platform/upstream/llvm.git
12 years agoAdd support for the 'd' mips inline asm output modifier.
Eric Christopher [Sat, 19 May 2012 00:51:56 +0000 (00:51 +0000)]
Add support for the 'd' mips inline asm output modifier.

Patch by Jack Carter.

llvm-svn: 157093

12 years agoSCEV: Add MarkPendingLoopPredicates to avoid recursive isImpliedCond.
Andrew Trick [Sat, 19 May 2012 00:48:25 +0000 (00:48 +0000)]
SCEV: Add MarkPendingLoopPredicates to avoid recursive isImpliedCond.

getUDivExpr attempts to simplify by checking for overflow.
isLoopEntryGuardedByCond then evaluates the loop predicate which
may lead to the same getUDivExpr causing endless recursion.

Fixes PR12868: clang 3.2 segmentation fault.

llvm-svn: 157092

12 years agobump to lldb-150.
Jason Molenda [Sat, 19 May 2012 00:32:19 +0000 (00:32 +0000)]
bump to lldb-150.

llvm-svn: 157090

12 years ago[analyzer] c++11: do not crash on namespace alias
Anna Zaks [Sat, 19 May 2012 00:22:11 +0000 (00:22 +0000)]
[analyzer] c++11: do not crash on namespace alias

llvm-svn: 157089

12 years ago[analyzer] For locations, use isGLValue() instead of isLValue().
Anna Zaks [Sat, 19 May 2012 00:22:07 +0000 (00:22 +0000)]
[analyzer] For locations, use isGLValue() instead of isLValue().

llvm-svn: 157088

12 years agoForgot to bump the local string buffer up in size after debugging to make sure long...
Greg Clayton [Sat, 19 May 2012 00:18:00 +0000 (00:18 +0000)]
Forgot to bump the local string buffer up in size after debugging to make sure long strings would be correctly read when the buffer is too small for the string.

llvm-svn: 157087

12 years agoSuggest adding 'typename' when it would make the compiler
Kaelyn Uhrain [Fri, 18 May 2012 23:42:49 +0000 (23:42 +0000)]
Suggest adding 'typename' when it would make the compiler
accept the template argument expression as a type.

llvm-svn: 157085

12 years agoAdd test cases where we start three worker threads, with a write watchpoint set....
Johnny Chen [Fri, 18 May 2012 23:30:57 +0000 (23:30 +0000)]
Add test cases where we start three worker threads, with a write watchpoint set.  As soon as a watchpoint is hit in either of the worker thread,
we delete the watchpoint.  The test succeeds when no more watchpoint hit event fires after the deletion of the watchpoint.

related to rdar://problem/11320188

llvm-svn: 157084

12 years agoFound a quick way to improve the speed with which we can read object files from memor...
Greg Clayton [Fri, 18 May 2012 23:20:01 +0000 (23:20 +0000)]
Found a quick way to improve the speed with which we can read object files from memory when they are in the shared cache: always read the symbol table strings from memory and let the process' memory cache do the work.

llvm-svn: 157083

12 years ago[analyzer] Fix a c++11 crash: xvalues can be locations (VisitMemberExpr)
Anna Zaks [Fri, 18 May 2012 22:47:43 +0000 (22:47 +0000)]
[analyzer] Fix a c++11 crash: xvalues can be locations (VisitMemberExpr)

llvm-svn: 157082

12 years ago[analyzer] Malloc checker: remove unnecessary comparisons.
Anna Zaks [Fri, 18 May 2012 22:47:40 +0000 (22:47 +0000)]
[analyzer] Malloc checker: remove unnecessary comparisons.

llvm-svn: 157081

12 years agoFix replacing all the users of objc weak runtime routines
Dan Gohman [Fri, 18 May 2012 22:17:29 +0000 (22:17 +0000)]
Fix replacing all the users of objc weak runtime routines
when deleting them. rdar://11434915.

llvm-svn: 157080

12 years agoModernize naming convention for class members.
Jakob Stoklund Olesen [Fri, 18 May 2012 22:10:15 +0000 (22:10 +0000)]
Modernize naming convention for class members.

