This is in response to silvas' post-commit suggestion.

llvm-svn: 260203

llvm-svn: 260202

This cache file can be used to generate a 3-stage clang build. You can configure using the following CMake command:

cmake -C <path to clang>/cmake/caches/3-stage.cmake -G Ninja <path to llvm>

You can then run "ninja stage3-clang" to build stage1, stage2 and stage3 clangs.

This is useful for finding non-determinism the compiler by verifying that stage2 and stage3 are identical.

llvm-svn: 260201

llvm-svn: 260200

This is because PyThreadState_Get() assumes a non-NULL thread state and crashes otherwise; but PyThreadState_GET is just a shortcut (in non-Python-debugging builds) for the global variable that holds the thread state

The behavior of CTRL+C is slightly more erratic than one would like. CTRL+C in the middle of execution of Python code will cause that execution to be interrupted (e.g. time.sleep(1000)), but a CTRL+C at the prompt will just cause a KeyboardInterrupt and not exit the interpreter - worse, it will only trigger the exception once one presses ENTER.

None of this is optimal, of course, but I don't have a lot of time to appease the Python deities with the proper spells right now, and fixing the crasher is already a good thing in and of itself

llvm-svn: 260199

llvm-svn: 260198

Summary:
Move the function renaming logic into the Function class, and the
MD5Hash routine into the MD5 header.

This will enable these routines to be shared with ThinLTO, which
will be changed to store the MD5 hash instead of full function name
in the combined index for significant size reductions. And using the same
function naming for locals in the function index facilitates future
integration with indirect call value profiles.

Reviewers: davidxl

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D17006

llvm-svn: 260197

Use std::forward to make ErrorOr<T> constructible from a value that has a user-defined conversion to T. No functionality change intended.

llvm-svn: 260196

Change the no_sanitize attribute to use the reserved spelling.

llvm-svn: 260195

Fixes PR26490

llvm-svn: 260194

In general, memory restrictions on a called function (e.g. readnone)
cannot be transferred to a CallSite that has operand bundles.  It is
possible to make this inference smarter, but lets fix the behavior to be
correct first.

llvm-svn: 260193

Reviewers: zturner

Subscribers: lldb-commits

Differential Revision: http://reviews.llvm.org/D17014

llvm-svn: 260192

Also added the defaults for whether to generate this load command, which
the cmdline options are able to override.

There was also a difference to ld64 which is fixed here in that ld64 will
generate an empty data in code command if requested.

rdar://problem/24472630

llvm-svn: 260191

compiler-specific issues. Instead, repeat an 'operator delete' definition in
each derived class that is actually deleted, and give up on the static type
safety of an error when sized delete is accidentally used on a type derived
from TrailingObjects.

llvm-svn: 260190

llvm-svn: 260189

This load command generates data in the LINKEDIT section which
is a list of ULEB128 delta's to all of the functions in the __text section.

It is then 0 terminated and pointer aligned to pad.

ld64 exposes the -function-starts and no-function-starts cmdline options
to override behaviour from the defaults based on file types.

rdar://problem/24472630

llvm-svn: 260188

Differential Revision: http://reviews.llvm.org/D17005

llvm-svn: 260186

This was a bug in our handling of these symbols compared to ld64.

Turns out that ld64 always marks these symbols as being not dead stripped.

llvm-svn: 260185

The atom content type enum is used as a tie breaker to sort atoms.
In that case, we want MachHeader to be before typeCode as it really will
be before the code in the final executable.

Test case to follow in the next commit or two.

llvm-svn: 260184

Summary:
Passes that call `getAnalysisIfAvailable<T>` also need to call
`addUsedIfAvailable<T>` in `getAnalysisUsage` to indicate to the
legacy pass manager that it uses `T`.  This contract was being
violated by passes that used `createLegacyPMAAResults`.  This change
fixes this by exposing a helper in AliasAnalysis.h,
`addUsedAAAnalyses`, that is complementary to createLegacyPMAAResults
and does the right thing when called from `getAnalysisUsage`.

Reviewers: chandlerc

Subscribers: mcrosier, llvm-commits

Differential Revision: http://reviews.llvm.org/D17010

llvm-svn: 260183

Summary:
createLegacyPMAAResults is only called by CGSCC and Module passes, so
the call to getAnalysisIfAvailable<SCEVAAWrapperPass>() never
succeeds (SCEVAAWrapperPass is a function pass).

