platform/upstream/llvm.git
10 years agoParse: Don't consider attributes of broken member declarators
David Majnemer [Thu, 1 Aug 2013 04:22:55 +0000 (04:22 +0000)]
Parse: Don't consider attributes of broken member declarators

ParseCXXClassMemberDeclaration was trying to use the result of
ActOnCXXMemberDeclarator to attach it to some late parsed attributes.

However when failures arise, we have no decl to attach to which
eventually leads us to a NULL pointer dereference.

While we are here, clean up the code a bit.

Fixes PR16765

llvm-svn: 187557

10 years agoFix assert when instantiating a default argument of a template defined in a
Richard Smith [Thu, 1 Aug 2013 04:12:04 +0000 (04:12 +0000)]
Fix assert when instantiating a default argument of a template defined in a
module.

llvm-svn: 187556

10 years agoFix some misc. issues with Mips16 fp stubs.
Reed Kotler [Thu, 1 Aug 2013 02:26:31 +0000 (02:26 +0000)]
Fix some misc. issues with Mips16 fp stubs.
1) They should never be inlined.
2) A naming inconsistency with gcc mips16
3) Stubs should not have the global attribute

llvm-svn: 187555

10 years agoFormatting.
Eric Christopher [Thu, 1 Aug 2013 01:38:16 +0000 (01:38 +0000)]
Formatting.

llvm-svn: 187554

10 years agoAdd an omitted IsCall=1.
Reed Kotler [Thu, 1 Aug 2013 00:59:06 +0000 (00:59 +0000)]
Add an omitted IsCall=1.

llvm-svn: 187553

10 years agoConstrain fill_n -> memset operations to include implicit convertibility to unsigned...
Howard Hinnant [Thu, 1 Aug 2013 00:41:55 +0000 (00:41 +0000)]
Constrain fill_n -> memset operations to include implicit convertibility to unsigned char.  This fixes llvm.org/bugs/show_bug.cgi?id=16764.  Also a drive-by fix on a chrono test suite bug.

llvm-svn: 187552

10 years agoRemove the SUPPORT_ALIASARGS define
Hans Wennborg [Thu, 1 Aug 2013 00:12:38 +0000 (00:12 +0000)]
Remove the SUPPORT_ALIASARGS define

It is not needed after LLVM r187546.

llvm-svn: 187551

10 years agoRemove the SUPPORT_ALIASARGS define
Hans Wennborg [Thu, 1 Aug 2013 00:12:13 +0000 (00:12 +0000)]
Remove the SUPPORT_ALIASARGS define

It is not needed after LLVM r187546.

llvm-svn: 187550

10 years agoclang-format: Make alignment of trailing comments optional ..
Daniel Jasper [Wed, 31 Jul 2013 23:55:15 +0000 (23:55 +0000)]
clang-format: Make alignment of trailing comments optional ..

.. in order to support WebKit style properly.

llvm-svn: 187549

10 years agoUse StringRef::equals_lower(). No functionality change.
Rui Ueyama [Wed, 31 Jul 2013 23:53:58 +0000 (23:53 +0000)]
Use StringRef::equals_lower(). No functionality change.

llvm-svn: 187548

10 years agoMake -m[no-]pascal-strings an alias of -f[no]pascal-strings in the .td file
Hans Wennborg [Wed, 31 Jul 2013 23:39:13 +0000 (23:39 +0000)]
Make -m[no-]pascal-strings an alias of -f[no]pascal-strings in the .td file

This way we don't have to translate it manually in Clang::ConstructJob.

Differential Revision: http://llvm-reviews.chandlerc.com/D1249

llvm-svn: 187547

10 years agoOption parsing: remove non-SUPPORT_ALIASARGS fall-back
Hans Wennborg [Wed, 31 Jul 2013 23:28:51 +0000 (23:28 +0000)]
Option parsing: remove non-SUPPORT_ALIASARGS fall-back

The clients of this code have been updated to all support AliasArgs.

This depends on Clang r187538 and lld r187541.

llvm-svn: 187546

10 years agoFix thread name updating in Linux. "thread list" should report correct names always...
Michael Sartain [Wed, 31 Jul 2013 23:27:46 +0000 (23:27 +0000)]
Fix thread name updating in Linux. "thread list" should report correct names always now.
Created new LinuxThread class inherited from POSIXThread and removed linux / freebsd ifdefs
Removed several un-needed set thread name calls

CR (and multiple suggestions): mkopec

llvm-svn: 187545

10 years ago[PECOFF][Driver] Prefer "/" over "-" as the indicator of the option.
Rui Ueyama [Wed, 31 Jul 2013 23:23:25 +0000 (23:23 +0000)]
[PECOFF][Driver] Prefer "/" over "-" as the indicator of the option.