No functional change.

llvm-svn: 157079

12 years agoWarn when we detect a valid dSYM file that is empty. This can happen when a dSYM...
Greg Clayton [Fri, 18 May 2012 21:47:20 +0000 (21:47 +0000)]
Warn when we detect a valid dSYM file that is empty. This can happen when a dSYM file is created from a binary with no debug info, that has been stripped, or when the .o files are not available when the dSYM is created.

llvm-svn: 157078

12 years agoCleanup a custom thread-format so we don't fail TestLoadUnload when running the
Filipe Cabecinhas [Fri, 18 May 2012 21:35:43 +0000 (21:35 +0000)]
Cleanup a custom thread-format so we don't fail TestLoadUnload when running the
test suite for two architectures (the full path to d.c would appear).

llvm-svn: 157077

12 years agoobjc: use "class extension" instead of "continuation class"
Fariborz Jahanian [Fri, 18 May 2012 21:22:49 +0000 (21:22 +0000)]
objc: use "class extension" instead of "continuation class"
to match documentation. // rdar://11309706

llvm-svn: 157074

12 years agoMove all work list processing to copyCoalesceWorkList().
Jakob Stoklund Olesen [Fri, 18 May 2012 21:09:40 +0000 (21:09 +0000)]
Move all work list processing to copyCoalesceWorkList().

This will make it possible to filter out erased instructions later.

llvm-svn: 157073

12 years agoallow LazyValueInfo::getEdgeValue() to reason about multiple edges from the same...
Nuno Lopes [Fri, 18 May 2012 21:02:10 +0000 (21:02 +0000)]
allow LazyValueInfo::getEdgeValue() to reason about multiple edges from the same switch instruction by doing union of ranges (which may still be conservative, but it's more aggressive than before)

llvm-svn: 157071

12 years agoOptimizations to the Python ObjC formatters - Benchmarking infrastructure
Enrico Granata [Fri, 18 May 2012 19:55:37 +0000 (19:55 +0000)]
Optimizations to the Python ObjC formatters - Benchmarking infrastructure

llvm-svn: 157066

12 years agoLexer::ReadToEndOfLine: Only build the string if it's actually used and do so in...
Benjamin Kramer [Fri, 18 May 2012 19:32:16 +0000 (19:32 +0000)]
Lexer::ReadToEndOfLine: Only build the string if it's actually used and do so in a less malloc-intensive way.

llvm-svn: 157064

12 years agoRefactor data-in-code annotations.
Jim Grosbach [Fri, 18 May 2012 19:12:01 +0000 (19:12 +0000)]
Refactor data-in-code annotations.

Use a dedicated MachO load command to annotate data-in-code regions.
This is the same format the linker produces for final executable images,
allowing consistency of representation and use of introspection tools
for both object and executable files.

Data-in-code regions are annotated via ".data_region"/".end_data_region"
directive pairs, with an optional region type.

data_region_directive := ".data_region" { region_type }
region_type := "jt8" | "jt16" | "jt32" | "jta32"
end_data_region_directive := ".end_data_region"

The previous handling of ARM-style "$d.*" labels was broken and has
been removed. Specifically, it didn't handle ARM vs. Thumb mode when
marking the end of the section.

rdar://11459456

llvm-svn: 157062

12 years agofix warnings when compiling with -Wshadow
Nick Kledzik [Fri, 18 May 2012 18:39:06 +0000 (18:39 +0000)]
fix warnings when compiling with -Wshadow

llvm-svn: 157061

12 years agoRemove duplicate code that we could just fallthrough to.
Eric Christopher [Fri, 18 May 2012 18:24:15 +0000 (18:24 +0000)]
Remove duplicate code that we could just fallthrough to.

llvm-svn: 157060

12 years agoSimplify RegisterCoalescer::copyCoalesceInMBB().
Jakob Stoklund Olesen [Fri, 18 May 2012 18:21:48 +0000 (18:21 +0000)]
Simplify RegisterCoalescer::copyCoalesceInMBB().

It is no longer necessary to separate VirtCopies, PhysCopies, and
ImpDefCopies. Implicitly defined copies are extremely rare after we
added the ProcessImplicitDefs pass, and physical register copies are not
joined any longer.

llvm-svn: 157059

12 years agoadd test case for bugfix in r157032
Nuno Lopes [Fri, 18 May 2012 17:44:58 +0000 (17:44 +0000)]
add test case for bugfix in r157032

llvm-svn: 157058

12 years agoAdd support for the mips 'x' inline asm modifier.
Eric Christopher [Fri, 18 May 2012 17:39:35 +0000 (17:39 +0000)]
Add support for the mips 'x' inline asm modifier.

