platform/upstream/llvm.git
7 years ago[ELF] Use late evaluation for ALIGN in expression
Petr Hosek [Tue, 30 May 2017 03:18:28 +0000 (03:18 +0000)]
[ELF] Use late evaluation for ALIGN in expression

While the following expression is handled fine:

  PROVIDE_HIDDEN(newsym = oldsym + address);

The following expression triggers an error because the expression
is evaluated as absolute:

  PROVIDE_HIDDEN(newsym = ALIGN(oldsym, CONSTANT(MAXPAGESIZE)) + address);

To avoid this error, we use late evaluation for ALIGN by making the
alignment an attribute of the expression itself.

Differential Revision: https://reviews.llvm.org/D33629

llvm-svn: 304185

7 years agoReverted r303602, as it will be fixed in gtest.
Galina Kistanova [Tue, 30 May 2017 03:17:10 +0000 (03:17 +0000)]
Reverted r303602, as it will be fixed in gtest.

llvm-svn: 304184

7 years ago[modules] When we #include a local submodule header that we've already built,
Richard Smith [Tue, 30 May 2017 02:03:19 +0000 (02:03 +0000)]
[modules] When we #include a local submodule header that we've already built,
and it has an include guard, produce callbacks for a module import, not for a
skipped non-modular header.

Fixes -E output when preprocessing a module to list these cases as a module
import, rather than suppressing the #include and losing the import side effect.

llvm-svn: 304183

7 years agoReplace a few more uses of OutputSections.
Rafael Espindola [Tue, 30 May 2017 01:36:48 +0000 (01:36 +0000)]
Replace a few more uses of OutputSections.

llvm-svn: 304182

7 years agoKeep a list of all OutputSectionCommands.
Rafael Espindola [Tue, 30 May 2017 01:30:14 +0000 (01:30 +0000)]
Keep a list of all OutputSectionCommands.

Now that we are trying to use the linker script representation as the
canonycal one, there are a few loops looking for just OutputSectionCommands.

Create a vector with just the OutputSectionCommands once that is
stable to simplify the rest of the code.

llvm-svn: 304181

7 years agoAdd missing 'requires coroutines' to module map
Eric Fiselier [Mon, 29 May 2017 23:17:28 +0000 (23:17 +0000)]
Add missing 'requires coroutines' to module map

llvm-svn: 304180

7 years agoRevert r303763, results in asserts i.e. while building Ruby.
Joerg Sonnenberger [Mon, 29 May 2017 22:52:17 +0000 (22:52 +0000)]
Revert r303763, results in asserts i.e. while building Ruby.

llvm-svn: 304179

7 years ago[TableGen] Use StringMap instead of DenseMap<StringRef> to unique CodeInit and String...
Craig Topper [Mon, 29 May 2017 21:49:37 +0000 (21:49 +0000)]
[TableGen] Use StringMap instead of DenseMap<StringRef> to unique CodeInit and StringInit objects. Override the allocator to keep using the BumpPtrAllocator. NFCI

StringMap is better suited to mapping strings than a DenseMap.

llvm-svn: 304178

7 years ago[TableGen] Introduce DagInit::getArgs that returns an ArrayRef. Use it to fix 80...
Craig Topper [Mon, 29 May 2017 21:49:34 +0000 (21:49 +0000)]
[TableGen] Introduce DagInit::getArgs that returns an ArrayRef. Use it to fix 80 column violations in arg_begin/arg_end. Remove DagInit::args and use getArgs instead. NFC

llvm-svn: 304177

7 years agoCGCoroutine.cpp: (NFC) clang-format misplaced brace
Gor Nishanov [Mon, 29 May 2017 21:15:31 +0000 (21:15 +0000)]
CGCoroutine.cpp: (NFC) clang-format misplaced brace

llvm-svn: 304176

7 years ago[ManagedStatic] Avoid putting function pointers in template args.
Benjamin Kramer [Mon, 29 May 2017 20:56:27 +0000 (20:56 +0000)]
[ManagedStatic] Avoid putting function pointers in template args.

This is super awkward, but GCC doesn't let us have template visible when
an argument is an inline function and -fvisibility-inlines-hidden is
used.

llvm-svn: 304175

7 years ago[GlobalIsel] Fix a warning with GCC 7 -Wpedantic. NFCI.
Davide Italiano [Mon, 29 May 2017 20:13:22 +0000 (20:13 +0000)]
[GlobalIsel] Fix a warning with GCC 7 -Wpedantic. NFCI.

llvm-svn: 304174

7 years agoFix <experimental/coroutine> in C++03
Eric Fiselier [Mon, 29 May 2017 19:46:16 +0000 (19:46 +0000)]
Fix <experimental/coroutine> in C++03

llvm-svn: 304173

7 years ago[coroutines] Make coroutine_handle<T>::from_address ill-formed for everything but...
Eric Fiselier [Mon, 29 May 2017 19:24:25 +0000 (19:24 +0000)]
[coroutines] Make coroutine_handle<T>::from_address ill-formed for everything but void*.

from_address requires that the provided pointer refer to the suspended coroutine,
which doesn't have a type, or at least not one knowable by the user. Therefore
every use of `from_address` with a typed pointer is almost certainly a bug.

This behavior is a part of the TS specification, but hopefully it will be
in the future.

llvm-svn: 304172

7 years ago[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC.
Zvi Rackover [Mon, 29 May 2017 19:00:57 +0000 (19:00 +0000)]
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC.

