llvm-svn: 308297

Coverage hooks that take less-than-64-bit-integers as parameters need the
zeroext parameter attribute (http://llvm.org/docs/LangRef.html#paramattrs)
to make sure they are properly extended by the x86_64 ABI.

llvm-svn: 308296

llvm-svn: 308295

Optimized IsRegIntersect by using MCRegAliasIterator

See Bug 33800: https://bugs.llvm.org//show_bug.cgi?id=33800

Reviewers: arsenm, artem.tamazov

Differential Revision: https://reviews.llvm.org/D35452

llvm-svn: 308294

Splitted from D35476.

llvm-svn: 308293

llvm-svn: 308292

Cleaned up the code in FastISel a bit.
Had to add make_range to MCInstrDesc as that was needed and seems missing.

Reviewed by: @t.p.northover
Differential Revision: https://reviews.llvm.org/D35494

llvm-svn: 308291

Annotate all public functions with the autoload magic cookie so that
update-directory-autoloads will find them.

Patch by Brendan O'Dea.

llvm-svn: 308290

find_library(lib) stores the result in the variable "lib", which is also
cached in CMakeCache.txt.  This could theoretically conflict if jsoncpp
required two libraries, which each would get cached as "lib".  Use a
more descriptive and disambiguative "jsoncpp_${libname}" for that.

llvm-svn: 308289

Use ${libname} instead of ${lib}. By a coincidence, this worked
because ${lib} also the variable used for finding the libjsoncpp.so
full path.

llvm-svn: 308288

pkg_search_module(JSONCPP) should set
JSONCPP_LIBDIR/JSONCPP_LIBRARY_DIRS to where the libjsoncpp.so can be
found. However, on Ubuntu 14.04 LTS (Trusty Tahr) it returns /usr/lib
while the libjsoncpp library can be found at
/usr/lib/x86_64-linux-gnu/libjsoncpp.so. Thus, while searching for
the full path of the jsoncpp library, it is not found.

JSONCPP_LIBDIR is correctly set to /usr/lib/x86_64-linux-gnu on e.g.,
Ubuntu 16.04 LTS (Xenial Xerus )

Fix by removing the NO_DEFAULT_PATH flag, in order to search the system
default paths even if the library is not found in
JSONCPP_LIBDIR/JSONCPP_LIBRARY_DIRS.

This fixes bug llvm.org/PR33798.

llvm-svn: 308287

llvm-svn: 308286

Widen to s32, and then do whatever Lowering/Custom/Libcall action the
subtarget wants.

llvm-svn: 308285

Summary:
Currently most tests for the loop interchange pass are in
test/Transforms/LoopInterchange/interchange.ll. This patch splits up the
large test file in smaller pieces, which makes debugging test failures
easier.

Reviewers: karthikthecool, blitz.opensource, hfinkel

Reviewed By: hfinkel

Subscribers: hfinkel, mcrosier, mkuper, mzolotukhin, mssimpso, llvm-commits

Differential Revision: https://reviews.llvm.org/D35488

llvm-svn: 308284

Summary:
Using 16 byte alignment is beneficial on Cortex-A73, similar to
Cortex-A72 (added in D34961).

Reviewers: mcrosier, t.p.northover, aadg, silviu.baranga

Reviewed By: t.p.northover

Subscribers: aemerson, rengolin, javed.absar, kristof.beyls, llvm-commits

Differential Revision: https://reviews.llvm.org/D35493

llvm-svn: 308283

Summary:
The usage of shared_from_this forces us to separate construction and
initialization phases, because shared_from_this() is not available in
the constructor (or destructor). The shared semantics are not necessary,
as we always have a clear owner of the native process class
(GDBRemoteCommunicationServerLLDB object). Even if we need shared
semantics in the future (which I think we should strongly avoid),
reverting this will not be necessary -- the owners can still easily
store the native process object in a shared pointer if they really want
to -- this just prevents the knowledge of that from leaking into the
class implementation.

After this a NativeThread object will hold a reference to the parent
process (instead of a weak_ptr) -- having a process instance always
available allows us to simplify some logic in this class (some of it was
already simplified because we were asserting that the process is
available, but this makes it obvious).

Reviewers: krytarowski, eugene, zturner

Subscribers: lldb-commits

Differential Revision: https://reviews.llvm.org/D35123

llvm-svn: 308282

Added support of the following opcodes:
  v_pk_sub_u16
  v_pk_mad_i16
  v_pk_mad_u16

See Bug 33593: https://bugs.llvm.org//show_bug.cgi?id=33593

Reviewers: vpykhtin, artem.tamazov, arsenm

Differential Revision: https://reviews.llvm.org/D34890

llvm-svn: 308281

This enables the suggestions of other mnemonics when invalid ones are
specified.