Patch by Jack Carter.

llvm-svn: 157057

12 years agoRemove support for PhysReg joining.
Jakob Stoklund Olesen [Fri, 18 May 2012 17:18:58 +0000 (17:18 +0000)]
Remove support for PhysReg joining.

This has been disabled for a while, and it is not a feature we want to
support. Copies between physical and virtual registers are eliminated by
good hinting support in the register allocator. Joining virtual and
physical registers is really a form of register allocation, and the
coalescer is not properly equipped to do that. In particular, it cannot
backtrack coalescing decisions, and sometimes that would cause it to
create programs that were impossible to register allocate, by exhausting
a small register class.

It was also very difficult to keep track of the live ranges of aliasing
registers when extending the live range of a physreg. By disabling
physreg joining, we can let fixed physreg live ranges remain constant
throughout the register allocator super-pass.

One type of physreg joining remains: A virtual register that has a
single value which is a copy of a reserved register can be merged into
the reserved physreg. This always lowers register pressure, and since we
don't compute live ranges for reserved registers, there are no problems
with aliases.

llvm-svn: 157055

12 years agoTypo.
Chad Rosier [Fri, 18 May 2012 16:56:52 +0000 (16:56 +0000)]
Typo.

llvm-svn: 157054

12 years agoFileCheck-ify, apropos of nothing
Joel Jones [Fri, 18 May 2012 16:24:01 +0000 (16:24 +0000)]
FileCheck-ify, apropos of nothing

llvm-svn: 157051

12 years agoFixed the test for the new process launch abbreviation.
Filipe Cabecinhas [Fri, 18 May 2012 13:21:05 +0000 (13:21 +0000)]
Fixed the test for the new process launch abbreviation.

llvm-svn: 157050

12 years agoProtect __shared_weak_count::__get_deleter declaration with _LIBCPP_NO_RTTI. Fixes...
Howard Hinnant [Fri, 18 May 2012 13:06:21 +0000 (13:06 +0000)]
Protect __shared_weak_count::__get_deleter declaration with _LIBCPP_NO_RTTI.  Fixes llvm.org/bugs/show_bug.cgi?id=12867

llvm-svn: 157049

12 years agotsan: check for overflow in malloc()
Dmitry Vyukov [Fri, 18 May 2012 09:41:52 +0000 (09:41 +0000)]
tsan: check for overflow in malloc()

llvm-svn: 157048

12 years agotsan: fix potential NULL deref
Dmitry Vyukov [Fri, 18 May 2012 08:53:16 +0000 (08:53 +0000)]
tsan: fix potential NULL deref

llvm-svn: 157047

12 years agoRecommited reworked r156804:
Stepan Dyatkovskiy [Fri, 18 May 2012 08:32:28 +0000 (08:32 +0000)]
Recommited reworked r156804:
SelectionDAGBuilder::Clusterify : main functinality was replaced with CRSBuilder::optimize, so big part of Clusterify's code was reduced.

llvm-svn: 157046

12 years agoSimplify code a bit. No functional change intended.
Craig Topper [Fri, 18 May 2012 07:07:36 +0000 (07:07 +0000)]
Simplify code a bit. No functional change intended.

llvm-svn: 157044

12 years agoSimplify handling of v16i8 shuffles and fix a missed optimization.
Craig Topper [Fri, 18 May 2012 06:42:06 +0000 (06:42 +0000)]
Simplify handling of v16i8 shuffles and fix a missed optimization.

llvm-svn: 157043

12 years ago<rdar://problem/11386214>
Greg Clayton [Fri, 18 May 2012 02:38:05 +0000 (02:38 +0000)]
<rdar://problem/11386214>
<rdar://problem/11455913>

"target symbol add" should flush the cached frames
"register write" should flush the thread state in case registers modifications change stack

llvm-svn: 157042

12 years agobump to lldb 149.
Jason Molenda [Fri, 18 May 2012 02:03:26 +0000 (02:03 +0000)]
bump to lldb 149.

