platform/upstream/llvm.git
11 years agoUpdate instructions for building and using libc++ on Mac OS
Howard Hinnant [Tue, 6 Nov 2012 21:31:37 +0000 (21:31 +0000)]
Update instructions for building and using libc++ on Mac OS

llvm-svn: 167490

11 years ago[libclang] Add some comments about the version constants for the libclang API.
Argyrios Kyrtzidis [Tue, 6 Nov 2012 21:21:49 +0000 (21:21 +0000)]
[libclang] Add some comments about the version constants for the libclang API.

llvm-svn: 167489

11 years agoCostModel: add another known vector trunc optimization.
Nadav Rotem [Tue, 6 Nov 2012 21:17:17 +0000 (21:17 +0000)]
CostModel: add another known vector trunc optimization.

llvm-svn: 167488

11 years agoFix a silly mistake in r167437.
Eli Friedman [Tue, 6 Nov 2012 21:10:22 +0000 (21:10 +0000)]
Fix a silly mistake in r167437.

llvm-svn: 167487

11 years agoProvide a way to disable use of extern templates in libc++. This is intended for...
Howard Hinnant [Tue, 6 Nov 2012 21:08:48 +0000 (21:08 +0000)]
Provide a way to disable use of extern templates in libc++.  This is intended for the clients of libc++, not the libc++ build.  The dylib should always contain the extern templates.  To disable the client needs to put -D'_LIBCPP_EXTERN_TEMPLATE(...)=' on the command line.

llvm-svn: 167486

11 years agoclang/test/Driver/fsanitize.c: Speculative fix to improve test on win32, not to mix...
NAKAMURA Takumi [Tue, 6 Nov 2012 20:44:29 +0000 (20:44 +0000)]
clang/test/Driver/fsanitize.c: Speculative fix to improve test on win32, not to mix emissions into stdout.

llvm-svn: 167485

11 years ago[c-index-test] When building with BUILD_CLANG_ONLY=YES, include c-index-test.
Argyrios Kyrtzidis [Tue, 6 Nov 2012 19:54:46 +0000 (19:54 +0000)]
[c-index-test] When building with BUILD_CLANG_ONLY=YES, include c-index-test.
It is part of libclang and has other uses besides running the clang tests.

llvm-svn: 167484

11 years agoRemove C++11-isms I just introduced in r167482
Douglas Gregor [Tue, 6 Nov 2012 19:41:11 +0000 (19:41 +0000)]
Remove C++11-isms I just introduced in r167482

llvm-svn: 167483

11 years agoIntroduce inferred framework modules into the module map file,
Douglas Gregor [Tue, 6 Nov 2012 19:39:40 +0000 (19:39 +0000)]
Introduce inferred framework modules into the module map file,
allowing a module map to be placed one level above the '.framework'
directories to specify that all .frameworks within that directory can
be inferred as framework modules. One can also specifically exclude
frameworks known not to work.

This makes explicit (and more restricted) behavior modules have had
"forever", where *any* .framework was assumed to be able to be built
as a module. That's not necessarily true, so we white-list directories
(with exclusions) when those directories have been audited.

llvm-svn: 167482

11 years agoTeach Clang parser to reject C++11 attributes that appertain to declaration specifiers.
Michael Han [Tue, 6 Nov 2012 19:34:54 +0000 (19:34 +0000)]
Teach Clang parser to reject C++11 attributes that appertain to declaration specifiers.

We don't support any C++11 attributes that appertain to declaration specifiers so reject
the attributes in parser until we support them; this also conforms to what g++ 4.8 is doing.

llvm-svn: 167481

11 years agoCost Model: add tables for some avx type-conversion hacks.
Nadav Rotem [Tue, 6 Nov 2012 19:33:53 +0000 (19:33 +0000)]
Cost Model: add tables for some avx type-conversion hacks.

llvm-svn: 167480

11 years agoUpdates to user's manual and release notes for -fsanitize= options.
Richard Smith [Tue, 6 Nov 2012 19:23:14 +0000 (19:23 +0000)]
Updates to user's manual and release notes for -fsanitize= options.

llvm-svn: 167479

11 years ago[c-index-test] When building with BUILD_CLANG_ONLY=YES, include c-index-test.
Argyrios Kyrtzidis [Tue, 6 Nov 2012 19:14:25 +0000 (19:14 +0000)]
[c-index-test] When building with BUILD_CLANG_ONLY=YES, include c-index-test.
It is part of libclang and has other uses besides running the clang tests.