To be improved by D33311.

llvm-svn: 304171

7 years ago[analyzer] Fix immutable map factory lifetime for partial taint.
Artem Dergachev [Mon, 29 May 2017 18:54:02 +0000 (18:54 +0000)]
[analyzer] Fix immutable map factory lifetime for partial taint.

This should fix the leaks found by asan buildbot in r304162.

Also don't store a reference to the factory with every map value,
which is the only difference between ImmutableMap and ImmutableMapRef.

llvm-svn: 304170

7 years ago[X86] Replace undef value in flaky test
Zvi Rackover [Mon, 29 May 2017 18:27:00 +0000 (18:27 +0000)]
[X86] Replace undef value in flaky test

D33311 exposes the flakiness in this test. Replacing the undef placed by
bugpoint, makes it more interesting and robust.

llvm-svn: 304168

7 years agoUnbreak long test after r304127.
Benjamin Kramer [Mon, 29 May 2017 18:11:11 +0000 (18:11 +0000)]
Unbreak long test after r304127.

llvm-svn: 304167

7 years ago[ManagedStatic] Make object_creator/object_deleter visible again.
Benjamin Kramer [Mon, 29 May 2017 18:00:33 +0000 (18:00 +0000)]
[ManagedStatic] Make object_creator/object_deleter visible again.

They're now exposed as template args, which creates complications when
ManagedStatics are used across .so boundaries.

llvm-svn: 304166

7 years agoDon't destroy ManagedStatics in a unit test.
Benjamin Kramer [Mon, 29 May 2017 17:25:37 +0000 (17:25 +0000)]
Don't destroy ManagedStatics in a unit test.

Turns out this is very hostile towards other unit tests running in the
same process, it unregisters all flags.

llvm-svn: 304165

7 years ago[wasm] Fix test after r304117.
Benjamin Kramer [Mon, 29 May 2017 16:32:52 +0000 (16:32 +0000)]
[wasm] Fix test after r304117.

llvm-svn: 304164

7 years ago[X86] Don't fold away the memory operand of an xchg.
Benjamin Kramer [Mon, 29 May 2017 16:25:20 +0000 (16:25 +0000)]
[X86] Don't fold away the memory operand of an xchg.

xchg with a mem operand has different locking semantics. If we unfold it
into a xchg r,r we will loose the implicit lock. Likewise we never want
to fold a register xchg into a memory one as it would be a lot slower.

This triggers during LLVM selfhost.

llvm-svn: 304163

7 years ago[analyzer] Support partially tainted records.
Artem Dergachev [Mon, 29 May 2017 15:42:56 +0000 (15:42 +0000)]
[analyzer] Support partially tainted records.

The analyzer's taint analysis can now reason about structures or arrays
originating from taint sources in which only certain sections are tainted.

In particular, it also benefits modeling functions like read(), which may
read tainted data into a section of a structure, but RegionStore is incapable of
expressing the fact that the rest of the structure remains intact, even if we
try to model read() directly.

Patch by Vlad Tsyrklevich!

Differential revision: https://reviews.llvm.org/D28445

llvm-svn: 304162

7 years ago[Docs] Add VectorizationPlan to docs/Proposals.
Ayal Zaks [Mon, 29 May 2017 15:36:23 +0000 (15:36 +0000)]
[Docs] Add VectorizationPlan to docs/Proposals.

Following the request made in https://reviews.llvm.org/D32871, the
general documentation of the Vectorization Plan is hereby placed
under docs/Proposals.

llvm-svn: 304161

7 years ago[analyzer] Initial commit for the upcoming refactoring of the IteratorChecker.
Artem Dergachev [Mon, 29 May 2017 15:03:20 +0000 (15:03 +0000)]
[analyzer] Initial commit for the upcoming refactoring of the IteratorChecker.

The new checker currently contains the very core infrastructure for tracking
the state of iterator-type objects in the analyzer: relating iterators to
their containers, tracking symbolic begin and end iterator values for
containers, and solving simple equality-type constraints over iterators.
A single specific check over this infrastructure is capable of finding usage of
out-of-range iterators in some simple cases.

Patch by Ádám Balogh!

Differential revision: https://reviews.llvm.org/D32592

llvm-svn: 304160

7 years ago[analyzer] PthreadLockChecker: model failed pthread_mutex_destroy() calls.
Artem Dergachev [Mon, 29 May 2017 14:51:39 +0000 (14:51 +0000)]
[analyzer] PthreadLockChecker: model failed pthread_mutex_destroy() calls.

pthread_mutex_destroy() may fail, returning a non-zero error number, and
keeping the mutex untouched. The mutex can be used on the execution branch
that follows such failure, so the analyzer shouldn't warn on using
a mutex that was previously destroyed, when in fact the destroy call has failed.

Patch by Malhar Thakkar!

Differential revision: https://reviews.llvm.org/D32449

llvm-svn: 304159

7 years agoReplace forward decl with include to unbreak the build.
Benjamin Kramer [Mon, 29 May 2017 14:40:07 +0000 (14:40 +0000)]
Replace forward decl with include to unbreak the build.

llvm-svn: 304158

7 years agoTry to work around MSVC being buggy. Attempt #1.
Benjamin Kramer [Mon, 29 May 2017 14:28:04 +0000 (14:28 +0000)]
Try to work around MSVC being buggy. Attempt #1.

error C2971: 'llvm::ManagedStatic': template parameter 'Creator': 'CreateDefaultTimerGroup': a variable with non-static storage duration cannot be used as a non-type argument

llvm-svn: 304157

7 years ago[Timer] Move DefaultTimerGroup into a ManagedStatic.
Benjamin Kramer [Mon, 29 May 2017 14:05:29 +0000 (14:05 +0000)]
[Timer] Move DefaultTimerGroup into a ManagedStatic.