llvm-svn: 157040

12 years agoCentralize the handling of the "attribute declaration must precede definition"
Rafael Espindola [Fri, 18 May 2012 01:47:00 +0000 (01:47 +0000)]
Centralize the handling of the "attribute declaration must precede definition"
warning. This also makes us warn on tags, which, ironically, is the only case
gcc warns on.

llvm-svn: 157039

12 years agoTeach two-address pass to update the "source" map so it doesn't perform a
Evan Cheng [Fri, 18 May 2012 01:33:51 +0000 (01:33 +0000)]
Teach two-address pass to update the "source" map so it doesn't perform a
non-profitable commute using outdated info. The test case would still fail
because of poor pre-RA schedule. That will be fixed by MI scheduler.

rdar://11472010

llvm-svn: 157038

12 years ago[analyzer]Malloc: refactor and report use after free by memory
Anna Zaks [Fri, 18 May 2012 01:16:10 +0000 (01:16 +0000)]
[analyzer]Malloc: refactor and report use after free by memory
allocating functions.

llvm-svn: 157037

12 years agordar://problem/11140741
Johnny Chen [Fri, 18 May 2012 00:51:36 +0000 (00:51 +0000)]
rdar://problem/11140741

For "process attach", make the success criterion as the inferior changes its state to eStateStopped.
Otherwise, mark it as a failure and say so.

llvm-svn: 157036

12 years agoTemporarily disabled the MCJIT tests for Darwin, because the RuntimeDyldMachO has...
Danil Malyshev [Fri, 18 May 2012 00:30:58 +0000 (00:30 +0000)]
Temporarily disabled the MCJIT tests for Darwin, because the RuntimeDyldMachO has a problems with relocations for 32bit x86.

llvm-svn: 157035

12 years agoAnother test for r157025 <rdar://problem/11460990>.
Fariborz Jahanian [Fri, 18 May 2012 00:19:25 +0000 (00:19 +0000)]
Another test for r157025  <rdar://problem/11460990>.

llvm-svn: 157034

12 years agoClarify comment.
Eric Christopher [Fri, 18 May 2012 00:16:22 +0000 (00:16 +0000)]
Clarify comment.

llvm-svn: 157033

12 years agofix corner case in ConstantRange::intersectWith().
Nuno Lopes [Fri, 18 May 2012 00:14:36 +0000 (00:14 +0000)]
fix corner case in ConstantRange::intersectWith().
this fixes the missed optimization I was seeing in the CorrelatedValuePropagation pass

llvm-svn: 157032

12 years agoFixed a bug in llvm-objdump when disassembling using -macho option for a binary
Kevin Enderby [Fri, 18 May 2012 00:13:56 +0000 (00:13 +0000)]
Fixed a bug in llvm-objdump when disassembling using -macho option for a binary
containing no symbols.  Fixed the crash and fixed it not disassembling anything.

llvm-svn: 157031

12 years agoRemove a test that was only testing for physreg joining.
Jakob Stoklund Olesen [Fri, 18 May 2012 00:07:14 +0000 (00:07 +0000)]
Remove a test that was only testing for physreg joining.

This is the same as the other tests: Clever tricks are required to make
the arguments and return value line up in a single-instruction function.
It rarely happens in real life.

We have plenty other examples of this behavior.

llvm-svn: 157030

12 years agoMake the debug output that comes as printf's from code called in the target for getti...
Jim Ingham [Fri, 18 May 2012 00:05:52 +0000 (00:05 +0000)]
Make the debug output that comes as printf's from code called in the target for getting ObjC class names and ObjC method implementations only come out when doing verbose logging.

llvm-svn: 157029

12 years ago<rdar://problem/11405850>
Greg Clayton [Fri, 18 May 2012 00:04:38 +0000 (00:04 +0000)]
<rdar://problem/11405850>

The "run" and "r" aliases were for gdb compatability, so make then do what GDB does by default: launch in a shell.

For those that don't want launching with a shell by default, add the following to your ~/.lldbinit file:

command unalias run
command unalias r
command alias r process launch --
command alias run process launch --

llvm-svn: 157028

12 years agoRemove -join-physregs from the test suite.
Jakob Stoklund Olesen [Thu, 17 May 2012 23:44:19 +0000 (23:44 +0000)]
Remove -join-physregs from the test suite.

