lib/Target/X86/X86ISelLowering.cpp:34613:25: error: enumeral mismatch in
conditional expression: 'llvm::ISD::NodeType' vs
'llvm::X86ISD::NodeType'

llvm-svn: 311580

The FXSAVE member `ftw` (FPU Tag Word) was given the wrong size (8-bit)
instead of the correct width (16-bit) as per the x87 Programmer's
Manual.  Adjust this to ensure that we print out the complete value for
the register.

llvm-svn: 311579

Remove some stray ';' that were in the source code.  NFC.

llvm-svn: 311577

GCC will interpret `__attribute__((__aligned__))` as 8-byte alignment on
ARM, but clang will not.  Explicitly specify the alignment.  This
mirrors the declaration in libunwind.

llvm-svn: 311576

Reviewers: arphaman

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D37072

llvm-svn: 311575

It seems that GCC interprets `__attribute__((__aligned__))` as 8-byte
alignment on ARM, but clang does not.  Explicitly specify the
double-word alignment value to ensure that the structure is properly
aligned.

llvm-svn: 311574

There are no 512-bit blend instructions so we shouldn't create SHRUNKBLEND for them.

On a side note, it looks like there may be a missed opportunity for constant folding TESTM when LHS and RHS are equal.

This fixes PR34139.

Differential Revision: https://reviews.llvm.org/D36992

llvm-svn: 311572

llvm-svn: 311571

Summary:
This adds shortcuts j and k to jump between changes.
It is especially useful in diffs with few changes.

Reviewers: arphaman

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D36685

llvm-svn: 311570

Summary:
This moves the data collection macro calls for Stmt nodes
to lib/AST/StmtDataCollectors.inc

Users can subclass ConstStmtVisitor and include StmtDataCollectors.inc
to define visitor methods for each Stmt subclass. This makes it also
possible to customize the visit methods as exemplified in
lib/Analysis/CloneDetection.cpp.

Move helper methods for data collection to a new module,
AST/DataCollection.

Add data collection for DeclRefExpr, MemberExpr and some literals.

Reviewers: arphaman, teemperor!

Subscribers: mgorny, xazax.hun, cfe-commits

Differential Revision: https://reviews.llvm.org/D36664

llvm-svn: 311569

There's no reason to have a target specific node with the same semantics as a target independent opcode.

This should simplify D36335 so that it doesn't need to touch X86ISelDAGToDAG.cpp

Differential Revision: https://reviews.llvm.org/D36983

llvm-svn: 311568

Signed-off-by: Yonghong Song <yhs@fb.com>
llvm-svn: 311567

This reverts commit r311559, which added a test containing raw string
literals in macros, which chokes gcc:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55971

llvm-svn: 311566

The lowering isn't really an optimization, so optnone shouldn't make a
difference. ARM relies on the pass running when using "-mthread-model
single", because in that mode, it doesn't run AtomicExpand. See bug for
more details.

Differential Revision: https://reviews.llvm.org/D37040

llvm-svn: 311565

LLVM_SVN_REVISION was used instead of LLVM_SVN_REV.
This caused a revision option to be ignored in Dockerfiles.

llvm-svn: 311564

The C++ ABI requires that the exception object (which under AEABI is the
`_Unwind_Control_Block`) is double-word aligned.  The attribute was
applied to the `_Unwind_Exception` type, but not the
`_Unwind_Control_Block`.  This should fix the libunwind test for the
alignment of the exception type.

llvm-svn: 311563

The C++ ABI requires that the exception object is double-word aligned.
The alignment attribute was applied to the `_Unwind_Exception` type
which is used on non-EHABI targets.  On EHABI, the exception object type
is `_Unwind_Control_Block`.  Apply the explicit maximal alignment on the
type to ensure that the allocation has the correct alignment.

Resolves PR33858!

llvm-svn: 311562

This makes -Wunreachable-code work for programs containing SEH (except for
__finally, which is still missing for now).

__try is modeled like try (but simpler since it can only have a single __except
or __finally), __except is fairly similar to catch (but simpler, since it can't
contain declarations). __leave is implemented similarly to break / continue.