This used to be just leaked. r295370 made it use magic statics. This adds
a global destructor, which is something we'd like to avoid. It also creates
a weird situation where the mutex used by TimerGroup is re-created during
global shutdown and leaked.

Using a ManagedStatic here is also subtle as it relies on the mutex
inside of ManagedStatic to be recursive. I've added a test for that
in a previous change.

llvm-svn: 304156

7 years ago[ManagedStatic] Add a way to pass custom creators/deleters.
Benjamin Kramer [Mon, 29 May 2017 14:05:26 +0000 (14:05 +0000)]
[ManagedStatic] Add a way to pass custom creators/deleters.

Also add a test case verifying that nested ManagedStatics work correctly.

llvm-svn: 304155

7 years ago[clang-tidy] Use getLocalOrGlobal for the StrictMode option
Alexander Kornienko [Mon, 29 May 2017 13:59:27 +0000 (13:59 +0000)]
[clang-tidy] Use getLocalOrGlobal for the StrictMode option

llvm-svn: 304154

7 years ago[DAGCombiner] fix load narrowing transform to exclude loads with extension
Sanjay Patel [Mon, 29 May 2017 13:24:58 +0000 (13:24 +0000)]
[DAGCombiner] fix load narrowing transform to exclude loads with extension

The extending load possibility was missed in:
https://reviews.llvm.org/rL304072

We might want to handle this cases as a follow-up, but bailing out for now
to avoid miscompiling.

llvm-svn: 304153

7 years ago[SystemZ] Improve buildVector() in SystemZISelLowering.cpp.
Jonas Paulsson [Mon, 29 May 2017 13:22:23 +0000 (13:22 +0000)]
[SystemZ]  Improve buildVector() in SystemZISelLowering.cpp.

Use VLREP when inserting one or more loads into a vector. This is more
efficient than to first load and then use a VLVGP.

Review: Ulrich Weigand
llvm-svn: 304152

7 years ago[ScopInfo] Do not add terminator & synthesizable instructions to the output instructions.
Michael Kruse [Mon, 29 May 2017 12:27:38 +0000 (12:27 +0000)]
[ScopInfo] Do not add terminator & synthesizable instructions to the output instructions.

Such instructions are generates on-demand by the CodeGenerator and thus
do not need representation in a statement.

Differential Revision: https://reviews.llvm.org/D33642

llvm-svn: 304151

7 years agoTest commit: fix typos
Mattias Eriksson [Mon, 29 May 2017 11:46:44 +0000 (11:46 +0000)]
Test commit: fix typos

Just fixing a few typos in comments to test commit access.

llvm-svn: 304149

7 years agoRevert "[NFC] Fix formatting & typecast issue. Build succeeds."
Siddharth Bhat [Mon, 29 May 2017 11:34:29 +0000 (11:34 +0000)]
Revert "[NFC] Fix formatting & typecast issue. Build succeeds."

Should not have 'fixed' the formatting issue, I did not have the most
recent version of `clang-format`.
This reverts commit 761b1268359e14e59142f253d77864a29d55c56c.

llvm-svn: 304148

7 years agoReplaced StructuredData::Type with eStructuredDataType
Abhishek Aggarwal [Mon, 29 May 2017 11:13:30 +0000 (11:13 +0000)]
Replaced StructuredData::Type with eStructuredDataType

...missing from r304138 "Added new API to SBStructuredData class"

llvm-svn: 304147

7 years ago[NFC] Fix formatting & typecast issue. Build succeeds.
Siddharth Bhat [Mon, 29 May 2017 11:00:31 +0000 (11:00 +0000)]
[NFC] Fix formatting & typecast issue. Build succeeds.

- Fix formatting in `RegisterPasses.cpp`.
- `assert` tried to compare `isl::boolean` against `long`. Explicitly
construct `bool` from `isl::boolean`. This allows the implicit cast of
`bool` to `long.

llvm-svn: 304146

7 years ago[Nios2] Target registration
Nikolai Bozhenov [Mon, 29 May 2017 09:48:30 +0000 (09:48 +0000)]
[Nios2] Target registration

Reviewers: craig.topper, hfinkel, joerg, lattner, zvi

Reviewed By: craig.topper

Subscribers: oren_ben_simhon, igorb, belickim, tvvikram, mgorny, llvm-commits, pavel.v.chupin, DavidKreitzer

Differential Revision: https://reviews.llvm.org/D32669
Patch by AndreiGrischenko <andrei.l.grischenko@intel.com>

llvm-svn: 304144

7 years ago[ARM] GlobalISel: Extract helper. NFCI.
Diana Picus [Mon, 29 May 2017 09:09:54 +0000 (09:09 +0000)]
[ARM] GlobalISel: Extract helper. NFCI.