Review: Ulrich Weigand
llvm-svn: 308280

    We extended kills in Polly to handle both `phi` nodes and scalars that
    are not used within the Scop. Update the comments and choice of
    variable names to reflect this.

llvm-svn: 308279

Treat widening G_SREM and G_UREM the same as G_SDIV and G_UDIV. This is
going to be used in the ARM backend (and that's when the test will come
too).

llvm-svn: 308278

LLVM_ENABLE_MODULES is sensitive of -D. Move them into config.h.

FIXME: It'd be better that they are #cmakedefine01 rather than #cmakedefine.
(#if FOO rather than #if defined(FOO))
Then we can find missing #include "clang/Config/config.h" in the future.

Differential Revision: https://reviews.llvm.org/D35527

llvm-svn: 308277

llvm/DebugInfo/CodeView/TypeStreamMerger.h: Prune a couple of \param(s), removed in r308212. [-Wdocumentation]

llvm-svn: 308276

llvm-svn: 308275

Differential Revision: https://reviews.llvm.org/D35497

llvm-svn: 308274

The commit r308100 updated WebAssembly tests for r308025. In one case it
merely made the test more resilient but in another case it made
a substantive update. Because r308025 was reverted in r308271, these
changes to the test also need to be reverted. They should be folded into
the recommit of r308025 when it is ready.

llvm-svn: 308273

Notably, this is failing on our PPC build bots:
http://lab.llvm.org:8011/builders/clang-ppc64le-linux/builds/8338/steps/ninja%20check%201/logs/FAIL%3A%20LLVM%3A%3Apr33772.ll

llvm-svn: 308272

with a minimal test case in http://llvm.org/PR33833.

Original commit message:
  Improve Aliasing of operations to static alloca

llvm-svn: 308271

Original commit log:
[AMDGPU] CodeGen: check dst operand type to determine if omod is supported for VOP3 instructions

llvm-svn: 308270

Summary:
This patch prevents getCanonicalDecl returning nullptr in case it finds
a canonical TemplateDeclaration with no attached TemplatedDecl.
Found by running the indexer over a version of the standard library deep inside
a template metaprogramming mess.

Reviewers: klimek, vsk

Reviewed By: vsk

Subscribers: vsk, arphaman, cfe-commits

Differential Revision: https://reviews.llvm.org/D35212

llvm-svn: 308269

Summary:
For example, this option is expected by ghc (haskell compiler). Currently, building with ghc will fail with:

```
clang: error: unknown argument: '-no-pie'
`gcc' failed in phase `Linker'. (Exit code: 1)
. /usr/share/haskell-devscripts/Dh_Haskell.sh && \
configure_recipe
```

This won't do anything (but won't fail with an error)


Reviewers: rafael, joerg

Reviewed By: joerg

Subscribers: joerg, cfe-commits

Differential Revision: https://reviews.llvm.org/D35462

llvm-svn: 308268

This isn't legal code, but we shouldn't crash on it. Now we just don't convert the gather intrinsic if the scale isn't constant and let it go through to isel where we'll report an isel failure.

Fixes PR33772.

llvm-svn: 308267

Reviewers: Anastasia

Reviewed By: Anastasia

Subscribers: bader, cfe-commits, yaxunl

Differential Revision: https://reviews.llvm.org/D35000

llvm-svn: 308266

optimizeMemoryInst contains a vector AddrModeInsts.
The only use of this vector is to check that all instructions are in the same
block as memory instruction. This check is guarded by PhiSeen flag,
so if we traversed through phi node then we do not need to keep information
in AddrModeInsts. AddModeInsts is set first time we found some addressing mode
and updated if we found new one later.
We can find next addressing mode only if we traverse phi node so all code
related to update of AddModeInsts can be safely removed.

Reviewers: loladiro, spatel, efriedma
Reviewed By: efriedma
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D35291

llvm-svn: 308265

In some particular cases eq/ne conditions can be turned into equivalent
slt/sgt conditions. This patch teaches parseLoopStructure to handle some
of these cases.

Differential Revision: https://reviews.llvm.org/D35010

llvm-svn: 308264

This reverts commit 223ef99f839f6b056272bcf4390841fcb26dda3c.

llvm-svn: 308263

This reverts commit f3aaaac609f801df6c12655ec203455be7094627.

llvm-svn: 308262

This reverts commit 66093fd60b848572f676023b8387bff69b151511.

llvm-svn: 308261

Summary:
Previously, we counted TotalMemInst by reading certain instruction counters before and after calling visit and then finding the difference. But that wouldn't be thread safe if this same pass was being ran on multiple threads.

This list of "memory instructions" doesn't make sense to me as it includes call/invoke and is missing atomics.

This patch removes the counter all together.

Reviewers: hfinkel, chandlerc, davide

Reviewed By: davide

Subscribers: davide, llvm-commits

Differential Revision: https://reviews.llvm.org/D33608

llvm-svn: 308260

Some shells seem to have trouble parsing non-alphanumeric symbols.

llvm-svn: 308259

These tests assume allocator_may_return_null=false
If allocator_may_return_null=true, gtest would not be able to switch it.
Tests needs to be re-implemented as lit tests.

llvm-svn: 308254