This patch does not change the behavior of LLD, but changes the output of the
help text. We want to show the help text with Windows style indicator rather
than Unix style indicator.

llvm-svn: 187544

10 years agoObjectiveC ARC: finishing off issuing error when
Fariborz Jahanian [Wed, 31 Jul 2013 23:19:34 +0000 (23:19 +0000)]
ObjectiveC ARC: finishing off issuing error when
retainable pointer is passed to an audited CF function
expecting CF type. // rdar://14569171

llvm-svn: 187543

10 years agoOptimize Host::GetThreadName() to read from /proc/$TID per Matt's suggestion.
Michael Sartain [Wed, 31 Jul 2013 23:19:14 +0000 (23:19 +0000)]
Optimize Host::GetThreadName() to read from /proc/$TID per Matt's suggestion.

CR: mkopec
llvm-svn: 187542

10 years agoOption parsing tables: pick up AliasArgs from the OPTION macro.
Hans Wennborg [Wed, 31 Jul 2013 23:17:41 +0000 (23:17 +0000)]
Option parsing tables: pick up AliasArgs from the OPTION macro.

This depends on LLVM r187537.

The SUPPORT_ALIASARGS macro will be removed once all option parsing
clients have been updated.

llvm-svn: 187541

10 years agoclang-format: Add more options to namespace indentation.
Daniel Jasper [Wed, 31 Jul 2013 23:16:02 +0000 (23:16 +0000)]
clang-format: Add more options to namespace indentation.

With this patch, clang-format can be configured to:
* not indent in namespace at all (former behavior).
* indent in namespace as in other blocks.
* indent only in inner namespaces (as required by WebKit style).

Also fix alignment of access specifiers in WebKit style.

Patch started by Marek Kurdej. Thank you!

llvm-svn: 187540

10 years ago[PECOFF][Driver] Rewrite the tablegen def file using multiclass to remove duplicates.
Rui Ueyama [Wed, 31 Jul 2013 23:12:06 +0000 (23:12 +0000)]
[PECOFF][Driver] Rewrite the tablegen def file using multiclass to remove duplicates.

llvm-svn: 187539

10 years agoOption parsing tables: pick up AliasArgs from the OPTION macro.
Hans Wennborg [Wed, 31 Jul 2013 23:07:21 +0000 (23:07 +0000)]
Option parsing tables: pick up AliasArgs from the OPTION macro.

This depends on LLVM r187537.

The SUPPORT_ALIASARGS macro will be removed once all option parsing
clients have been updated.

llvm-svn: 187538

10 years agoOption parsing: add support for alias arguments.
Hans Wennborg [Wed, 31 Jul 2013 22:44:41 +0000 (22:44 +0000)]
Option parsing: add support for alias arguments.

This makes option aliases more powerful by enabling them to
pass along arguments to the option they're aliasing.

For example, if we have a joined option "-foo=", we can now
specify a flag option "-bar" to be an alias of that, with the
argument "baz".

This is especially useful for the cl.exe compatible clang driver,
where many options are aliases. For example, this patch enables
us to alias "/Ox" to "-O3" (-O is a joined option), and "/WX" to
"-Werror" (again, -W is a joined option).

Differential Revision: http://llvm-reviews.chandlerc.com/D1245

llvm-svn: 187537

10 years agoRemove dead #includes.
Hans Wennborg [Wed, 31 Jul 2013 22:28:46 +0000 (22:28 +0000)]
Remove dead #includes.

llvm-svn: 187536

10 years ago80-col
Nadav Rotem [Wed, 31 Jul 2013 22:17:45 +0000 (22:17 +0000)]
80-col

llvm-svn: 187535

10 years agoRevert "[PECOFF][Driver] Remove quotes from command line arguments."
Rui Ueyama [Wed, 31 Jul 2013 22:13:15 +0000 (22:13 +0000)]
Revert "[PECOFF][Driver] Remove quotes from command line arguments."

This reverts commit r187390 because we should not handle argv's quotes ourselves.

In Windows, unlike Unix, quotes are not processed by the shell. Instead the C
startup routine parses it as described in
http://msdn.microsoft.com/en-us/library/a1y7w461.aspx and pass the results to
main(). So, at the time when the control reaches main(), quotes that should be
removed has already been removed.

We still need to handle quotes in the response file and in .drectve section
ourselves. That will be addressed in different patches.

llvm-svn: 187534

10 years agoMark some tests as explicitly failing on clang/gcc as they pass on icc.
Matt Kopec [Wed, 31 Jul 2013 21:52:25 +0000 (21:52 +0000)]
Mark some tests as explicitly failing on clang/gcc as they pass on icc.
Also, rework the signed types test to check for signed or char type in the output as char is signed by default.

llvm-svn: 187533

10 years agoObjectiveC ARC: Do not issue bridge cast diagnostic when
Fariborz Jahanian [Wed, 31 Jul 2013 21:40:51 +0000 (21:40 +0000)]
ObjectiveC ARC: Do not issue bridge cast diagnostic when
passing a retainable object arg to a CF audited function
expecting a CF object type. Issue a normal type mismatch
diagnostic. This is wip // rdar://14569171

llvm-svn: 187532

10 years agocomment typo.
Andrew Trick [Wed, 31 Jul 2013 21:05:54 +0000 (21:05 +0000)]
comment typo.

