llvm-svn: 272302

Previously we eliminated the randomized scheme for finding memory when the
underlying process cannot allocate memory, and replaced it with an algorithm
that starts the allocations at 00x.

This was more determinstic, but runs into problems on embedded targets where the
pages near 0x0 are in fact interesting memory.  To deal with those cases, this
patch does two things:

- It makes the default fallback be an address that is less likely than 0x0 to
  contain interesting information.

- Before falling back to this, it adds an algorithm that consults the
  GetMemoryRegionInfo() API to see if it can find an unmapped area.

This should eliminate the randomness (and unpredictable memory accesseas) of the
previous scheme while making expressions more likely to return correct results.

<rdar://problem/25545573>

llvm-svn: 272301

512-bit VPSLLDQ/VPSRLDQ can only be used for avx512bw targets so lowerVectorShuffleAsShift had to be adjusted to include the subtarget

llvm-svn: 272300

Summary: Clang changes to make use of the LLVM intrinsics added in D21160.

Reviewers: tra

Subscribers: jholewinski, cfe-commits

Differential Revision: http://reviews.llvm.org/D21162

llvm-svn: 272299

Summary:
Currently clang emits these instructions via inline (volatile) asm in
the CUDA headers.  Switching to intrinsics will let the optimizer reason
across calls to these intrinsics.

Reviewers: tra

Subscribers: llvm-commits, jholewinski

Differential Revision: http://reviews.llvm.org/D21160

llvm-svn: 272298

 - Alphabetically sort the initializeXXX calls (this was brought up in
   D21115)
 - Remove repeated function names from doxygen comments

llvm-svn: 272297

These ExprWithCleanups are added for holding a RunCleanupsScope not
for destructor calls; rather, they are for lifetime marks. This requires
ExprWithCleanups to keep a bit to indicate whether it have cleanups with
side effects (e.g. dtor calls).

Differential Revision: http://reviews.llvm.org/D20498

llvm-svn: 272296

Summary:
__syncthreads, which corresponds to bar.sync 0, is already convergent.
This makes the more general bar.sync n likewise convergent.

Reviewers: tra

Subscribers: llvm-commits, jholewinski

Differential Revision: http://reviews.llvm.org/D21161

llvm-svn: 272295

Differential Revision: http://reviews.llvm.org/D21090

llvm-svn: 272294

This reverts commit r272290. It breaks a test that depends on being able
to seek the /dev/null equivalent on Windows:

http://bb.pgr.jp/builders/ninja-clang-x64-mingw64-RA/builds/11360

llvm-svn: 272293

We were using the fast fdiv lowering for all division, implementation of
IEEE754 fdiv is added.

http://reviews.llvm.org/D20557

llvm-svn: 272292

llvm-svn: 272291

Use os.devnull instead of tempfiles when substituting '/dev/null' on
Windows machines. This should make the bots just a bit speedier.

Thanks to Yunzhong Gao for testing this patch on Windows!

Differential Revision: http://reviews.llvm.org/D20549

llvm-svn: 272290

Make the comparison objects that we pass in for various tests look more like actual comparison objects. No functional change.

llvm-svn: 272288

llvm-svn: 272287

Refactored __kmp_execute_tasks_template to shorten and remove code redundancy.
The original code for __kmp_execute_tasks_template was very redundant with
large sections of repeated code that needed to be kept consistent, and goto
statements that made the control flow difficult to discern. This refactoring
removes all gotos and redundancy.

Patch by Terry Wilmarth

Differential Revision: http://reviews.llvm.org/D20879

llvm-svn: 272286

llvm-svn: 272284

Previously, we materialized secondary vector IVs from the primary scalar IV,
by offseting the primary to match the correct start value, and then broadcasting
it - inside the loop body. Instead, we can use a real vector IV, like we do for
the primary.

This enables using vector IVs for secondary integer IVs whose type matches the
type of the primary.

Differential Revision: http://reviews.llvm.org/D20932

llvm-svn: 272283

We can now use this to decide whether to emit a verneed during the final
pass over the symbols. We were previously wrongly creating a verneed entry
in the case where all references to a DSO's symbols were weak.

In a future change we may also want to use the used bit to control whether
shared symbols are preemptible and appear in the dynsym. This seems a little
tricky to do at the moment because isNeeded() is templated.

