This commit catches up to a matching bugfix in patmos-llvm.

The former implementation passed the platina type and expression
infromation by means of two metadata items attached to the intrinsic
call instruction. However, the tail-call-optimization dropped this
metadata (in sync with the docs):

  A transformation is required to drop any metadata attachment that it
  does not know or know it can’t preserve.

Passing this data by means of operands to the call instruction should
be more robust.

What it says on the tin. Parsing only.

What it says on the tin. Parsing only.

This commit refactors clangs handling of `#pragma platina` to be truly
generic for any combination of type and string-expression.
Adding a new `#pragma platina <subcommand>` to the C/C++ frontend
should be as subclassing PragmaPlatinaHandler,
registering/deregistering it with the Parser in
initialize/resetPragmaHandlers, adapting `include/clang/Basic/Attr.td`
with the additional invocation and adding the actual handlers in
patmos-llvm's PMLExport.

The platina intrinsic now takes two metadata items, "platina.type" and
"platina.expr". This commit changes the handling to provide both.

This commit adds a custom pragma to the clang frontend:
  `#pragma platina guard "..."`
It degrades to the llvm's intrinisc::platina, which stores the
annotation string in the associated metadata node.

Rational is to use that pragma to mark viable, but effectively dead
code paths, e.g. code paths that are only entered during an
initialisation phase.

Currently, it is required to state the export filename explicitly via
"-mserialize=out.pml". This makes it hard to easily integrate it into a
Makefile e.g. in the kernel context.
Now it is possible to use "-mserialize-auto", which determines the
output filenames based on the value of the "-o outfile" parameter by
using "outfile.pml" and "outfile.bc" respectively

This reverts commit c6069fb2.

CFRGs can't handle several optimizations. Consequently, those
optimizations (already disabled for patmos) need to be disabled
for the ARM backend, too.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@262836 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@262544 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@262523 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@262365 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r261669 | aaronballman | 2016-02-23 10:55:15 -0800 (Tue, 23 Feb 2016) | 1 line

Amends r252104 to evaluate the controlling expression in an unevaluated context. This eliminates false-positive diagnostics about null pointer dereferences (etc) in the controlling expression.
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261684 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r261422 | rdivacky | 2016-02-20 00:31:24 -0800 (Sat, 20 Feb 2016) | 10 lines

Fix handling of vaargs on PPC32 when going from regsave to overflow.

It can happen that when we only have 1 more register left in the regsave
area we need to store a value bigger than 1 register and therefore we
go to the overflow area. In this case we have to leave the last slot
in the regsave area unused and keep using overflow area. Do this
by storing a limit value to the used register counter in the overflow block.

Issue diagnosed by and solution tested by Mark Millard!

------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261553 91177308-0d34-0410-b5e6-96231b3b80d8

http://reviews.llvm.org/D17323

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261526 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261357 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261343 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r261209 | abataev | 2016-02-18 05:48:15 -0800 (Thu, 18 Feb 2016) | 4 lines

[OPENMP] Fix codegen for lastprivate loop counters.

Patch fixes bug with codegen for lastprivate loop counters. Also it may
improve performance for lastprivates calculations in some cases.
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261257 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r261080 | abataev | 2016-02-17 02:29:05 -0800 (Wed, 17 Feb 2016) | 3 lines

[OPENMP] Fix handling loop-based directives with arrays.
Patch fixes possible problems with correct handling arrays as
expressions in initialization, conditions etc in loop-based constructs.
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261256 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r257763 | djasper | 2016-01-14 05:36:46 -0800 (Thu, 14 Jan 2016) | 8 lines

clang-format: Fix incorrectly enforced linebreak with ColumnLimit 0.

Before:
  aaaa[bbbb]
      .cccc();

After:
  aaaa[bbbb].cccc();
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261225 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r259874 | compnerd | 2016-02-04 20:12:40 -0800 (Thu, 04 Feb 2016) | 8 lines

CodeGen: correct Windows ARM C++ assertion

Because the Decl is explicitly passed as nullptr further up the call chain, it
is possible to invoke isa on a nullptr, which will assert.  Guard against the
nullptr.

Take the opportunity to reuse the helper method rather than re-implementing this
logic.
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261158 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r260851 | dim | 2016-02-14 17:08:20 +0100 (Sun, 14 Feb 2016) | 13 lines

As reported in https://llvm.org/bugs/show_bug.cgi?id=25496, on FreeBSD,
C++ programs compiled for profiling (using `-pg`) should be linked with
`-lc++_p` (or `-lstdc++_p`, depending on the `-stdlib=` setting), not
with the regular C++ libraries.

Add a `FreeBSD::AddCXXStdlibLibArgs()` override to handle this, and add
a test case for it.  While here, extend the test case for the proper
passing of -lm and -lm_p.

Reviewers: compnerd, davide, dws, emaste
Reviewed By: compnerd
Differential Revision: http://reviews.llvm.org/D16264

------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261003 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r260370 | abataev | 2016-02-10 02:50:12 -0800 (Wed, 10 Feb 2016) | 1 line

Fix PR26543: add a check for definition in CXXRecordDecl.
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@261002 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260997 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260992 91177308-0d34-0410-b5e6-96231b3b80d8

By Anastasia Stulova!

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260988 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260984 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r260616 | joerg | 2016-02-11 15:18:36 -0800 (Thu, 11 Feb 2016) | 4 lines

Now that Sparc/Sparc64 backend is mostly usable, provide the same
linking defaults as other NetBSD targets, i.e. compiler_rt-in-libc and
libc++ as STL.

------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260756 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260718 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260716 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260642 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r260637 | hans | 2016-02-11 17:01:37 -0800 (Thu, 11 Feb 2016) | 1 line

UsersManual: update clang-cl commands
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260638 91177308-0d34-0410-b5e6-96231b3b80d8

I ran:
$ bin/clang-tblgen -gen-attr-docs -I../cfe.src/include ../cfe.src/include/clang/Basic/Attr.td -o ../cfe.src/docs/AttributeReference.rst

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260636 91177308-0d34-0410-b5e6-96231b3b80d8

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@260635 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r259931 | uweigand | 2016-02-05 13:34:28 -0800 (Fri, 05 Feb 2016) | 33 lines

[SystemZ] Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP macros

Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_[1248] macros on SystemZ.

This fixes a miscompile of GCC C++11 standard library headers
due to use of those macros in an ABI-changing manner.

See e.g. /usr/include/c++/4.8.5/ext/concurrence.h:

  // Compile time constant that indicates prefered locking policy in
  // the current configuration.
  static const _Lock_policy __default_lock_policy =
#ifdef __GTHREADS
#if (defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_2) \
     && defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4))
  _S_atomic;
#else
  _S_mutex;
#endif
#else
  _S_single;
#endif

A different choice of __default_lock_policy causes different
sizes of several of the C++11 data structures, which are then
incompatible when inlined in clang-compiled code with what the
(GCC-compiled) external library expects.

This in turn leads to various crashes when using std::thread
in code compiled with clang, as see e.g. via the ThreadPool
unit tests.  See PR 26473 for an example.

------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@259939 91177308-0d34-0410-b5e6-96231b3b80d8

------------------------------------------------------------------------
r259183 | uweigand | 2016-01-29 02:45:23 -0800 (Fri, 29 Jan 2016) | 8 lines

Add target triple to CodeGenOpenCL/pipe_types.cl test case

The test is failing on SystemZ since different IR is being
generated due to platform ABI differences.  Add a target triple.

Fix suggested by Anastasia Stulova.

------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_38@259808 91177308-0d34-0410-b5e6-96231b3b80d8