llvm-svn: 187531

10 years agoAdded the B9.3.19 SUBS PC, LR, #imm (Thumb2) system instruction.
Kevin Enderby [Wed, 31 Jul 2013 21:05:30 +0000 (21:05 +0000)]
Added the B9.3.19 SUBS PC, LR, #imm (Thumb2) system instruction.
While the .td entry is nice and all, it takes a pretty gross hack in
ARMAsmParser::ParseInstruction() because of handling of other "subs"
instructions to get it to match.  Ran it by Jim Grosbach and he said it was
about what he expected to make this work given the existing code.

rdar://14214063

llvm-svn: 187530

10 years agoImplement constexpr (n3302) and fix operator *= and /=
Marshall Clow [Wed, 31 Jul 2013 21:02:34 +0000 (21:02 +0000)]
Implement constexpr (n3302) and fix operator *= and /=

llvm-svn: 187529

10 years agoFix declaring class template methods with an attributed typedef
Reid Kleckner [Wed, 31 Jul 2013 21:00:18 +0000 (21:00 +0000)]
Fix declaring class template methods with an attributed typedef

This change unifies the logic for template instantiation of methods and
functions declared with typedefs.

It ensures that SubstFunctionType() always fills the Params out param
with non-null ParmVarDecls or returns null.

Reviewers: rsmith

Differential Revision: http://llvm-reviews.chandlerc.com/D1135

llvm-svn: 187528

10 years agoclang-cl: add the /c, /W0 and /W1 options
Hans Wennborg [Wed, 31 Jul 2013 20:51:53 +0000 (20:51 +0000)]
clang-cl: add the /c, /W0 and /W1 options

This adds a few more clang-cl options. It also exposes two core clang
options to the clang-cl mode: we need to be able to claim --driver_mode
so it doesn't show up as unused in cl mode, and we need -### for tests.

Differential Revision: http://llvm-reviews.chandlerc.com/D1232

llvm-svn: 187527

10 years agoRevert "R600: Non vector only instruction can be scheduled on trans unit"
Tom Stellard [Wed, 31 Jul 2013 20:43:27 +0000 (20:43 +0000)]
Revert "R600: Non vector only instruction can be scheduled on trans unit"

This reverts commit 98ce62780ea7185ba710868bf83c8077e8d7f6d6.

llvm-svn: 187526

10 years agoRevert "TableGen: Enumerate Schedule Model too."
Tom Stellard [Wed, 31 Jul 2013 20:43:08 +0000 (20:43 +0000)]
Revert "TableGen: Enumerate Schedule Model too."

This reverts commit 2ca1e4a39c7e0d7a00e66ff5437c6d7ace2404a0.

llvm-svn: 187525

10 years agoRevert "R600: Use SchedModel enum for is{Trans,Vector}Only functions"
Tom Stellard [Wed, 31 Jul 2013 20:43:03 +0000 (20:43 +0000)]
Revert "R600: Use SchedModel enum for is{Trans,Vector}Only functions"

This reverts commit 3f1de26cb5cc0543a6a1d71259a7a39d97139051.

llvm-svn: 187524

10 years agoFix lock hierarchy violation in Process (lock ordering of ThreadList mutex and StackF...
Daniel Malea [Wed, 31 Jul 2013 20:21:20 +0000 (20:21 +0000)]
Fix lock hierarchy violation in Process (lock ordering of ThreadList mutex and StackFrameList mutex)
- this fix ensures the ThreadList mutex is always locked before the StackFrameList mutex

Situation where deadlock could occur (without this fix):
Thread 1 is in Process::WillResume and locks the ThreadList mutex (on entry), and subsequently calls StackFrameList::Clear() which locks the StackFrameList mutex.
Meanwhile, thread 2 is in Process::RunThreadPlan and calls Thread::SetSelectedFrame() (which locks the StackFrameList mutex) before calling GetSelectedThread (which attempts to lock the ThreadList mutex)

In my testing on both Linux and Mac OS X, I was unable to reproduce any hangs with this patch applied.

llvm-svn: 187522

10 years agoA few small cleanups to r187504. Thanks to dblaikie for the assist.
Kaelyn Uhrain [Wed, 31 Jul 2013 20:16:17 +0000 (20:16 +0000)]
A few small cleanups to r187504. Thanks to dblaikie for the assist.

llvm-svn: 187521

10 years agoAdd the LD_LIBRARY_PATH to target.env-vars for tests expecting libimf.so when testing...
Matt Kopec [Wed, 31 Jul 2013 20:13:04 +0000 (20:13 +0000)]
Add the LD_LIBRARY_PATH to target.env-vars for tests expecting libimf.so when testing with ICC.