llvm-svn: 167478

11 years agoFix build error from previous commit.
Andrew Kaylor [Tue, 6 Nov 2012 19:06:46 +0000 (19:06 +0000)]
Fix build error from previous commit.

llvm-svn: 167477

11 years agoCleanup: 80-column violation
Manman Ren [Tue, 6 Nov 2012 19:05:29 +0000 (19:05 +0000)]
Cleanup: 80-column violation

llvm-svn: 167476

11 years agoAdd interface for object-based JIT events.
Andrew Kaylor [Tue, 6 Nov 2012 18:51:59 +0000 (18:51 +0000)]
Add interface for object-based JIT events.

This patch adds the interface to expose events from MCJIT when an object is emitted or freed and implements the MCJIT functionality to send those events.  The IntelJITEventListener implementation is left empty for now.  It will be fleshed out in a future patch.

llvm-svn: 167475

11 years agoAdded unnamed bitfields to our bitfield test.
Greg Clayton [Tue, 6 Nov 2012 18:22:59 +0000 (18:22 +0000)]
Added unnamed bitfields to our bitfield test.

llvm-svn: 167473

11 years agoUse the -fsanitize=thread flag to unbreak buildbot.
Chad Rosier [Tue, 6 Nov 2012 18:16:37 +0000 (18:16 +0000)]
Use the -fsanitize=thread flag to unbreak buildbot.

llvm-svn: 167472

11 years agoMake sure to check the module is valid after trying to find a shared module.
Greg Clayton [Tue, 6 Nov 2012 17:50:15 +0000 (17:50 +0000)]
Make sure to check the module is valid after trying to find a shared module.

llvm-svn: 167471

11 years agoAdd unit tests for source locations of AST nodes.
Manuel Klimek [Tue, 6 Nov 2012 17:31:40 +0000 (17:31 +0000)]
Add unit tests for source locations of AST nodes.

Patch by Philip Craig.

llvm-svn: 167470

11 years agoModern objective-C translator: Start adding line info to the
Fariborz Jahanian [Tue, 6 Nov 2012 17:30:23 +0000 (17:30 +0000)]
Modern objective-C translator: Start adding line info to the
translated source where it is needed. wip.

llvm-svn: 167469

11 years agoBasic: Windows doesn't define S_IFIFO.
Daniel Dunbar [Tue, 6 Nov 2012 17:08:24 +0000 (17:08 +0000)]
Basic: Windows doesn't define S_IFIFO.

llvm-svn: 167468

11 years agoMemoryBuffer: Windows doesn't define S_IFIFO.
Daniel Dunbar [Tue, 6 Nov 2012 17:08:09 +0000 (17:08 +0000)]
MemoryBuffer: Windows doesn't define S_IFIFO.

llvm-svn: 167467

11 years agotsan: slightly relax requirements for lazy shadow memory (can overlap and may not...
Dmitry Vyukov [Tue, 6 Nov 2012 16:48:46 +0000 (16:48 +0000)]
tsan: slightly relax requirements for lazy shadow memory (can overlap and may not be properly aligned)
it's problematic on windows where allocation granularity is much larger than page size

llvm-svn: 167466

11 years agoUse -fsanitize= flag in docs for ASan/TSan
Alexey Samsonov [Tue, 6 Nov 2012 16:19:11 +0000 (16:19 +0000)]
Use -fsanitize= flag in docs for ASan/TSan

llvm-svn: 167465

11 years agotsan: lazily allocate shadow for Go
Dmitry Vyukov [Tue, 6 Nov 2012 16:00:16 +0000 (16:00 +0000)]
tsan: lazily allocate shadow for Go

llvm-svn: 167464

11 years agotsan: fix debug output
Dmitry Vyukov [Tue, 6 Nov 2012 15:54:34 +0000 (15:54 +0000)]
tsan: fix debug output

llvm-svn: 167463

11 years agotsan: better diagnostics for failed mmap()
Dmitry Vyukov [Tue, 6 Nov 2012 15:39:16 +0000 (15:39 +0000)]
tsan: better diagnostics for failed mmap()

llvm-svn: 167462

11 years agoFollow-up for r167411 to un-break ASan on Mac. Move SanitizerArgs to a header file...
Alexey Samsonov [Tue, 6 Nov 2012 15:09:03 +0000 (15:09 +0000)]
Follow-up for r167411 to un-break ASan on Mac. Move SanitizerArgs to a header file and use it on Darwin toolchain.