Create a helper to deal with the common code for merging incoming values
together after they've been split during call lowering. There's likely
more stuff that can be commoned up here, but we'll leave that for later.

llvm-svn: 304143

7 years agoMore StructuredData::Type::eTypeDictionary -> lldb::eStructuredDataTypeDictionary
Stephan Bergmann [Mon, 29 May 2017 08:51:58 +0000 (08:51 +0000)]
More StructuredData::Type::eTypeDictionary -> lldb::eStructuredDataTypeDictionary

...missing from previous r304138 "Added new API to SBStructuredData class"

llvm-svn: 304142

7 years agoclang-format: [JS] do not clean up duplicated commas.
Martin Probst [Mon, 29 May 2017 08:41:11 +0000 (08:41 +0000)]
clang-format: [JS] do not clean up duplicated commas.

Summary:
In JavaScript, duplicated commas have semantic meaning.
    x = [a,,b];

The statement above creates an array with three entries, the middle being undefined. Because clang-format should not change semantics, disable this cleanup in JS.

Reviewers: djasper

Subscribers: klimek

Differential Revision: https://reviews.llvm.org/D33641

llvm-svn: 304141

7 years ago[ELF] - Do not allow -r to eat comdats.
George Rimar [Mon, 29 May 2017 08:37:50 +0000 (08:37 +0000)]
[ELF] - Do not allow -r to eat comdats.

This is PR33052, "Bug 33052 - -r eats comdats ".

To fix it I stop removing group section from out when -r is given
and fixing SHT_GROUP content when writing it just like we do some
other fixup, e.g. for Rel[a]. (it needs fix for section indices that
are in group).

Differential revision: https://reviews.llvm.org/D33485

llvm-svn: 304140

7 years ago[trivial] fix a typo in comment, NFC
Hiroshi Inoue [Mon, 29 May 2017 08:37:42 +0000 (08:37 +0000)]
[trivial] fix a typo in comment, NFC

llvm-svn: 304139

7 years agoAdded new API to SBStructuredData class
Abhishek Aggarwal [Mon, 29 May 2017 08:25:46 +0000 (08:25 +0000)]
Added new API to SBStructuredData class

Summary:
 - Added API to access data types
    -- integer, double, array, string, boolean and dictionary data types
    -- Earlier user had to parse through the string output to get these
       values

 - Added Test cases for API testing

 - Added new StructuredDataType enum in public include file
   -- Replaced locally-defined enum in StructuredData.h with this new
      one
   -- Modified other internal files using this locally-defined enum

Signed-off-by: Abhishek Aggarwal <abhishek.a.aggarwal@intel.com>
Reviewers: clayborg, lldb-commits

Reviewed By: clayborg

Subscribers: labath

Differential Revision: https://reviews.llvm.org/D33434

llvm-svn: 304138

7 years ago[ARM] GlobalISel: Support array returns
Diana Picus [Mon, 29 May 2017 08:19:19 +0000 (08:19 +0000)]
[ARM] GlobalISel: Support array returns

These are a bit rare in practice, but they don't require anything
special compared to array parameters, so support them as well.

llvm-svn: 304137

7 years agoAdapt to recent clang-format changes
Tobias Grosser [Mon, 29 May 2017 08:06:29 +0000 (08:06 +0000)]
Adapt to recent clang-format changes

llvm-svn: 304136

7 years agoclang-format: [JS] fix indenting bound functions.
Martin Probst [Mon, 29 May 2017 07:50:52 +0000 (07:50 +0000)]
clang-format: [JS] fix indenting bound functions.

Summary:
The previous fix to force build style wrapping if the previous token is a closing parenthesis broke a peculiar pattern where users parenthesize the function declaration in a bind call:
    fn((function() { ... }).bind(this));

This restores the previous behaviour by reverting that change, but narrowing the special case for unindenting closing parentheses to those followed by semicolons and opening braces, i.e. immediate calls and function declarations.

Reviewers: djasper

Subscribers: cfe-commits, klimek

Differential Revision: https://reviews.llvm.org/D33640

llvm-svn: 304135

7 years ago[OpenCL] Test on half immediate support.
Egor Churaev [Mon, 29 May 2017 07:44:22 +0000 (07:44 +0000)]
[OpenCL] Test on half immediate support.

Reviewers: Anastasia

Reviewed By: Anastasia

Subscribers: yaxunl, cfe-commits, bader

Differential Revision: https://reviews.llvm.org/D33592

llvm-svn: 304134

7 years ago[PPC] Fix assertion failure during binary encoding with -mcpu=pwr9
Hiroshi Inoue [Mon, 29 May 2017 07:12:39 +0000 (07:12 +0000)]
[PPC] Fix assertion failure during binary encoding with -mcpu=pwr9

Summary
clang -c -mcpu=pwr9 test/CodeGen/PowerPC/build-vector-tests.ll causes an assertion failure during the binary encoding.
The failure occurs when a D-form load instruction takes two register operands instead of a register + an immediate.

This patch fixes the problem and also adds an assertion to catch this failure earlier before the binary encoding (i.e. during lit test).
The fix is from Nemanja Ivanovic @nemanjai.

Differential Revision: https://reviews.llvm.org/D33482

llvm-svn: 304133

7 years ago[ARM] GlobalISel: Support array parameters/arguments
Diana Picus [Mon, 29 May 2017 07:01:52 +0000 (07:01 +0000)]
[ARM] GlobalISel: Support array parameters/arguments

Clang coerces structs into arrays, so it's a good idea to support them.
Most of the support boils down to getting the splitToValueTypes helper
to actually split types. We then use G_INSERT/G_EXTRACT to deal with the
parts.

llvm-svn: 304132

7 years agoFix coroutine test failures caused by API misusages.
Eric Fiselier [Mon, 29 May 2017 06:42:01 +0000 (06:42 +0000)]
Fix coroutine test failures caused by API misusages.