Patch from Andy Kaylor.

llvm-svn: 187520

10 years agoAdds a test for 'target module dump symfile' to the LLDB suite.
Ashok Thirumurthi [Wed, 31 Jul 2013 20:01:04 +0000 (20:01 +0000)]
Adds a test for 'target module dump symfile' to the LLDB suite.

TODO: Improve coverage of SBTypeMember and of 'target module dump'.
llvm-svn: 187519

10 years agoBackwards!
Marshall Clow [Wed, 31 Jul 2013 19:39:37 +0000 (19:39 +0000)]
Backwards!

llvm-svn: 187518

10 years agoImplement n3469 - constexpr for chrono
Marshall Clow [Wed, 31 Jul 2013 19:32:19 +0000 (19:32 +0000)]
Implement n3469 - constexpr for chrono

llvm-svn: 187517

10 years agoR600: Do not mergevector after a vector reg is used
Vincent Lejeune [Wed, 31 Jul 2013 19:32:12 +0000 (19:32 +0000)]
R600: Do not mergevector after a vector reg is used

If we merge vector when a vector is used, it will generate an artificial
antidependency that can prevent 2 tex/vtx instructions to use the same
clause and thus generate extra clauses that reduce performance.

There is no test case as such situation is really hard to predict.

llvm-svn: 187516

10 years agoR600: Avoid more than 4 literals in the same instruction group at scheduling
Vincent Lejeune [Wed, 31 Jul 2013 19:32:07 +0000 (19:32 +0000)]
R600: Avoid more than 4 literals in the same instruction group at scheduling

llvm-svn: 187515

10 years agoR600: Non vector only instruction can be scheduled on trans unit
Vincent Lejeune [Wed, 31 Jul 2013 19:31:56 +0000 (19:31 +0000)]
R600: Non vector only instruction can be scheduled on trans unit

llvm-svn: 187514

10 years agoR600: Don't mix LDS and non-LDS instructions in the same group
Vincent Lejeune [Wed, 31 Jul 2013 19:31:41 +0000 (19:31 +0000)]
R600: Don't mix LDS and non-LDS instructions in the same group

There are a lot of restrictions on instruction groups that contain
LDS instructions, so for now we will be conservative and not packetize
anything else with them.

llvm-svn: 187513

10 years agoR600: Use SchedModel enum for is{Trans,Vector}Only functions
Vincent Lejeune [Wed, 31 Jul 2013 19:31:35 +0000 (19:31 +0000)]
R600: Use SchedModel enum for is{Trans,Vector}Only functions

llvm-svn: 187512

10 years agoTableGen: Enumerate Schedule Model too.
Vincent Lejeune [Wed, 31 Jul 2013 19:31:20 +0000 (19:31 +0000)]
TableGen: Enumerate Schedule Model too.

llvm-svn: 187511

10 years agoR600: Remove predicated_break inst
Vincent Lejeune [Wed, 31 Jul 2013 19:31:14 +0000 (19:31 +0000)]
R600: Remove predicated_break inst

We were using two instructions for similar purpose : break and
predicated break. Only predicated_break was emitted and it was
lowered at R600ControlFlowFinalizer to JUMP;CF_BREAK;POP.
This commit simplify the situation by making AMDILCFGStructurizer
emit IF_PREDICATE;BREAK;ENDIF; instead of predicated_break (which
is now removed).

There is no functionality change.

llvm-svn: 187510

10 years agoObjectiveC arc: minor refactoring in my last patch
Fariborz Jahanian [Wed, 31 Jul 2013 18:39:08 +0000 (18:39 +0000)]
ObjectiveC arc: minor refactoring in my last patch
to avoid future false positives. // rdar://14569171

llvm-svn: 187509

10 years agoObjectiveC arc: Introduce a new initialization kind
Fariborz Jahanian [Wed, 31 Jul 2013 18:21:45 +0000 (18:21 +0000)]
ObjectiveC arc: Introduce a new initialization kind
for parameters passed to CF audited functions
to be used for better diagnostics. Current set but
unused. // rdar://14569171

llvm-svn: 187508

10 years agoMark additional ICC failing tests as expected fail.
Matt Kopec [Wed, 31 Jul 2013 17:56:22 +0000 (17:56 +0000)]
Mark additional ICC failing tests as expected fail.

llvm-svn: 187507

10 years agoReject bitcasts between address spaces with different sizes
Matt Arsenault [Wed, 31 Jul 2013 17:49:08 +0000 (17:49 +0000)]
Reject bitcasts between address spaces with different sizes

llvm-svn: 187506

10 years ago[PECOFF] Add a comment about .bss section.
Rui Ueyama [Wed, 31 Jul 2013 17:45:48 +0000 (17:45 +0000)]
[PECOFF] Add a comment about .bss section.

llvm-svn: 187505

10 years agoImprove the diagnostic experience, including adding recovery, for
Kaelyn Uhrain [Wed, 31 Jul 2013 17:38:24 +0000 (17:38 +0000)]
Improve the diagnostic experience, including adding recovery, for
changing '->' to '.' when there is no operator-> defined for a class.