llvm-svn: 167460

11 years agodocs: use code font for console commands in phabricator manual
Alexey Samsonov [Tue, 6 Nov 2012 15:04:37 +0000 (15:04 +0000)]
docs: use code font for console commands in phabricator manual

llvm-svn: 167459

11 years agotsan: fix compiler warning
Dmitry Vyukov [Tue, 6 Nov 2012 14:05:20 +0000 (14:05 +0000)]
tsan: fix compiler warning

llvm-svn: 167458

11 years agotsan: windows platform support
Dmitry Vyukov [Tue, 6 Nov 2012 13:35:02 +0000 (13:35 +0000)]
tsan: windows platform support

llvm-svn: 167457

11 years agotsan: update script to support windows
Dmitry Vyukov [Tue, 6 Nov 2012 13:32:53 +0000 (13:32 +0000)]
tsan: update script to support windows

llvm-svn: 167456

11 years agotsan: add shadow mapping for windows
Dmitry Vyukov [Tue, 6 Nov 2012 13:26:57 +0000 (13:26 +0000)]
tsan: add shadow mapping for windows

llvm-svn: 167455

11 years agotsan: add missing function for windows
Dmitry Vyukov [Tue, 6 Nov 2012 13:25:05 +0000 (13:25 +0000)]
tsan: add missing function for windows

llvm-svn: 167454

11 years agotsan: fix constant types
Dmitry Vyukov [Tue, 6 Nov 2012 13:21:06 +0000 (13:21 +0000)]
tsan: fix constant types

llvm-svn: 167453

11 years agotsan: windows: less includes, better diagnostics for failed VirtualAlloc(), implement...
Dmitry Vyukov [Tue, 6 Nov 2012 13:19:59 +0000 (13:19 +0000)]
tsan: windows: less includes, better diagnostics for failed VirtualAlloc(), implement sched_yield()

llvm-svn: 167452

11 years agotsan: don't release disabled clocks
Dmitry Vyukov [Tue, 6 Nov 2012 13:16:25 +0000 (13:16 +0000)]
tsan: don't release disabled clocks

llvm-svn: 167451

11 years agotsan: define missing compier macros for windows
Dmitry Vyukov [Tue, 6 Nov 2012 12:54:16 +0000 (12:54 +0000)]
tsan: define missing compier macros for windows

llvm-svn: 167450

11 years agotsan: fix non-msvc windows build
Dmitry Vyukov [Tue, 6 Nov 2012 12:50:13 +0000 (12:50 +0000)]
tsan: fix non-msvc windows build

llvm-svn: 167449

11 years agotasn: improve sanitizer_internal_defs.h
Dmitry Vyukov [Tue, 6 Nov 2012 12:49:22 +0000 (12:49 +0000)]
tasn: improve sanitizer_internal_defs.h
The file uses namespaces and extern "LANG", so it's kinda senseless to use #ifdef __cplusplus

llvm-svn: 167448

11 years agotsan: fix windows build
Dmitry Vyukov [Tue, 6 Nov 2012 12:47:42 +0000 (12:47 +0000)]
tsan: fix windows build
__declspec is not _WIN32, it's _MSC_VER

llvm-svn: 167447

11 years agotsan: fix Go build
Dmitry Vyukov [Tue, 6 Nov 2012 12:46:26 +0000 (12:46 +0000)]
tsan: fix Go build

llvm-svn: 167446

11 years agoRemove tailing whitespaces
Michael Liao [Tue, 6 Nov 2012 08:06:35 +0000 (08:06 +0000)]
Remove tailing whitespaces

llvm-svn: 167445

11 years agomisched: TargetSchedule interface for machine resources.
Andrew Trick [Tue, 6 Nov 2012 07:10:38 +0000 (07:10 +0000)]
misched: TargetSchedule interface for machine resources.

Expose the processor resources defined by the machine model to the
scheduler and other clients through the TargetSchedule interface.

Normalize each resource count with respect to other kinds of
resources. This allows scheduling heuristics to balance resources
against other kinds of resources and latency.

llvm-svn: 167444

11 years agomisched: Rename RemainingCount to avoid confusion with remaining resources.
Andrew Trick [Tue, 6 Nov 2012 07:10:34 +0000 (07:10 +0000)]
misched: Rename RemainingCount to avoid confusion with remaining resources.

llvm-svn: 167443

11 years agoRemove completely irrelevant code from USR generation.
Ted Kremenek [Tue, 6 Nov 2012 06:36:45 +0000 (06:36 +0000)]
Remove completely irrelevant code from USR generation.