This option has been disabled for a while, and it is going away so I can
clean up the coalescer code.

The tests that required physreg joining to be enabled were almost all of
the form "tiny function with interference between arguments and return
value". Such functions are usually inlined in the real world.

The problem exposed by phys_subreg_coalesce-3.ll is real, but fairly
rare.

llvm-svn: 157027

12 years agoI have updated Clang to include support for Objective-C
Sean Callanan [Thu, 17 May 2012 23:29:56 +0000 (23:29 +0000)]
I have updated Clang to include support for Objective-C
boxed expressions returning numbers and strings.

I also added boxed expressions to our testcases, and
enabled boxed expressions when libarclite is linked into
the inferior.

llvm-svn: 157026

12 years agoA selector match between two Objective-C methods does *not* guarantee
Douglas Gregor [Thu, 17 May 2012 23:13:29 +0000 (23:13 +0000)]
A selector match between two Objective-C methods does *not* guarantee
that the methods have the same number of parameters, although we
certainly assumed this in many places. Objective-C can be insane
sometimes. Fixes <rdar://problem/11460990>.

llvm-svn: 157025

12 years agominor simplification in the call to ConstantRange constructor
Nuno Lopes [Thu, 17 May 2012 23:04:08 +0000 (23:04 +0000)]
minor simplification in the call to ConstantRange constructor

llvm-svn: 157024

12 years agoRemove trailing periods which snuck into a few diagnostic strings.
Matt Beaumont-Gay [Thu, 17 May 2012 22:46:56 +0000 (22:46 +0000)]
Remove trailing periods which snuck into a few diagnostic strings.

llvm-svn: 157022

12 years agoIn the override search for Objective-C methods, protect against ASTs that have NULL...
Douglas Gregor [Thu, 17 May 2012 22:39:14 +0000 (22:39 +0000)]
In the override search for Objective-C methods, protect against ASTs that have NULL interfaces behind a category, which can happen in invalid code. Fixes <rdar://problem/11478173>, a recent regression

llvm-svn: 157021

12 years agocomments
Andrew Trick [Thu, 17 May 2012 22:37:09 +0000 (22:37 +0000)]
comments

llvm-svn: 157020

12 years agoFix the encoding of the armv7m (MClass) for MSR APSR writes which was missing
Kevin Enderby [Thu, 17 May 2012 22:18:01 +0000 (22:18 +0000)]
Fix the encoding of the armv7m (MClass) for MSR APSR writes which was missing
the 0b10 mask encoding bits.  Make MSR APSR writes without a _<bits> qualifier
an alias for MSR APSR_nzcvq even though ARM as deprecated it use.  Also add
support for suffixes (_nzcvq, _g, _nzcvqg) for APSR versions.  Some FIXMEs in
the code for better error checking when versions shouldn't be used.
rdar://11457025

llvm-svn: 157019

12 years agoDrop the question mark when complaining about a non-existing -Wno- flag.
Benjamin Kramer [Thu, 17 May 2012 21:14:26 +0000 (21:14 +0000)]
Drop the question mark when complaining about a non-existing -Wno- flag.

We're certain that it doesn't exist.

llvm-svn: 157017

12 years ago- Added ExecutionEngine/MCJIT tests
Danil Malyshev [Thu, 17 May 2012 21:07:47 +0000 (21:07 +0000)]
- Added ExecutionEngine/MCJIT tests
- Added HOST_ARCH to Makefile.config.in
The HOST_ARCH will be used by MCJIT tests filter, because MCJIT supported only x86 and ARM architectures now.

llvm-svn: 157015

12 years agoAdd -mno-implicit-float to the M group.
Chad Rosier [Thu, 17 May 2012 20:53:37 +0000 (20:53 +0000)]
Add -mno-implicit-float to the M group.

llvm-svn: 157014

12 years ago<rdar://11477301>
Han Ming Ong [Thu, 17 May 2012 20:41:48 +0000 (20:41 +0000)]
<rdar://11477301>