llvm-svn: 187504

10 years agoObjectiveC arc: Move check for type conversions in arc
Fariborz Jahanian [Wed, 31 Jul 2013 17:12:26 +0000 (17:12 +0000)]
ObjectiveC arc: Move check for type conversions in arc
out of ImpCastExprToType and to the caller site
as appropriate. This is in prep. to do more work for
// rdar://14569171

llvm-svn: 187503

10 years agoFix a comment.
Hans Wennborg [Wed, 31 Jul 2013 16:57:56 +0000 (16:57 +0000)]
Fix a comment.

llvm-svn: 187502

10 years agoUpdate break conditions test to have consistent behaviour on all test suite compilers.
Matt Kopec [Wed, 31 Jul 2013 16:27:27 +0000 (16:27 +0000)]
Update break conditions test to have consistent behaviour on all test suite compilers.
Also update comment in const variables test to reflect ICC status.

llvm-svn: 187501

10 years agoSet an extra debug flag when testing with ICC so that it generates the correct debug...
Matt Kopec [Wed, 31 Jul 2013 16:10:03 +0000 (16:10 +0000)]
Set an extra debug flag when testing with ICC so that it generates the correct debug info for inlined tests.

llvm-svn: 187500

10 years agoUpdated the LLDB rvalue reference test to cross-reference with a bugzilla report,
Ashok Thirumurthi [Wed, 31 Jul 2013 15:58:01 +0000 (15:58 +0000)]
Updated the LLDB rvalue reference test to cross-reference with a bugzilla report,
provide more detail on compiler compatibility, and to illustrate that this is
an issue with expression evaluation.

- Note that clang doesn't emit DW_TAG_const_type, which might be okay if there's
no such thing as a non-const rvalue reference.  How about foo(make_int())?

llvm-svn: 187499

10 years agoDependence: Add DEBUG support.
Tobias Grosser [Wed, 31 Jul 2013 14:35:17 +0000 (14:35 +0000)]
Dependence: Add DEBUG support.

Contributed-by: Star Tan <tanmx_star@yeah.net>
llvm-svn: 187498

10 years ago[SystemZ] Implement isLegalAddressingMode()
Richard Sandiford [Wed, 31 Jul 2013 12:58:26 +0000 (12:58 +0000)]
[SystemZ] Implement isLegalAddressingMode()

The loop optimizers were assuming that scales > 1 were OK.  I think this
is actually a bug in TargetLoweringBase::isLegalAddressingMode(),
since it seems to be trying to reject anything that isn't r+i or r+r,
but it has no default case for scales other than 0, 1 or 2.  Implementing
the hook for z means that z can no longer test any change there though.

llvm-svn: 187497

10 years ago[SystemZ] Be more careful about inverting CC masks (conditional loads)
Richard Sandiford [Wed, 31 Jul 2013 12:38:08 +0000 (12:38 +0000)]
[SystemZ] Be more careful about inverting CC masks (conditional loads)

Extend r187495 to conditional loads.  I split this out because the
easiest way seemed to be to force a particular operand order in
SystemZISelDAGToDAG.cpp.

llvm-svn: 187496

10 years ago[SystemZ] Be more careful about inverting CC masks
Richard Sandiford [Wed, 31 Jul 2013 12:30:20 +0000 (12:30 +0000)]
[SystemZ] Be more careful about inverting CC masks

System z branches have a mask to select which of the 4 CC values should
cause the branch to be taken.  We can invert a branch by inverting the mask.
However, not all instructions can produce all 4 CC values, so inverting
the branch like this can lead to some oddities.  For example, integer
comparisons only produce a CC of 0 (equal), 1 (less) or 2 (greater).
If an integer EQ is reversed to NE before instruction selection,
the branch will test for 1 or 2.  If instead the branch is reversed
after instruction selection (by inverting the mask), it will test for
1, 2 or 3.  Both are correct, but the second isn't really canonical.
This patch therefore keeps track of which CC values are possible
and uses this when inverting a mask.

Although this is mostly cosmestic, it fixes undefined behavior
for the CIJNLH in branch-08.ll.  Another fix would have been
to mask out bit 0 when generating the fused compare and branch,
but the point of this patch is that we shouldn't need to do that
in the first place.

The patch also makes it easier to reuse CC results from other instructions.

llvm-svn: 187495

10 years ago[SystemZ] Move compare-and-branch generation even later
Richard Sandiford [Wed, 31 Jul 2013 12:11:07 +0000 (12:11 +0000)]
[SystemZ] Move compare-and-branch generation even later

r187116 moved compare-and-branch generation from the instruction-selection
pass to the peephole optimizer (via optimizeCompare).  It turns out that even
this is a bit too early.  Fused compare-and-branch instructions don't
interact well with predication, where a CC result is needed.  They also
make it harder to reuse the CC side-effects of earlier instructions
(not yet implemented, but the subject of a later patch).