More tests to come. I think that from_address overload should be deleted
or ill-formed, except for the 'void*' one; The user cannot possibly
have a typed pointer to the coroutine state.

llvm-svn: 304131

7 years agoDebugInfo: Include .dwo file name when hashing multiple CUs in a single file
Mehdi Amini [Mon, 29 May 2017 06:32:34 +0000 (06:32 +0000)]
DebugInfo: Include .dwo file name when hashing multiple CUs in a single file

This is really a workaround for ThinLTO in particular - since it can
import partial CUs that may end up looking very similar/the same as
the same partial import in another ThinLTO compile.

An alternative fix would be to change the DICompileUnit metadata to
include a "primary file" or the like - and when importing for ThinLTO
set the primary file to the name of the DICompileUnit that is being
imported into. This involves changing the schema and would reduce the
excessive uniqueness in the hash that this change creates - allowing
diagnosing of more duplicate CUs than will be caught with this change.

But duplicate CUs can still be caught in non-ThinLTO builds & are mostly
a nuisance rather than a particularly deliberate/effective tool for
finding broken code. (arguably the hash could always include the dwo
file and nothing in fission would break, I think..)

Reapply of r304119 after adding a triple to the test and moving it
to the X86 directory.

llvm-svn: 304130

7 years agoDebugInfo: Omit an empty CU when a subprogram was moved into its use
Mehdi Amini [Mon, 29 May 2017 06:25:30 +0000 (06:25 +0000)]
DebugInfo: Omit an empty CU when a subprogram was moved into its use

When the only use of a CU is for a subprogram that's only emitted into
the using CU (to avoid cross-CU references in DWO files), avoid creating
that CU at all.

Reapply of r304111 after adding a triple to the test and moving it
to the X86 directory.

llvm-svn: 304129

7 years agoRevert "[IfConversion] Keep the CFG updated incrementally in IfConvertTriangle"
Tobias Grosser [Mon, 29 May 2017 06:12:18 +0000 (06:12 +0000)]
Revert "[IfConversion] Keep the CFG updated incrementally in IfConvertTriangle"

The reverted change introdued assertions ala:

"MachineBasicBlock::succ_iterator
llvm::MachineBasicBlock::removeSuccessor(succ_iterator, bool): Assertion
`I != Successors.end() && "Not a current successor!"'

Mikael, the original committer, wrote me that he is working on a fix, but that
it likely will take some time to get this resolved. As this bug is one of the
last two issues that keep the AOSP buildbot from turning green, I revert the
original commit r302876.

I am looking forward to see this recommitted after the assertion has been
resolved.

llvm-svn: 304128

7 years agoIRGen: Add optnone attribute on function during O0
Mehdi Amini [Mon, 29 May 2017 05:38:20 +0000 (05:38 +0000)]
IRGen: Add optnone attribute on function during O0

Amongst other, this will help LTO to correctly handle/honor files
compiled with O0, helping debugging failures.
It also seems in line with how we handle other options, like how
-fnoinline adds the appropriate attribute as well.

Differential Revision: https://reviews.llvm.org/D28404

llvm-svn: 304127

7 years agoRevert "DebugInfo: Omit an empty CU when a subprogram was moved into its use"
Mehdi Amini [Mon, 29 May 2017 05:17:57 +0000 (05:17 +0000)]
Revert "DebugInfo: Omit an empty CU when a subprogram was moved into its use"

This reverts commit r304111.
GreenDragon is broken.

llvm-svn: 304126

7 years agoRevert "DebugInfo: Include .dwo file name when hashing multiple CUs in a single file"
Mehdi Amini [Mon, 29 May 2017 05:17:54 +0000 (05:17 +0000)]
Revert "DebugInfo: Include .dwo file name when hashing multiple CUs in a single file"

This reverts commit r304119 and r304118. GreenDragon is broken.

llvm-svn: 304125

7 years agoFix multiple bugs in coroutine tests.
Eric Fiselier [Mon, 29 May 2017 05:00:24 +0000 (05:00 +0000)]
Fix multiple bugs in coroutine tests.

llvm-svn: 304124

7 years agoDon't capture a temporary std::string in a StringRef.
Zachary Turner [Mon, 29 May 2017 02:20:12 +0000 (02:20 +0000)]
Don't capture a temporary std::string in a StringRef.

This fixes the breakages in llvm-tblgen.

llvm-svn: 304123

7 years agoResubmit "[X86] Adding new LLVM TableGen backend that generates the X86 backend memor...
Zachary Turner [Mon, 29 May 2017 02:19:37 +0000 (02:19 +0000)]
Resubmit "[X86] Adding new LLVM TableGen backend that generates the X86 backend memory folding tables."

This was reverted due to buildbot breakages and I was not familiar
with this code to investigate it.  But while trying to get a
useful backtrace for the author, it turns out the fix was very
obvious.  Resubmitting this patch as is, and will submit the
fix in a followup so that the fix is not hidden in the larger
CL.

llvm-svn: 304122

7 years agoRevert "[X86] Adding new LLVM TableGen backend that generates the X86 backend memory...
Zachary Turner [Mon, 29 May 2017 01:48:53 +0000 (01:48 +0000)]
Revert "[X86] Adding new LLVM TableGen backend that generates the X86 backend memory folding tables."