Thanks to Richard Smith for pointing this out.  This code stopped
serving its purpose during r103212 in a refactoring.  My initial
fix was to add back the logic to abort the USR generation for
InternalLinkage, but enough tests broke suspiciously that I fear
that USR generation for cursors with InternalLinkage is now expected
by some clients (where it wasn't the case when the refactoring
took place).  I don't own this code anymore and have not looked
at it for some time, but clearly this code is dead and can be removed
pending further review on the proper logic here.

llvm-svn: 167442

11 years agoDelete an outdated comment.
Nico Weber [Tue, 6 Nov 2012 05:20:23 +0000 (05:20 +0000)]
Delete an outdated comment.

llvm-svn: 167441

11 years agoARM byval: when type alignment is bigger than ABI alignment, instead of
Manman Ren [Tue, 6 Nov 2012 04:58:01 +0000 (04:58 +0000)]
ARM byval: when type alignment is bigger than ABI alignment, instead of
disabling byval, we set realign to true.

It will perform an aligned alloca, and call memcpy to copy the byval
argument to the local variable.
Change the size threshold back to 64 bytes.

rdar://12596507

llvm-svn: 167440

11 years ago[analyzer] Add symbol escapes logic to the SimpleStreamChecker.
Anna Zaks [Tue, 6 Nov 2012 04:20:57 +0000 (04:20 +0000)]
[analyzer] Add symbol escapes logic to the SimpleStreamChecker.

llvm-svn: 167439

11 years ago[analyzer] Remove isWithinInlined. It's been replaced with inTopFrame().
Anna Zaks [Tue, 6 Nov 2012 04:20:54 +0000 (04:20 +0000)]
[analyzer] Remove isWithinInlined. It's been replaced with inTopFrame().

Thanks Jordan.

llvm-svn: 167438

11 years agoPropagate CharUnits through CGObjCMac.cpp.
Eli Friedman [Tue, 6 Nov 2012 03:41:04 +0000 (03:41 +0000)]
Propagate CharUnits through CGObjCMac.cpp.

llvm-svn: 167437

11 years agoMinor fix to ObjC layout bitmap metadata. Found while I was trying to
Eli Friedman [Tue, 6 Nov 2012 03:38:02 +0000 (03:38 +0000)]
Minor fix to ObjC layout bitmap metadata.  Found while I was trying to
refactor the code.

llvm-svn: 167436

11 years agoScheduleDAG interface. Added OrderKind to distinguish nonregister dependencies.
Andrew Trick [Tue, 6 Nov 2012 03:13:46 +0000 (03:13 +0000)]
ScheduleDAG interface. Added OrderKind to distinguish nonregister dependencies.

This is in preparation for adding "weak" DAG edges, but generally
simplifies the design.

llvm-svn: 167435

11 years agoUpdate compiler-rt tests to match flag renaming/deprecation in Clang.
Richard Smith [Tue, 6 Nov 2012 02:31:42 +0000 (02:31 +0000)]
Update compiler-rt tests to match flag renaming/deprecation in Clang.

llvm-svn: 167434

11 years agoClassify the INT_MIN/-1 check as -fsanitize=signed-integer-overflow, not as -fsanitiz...
Richard Smith [Tue, 6 Nov 2012 02:30:30 +0000 (02:30 +0000)]
Classify the INT_MIN/-1 check as -fsanitize=signed-integer-overflow, not as -fsanitize=divide-by-zero.

llvm-svn: 167433

11 years ago[c-index-test] Do install c-index-test because it's not just for tests,
Argyrios Kyrtzidis [Tue, 6 Nov 2012 02:12:21 +0000 (02:12 +0000)]
[c-index-test] Do install c-index-test because it's not just for tests,
it's also setup as 'INTERNAL_TOOL'.

llvm-svn: 167432

11 years agoPropagate CharUnits into ObjC CodeGen. No intended functional change.
Eli Friedman [Tue, 6 Nov 2012 01:51:20 +0000 (01:51 +0000)]
Propagate CharUnits into ObjC CodeGen.  No intended functional change.

llvm-svn: 167431

11 years agoAdd a workaround to problems with the clang debug info for
Jim Ingham [Tue, 6 Nov 2012 01:14:52 +0000 (01:14 +0000)]
Add a workaround to problems with the clang debug info for
inlined subroutine ranges.