Use the existing addTryDispatchBlock infrastructure (which
FindUnreachableCode() in ReachableCode.cpp uses via cfg->try_blocks_begin()) to
mark things in the __except blocks as reachable.

Re-use TryTerminatedBlock. This means we add EH edges from calls to the __try
block, but not from all other statements. While this is incomplete, it matches
LLVM's SEH codegen support. Also, in practice, BuildOpts.AddEHEdges is always
false in practice from what I can tell, so we never even insert the call EH
edges either.

https://reviews.llvm.org/D36914

llvm-svn: 311561

http://lab.llvm.org:8011/builders/llvm-clang-x86_64-expensive-checks-win/builds/4394

llvm-svn: 311560

Summary:
This patch makes the splits emitted for the beginning of comment lines during
reformatting absolute. Previously, they were relative to the start of the
non-whitespace content of the line, which messes up further TailOffset
calculations in breakProtrudingToken. This fixes an assertion failure reported
in bug 34236: https://bugs.llvm.org/show_bug.cgi?id=34236.

Reviewers: djasper

Reviewed By: djasper

Subscribers: klimek, cfe-commits

Differential Revision: https://reviews.llvm.org/D36956

llvm-svn: 311559

It was marked as unsupported on Windows in r311230 because on some Win10 
machines it failed or caused hang. The problem was that on these machines
system bash (C:\Windows\System32\bash.exe) was used which requires paths to be
passed like '/mnt/c/path/to/my/script' instead of 'C:\path\to\my\script'.

TODO: we should make lit detect if system bash is used instead of msys and set
appropriate path format.

llvm-svn: 311558

This reverts commit r311552 because it broke ubsan and asan bots.

llvm-svn: 311557

Summary:
If a coroutine outer calls another coroutine inner and the inner coroutine body is inlined into the outer, coro.begin from the inner coroutine should be considered for spilling if accessed across suspends.

Prior to this change, coroutine frame building code was not considering any coro.begins for spilling.
With this change, we only ignore coro.begin for the current coroutine, but, any coro.begins that were inlined into the current coroutine are eligible for spills.

Fixes PR34267

Reviewers: GorNishanov

Subscribers: qcolombet, llvm-commits, EricWF

Differential Revision: https://reviews.llvm.org/D37062

llvm-svn: 311556

[ARM][Compiler-rt] Fix AEABI builtins to correctly pass arguments to non-AEABI functions on HF targets

Summary:
This is a patch for PR34167.

On HF targets functions like `__{eq,lt,le,ge,gt}df2` and `__{eq,lt,le,ge,gt}sf2` expect their arguments to be passed in d/s registers, while some of the AEABI builtins pass them in r registers.

Reviewers: compnerd, peter.smith, asl

Reviewed By: peter.smith, asl

Subscribers: peter.smith, aemerson, dberris, javed.absar, llvm-commits, asl, kristof.beyls

Differential Revision: https://reviews.llvm.org/D36675

llvm-svn: 311555

..if the resulting subtract will be broken up later.  This can cause us to get
into an infinite loop.

x + (-5.0 * y)      -> x - (5.0 * y)       ; Canonicalize neg const
x - (5.0 * y)       -> x + (0 - (5.0 * y)) ; Break up subtract
x + (0 - (5.0 * y)) -> x + (-5.0 * y)      ; Replace 0-X with X*-1.

PR34078

llvm-svn: 311554

Add statistics about
- Which optimizations are applied
- Number of loops in Scops at various stages
- Number of scalar/singleton writes at various stages representative
  for scalar false dependencies
- Number of parallel loops

These will be useful to find regressions due to moving Polly further
down of LLVM's pass pipeline.

Differential Revision: https://reviews.llvm.org/D37049

llvm-svn: 311553

Summary:
This is a patch for clang autocomplete feature.