This reverts commit 28cb1003507f287726f43c771024a1dc102c45fe as well
as all subsequent followups.  llvm-tblgen currently segfaults with
this change, and it seems it has been broken on the bots all
day with no fixes in preparation.  See, for example:

http://lab.llvm.org:8011/builders/clang-x86-windows-msvc2015/

llvm-svn: 304121

7 years agoDisabled implicit-fallthrough warnings for ConvertUTF.cpp.
Galina Kistanova [Mon, 29 May 2017 01:34:26 +0000 (01:34 +0000)]
Disabled implicit-fallthrough warnings for ConvertUTF.cpp.

ConvertUTF.cpp has a little dependency on LLVM, and since the code extensively uses fall-through switches,
I prefer disabling the warning for the whole file, rather than adding attributes for each case.

llvm-svn: 304120

7 years agoDebugInfo: Include .dwo file name when hashing multiple CUs in a single file
David Blaikie [Mon, 29 May 2017 00:48:45 +0000 (00:48 +0000)]
DebugInfo: Include .dwo file name when hashing multiple CUs in a single file

This is really a workaround for ThinLTO in particular - since it can
import partial CUs that may end up looking very similar/the same as
the same partial import in another ThinLTO compile.

An alternative fix would be to change the DICompileUnit metadata to
include a "primary file" or the like - and when importing for ThinLTO
set the primary file to the name of the DICompileUnit that is being
imported into. This involves changing the schema and would reduce the
excessive uniqueness in the hash that this change creates - allowing
diagnosing of more duplicate CUs than will be caught with this change.

But duplicate CUs can still be caught in non-ThinLTO builds & are mostly
a nuisance rather than a particularly deliberate/effective tool for
finding broken code. (arguably the hash could always include the dwo
file and nothing in fission would break, I think..)

llvm-svn: 304119

7 years agoAttempt to fix buildbots...
David Blaikie [Mon, 29 May 2017 00:24:01 +0000 (00:24 +0000)]
Attempt to fix buildbots...

llvm-svn: 304118

7 years agoSupport: adjust the default obj format for wasm
Saleem Abdulrasool [Mon, 29 May 2017 00:14:57 +0000 (00:14 +0000)]
Support: adjust the default obj format for wasm

WebAssemly uses a custom object file format.  For the wasm targets,
default to the `Wasm` object file format.

llvm-svn: 304117

7 years ago[AVR] Remove SREG from CPI's Uses; authored by Florian Zeitz
Dylan McKay [Mon, 29 May 2017 00:10:14 +0000 (00:10 +0000)]
[AVR] Remove SREG from CPI's Uses; authored by Florian Zeitz

Summary: CPI does not read the status register, but only writes it.

Reviewers: dylanmckay

Reviewed By: dylanmckay

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D33223

llvm-svn: 304116

7 years ago[TableGen][X86] Fix formatting I accidentally messed up in r304099. NFC
Craig Topper [Sun, 28 May 2017 23:47:17 +0000 (23:47 +0000)]
[TableGen][X86] Fix formatting I accidentally messed up in r304099. NFC

llvm-svn: 304115

7 years ago[ItaniumDemangle] Fix a exponential string copying bug
Erik Pilkington [Sun, 28 May 2017 23:24:52 +0000 (23:24 +0000)]
[ItaniumDemangle] Fix a exponential string copying bug

This is a port of libcxxabi's r304113.

llvm-svn: 304114

7 years ago[demangler] Fix a exponential string copying bug
Erik Pilkington [Sun, 28 May 2017 23:15:50 +0000 (23:15 +0000)]
[demangler] Fix a exponential string copying bug

The problem was that if base_name() was called from a context without
an actual base name, it could gulp up the entire string, which can
result in recursive duplications. The fix is to be more strict as to
what qualifies as a base name.

Differential revision: https://reviews.llvm.org/D33637

llvm-svn: 304113

7 years agoPrune trailing whitespace. (To regenerate makefiles)
NAKAMURA Takumi [Sun, 28 May 2017 22:54:25 +0000 (22:54 +0000)]
Prune trailing whitespace. (To regenerate makefiles)

llvm-svn: 304112

7 years agoDebugInfo: Omit an empty CU when a subprogram was moved into its use
David Blaikie [Sun, 28 May 2017 22:51:37 +0000 (22:51 +0000)]
DebugInfo: Omit an empty CU when a subprogram was moved into its use

When the only use of a CU is for a subprogram that's only emitted into
the using CU (to avoid cross-CU references in DWO files), avoid creating
that CU at all.

llvm-svn: 304111

7 years ago[libcxxabi] Disable DLL annotations on static
Martell Malone [Sun, 28 May 2017 22:46:50 +0000 (22:46 +0000)]
[libcxxabi] Disable DLL annotations on static

rL288692 renames
_LIBCXXABI_DISABLE_DLL_IMPORT_EXPORT to
_LIBCXXABI_DISABLE_VISIBILITY_ANNOTATIONS

llvm-svn: 304110

7 years ago[AArch64][Falkor] Combine sched details files into one. NFC.
Geoff Berry [Sun, 28 May 2017 22:20:44 +0000 (22:20 +0000)]
[AArch64][Falkor] Combine sched details files into one. NFC.

llvm-svn: 304109

7 years ago[AArch64][Falkor] Fix some sched details.
Geoff Berry [Sun, 28 May 2017 21:48:31 +0000 (21:48 +0000)]
[AArch64][Falkor] Fix some sched details.