Another problem was that the AnalyzeBranch family of routines weren't
handling compares and branches, so we weren't able to reverse the fused
form in cases where we would reverse a separate branch.  This could have
been fixed by extending AnalyzeBranch, but given the other problems,
I've instead moved the fusing to the long-branch pass, which is also
responsible for the opposite transformation: splitting out-of-range
compares and branches into separate compares and long branches.

I've added a test for the AnalyzeBranch problem.  A test for the
predication problem is included in the next patch, which fixes a bug
in the choice of CC mask.

llvm-svn: 187494

10 years agoFixed assertion in Extract128BitVector()
Elena Demikhovsky [Wed, 31 Jul 2013 12:03:08 +0000 (12:03 +0000)]
Fixed assertion in Extract128BitVector()

llvm-svn: 187493

10 years ago[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress()
Richard Sandiford [Wed, 31 Jul 2013 11:36:35 +0000 (11:36 +0000)]
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress()

r186399 aggressively used the RISBG instruction for immediate ANDs,
both because it can handle some values that AND IMMEDIATE can't,
and because it allows the destination register to be different from
the source.  I realized later while implementing the distinct-ops
support that it would be better to leave the choice up to
convertToThreeAddress() instead.  The AND IMMEDIATE form is shorter
and is less likely to be cracked.

This is a problem for 32-bit ANDs because we assume that all 32-bit
operations will leave the high word untouched, whereas RISBG used in
this way will either clear the high word or copy it from the source
register.  The patch uses the z196 instruction RISBLG for this instead.

This means that z10 will be restricted to NILL, NILH and NILF for
32-bit ANDs, but I think that should be OK for now.  Although we're
using z10 as the base architecture, the optimization work is going
to be focused more on z196 and zEC12.

llvm-svn: 187492

10 years agoAdded INSERT and EXTRACT intructions from AVX-512 ISA.
Elena Demikhovsky [Wed, 31 Jul 2013 11:35:14 +0000 (11:35 +0000)]
Added INSERT and EXTRACT intructions from AVX-512 ISA.
All insertf*/extractf* functions replaced with insert/extract since we have insertf and inserti forms.
Added lowering for INSERT_VECTOR_ELT / EXTRACT_VECTOR_ELT for 512-bit vectors.
Added lowering for EXTRACT/INSERT subvector for 512-bit vectors.
Added a test.

llvm-svn: 187491

10 years ago[SystemZ] Add RISBLG and RISBHG instruction definitions
Richard Sandiford [Wed, 31 Jul 2013 11:17:35 +0000 (11:17 +0000)]
[SystemZ] Add RISBLG and RISBHG instruction definitions

The next patch will make use of RISBLG for codegen.

llvm-svn: 187490

10 years agoWith help from chapuni on IRC, I think this is right now. =] The issue
Chandler Carruth [Wed, 31 Jul 2013 11:14:13 +0000 (11:14 +0000)]
With help from chapuni on IRC, I think this is right now. =] The issue
is that the command is quoted differently from the arguments. The
command has '\' and the argument has '\\'. This is made unclear because
FileCheck escapes the single matched '\' when it prints the contents of
the variable, thus fooling me into thinking it had matched '\\' as
intended. The solution is to bind the gcc_install variable in the
argument list rather than out of the command. To do so we also have to
be a bit more careful so that we don't get stray other things into the
'.*' regex.

Also, because of the argument difference, '\\\\' is the correct
formulation before crtbegin, go back to that.

llvm-svn: 187489

10 years agoRun an experiment to try to understand the windows failure better by
Chandler Carruth [Wed, 31 Jul 2013 10:54:11 +0000 (10:54 +0000)]
Run an experiment to try to understand the windows failure better by
replacing one variable with the regex. This won't fix anything, but will
hopefully shed light on the nature of the failure.

llvm-svn: 187488

10 years agoTry only two slashes as that seems to be working elsewhere in this
Chandler Carruth [Wed, 31 Jul 2013 10:34:48 +0000 (10:34 +0000)]
Try only two slashes as that seems to be working elsewhere in this
testcase.

llvm-svn: 187487

10 years agoRe-enable the cross-linux test on windows after making it tolerant of
Chandler Carruth [Wed, 31 Jul 2013 10:02:58 +0000 (10:02 +0000)]
Re-enable the cross-linux test on windows after making it tolerant of
the path separator used when locating crtbegin.o.