Restore Xcode as a valid white list client using the XPC root launcher

llvm-svn: 157012

12 years agoRemove extraneous ';'.
Bill Wendling [Thu, 17 May 2012 20:27:58 +0000 (20:27 +0000)]
Remove extraneous ';'.

llvm-svn: 157011

12 years agoReduce the timeout value for the "get class name" and "po" functions to .1 second...
Jim Ingham [Thu, 17 May 2012 18:51:37 +0000 (18:51 +0000)]
Reduce the timeout value for the "get class name" and "po" functions to .1 second.  1 second (what they were before) is way too long.

llvm-svn: 157009

12 years agoIf we notice that a module with a given file path is replaced by another with the...
Jim Ingham [Thu, 17 May 2012 18:38:42 +0000 (18:38 +0000)]
If we notice that a module with a given file path is replaced by another with the same file
path on rerunning, evict the old module from the target module list, inform the breakpoints
about this so they can do something intelligent as well.

rdar://problem/11273043

llvm-svn: 157008

12 years agomisched: trace ReadyQ.
Andrew Trick [Thu, 17 May 2012 18:35:13 +0000 (18:35 +0000)]
misched: trace ReadyQ.

llvm-svn: 157007

12 years agomisched: Added 3-level regpressure back-off.
Andrew Trick [Thu, 17 May 2012 18:35:10 +0000 (18:35 +0000)]
misched: Added 3-level regpressure back-off.

Introduce the basic strategy for register pressure scheduling.

1) Respect target limits at all times.

2) Indentify critical register classes (pressure sets).
   Track pressure within the scheduled region.
   Avoid increasing scheduled pressure for critical registers.

3) Avoid exceeding the max pressure of the region prior to scheduling.

Added logic for picking between the top and bottom ready Q's based on
regpressure heuristics.

Status: functional but needs to be asjusted to achieve good results.
llvm-svn: 157006

12 years agocomment
Andrew Trick [Thu, 17 May 2012 18:35:07 +0000 (18:35 +0000)]
comment

llvm-svn: 157005

12 years agoregpressure: Fix getMaxUpwardPressureDelta.
Andrew Trick [Thu, 17 May 2012 18:35:05 +0000 (18:35 +0000)]
regpressure: Fix getMaxUpwardPressureDelta.

llvm-svn: 157004

12 years agomisched: fix liveness iterators
Andrew Trick [Thu, 17 May 2012 18:35:03 +0000 (18:35 +0000)]
misched: fix liveness iterators

llvm-svn: 157003

12 years agowhitespace
Andrew Trick [Thu, 17 May 2012 18:35:00 +0000 (18:35 +0000)]
whitespace

llvm-svn: 157002

12 years agoNever clear <undef> flags on already joined copies.
Jakob Stoklund Olesen [Thu, 17 May 2012 18:32:42 +0000 (18:32 +0000)]
Never clear <undef> flags on already joined copies.

RegisterCoalescer set <undef> flags on all operands of copy instructions
that are scheduled to be removed. This is so they won't affect
shrinkToUses() by introducing false register reads.

Make sure those <undef> flags are never cleared, or shrinkToUses() could
cause live intervals to end at instructions about to be deleted.

This would be a lot simpler if RegisterCoalescer could just erase joined
copies immediately instead of keeping all the to-be-deleted instructions
around.

This fixes PR12862. Unfortunately, bugpoint can't create a sane test
case for this. Like many other coalescer problems, this failure depends
of a very fragile series of events.

<rdar://problem/11474428>

llvm-svn: 157001

12 years agoFix a verifier bug.
Jakob Stoklund Olesen [Thu, 17 May 2012 18:32:40 +0000 (18:32 +0000)]
Fix a verifier bug.

Make sure useless (def-only) intervals also get verified.

llvm-svn: 157000

12 years ago[arcmt] Remove the "it is not safe to remove an unused 'autorelease' message" ARC
Argyrios Kyrtzidis [Thu, 17 May 2012 18:16:05 +0000 (18:16 +0000)]
[arcmt] Remove the "it is not safe to remove an unused 'autorelease' message" ARC
migration error.