- Remove all uses of base sched model entries and set them all to
  Unsupported so all the opcodes are described in
  AArch64SchedFalkorDetails.td.
- Remove entries for unsupported half-float opcodes.
- Remove entries for unsupported LSE extension opcodes.
- Add entry for MOVbaseTLS (and set Sched in base td file entry to
  WriteSys) and a few other pseudo ops.
- Fix a few FP load/store with reg offset entries to use the LSLfast
  predicates.
- Add Q size BIF/BIT/BSL entries.
- Fix swapped Q/D sized CLS/CLZ/CNT/RBIT entires.
- Fix pre/post increment address register latency (this operand is
  always dest 0).
- Fix swapped FCVTHD/FCVTHS/FCVTDH/FCVTDS entries.
- Fix XYZ resource over usage on LD[1-4] opcodes.

llvm-svn: 304108

7 years ago[coroutines] Support "coroutines" feature in module map requires clause
Eric Fiselier [Sun, 28 May 2017 21:07:22 +0000 (21:07 +0000)]
[coroutines] Support "coroutines" feature in module map requires clause

Summary: In order for libc++ to add `<experimental/coroutine>` to its module map, there has to be a feature that can be used to detect if coroutines support is enabled in Clang.

Reviewers: rsmith

Reviewed By: rsmith

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D33538

llvm-svn: 304107

7 years agoCorrect XFAIL's in coroutine tests to only list ubsan
Eric Fiselier [Sun, 28 May 2017 20:56:16 +0000 (20:56 +0000)]
Correct XFAIL's in coroutine tests to only list ubsan

llvm-svn: 304106

7 years agoFix use of uninitialized memory
Eric Fiselier [Sun, 28 May 2017 20:49:13 +0000 (20:49 +0000)]
Fix use of uninitialized memory

llvm-svn: 304105

7 years agoXFAIL tests which trigger coroutine debug info crash in Clang
Eric Fiselier [Sun, 28 May 2017 20:49:11 +0000 (20:49 +0000)]
XFAIL tests which trigger coroutine debug info crash in Clang

llvm-svn: 304104

7 years agoAdd hack to allow test to run w/o C++17 copy elision
Eric Fiselier [Sun, 28 May 2017 20:30:18 +0000 (20:30 +0000)]
Add hack to allow test to run w/o C++17 copy elision

llvm-svn: 304103

7 years agoRemove XFAIL's for recently fixed bug
Eric Fiselier [Sun, 28 May 2017 20:25:41 +0000 (20:25 +0000)]
Remove XFAIL's for recently fixed bug

llvm-svn: 304102

7 years ago[coroutines] Add end-to-end tests within libc++
Eric Fiselier [Sun, 28 May 2017 19:38:21 +0000 (19:38 +0000)]
[coroutines] Add end-to-end tests within libc++

This patch adds end-to-end/breathing tests for coroutines
into libc++. The tests aren't specifically to test libc++ requirements
but instead are intented to ensure coroutines are working fine in general.

Although libc++ isn't exactly the most correct place for these tests
to live, there is one major advantage. The libc++ test suite is also
used by MSVC and by adding the tests here it ensures they will be
run against all currently available coroutine implementations.

llvm-svn: 304101

7 years ago[TableGen][X86] Use CHAR_BIT with sizeof instead of hardcoded 8. NFC
Craig Topper [Sun, 28 May 2017 18:24:43 +0000 (18:24 +0000)]
[TableGen][X86] Use CHAR_BIT with sizeof instead of hardcoded 8. NFC

llvm-svn: 304100

7 years ago[TableGen][X86] Mark a couple global tables as const. NFC
Craig Topper [Sun, 28 May 2017 18:24:41 +0000 (18:24 +0000)]
[TableGen][X86] Mark a couple global tables as const. NFC

llvm-svn: 304099

7 years ago[TableGen][X86] Improve formatting of the fold table output by indenting the body...
Craig Topper [Sun, 28 May 2017 18:24:39 +0000 (18:24 +0000)]
[TableGen][X86] Improve formatting of the fold table output by indenting the body of the table and adding blank lines between tables. NFC

llvm-svn: 304098

7 years ago[TableGen][X86] Add an llvm_unreachable to a switch so we get an error if we need...
Craig Topper [Sun, 28 May 2017 18:24:37 +0000 (18:24 +0000)]
[TableGen][X86] Add an llvm_unreachable to a switch so we get an error if we need expansion in the future.

llvm-svn: 304097

7 years ago[TableGen][X86] Remove unnecessary std::string creations. NFC
Craig Topper [Sun, 28 May 2017 18:24:35 +0000 (18:24 +0000)]
[TableGen][X86] Remove unnecessary std::string creations. NFC

llvm-svn: 304096

7 years ago[TableGen][X86] Replace a global std::vector with a regular array. llvm::find works...
Craig Topper [Sun, 28 May 2017 18:24:32 +0000 (18:24 +0000)]
[TableGen][X86] Replace a global std::vector with a regular array. llvm::find works on arrays, just need to use std::end to check the result.

llvm-svn: 304095

7 years ago[coroutines] Diagnose invalid result types for `await_resume` and `await_suspend...
Eric Fiselier [Sun, 28 May 2017 18:21:12 +0000 (18:21 +0000)]
[coroutines] Diagnose invalid result  types for `await_resume` and `await_suspend` and add missing conversions.

Summary:
The expression `await_ready` is required to be contextually convertible to bool and `await_suspend` must be a prvalue of either `void` or `bool`.
This patch adds diagnostics for when those requirements are violated.