I'll watch the bots to see if there are other issues lurking here.

llvm-svn: 187486

10 years ago[msan] Fix msan tests with pipefail option.
Evgeniy Stepanov [Wed, 31 Jul 2013 09:14:55 +0000 (09:14 +0000)]
[msan] Fix msan tests with pipefail option.

llvm-svn: 187485

10 years ago[PECOFF] Add a comment on the idata section fragments.
Rui Ueyama [Wed, 31 Jul 2013 05:10:20 +0000 (05:10 +0000)]
[PECOFF] Add a comment on the idata section fragments.

llvm-svn: 187484

10 years agoclang/test/Driver/cross-linux.c is suppressed on win32 for now. Investigating.
NAKAMURA Takumi [Wed, 31 Jul 2013 04:56:19 +0000 (04:56 +0000)]
clang/test/Driver/cross-linux.c is suppressed on win32 for now. Investigating.

llvm-svn: 187483

10 years agoAdd parentheses to silence gcc warning.
Richard Trieu [Wed, 31 Jul 2013 04:07:28 +0000 (04:07 +0000)]
Add parentheses to silence gcc warning.

llvm-svn: 187482

10 years agoFixed path differences when using include/exclude headers
Ariel J. Bernal [Wed, 31 Jul 2013 04:00:28 +0000 (04:00 +0000)]
Fixed path differences when using include/exclude headers

Added function for removing relative operators from input paths.

llvm-svn: 187481

10 years agoReverts r187449 (report_fatal_error) in favor of a log message since
Ashok Thirumurthi [Wed, 31 Jul 2013 03:56:45 +0000 (03:56 +0000)]
Reverts r187449 (report_fatal_error) in favor of a log message since
the extra check introduces 22 new test failures with the LLDB clang buildbot.

Note that the unhandled DWARF_OP codes in DWARFExpression::Evaluate don't cause test failures if the check is ignored.

llvm-svn: 187480

10 years agoFix register pressure tables on ARM.
Andrew Trick [Wed, 31 Jul 2013 03:24:31 +0000 (03:24 +0000)]
Fix register pressure tables on ARM.

The heuristic that merges register pressure sets was bogus for ARM's S/D regs.

llvm-svn: 187479

10 years agoAdd tracing to the tblgen register pressure table generator.
Andrew Trick [Wed, 31 Jul 2013 03:24:28 +0000 (03:24 +0000)]
Add tracing to the tblgen register pressure table generator.

llvm-svn: 187478

10 years agoIncrement arg_count inside the loop in printInline. Patch by Joe Matarazzo.
Craig Topper [Wed, 31 Jul 2013 03:22:07 +0000 (03:22 +0000)]
Increment arg_count inside the loop in printInline. Patch by Joe Matarazzo.

llvm-svn: 187477

10 years agoChanged register names (and pointer keywords) to be lower case when using Intel X86...
Craig Topper [Wed, 31 Jul 2013 02:47:52 +0000 (02:47 +0000)]
Changed register names (and pointer keywords) to be lower case when using Intel X86 assembler syntax.

Patch by Richard Mitton.

llvm-svn: 187476

10 years agoFix a severe compile time problem when forming large SCEV expressions.
Andrew Trick [Wed, 31 Jul 2013 02:43:40 +0000 (02:43 +0000)]
Fix a severe compile time problem when forming large SCEV expressions.

This fix is very lightweight. The same fix already existed for AddRec
but was missing for NAry expressions.

This is obviously an improvement and I'm unsure how to test compile
time problems.

Patch by Xiaoyi Guo!

llvm-svn: 187475

10 years agoFixing an unused variable warning.
Aaron Ballman [Wed, 31 Jul 2013 02:20:22 +0000 (02:20 +0000)]
Fixing an unused variable warning.

llvm-svn: 187474

10 years agoThe DisassemblerLLVMC has a retain cycle - the InstructionLLVMC's contained in its...
Jim Ingham [Wed, 31 Jul 2013 02:19:15 +0000 (02:19 +0000)]
The DisassemblerLLVMC has a retain cycle - the InstructionLLVMC's contained in its instruction
list have a shared pointer back to their DisassemblerLLVMC.  This checkin force clears the InstructionList
in all the places we use the DisassemblerSP to stop the leaking for now.  I'll go back and fix this
for real when I have time to do so.

<rdar://problem/14581918>

llvm-svn: 187473

10 years agoRemove trailing whitespace and some tab characters.
Craig Topper [Wed, 31 Jul 2013 02:00:15 +0000 (02:00 +0000)]
Remove trailing whitespace and some tab characters.

llvm-svn: 187472

10 years agoFixed incorrect disassembly for MOV16o16a when using Intel syntax.
Craig Topper [Wed, 31 Jul 2013 01:50:26 +0000 (01:50 +0000)]
Fixed incorrect disassembly for MOV16o16a when using Intel syntax.

Patch by Richard Mitton.

llvm-svn: 187471

10 years agoFix crashing on invalid inline asm with matching constraints.
Eric Christopher [Wed, 31 Jul 2013 01:26:24 +0000 (01:26 +0000)]
Fix crashing on invalid inline asm with matching constraints.

For a testcase like the following:

 typedef unsigned long uint64_t;

 typedef struct {
   uint64_t lo;
   uint64_t hi;
 } blob128_t;

 void add_128_to_128(const blob128_t *in, blob128_t *res) {
   asm ("PAND %1, %0" : "+Q"(*res) : "Q"(*in));
 }

where we'll fail to allocate the register for the output constraint,
our matching input constraint will not find a register to match,
and could try to search past the end of the current operands array.