It will collect values which -analyzer-checker takes, which is defined in
clang/StaticAnalyzer/Checkers/Checkers.inc, dynamically.
First, from ValuesCode class in Options.td, TableGen will generate C++
code in Options.inc. Options.inc will be included in DriverOptions.cpp, and
calls OptTable's addValues function. addValues function will add second
argument to Option's Values class. Values contains string like "foo,bar,.."
which is handed to Values class
in OptTable.

Reviewers: v.g.vassilev, teemperor, ruiu

Subscribers: hiraditya, cfe-commits

Differential Revision: https://reviews.llvm.org/D36782

llvm-svn: 311552

Loop with zero iteration are, syntactically, loops. They have been
excluded from the loop counter even for the non-profitable counters.
This seems to be unintentially as the sentinel value of '0' minimal
iterations does exclude such loops.

Fix by never considering the iteration count when the sentinel
value of 0 is found.

This makes the recently added NumTotalLoops couter redundant
with NumLoopsOverall, which now is equivalent. Hence, NumTotalLoops
is removed as well.

Note: The test case 'ScopDetect/statistics.ll' effectively does not
check profitability, because -polly-process-unprofitable is passed
to all test cases.

llvm-svn: 311551

By exposing the the hidden member, but as private.

llvm-svn: 311550

Conditionally compile function only used in an assert().

llvm-svn: 311549

MSVC warns about comparison between a signed and unsigned integer.
The rules of C(++) define that an unsigned comparison has to be
carried-out in this case. This is unlikely to be intended.

Fix by assigning the loop's upper bound to a signed integer first.
This also avoids repeated evaluation of the invariant upper bound.

llvm-svn: 311548

llvm-svn: 311547

Summary:
This patch adds support for predicates on imm nodes but only for ImmLeaf and not for PatLeaf or PatFrag and only where the value does not need to be transformed before being rendered into the instruction.

The limitation on PatLeaf/PatFrag/SDNodeXForm is due to differences in the necessary target-supplied C++ for GlobalISel.

Depends on D36085

Reviewers: ab, t.p.northover, qcolombet, rovka, aditya_nandakumar

Reviewed By: rovka

Subscribers: kristof.beyls, javed.absar, igorb, llvm-commits

Differential Revision: https://reviews.llvm.org/D36086

llvm-svn: 311546

Currently this test causes test failures on some machines, due to isel not being registered. Update the test to run all passes and check emitted assembly instructions for now. 

llvm-svn: 311545

llvm-svn: 311544

Summary: In some cases, shufflevector instruction can be transformed involving insert_subvector instructions. The ARM backend was missing some insert_subvector patterns, causing a failure during instruction selection. AArch64 has similar patterns.

Reviewers: t.p.northover, olista01, javed.absar, rengolin

Reviewed By: javed.absar

Subscribers: aemerson, kristof.beyls, llvm-commits

Differential Revision: https://reviews.llvm.org/D36796

llvm-svn: 311543

llvm-svn: 311542

lld was broken in this regard (PR33097). The gold plugin gets this
right so, no changes needed, but better adding a test.

llvm-svn: 311541

InstCombine folds instructions with irrelevant conditions to undef.
This, as Nuno confirmed is a bug.
(see https://bugs.llvm.org/show_bug.cgi?id=33409#c1 )

Given the original motivation for the change is that of removing an
USE, we now fold to false instead (which reaches the same goal
without undesired side effects).

Fixes PR33409.

Differential Revision:  https://reviews.llvm.org/D36975

llvm-svn: 311540

- recommitting after fixing a test failure on MacOS

On PPC64, OR (XOR) with a 32-bit immediate can be done with only two instructions, i.e. ori + oris.
But the current LLVM generates three or four instructions for this purpose (and also it clobbers one GPR).

This patch makes PPC backend generate ori + oris (xori + xoris) for OR (XOR) with a 32-bit immediate.

e.g. (x | 0xFFFFFFFF) should be

	ori 3, 3, 65535
	oris 3, 3, 65535

but LLVM generates without this patch

	li 4, 0
	oris 4, 4, 65535
	ori 4, 4, 65535
	or 3, 3, 4

Differential Revision: https://reviews.llvm.org/D34757

llvm-svn: 311538