It also correctly performs the contextual conversion to bool on the result of `await_ready`

Reviewers: GorNishanov, rsmith

Reviewed By: GorNishanov

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D33625

llvm-svn: 304094

7 years ago[TableGen][X86] getValueAsString returns a std::string not a StringRef. Capture it...
Craig Topper [Sun, 28 May 2017 17:48:41 +0000 (17:48 +0000)]
[TableGen][X86] getValueAsString returns a std::string not a StringRef. Capture it that way to avoid a StringRef to a temporary.

llvm-svn: 304093

7 years ago[coroutines] www/cxx_status.html: add non-breaking hyphen
Gor Nishanov [Sun, 28 May 2017 17:35:23 +0000 (17:35 +0000)]
[coroutines] www/cxx_status.html: add non-breaking hyphen

Reviewers: GorNishanov

Reviewed By: GorNishanov

Subscribers: EricWF, rsmith, cfe-commits

Differential Revision: https://reviews.llvm.org/D33632

llvm-svn: 304092

7 years ago[coroutines] www/cxx_status.html: add non-breaking hyphen
Gor Nishanov [Sun, 28 May 2017 16:25:31 +0000 (16:25 +0000)]
[coroutines] www/cxx_status.html: add non-breaking hyphen

Reviewers: GorNishanov

Reviewed By: GorNishanov

Subscribers: EricWF, rsmith, cfe-commits

Differential Revision: https://reviews.llvm.org/D33632

llvm-svn: 304091

7 years ago[x86] auto-generate better checks; NFC
Sanjay Patel [Sun, 28 May 2017 13:57:59 +0000 (13:57 +0000)]
[x86] auto-generate better checks; NFC

llvm-svn: 304090

7 years ago[InstrProf] Use more ArrayRef/StringRef.
Benjamin Kramer [Sun, 28 May 2017 13:23:02 +0000 (13:23 +0000)]
[InstrProf] Use more ArrayRef/StringRef.

No functional change intended.

llvm-svn: 304089

7 years ago[X86] Adding new LLVM TableGen backend that generates the X86 backend memory folding...
Ayman Musa [Sun, 28 May 2017 12:55:36 +0000 (12:55 +0000)]
[X86] Adding new LLVM TableGen backend that generates the X86 backend memory folding tables.

X86 backend holds huge tables in order to map between the register and memory forms of each instruction.
This TableGen Backend automatically generated all these tables with the appropriate flags for each entry.

Differential Revision: https://reviews.llvm.org/D32684

llvm-svn: 304088

7 years ago[X86] Adding FoldGenRegForm helper field (for memory folding tables tableGen backend...
Ayman Musa [Sun, 28 May 2017 12:39:37 +0000 (12:39 +0000)]
[X86] Adding FoldGenRegForm helper field (for memory folding tables tableGen backend) to X86Inst class and set its value for the relevant instructions.

Some register-register instructions can be encoded in 2 different ways, this happens when 2 register operands can be folded (separately).
For example if we look at the MOV8rr and MOV8rr_REV, both instructions perform exactly the same operation, but are encoded differently. Here is the relevant information about these instructions from Intel's 64-ia-32-architectures-software-developer-manual:

Opcode  Instruction  Op/En  64-Bit Mode  Compat/Leg Mode  Description
8A /r   MOV r8,r/m8  RM     Valid        Valid            Move r/m8 to r8.
88 /r   MOV r/m8,r8  MR     Valid        Valid            Move r8 to r/m8.
Here we can see that in order to enable the folding of the output and input registers, we had to define 2 "encodings", and as a result we got 2 move 8-bit register-register instructions.

In the X86 backend, we define both of these instructions, usually one has a regular name (MOV8rr) while the other has "_REV" suffix (MOV8rr_REV), must be marked with isCodeGenOnly flag and is not emitted from CodeGen.

Automatically generating the memory folding tables relies on matching encodings of instructions, but in these cases where we want to map both memory forms of the mov 8-bit (MOV8rm & MOV8mr) to MOV8rr (not to MOV8rr_REV) we have to somehow point from the MOV8rr_REV to the "regular" appropriate instruction which in this case is MOV8rr.

This field enable this "pointing" mechanism - which is used in the TableGen backend for generating memory folding tables.

Differential Revision: https://reviews.llvm.org/D32683

llvm-svn: 304087

7 years ago[X86] Fixing VPOPCNTDQ feature set lookup.
Oren Ben Simhon [Sun, 28 May 2017 11:26:11 +0000 (11:26 +0000)]
[X86] Fixing VPOPCNTDQ feature set lookup.

llvm-svn: 304086

7 years agobaremetal.cpp: Fix tests where lib suffix is 64
Ismail Donmez [Sun, 28 May 2017 06:15:42 +0000 (06:15 +0000)]
baremetal.cpp: Fix tests where lib suffix is 64

llvm-svn: 304085

7 years agoReverted r304083 as it seems there is a desire to address this in the googletest.
Galina Kistanova [Sun, 28 May 2017 05:50:22 +0000 (05:50 +0000)]
Reverted r304083 as it seems there is a desire to address this in the googletest.

llvm-svn: 304084

7 years agoAdded braces to address gcc warning: suggest explicit braces to avoid ambiguous ...
Galina Kistanova [Sun, 28 May 2017 03:50:52 +0000 (03:50 +0000)]
Added braces to address gcc warning: suggest explicit braces to avoid ambiguous 'else' [-Wdangling-else]. NFC.

llvm-svn: 304083