSubtargetImpl() protected (and accessing non-const members makes
abusing getSubtarget() unsuitable). Making it public in the base class
allows this forwarding pattern.

llvm-svn: 171976

Some linux distibutions (for example, Mageia 2, Fedora 17) ship Clang that is
essentially broken for the end user.  Clang can not find or compile libstdc++
headers.

The issue is that our configure prefers clang over gcc, thus selecting a broken
Clang when a working GCC is available.

Now we detect this issue by compiling a simple program.  If it does not
compile, configure stops with an error suggesting the user to select a
different compiler.

llvm-svn: 171975

Previously, we'd always start at indent level 0 after a preprocessor
directive, now we layout the following snippet (column limit 69) as
follows:

functionCallTo(someOtherFunction(
    withSomeParameters, whichInSequence,
    areLongerThanALine(andAnotherCall,
  B
                       withMoreParamters,
                       whichStronglyInfluenceTheLayout),
    andMoreParameters),
               trailing);

Note that the different jumping indent is a different issue that will be
addressed separately.

This is the first step towards handling #ifdef->#else->#endif chains
correctly.

llvm-svn: 171974

[asan] make the slow unwinder a bit more robust. The unittests pass with fast_unwind_on_fatal=0, but I still observe some differences between the two unwinders

llvm-svn: 171973

When processing possible aliases, TableGen assumes that if an operand *can* be
an immediate, then it always *will* be. This is incorrect for the AArch64
backend. This patch inserts a check in the generated code to make sure isImm is
true first.

llvm-svn: 171972

fp128 is almost but not quite completely illegal as a type on AArch64. As a
result it needs to have a register class (for argument passing mainly), but all
operations need to be lowered to runtime calls. Currently there's no way for
targets to do this (without duplicating code), as the relevant functions are
hidden in SelectionDAG. This patch changes that.

llvm-svn: 171971

llvm-svn: 171970

llvm-svn: 171969

llvm-svn: 171968

llvm-svn: 171967

This fixes llvm.org/PR14870 and we no longer mess up:
template <typename T1, typename T2 = char, typename T3 = char,
          typename T4 = char>
void f();

It removes the nice aligment for assignments inside other expressions,
but I am not sure those are actually practically relevant. If so, we can
fix those later.

llvm-svn: 171966

We need this to correctly fix ASTMatchers/ASTTypeTraits.h in clang.

llvm-svn: 171965

This addresses llvm.org/PR14847.

We can now format something like:
int aaaaaaaaaaaaaaaaaaaaaaaaaaa =
    // aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaiaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
    + bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb;

clang-format unavoidably exceeds the column limit, but does not just
flush everything into a single line. Moreover, it tries to minimize the
number of characters beyond the column limit.

llvm-svn: 171964

This fixes llvm.org/PR14846.

Before:
template <typename T>
void aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa(
        int aaaaaaaaaaaaaaaaa);

After:
template <typename T>
void aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa(
    int aaaaaaaaaaaaaaaaa);

llvm-svn: 171963

docs.

llvm-svn: 171962

This fixes llvm.org/PR14860.

Before, we messed up the format of:
if (DeclaratorInfo.isFunctionDeclarator() &&
    //getDeclSpecContextFromDeclaratorContext(Context) == DSC_top_level &&
    Tok.is(tok::semi) && NextToken().is(tok::l_brace)) {
}

llvm-svn: 171961

llvm-svn: 171960

Before: int ( ^ Block1) (int, int) = ^ (int i, int j)
After:  int (^Block1) (int, int) = ^(int i, int j)
llvm-svn: 171959

llvm-svn: 171958

This addresses llvm.org/PR14864.

We used to completely mess this up and now format as:
Diag(NewFD->getLocation(),
     getLangOpts().MicrosoftExt ? diag::ext_function_specialization_in_class :
         diag::err_function_specialization_in_class)
    << NewFD->getDeclName();

llvm-svn: 171957

llvm-svn: 171956

Need a more reliable way to check for support of this intrinsic.

llvm-svn: 171955

llvm-svn: 171954

PR 14848. The lowered sequence is based on the existing sequence the target-independent
DAG Combiner creates for the scalar case.

Patch by Zvi Rackover.

llvm-svn: 171953

Clang doesn't understand it.

If someone wants to teach clang about that flag instead, please also add
support for -fno-diagnostics-parseable-fixits for consistency.

Until then, let the documentation match the current behavior.

llvm-svn: 171952

llvm-svn: 171951

address space. Reordered the EmitULEB128IntValue arguments to
make this easier.

llvm-svn: 171949

llvm-svn: 171948

llvm-svn: 171947

This was an experimental option, but needs to be defined
per-target. e.g. PPC A2 needs to aggressively hide latency.

I converted some in-order scheduling tests to A2. Hal is working on
more test cases.

llvm-svn: 171946

Add a "--reverse" or "-r" option to the "list" with no options command.  This will list backwards from the
last source point listed.
Also fix the setting of the default file & line to the file containing main, when you do a plain "list".

<rdar://problem/12685226>

llvm-svn: 171945

Make sure clang puts tokens from different files on separate lines in "-E -P" mode.  <rdar://problem/12774044>
 

llvm-svn: 171944

test/DebugInfo/member-pointers.ll would not fail in targetting BE any more.

llvm-svn: 171943

Now BangOperator should be nicely hyperlinked.

Pointed out by Joel Jones.

llvm-svn: 171942

Suggested by Joel Jones.

llvm-svn: 171941

llvm-svn: 171940

llvm-svn: 171939