On the idea that we'd like to attempt to keep compilation going
to find more errors in the module, change the error cases when
we're visiting inline asm IR to return immediately and avoid
trying to create a node in the DAG. This leaves us with only
a single error message per inline asm instruction, but allows us
to safely keep going in the general case.

llvm-svn: 187470

10 years ago[mips] Rename instruction DANDi to ANDi64.
Akira Hatanaka [Wed, 31 Jul 2013 00:57:41 +0000 (00:57 +0000)]
[mips] Rename instruction DANDi to ANDi64.

No functionality change.

llvm-svn: 187469

10 years ago[mips] Define instruction itineraries IIArith and IILogic.
Akira Hatanaka [Wed, 31 Jul 2013 00:55:34 +0000 (00:55 +0000)]
[mips] Define instruction itineraries IIArith and IILogic.

No functionality change.

llvm-svn: 187468

10 years agoFix a crasher than manifests when typo correction suggests a function template.
Richard Trieu [Wed, 31 Jul 2013 00:48:10 +0000 (00:48 +0000)]
Fix a crasher than manifests when typo correction suggests a function template.

llvm-svn: 187467

10 years agoFind a better compromise with the default library search paths used by
Chandler Carruth [Wed, 31 Jul 2013 00:37:07 +0000 (00:37 +0000)]
Find a better compromise with the default library search paths used by
Clang when linking and using a GCC installation from a GCC
cross-compiler.

This was desired already by two special case platforms (Android and
Mips), and turns out to be generally (if frustratingly) true. I've added
a substantial comment to the code clarifying the underlying assumptions
of doing actual cross compiles with Clang (or GCC for that matter!) and
help avoid further confusion here.

The end result is to realize that fully general form of PR12478 cannot
be resolved while we support existing cross-compiling GCC toolchains,
and linking with them (namely, linking against their libgcc and
libstdc++ installs). GCC installs these target libraries under
a target-specific prefix but one that may not be available within the
actual sysroot in use. When linking in this world, GCC works and Clang
should as well, but caveat emptor: DSOs from this tree must be
replicated and rpath-fixed to be found at runtime within the sysroot.

I've extended the cross compile test cases to cover these issues by
pointing them at a sysroot and actually checking the library search
paths.

llvm-svn: 187466

10 years agoFix read of uninitialized enum value in test, caught by UBSan. No functionality
Richard Smith [Wed, 31 Jul 2013 00:26:46 +0000 (00:26 +0000)]
Fix read of uninitialized enum value in test, caught by UBSan. No functionality
change, other than removal of undefined behavior.

llvm-svn: 187465

10 years agoFix ptr vector inconsistency in CreatePointerCast
Matt Arsenault [Wed, 31 Jul 2013 00:17:33 +0000 (00:17 +0000)]
Fix ptr vector inconsistency in CreatePointerCast

One form would accept a vector of pointers, and the other did not.
Make both accept vectors of pointers, and add an assertion
for the number of elements.

llvm-svn: 187464

10 years agoFix windows' implementation of status when a file doesn't exist.
Rafael Espindola [Wed, 31 Jul 2013 00:10:25 +0000 (00:10 +0000)]
Fix windows' implementation of status when a file doesn't exist.

The unix one was returning no_such_file_or_directory, but the windows one
was return success.

Update the one one caller that was depending on the old behavior.

llvm-svn: 187463

10 years agoPreserve fast-math flags when folding (fsub x, (fneg y)) to (fadd x, y).
Owen Anderson [Tue, 30 Jul 2013 23:53:17 +0000 (23:53 +0000)]
Preserve fast-math flags when folding (fsub x, (fneg y)) to (fadd x, y).

llvm-svn: 187462

10 years agoFix comment.
Rui Ueyama [Tue, 30 Jul 2013 23:20:10 +0000 (23:20 +0000)]
Fix comment.

llvm-svn: 187461

10 years ago[PECOFF] Emit BSS section.
Rui Ueyama [Tue, 30 Jul 2013 22:56:46 +0000 (22:56 +0000)]
[PECOFF] Emit BSS section.

llvm-svn: 187460

10 years agoReflow this to be easier to read.
Eric Christopher [Tue, 30 Jul 2013 22:50:44 +0000 (22:50 +0000)]
Reflow this to be easier to read.

llvm-svn: 187459

10 years agoclang-format: Improve detection of templates.
Daniel Jasper [Tue, 30 Jul 2013 22:37:19 +0000 (22:37 +0000)]
clang-format: Improve detection of templates.

Before:
  template <typename... Types>
        typename enable_if < 0<sizeof...(Types)>::type Foo() {}
After:
  template <typename... Types>
  typename enable_if<0 < sizeof...(Types)>::type Foo() {}

llvm-svn: 187458

10 years agoMake these just inline, not static inline.
Eric Christopher [Tue, 30 Jul 2013 22:35:06 +0000 (22:35 +0000)]
Make these just inline, not static inline.

llvm-svn: 187457