<rdar://problem/12588579>

llvm-svn: 167430

11 years agoPer discussion on cfe-commits, treat -faddress-sanitizer,
Richard Smith [Tue, 6 Nov 2012 01:12:02 +0000 (01:12 +0000)]
Per discussion on cfe-commits, treat -faddress-sanitizer,
-fno-address-sanitizer, -fthread-sanitizer, -fno-thread-sanitizer, and
-fcatch-undefined-behavior as deprecated: produce a warning if they are used
pointing to the corresponding -fsanitize= option. In passing add the missing
'-' to some diagnostics.

llvm-svn: 167429

11 years ago[PCH] Take into account the 'include' source location of a SM_SLOC_BUFFER_ENTRY.
Argyrios Kyrtzidis [Tue, 6 Nov 2012 00:35:04 +0000 (00:35 +0000)]
[PCH] Take into account the 'include' source location of a SM_SLOC_BUFFER_ENTRY.

llvm-svn: 167426

11 years ago[PCH] Write out the ClassTemplateDecl::Common::InjectedClassNameType type
Argyrios Kyrtzidis [Tue, 6 Nov 2012 00:35:02 +0000 (00:35 +0000)]
[PCH] Write out the ClassTemplateDecl::Common::InjectedClassNameType type
reference instead of relying on computing it.

In general, if storage is no issue, it is preferable to deserialize info from
the PCH instead of trying to recompute it after the PCH was loaded.

The incentive to change this now was due to r155303 changing how friend template
classes in dependent contexts are handled; such classes can now be chained to
a previous template class but the computed InjectedClassNameType may be different
due to the extra template parameters from the dependent context.

The new handling requires more investigation but, in the meantime, writing out
InjectedClassNameType fixes PCH issue in rdar://12627738.

llvm-svn: 167425

11 years ago<rdar://problem/12582031>
Greg Clayton [Tue, 6 Nov 2012 00:20:41 +0000 (00:20 +0000)]
<rdar://problem/12582031>

Unnamed bitfields cause struct layout problems

Synthesize unnamed bitfields when required. Most compilers don't mention unnamed bitfields in the DWARF, so we need to create them to keep clang happy with the types we create from the DWARF. We currently can't do this for ObjC since the DW_AT_bit_offset value for any direct ivars of ObjC classes as the values for these attributes are bogus. A bug has been filed on Clang to fix this, and another bug has been filed on LLDB to make sure we fix the DWARF parser once the clang fix is in by looking the the DW_AT_producer in the compile unit attributes and finding the compiler version and only enabling it for newer versions of clang.

llvm-svn: 167424

11 years agoHave the parser initialize Sema before it consumes the first
Douglas Gregor [Mon, 5 Nov 2012 23:58:27 +0000 (23:58 +0000)]
Have the parser initialize Sema before it consumes the first
token. This is important because the first token could actually be
after an #include that triggers a module import, which might use
either Sema or the AST consumer before it would have been initialized.

llvm-svn: 167423

11 years agoRefactor the getTypeLegalizationCost interface. No functionality change.
Nadav Rotem [Mon, 5 Nov 2012 23:57:45 +0000 (23:57 +0000)]
Refactor the getTypeLegalizationCost interface. No functionality change.

llvm-svn: 167422

11 years agoCostModel: Add tables for the common x86 compares.
Nadav Rotem [Mon, 5 Nov 2012 23:48:20 +0000 (23:48 +0000)]
CostModel: Add tables for the common x86 compares.

llvm-svn: 167421

11 years agoSplit the instrinsic header wmmintrin.h into AES and PCLMUL parts, so
Douglas Gregor [Mon, 5 Nov 2012 23:30:26 +0000 (23:30 +0000)]
Split the instrinsic header wmmintrin.h into AES and PCLMUL parts, so
that we can model them as separate submodules.

llvm-svn: 167420

11 years agoFrontend: Add support for reading named pipes as the main file.
Daniel Dunbar [Mon, 5 Nov 2012 22:53:33 +0000 (22:53 +0000)]
Frontend: Add support for reading named pipes as the main file.