This is more trouble that it is worth; autoreleasing a value without holding on it
is a valid use-case, we should not "punish" correct code for the minority of
broken/fragile programs that depend on the behavior of  -autorelease.

rdar://9914061

llvm-svn: 156999

12 years agoRelax the requirement that the exception object must be an instruction. During
Bill Wendling [Thu, 17 May 2012 17:59:51 +0000 (17:59 +0000)]
Relax the requirement that the exception object must be an instruction. During
bugpoint-ing, it may turn into something else.

llvm-svn: 156998

12 years ago[Hexagon] Clean up Hexagon ELF definition.
Evandro Menezes [Thu, 17 May 2012 16:46:46 +0000 (16:46 +0000)]
[Hexagon] Clean up Hexagon ELF definition.

llvm-svn: 156996

12 years agoenhance the intrinsic info stuff to emit encodings that don't fit in 32-bits into a
Chris Lattner [Thu, 17 May 2012 15:55:41 +0000 (15:55 +0000)]
enhance the intrinsic info stuff to emit encodings that don't fit in 32-bits into a
separate side table, using the handy SequenceToOffsetTable class.  This encodes all
these weird things into another 256 bytes, allowing all intrinsics to be encoded this way.

llvm-svn: 156995

12 years agoWe shouldn't save g_dummy_target_sp. Other code will simply call Destroy() on it.
Filipe Cabecinhas [Thu, 17 May 2012 15:48:02 +0000 (15:48 +0000)]
We shouldn't save g_dummy_target_sp. Other code will simply call Destroy() on it.

TestBackticksWithoutATarget.BackticksWithNoTargetTestCase was calling
GetDummyTarget() when executing for x86_64. When performing session
tearDown, it would get destroyed (and everything would be invalid (arch,
etc).

Then the test would run for i386. The dummy target wasn't being
reinitialized and was invalid. lldb complained that 'current process state
is unsuitable for expression parsing'.

llvm-svn: 156994

12 years ago[driver] Reword the warning message for missing value to a joined argument.
Chad Rosier [Thu, 17 May 2012 15:45:13 +0000 (15:45 +0000)]
[driver] Reword the warning message for missing value to a joined argument.

llvm-svn: 156993

12 years agotsan: remove shutdown code
Dmitry Vyukov [Thu, 17 May 2012 15:00:27 +0000 (15:00 +0000)]
tsan: remove shutdown code
tsan runtime shutdown is problematic for 2 reasons:
1. others crash during shutdown
2. we have to override user exit status (don't know it and can't return from atexit handler)

llvm-svn: 156991

12 years agotsan: detect accesses to freed memory
Dmitry Vyukov [Thu, 17 May 2012 14:17:51 +0000 (14:17 +0000)]
tsan: detect accesses to freed memory
http://codereview.appspot.com/6214052

llvm-svn: 156990

12 years agoRemove incorrect pattern for ARM SMML instruction.
Tim Northover [Thu, 17 May 2012 13:12:13 +0000 (13:12 +0000)]
Remove incorrect pattern for ARM SMML instruction.

Patch by Meador Inge.

llvm-svn: 156989

12 years agoAdded two missing const qualifiers.
Abramo Bagnara [Thu, 17 May 2012 12:44:05 +0000 (12:44 +0000)]
Added two missing const qualifiers.

llvm-svn: 156988

12 years agoCXXThisScopeRAII objects aren't free, don't compute one if it's unused.
Benjamin Kramer [Thu, 17 May 2012 12:01:52 +0000 (12:01 +0000)]
CXXThisScopeRAII objects aren't free, don't compute one if it's unused.

llvm-svn: 156987

12 years agoFix compile error.
Manuel Klimek [Thu, 17 May 2012 09:32:05 +0000 (09:32 +0000)]
Fix compile error.

llvm-svn: 156986

12 years agoSelectionDAGBuilder: CaseBlock, CaseRanges and CaseCmp changed representation of...
Stepan Dyatkovskiy [Thu, 17 May 2012 08:56:30 +0000 (08:56 +0000)]
SelectionDAGBuilder: CaseBlock, CaseRanges and CaseCmp changed representation of Low and High from signed to unsigned. Since unsigned ints usually simpler, faster and allows to reduce some extra signed bit checks needed before <,>,<=,>= comparisons.