The only other functional change here is that we emit a DT_NEEDED for DSOs
whose symbols are all preempted by objects that appear later in the link. But
that doesn't seem too important to me.

Differential Revision: http://reviews.llvm.org/D21171

llvm-svn: 272282

Some core files on MacOSX don't have permissions setup correctly on the LC_SEGMENT load commands. Assume read + execute if the permissions are not set.

<rdar://problem/26720522> 

llvm-svn: 272281

This reverts r272275. This actually wasn't the right way to fix the problem. The correct solution is in r272279.

Applying the fix to LLVM as done in r272279, means this fix will get picked up by all projects building out of tree using LLVM's CMake modules. As opposed to the fix I had in r272275, which would require each project to change.

llvm-svn: 272280

This is the more-correct fix to out-of-tree building. AddLLVM.cmake relies on this variable being set, so we should make sure it is set in LLVMConfig.cmake.

llvm-svn: 272279

llvm-svn: 272278

Thanks to Rafael for the suggestion.

llvm-svn: 272277

Since our expression parser needs to locate areas of memory that are not in use when you have a process that can't JIT code, like core file debugging, the core file process plug-ins should be able to override the Process::GetMemoryRegionInfo(...) function.

In order to make this happen, I have added permissions to sections so that we can know what the permissions are for a given section, and modified both core file plug-ins to override Process::GetMemoryRegionInfo() and answer things correctly.

llvm-svn: 272276

The out-of-tree build needs to read LLVM_TOOLS_INSTALL_DIR out of TOOLS_BINARY_DIR because LLVM_TOOLS_INSTALL_DIR is used by AddLLVM.cmake

llvm-svn: 272275

So we can match sizeof expressions more accurately than with UnexposedExpr

Differential Revision: http://reviews.llvm.org/D18081

llvm-svn: 272274

Differential Revision: http://reviews.llvm.org/D18080

llvm-svn: 272273

Fixes {u,}long_{min,max,clamp} opencl piglit regressions on EG.

Reviewers: arsenm
Differential Revision: http://reviews.llvm.org/D17898

llvm-svn: 272272

MSVC doesn't allow std::atomic<>s in a union since they don't have trivial
copy constructor. Replacing them with e.g. std::atomic_int works, but that
breaks the GCC build on Linux, because then calls to e.g. std::atomic_load_explicit
fail, as they expect a real std::atomic<> pointer.

Fixing this with an #ifdef to unbreak the build for now.

llvm-svn: 272271

We should not ignore the symbol, just the version.

Fixes error with -z defs.

llvm-svn: 272270

llvm-svn: 272269

This resolves PR28066.

Patch by David Binderman!

llvm-svn: 272268

This reapplies commit r271930, r271915, r271923.  They hit a bug in
Thumb which is fixed in r272258 now.

The original message:

The code layout that TailMerging (inside BranchFolding) works on is not the
final layout optimized based on the branch probability. Generally, after
BlockPlacement, many new merging opportunities emerge.

This patch calls Tail Merging after MBP and calls MBP again if Tail Merging
merges anything.

llvm-svn: 272267

This enables use of the 'S' constraint for inline ASM operands on
SystemZ, which allows for a memory reference with a signed 20-bit
immediate displacement. This patch includes corresponding documentation
and test case updates.

I've changed the 'T' constraint to match the new behavior for 'S', as
'T' also uses a long displacement (though index constraints are still
not implemented). I also changed 'm' to match the behavior for 'S' as
this will allow for a wider range of displacements for 'm', though
correct me if that's not the right decision.

Author: colpell
Differential Revision: http://reviews.llvm.org/D21097

llvm-svn: 272266

llvm-svn: 272265

llvm-svn: 272264

llvm-svn: 272263

Fixes a crash in the backend during an LTO build of rtld(1) in
FreeBSD.

llvm-svn: 272262

Summary:
we want to exclude friend declaration, but the `DeclContext` of a
friend function declaration is not the class in which it is declared, so we need
to explicitly check if the parent is a `friendDecl`.

Reviewers: bkramer

Subscribers: cfe-commits

Differential Revision: http://reviews.llvm.org/D21175

llvm-svn: 272261