 - The whole {File,Source}Manager is built around wanting to pre-determine the
   size of files, so we can't fit this in naturally. Instead, we handle it like
   we do STDIN, where we just replace the main file contents upfront.

llvm-svn: 167419

11 years agoDelete comment I forgot to delete in my last change.
Nico Weber [Mon, 5 Nov 2012 22:50:26 +0000 (22:50 +0000)]
Delete comment I forgot to delete in my last change.

llvm-svn: 167418

11 years agoUse Richard's BE_THE_HEADER trick to simplify a test. No intended behavior change.
Nico Weber [Mon, 5 Nov 2012 22:49:03 +0000 (22:49 +0000)]
Use Richard's BE_THE_HEADER trick to simplify a test. No intended behavior change.

llvm-svn: 167417

11 years agoARM byval: when type alignment is bigger than ABI alignment, we can't guarantee
Manman Ren [Mon, 5 Nov 2012 22:42:46 +0000 (22:42 +0000)]
ARM byval: when type alignment is bigger than ABI alignment, we can't guarantee
the type alignment of the byval argument. This patch will disable byval in this case,
it also increases the size threshold for turning on byval.

A backend fix will be attempted.

rdar://12596507

llvm-svn: 167416

11 years agoUse the individual -fsanitize=<...> arguments to control which of the UBSan
Richard Smith [Mon, 5 Nov 2012 22:21:05 +0000 (22:21 +0000)]
Use the individual -fsanitize=<...> arguments to control which of the UBSan
checks to enable. Remove frontend support for -fcatch-undefined-behavior,
-faddress-sanitizer and -fthread-sanitizer now that they don't do anything.

llvm-svn: 167413

11 years agoCode Model: Improve the accuracy of the zext/sext/trunc vector cost estimation.
Nadav Rotem [Mon, 5 Nov 2012 22:20:53 +0000 (22:20 +0000)]
Code Model: Improve the accuracy of the zext/sext/trunc vector cost estimation.

llvm-svn: 167412

11 years agoAdd -fsanitize=<sanitizers> argument to driver and frontend, and add
Richard Smith [Mon, 5 Nov 2012 22:04:41 +0000 (22:04 +0000)]
Add -fsanitize=<sanitizers> argument to driver and frontend, and add
-fno-sanitize=<sanitizers> argument to driver. These allow ASan, TSan, and the
various UBSan checks to be enabled and disabled separately. Right now, the
different modes can't be combined, but the intention is that combining UBSan
and the other sanitizers will be permitted in the near future.

Currently, the UBSan checks will all be enabled if any of them is; that will be
fixed by the next patch.

llvm-svn: 167411

11 years agoSuppress signed/unsigned comparison warning.
Richard Smith [Mon, 5 Nov 2012 22:01:44 +0000 (22:01 +0000)]
Suppress signed/unsigned comparison warning.

llvm-svn: 167410

11 years agoFix for PR14264 cause by commit r167237 which did not take into account a
Kevin Enderby [Mon, 5 Nov 2012 21:55:41 +0000 (21:55 +0000)]
Fix for PR14264 cause by commit r167237 which did not take into account a
possible buffer change with a .macro directive.

rdar://12637628

llvm-svn: 167408

11 years agoMemoryBuffer: Support reading named pipes in getFile().
Daniel Dunbar [Mon, 5 Nov 2012 21:55:40 +0000 (21:55 +0000)]
MemoryBuffer: Support reading named pipes in getFile().

 - We only support this when the client didn't claim to know the file size.

llvm-svn: 167407

11 years agoRename LangOptions members for address sanitizer and thread sanitizer from
Richard Smith [Mon, 5 Nov 2012 21:48:12 +0000 (21:48 +0000)]
Rename LangOptions members for address sanitizer and thread sanitizer from
*Sanitizer to Sanitize* in preparation for later patches.

llvm-svn: 167405

11 years agoCost Model: Normalize the insert/extract index when splitting types
Nadav Rotem [Mon, 5 Nov 2012 21:12:13 +0000 (21:12 +0000)]
Cost Model: Normalize the insert/extract index when splitting types

llvm-svn: 167402

11 years agoCost Model: teach the cost model about expanding integers.
Nadav Rotem [Mon, 5 Nov 2012 21:11:10 +0000 (21:11 +0000)]
Cost Model: teach the cost model about expanding integers.

llvm-svn: 167401

11 years agoAdd a method to indicate section address re-assignment is finished.
Andrew Kaylor [Mon, 5 Nov 2012 20:57:16 +0000 (20:57 +0000)]
Add a method to indicate section address re-assignment is finished.