llvm-svn: 156985

12 years ago[tsan] grammar fixes
Kostya Serebryany [Thu, 17 May 2012 08:49:14 +0000 (08:49 +0000)]
[tsan] grammar fixes

llvm-svn: 156984

12 years ago[tsan] fix dependency rules in Makefile.old
Kostya Serebryany [Thu, 17 May 2012 08:33:14 +0000 (08:33 +0000)]
[tsan] fix dependency rules in Makefile.old

llvm-svn: 156983

12 years agotsan: ValgrindSlowdown() should be weak for some time
Dmitry Vyukov [Thu, 17 May 2012 08:31:47 +0000 (08:31 +0000)]
tsan: ValgrindSlowdown() should be weak for some time

llvm-svn: 156982

12 years agotsan: add ValgrindSlowdown() "dynamic annotation"
Dmitry Vyukov [Thu, 17 May 2012 08:04:41 +0000 (08:04 +0000)]
tsan: add ValgrindSlowdown() "dynamic annotation"

llvm-svn: 156981

12 years agoGenericize the intrinsics descriptor decoding a bit to make room
Chris Lattner [Thu, 17 May 2012 05:13:57 +0000 (05:13 +0000)]
Genericize the intrinsics descriptor decoding a bit to make room
for future expansion, no functionality change yet though.

llvm-svn: 156979

12 years agofinish encoding all of the interesting details of intrinsics. Now intrinsics
Chris Lattner [Thu, 17 May 2012 05:03:24 +0000 (05:03 +0000)]
finish encoding all of the interesting details of intrinsics.  Now intrinsics
are only rejected because they can't be encoded into a 32-bit unit, not because
they contain an unencodable feature.

llvm-svn: 156978

12 years agostrengthen the intrinsic descriptor stuff to be able to handle sin, cos and other
Chris Lattner [Thu, 17 May 2012 04:30:58 +0000 (04:30 +0000)]
strengthen the intrinsic descriptor stuff to be able to handle sin, cos and other
intrinsics that use passed-in arguments.

llvm-svn: 156977

12 years agosimplify code generated by tblgen that is not necessary since we dropped
Chris Lattner [Thu, 17 May 2012 04:07:48 +0000 (04:07 +0000)]
simplify code generated by tblgen that is not necessary since we dropped
compatibility with LLVM 2.x bitcode files.

llvm-svn: 156976

12 years agoI forgot the #ifdef _MSC_VER guard in my last commit.
Francois Pichet [Thu, 17 May 2012 04:00:03 +0000 (04:00 +0000)]
I forgot the #ifdef _MSC_VER guard in my last commit.

llvm-svn: 156975

12 years agoMake sure to subtract one from the PC when doing the symbolication of stack frames...
Greg Clayton [Thu, 17 May 2012 03:58:23 +0000 (03:58 +0000)]
Make sure to subtract one from the PC when doing the symbolication of stack frames when it isn't the zero'th frame.

llvm-svn: 156974

12 years agoFix the MSVC 2010 build: disable the optimizer for a problematic function.
Francois Pichet [Thu, 17 May 2012 03:38:19 +0000 (03:38 +0000)]
Fix the MSVC 2010 build: disable the optimizer for a problematic function.

llvm-svn: 156973

12 years agoBump to version 148.
Jason Molenda [Thu, 17 May 2012 01:49:53 +0000 (01:49 +0000)]
Bump to version 148.

llvm-svn: 156971

12 years agoUse the argument location instead of the format string location when warning
Matt Beaumont-Gay [Thu, 17 May 2012 00:03:16 +0000 (00:03 +0000)]
Use the argument location instead of the format string location when warning
about argument type mismatch.

This gives a nicer diagnostic in cases like
  printf(fmt,
         i);
where previously the snippet just pointed at 'fmt' (with a note at the
definition of fmt).

It's a wash for cases like
  printf("%f",
         i);
where previously we snippeted the offending portion of the format string,
but didn't indicate which argument was at fault.

llvm-svn: 156968

12 years ago[libclang/AST] When declaring a local class, don't neglect to set the end location
Argyrios Kyrtzidis [Wed, 16 May 2012 23:49:15 +0000 (23:49 +0000)]
[libclang/AST] When declaring a local class, don't neglect to set the end location
of the DeclStmt node, otherwise libclang will not work for anything inside that
class.

rdar://10837710

llvm-svn: 156966