Reviewers: chandlerc

Subscribers: mcrosier, llvm-commits

Differential Revision: http://reviews.llvm.org/D17009

llvm-svn: 260182

being called with the wrong size: convert CGFunctionInfo to use TrailingObjects
and ask TrailingObjects to provide a working 'operator delete' for us.

llvm-svn: 260181

This fixes undefined behavior in C++14 due to the size of the object being
deleted being different from sizeof(dynamic type) when it is allocated with
trailing objects.

MSVC seems to have several bugs around using-declarations changing the access
of a member inherited from a base class, so use forwarding functions instead of
using-declarations to make TrailingObjects::operator delete accessible where
desired.

llvm-svn: 260180

llvm-svn: 260179

This removes the following decorators:

* skipIfI386
* expectedFailureI386
* expectedFailurex86_64
* skipIfArch
* skipUnlessArch
* skipUnlessI386

And other related decorators.  All code using those decorators
is updated to use expectedFailureAll and skipIf

llvm-svn: 260178

llvm-svn: 260177

Convert everything over to using skipIf.

llvm-svn: 260176

llvm-svn: 260175

IndVarSimplify assumes scAddRecExpr to be expanded in literal form instead of
canonical form by calling disableCanonicalMode after it creates SCEVExpander.
When CanonicalMode is disabled, SCEVExpander::expand should always return PHI
node for scAddRecExpr. r259736 broke the assumption.

The fix is to let SCEVExpander::expand skip the reuse Value logic if
CanonicalMode is false.

In addition, Besides IndVarSimplify, LSR pass also calls disableCanonicalMode
before doing rewrite. We can remove the original check of LSRMode in reuse
Value logic and use CanonicalMode instead.

llvm-svn: 260174

llvm-svn: 260173

llvm-svn: 260172

r260146 | xur | 2016-02-08 13:07:46 -0800 (Mon, 08 Feb 2016) | 13 lines
[PGO] Differentiate Clang instrumentation and IR level instrumentation profiles

llvm-svn: 260170

Summary:
Unrolling Analyzer is already pretty complicated, and it becomes harder and harder to exercise it with usual IR tests, as with them we can only check the final decision: whether the loop is unrolled or not. This change factors this framework out from LoopUnrollPass to analyses, which allows to use unit tests.
The change itself is supposed to be NFC, except adding a couple of tests.

I plan to add more tests as I add new functionality and find/fix bugs.

Reviewers: chandlerc, hfinkel, sanjoy

Subscribers: zzheng, sanjoy, llvm-commits

Differential Revision: http://reviews.llvm.org/D16623

llvm-svn: 260169

As discussed on PR26491, this patch adds support for lowering v4f32 shuffles to the MOVLHPS/MOVHLPS instructions. It also adds support for memory folding with their MOVLPS/MOVHPS load equivalents.

This first patch only really helps SSE1 targets as SSE2+ targets will widen the shuffle mask and use v2f64 equivalents (although they still combine to MOVLHPS/MOVHLPS for v2f64 splats). This will have to be addressed in a future patch, most likely when we add support for binary target shuffle combines.

Differential Revision: http://reviews.llvm.org/D16956

llvm-svn: 260168

llvm-svn: 260167

llvm-svn: 260166

llvm-svn: 260165

Differential Revision: http://reviews.llvm.org/D16831

llvm-svn: 260164

llvm-svn: 260163

In order for recent gcov versions to read the coverage data, you have
to use UseCfgChecksum=true and FunctionNamesInData=false options for
coverage profiling pass. This is because gcov is expecting the
function section in .gcda to be exactly 3 words in size, containing
ident and two checksums.

While llvm-cov is compatible with UseCfgChecksum=true, it always
expects a function name in .gcda function sections (it's not
compatible with FunctionNamesInData=false). Thus it's currently
impossible to generate one set of coverage files that works with both
gcov and llvm-cov.

This change fixes the reading of coverage information to only read the
function name if it's present.

Patch by Arseny Kapoulkine. Thanks!

llvm-svn: 260162