Prior to this patch RuntimeDyld attempted to re-apply relocations every time reassignSectionAddress was called (via MCJIT::mapSectionAddress).  In addition to being inefficient and redundant, this led to a problem when a section was temporarily moved too far away from another section with a relative relocation referencing the section being moved.  To fix this, I'm adding a new method (finalizeObject) which the client can call to indicate that it is finished rearranging section addresses so the relocations can safely be applied.

llvm-svn: 167400

11 years agoFix module map for SSE4a builtins
Douglas Gregor [Mon, 5 Nov 2012 20:41:30 +0000 (20:41 +0000)]
Fix module map for SSE4a builtins

llvm-svn: 167399

11 years agoMake cpuid.h actually work with -std=c99 <rdar://problem/12552716>.
Douglas Gregor [Mon, 5 Nov 2012 20:11:10 +0000 (20:11 +0000)]
Make cpuid.h actually work with -std=c99 <rdar://problem/12552716>.

While we're here, extend the module map to cover most of the
newly-added instrinsic headers. Only wmmintrin.h is missing, because
it needs to be split into AES/PCLMUL subheaders (as a separate commit).

llvm-svn: 167398

11 years agoHash the various compiler version, target, preprocessor, and
Douglas Gregor [Mon, 5 Nov 2012 19:45:09 +0000 (19:45 +0000)]
Hash the various compiler version, target, preprocessor, and
header-search options into the module hash. We're just using
ADT/Hashing.hpp for this, which isn't as cryptographically strong as
I'd like, but it'll do. If someone contributes (say) and MD4
implementation, we'd happily switch to that.

llvm-svn: 167397

11 years agoOn PowerPC64, integer return values (as well as arguments) are supposed
Ulrich Weigand [Mon, 5 Nov 2012 19:39:45 +0000 (19:39 +0000)]
On PowerPC64, integer return values (as well as arguments) are supposed
to be extended to a full register.   This is modeled in the IR by marking
the return value (or argument) with a signext or zeroext attribute.

However, while these attributes are respected for function arguments,
they are currently ignored for function return values by the PowerPC
back-end.  This patch updates PPCCallingConv.td to ask for the promotion
to i64, and fixes LowerReturn and LowerCallResult to implement it.

The new test case verifies that both arguments and return values are
properly extended when passing them; and also that the optimizers
understand incoming argument and return values are in fact guaranteed
by the ABI to be extended.

The patch caused a spurious breakage in CodeGen/PowerPC/coalesce-ext.ll,
since the test case used a "ret" instruction to create a use of an i32
value at the end of the function (to set up data flow as required for
what the test is intended to test).  Since there's now an implicit
promotion to i64, that data flow no longer works as expected.  To fix
this, this patch now adds an extra "add" to ensure we have an appropriate
use of the i32 value.

llvm-svn: 167396

11 years agoImplement the cost of abnormal x86 instruction lowering as a table.
Nadav Rotem [Mon, 5 Nov 2012 19:32:46 +0000 (19:32 +0000)]
Implement the cost of abnormal x86 instruction lowering as a table.

llvm-svn: 167395

11 years agoAdd absolute symbol scope recognition in lld/Core
Hemant Kulkarni [Mon, 5 Nov 2012 19:13:54 +0000 (19:13 +0000)]
Add absolute symbol scope recognition in lld/Core

llvm-svn: 167394

11 years agoOn PowerPC64, integer arguments and return values need to be sign- or
Ulrich Weigand [Mon, 5 Nov 2012 19:13:42 +0000 (19:13 +0000)]
On PowerPC64, integer arguments and return values need to be sign- or
zero-extended to 64 bits.  This information is currently provided to
the back end by setting "signext" or "zeroext" attributes.  However,
this is done only for integer types *smaller* than i32, not for i32
itself.  This causes clang to generate code violating the ABI, which
results in a failure of the tramp3d-v4 test case (due to calling a
system library routine without ABI-required extension).

This patch implements custom versions of classifyArgumentType and
classifyReturnType for PPC64_SVR4_ABIInfo, which are the same as the
default versions except that they also classify "int" and "unsigned int"
as types needing extending.  This fixed tramp3d-v4 on PowerPC64.

llvm-svn: 167393

11 years agolli: Initialize the native asm parser for inline assembly.
Jim Grosbach [Mon, 5 Nov 2012 19:06:05 +0000 (19:06 +0000)]
lli: Initialize the native asm parser for inline assembly.

MCJIT supports inline assembly, but requires the asm parser to do so.
Make sure to link it in and initialize it.

llvm-svn: 167392

11 years agoAvoid potentially disastrous mistakes when removing a file.
Filipe Cabecinhas [Mon, 5 Nov 2012 18:41:33 +0000 (18:41 +0000)]
Avoid potentially disastrous mistakes when removing a file.

llvm-svn: 167391

11 years agoFollowing the example from loop-convert tests, simplifying the remove-cstr-calls...
Manuel Klimek [Mon, 5 Nov 2012 18:28:45 +0000 (18:28 +0000)]
Following the example from loop-convert tests, simplifying the remove-cstr-calls lit script.

Patch by Edwin Vane.

llvm-svn: 167390

11 years agoCleaned up the code signing instructions:
Greg Clayton [Mon, 5 Nov 2012 18:23:25 +0000 (18:23 +0000)]
Cleaned up the code signing instructions:
- make sure to delet the .cer file on the desktop
- added section that describes what to do when you re-install a new OS where
  you still have an old code signing certificate around.

llvm-svn: 167389

11 years agoAdd support for the PowerPC-specific inline asm Z constraint and y modifier.
Hal Finkel [Mon, 5 Nov 2012 18:18:42 +0000 (18:18 +0000)]
Add support for the PowerPC-specific inline asm Z constraint and y modifier.

The Z constraint specifies an r+r memory address, and the y modifier expands
to the "r, r" in the asm string. For this initial implementation, the base
register is forced to r0 (which has the special meaning of 0 for r+r addressing
on PowerPC) and the full address is taken in the second register. In the
future, this should be improved.

llvm-svn: 167388

11 years agoTests for 32bit version of block ivar layout metadata.
Fariborz Jahanian [Mon, 5 Nov 2012 18:14:34 +0000 (18:14 +0000)]
Tests for 32bit version of block ivar layout metadata.

llvm-svn: 167387

11 years ago[PATCH] PowerPC: Expand load extend vector operations
Adhemerval Zanella [Mon, 5 Nov 2012 17:15:56 +0000 (17:15 +0000)]
[PATCH] PowerPC: Expand load extend vector operations

This patch expands the SEXTLOAD, ZEXTLOAD, and EXTLOAD operations for
vector types when altivec is enabled.

llvm-svn: 167386

11 years ago[analyzer] Move convenience REGISTER_*_WITH_PROGRAMSTATE to CheckerContext.h
Jordan Rose [Mon, 5 Nov 2012 16:58:00 +0000 (16:58 +0000)]
[analyzer] Move convenience REGISTER_*_WITH_PROGRAMSTATE to CheckerContext.h

As Anna pointed out, ProgramStateTrait.h is a relatively obscure header,
and checker writers may not know to look there to add their own custom
state.

The base macro that specializes the template remains in ProgramStateTrait.h
(REGISTER_TRAIT_WITH_PROGRAMSTATE), which allows the analyzer core to keep
using it.

llvm-svn: 167385

11 years agoAdd missing this->. Fixes pr14238.
Rafael Espindola [Mon, 5 Nov 2012 14:57:21 +0000 (14:57 +0000)]
Add missing this->. Fixes pr14238.

llvm-svn: 167383

11 years agoDon't infer whether a value is captured in the current function from the
Richard Osborne [Mon, 5 Nov 2012 10:48:24 +0000 (10:48 +0000)]
Don't infer whether a value is captured in the current function from the
'nocapture' attribute.

The nocapture attribute only specifies that no copies are made that
outlive the function. This isn't the same as there being no copies at all.
This fixes PR14045.

llvm-svn: 167381

11 years agoAdd a couple of stubs to the release notes for things I noticed while
Chandler Carruth [Mon, 5 Nov 2012 10:17:00 +0000 (10:17 +0000)]
Add a couple of stubs to the release notes for things I noticed while
clearing out my backlog of commit mail.

llvm-svn: 167380

11 years agoOur style for diagnostic messages is to not include a trailing dot.
Matt Beaumont-Gay [Mon, 5 Nov 2012 05:32:00 +0000 (05:32 +0000)]
Our style for diagnostic messages is to not include a trailing dot.

llvm-svn: 167379