Dmitry Vyukov [Tue, 16 Jul 2013 11:28:04 +0000 (11:28 +0000)]
tsan: support sigsuspend() call
Intercepting it makes it process pending signal before return.
llvm-svn: 186400
Richard Sandiford [Tue, 16 Jul 2013 11:02:24 +0000 (11:02 +0000)]
[SystemZ] Use RISBG for (shift (and ...))
Another patch in the series to make more use of R.SBG. This one extends
r186072 and r186073 to handle cases where the AND is inside the shift.
llvm-svn: 186399
Tim Northover [Tue, 16 Jul 2013 10:22:19 +0000 (10:22 +0000)]
Fix test on release builds.
Unfortunately I don't think there's a good way to validate branch targets on
release builds. Fortunately it's a minor part of this test (and based on
generic code) so I don't mind dropping it.
llvm-svn: 186398
Vladimir Medic [Tue, 16 Jul 2013 10:07:14 +0000 (10:07 +0000)]
This patch represents Mips utilization of r186388 code that alows asm matcher to emit mnemonics contain '.' characters. This makes asm parser code simpler and more efficient.
llvm-svn: 186397
NAKAMURA Takumi [Tue, 16 Jul 2013 09:59:51 +0000 (09:59 +0000)]
PPCJITInfo.cpp: Tweak r186252 with s/__ppc/__powerpc/ to work on powerpc-linux Fedora 12.
g++ (GCC) 4.4.4
20100630 (Red Hat 4.4.4-10)
llvm-svn: 186396
Tim Northover [Tue, 16 Jul 2013 09:47:53 +0000 (09:47 +0000)]
ARM: implement low-level intrinsics for the atomic exclusive operations.
This adds three overloaded intrinsics to Clang:
T __builtin_arm_ldrex(const volatile T *addr)
int __builtin_arm_strex(T val, volatile T *addr)
void __builtin_arm_clrex()
The intent is that these do what users would expect when given most sensible
types. Currently, "sensible" translates to ints, floats and pointers.
llvm-svn: 186394
Timur Iskhodzhanov [Tue, 16 Jul 2013 09:47:39 +0000 (09:47 +0000)]
[ASan] Use less shadow on Win 32-bit
llvm-svn: 186393
Tim Northover [Tue, 16 Jul 2013 09:46:55 +0000 (09:46 +0000)]
ARM: implement ldrex, strex and clrex intrinsics
Intrinsics already existed for the 64-bit variants, so these support operations
of size at most 32-bits.
llvm-svn: 186392
Renato Golin [Tue, 16 Jul 2013 09:32:17 +0000 (09:32 +0000)]
ARM EABI divmod support
This patch enables calls to __aeabi_idivmod when in EABI mode,
by using the remainder value returned on registers (R1),
enabled by the ARM triple "none-eabi". Note that Darwin and
GNUEABI triples will continue lowering on GNU style, that is,
using the stack for the remainder.
Still need to add SREM/UREM support fix for 64-bit lowering.
llvm-svn: 186390
Alexander Potapenko [Tue, 16 Jul 2013 09:29:48 +0000 (09:29 +0000)]
[ASan] Cache the OSX version to avoid calling sysctl() on every GetMacosVersion() call.
llvm-svn: 186389
Vladimir Medic [Tue, 16 Jul 2013 09:22:38 +0000 (09:22 +0000)]
This patch allows targets to define weather the instruction mnemonics in asm matcher tables will contain '.' character.
llvm-svn: 186388
NAKAMURA Takumi [Tue, 16 Jul 2013 09:06:47 +0000 (09:06 +0000)]
llvm/test/Object/directory.ll: Mark it as XFAIL:cygwin. Directories can be opened on cygwin.
llvm-svn: 186387
Alexander Potapenko [Tue, 16 Jul 2013 08:35:42 +0000 (08:35 +0000)]
[ASan] Add support for OS X Mavericks to GetMacosVersion.
llvm-svn: 186386
Serge Pavlov [Tue, 16 Jul 2013 07:14:18 +0000 (07:14 +0000)]
Limit number of bits in size representation so that bit size fit 64 bits.
This fixes PR8256 and some others.
llvm-svn: 186385
Rui Ueyama [Tue, 16 Jul 2013 06:21:39 +0000 (06:21 +0000)]
[PECOFF][Writer] Replace magic numbers with sizeof().
This is a follow up patch for r186336.
Reviewers: LegalizeAdulthood
CC: silvas, llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1144
llvm-svn: 186384
Craig Topper [Tue, 16 Jul 2013 05:03:10 +0000 (05:03 +0000)]
Add 'const' qualifiers to static const char* variables.
llvm-svn: 186383
Rui Ueyama [Tue, 16 Jul 2013 04:47:02 +0000 (04:47 +0000)]
[PECOFF][Writer] Use defined constants instead of magic numbers.
llvm-svn: 186382
Rafael Espindola [Tue, 16 Jul 2013 03:34:31 +0000 (03:34 +0000)]
Use open+fstat instead of stat+open.
llvm-svn: 186381
Rafael Espindola [Tue, 16 Jul 2013 03:30:10 +0000 (03:30 +0000)]
Remember that we have a null terminated string.
This is a micro optimization. Instead of going char*->StringRef->Twine->char*,
go char*->Twine->char* and avoid having to copy the filename on the stack.
llvm-svn: 186380
Rui Ueyama [Tue, 16 Jul 2013 03:23:55 +0000 (03:23 +0000)]
[Object/COFF] Add import_directory_table_entry.
Summary: Add import_directory_table_entry to use for .idata section.
Reviewers: Bigcheese
CC: llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1059
llvm-svn: 186379
Rafael Espindola [Tue, 16 Jul 2013 03:20:13 +0000 (03:20 +0000)]
Add a version of sys::fs::status that uses fstat.
llvm-svn: 186378
Rui Ueyama [Tue, 16 Jul 2013 03:11:55 +0000 (03:11 +0000)]
COFF: Add constants for optional data directory.
llvm-svn: 186377
Rafael Espindola [Tue, 16 Jul 2013 02:55:33 +0000 (02:55 +0000)]
Instead friending status, provide windows and posix constructors to file_status.
This opens the way of having static helpers in the .inc files that can
construct a file_status.
llvm-svn: 186376
NAKAMURA Takumi [Tue, 16 Jul 2013 02:44:23 +0000 (02:44 +0000)]
unittests/Support: Add TimeValue.Win32FILETIME, corresponding to r186374.
llvm-svn: 186375
NAKAMURA Takumi [Tue, 16 Jul 2013 02:43:51 +0000 (02:43 +0000)]
Fix TimeValue::toWin32Time() to be symmetric to fromWin32Time() and compatible to Win32's FILETIME.
llvm-ar is the only user of toWin32Time() (via setLastModificationAndAccessTime), and r186298 can be reverted.
It had been buggy since the initial commit.
FIXME: Could we rename {from|to}Win32Time as {from|to}Win32FILETIME in TimeValue?
llvm-svn: 186374
Eli Friedman [Tue, 16 Jul 2013 02:07:49 +0000 (02:07 +0000)]
Merge attributes on typedef decls.
Not completely sure this is right, but it's clearly better than what
we did before this commit (effectively dropping the attribute).
<rdar://problem/
14413117>
llvm-svn: 186373
NAKAMURA Takumi [Tue, 16 Jul 2013 02:03:32 +0000 (02:03 +0000)]
Rename Support.TimeValue to TimeValue.time_t in unittests/Support.
llvm-svn: 186372
Craig Topper [Tue, 16 Jul 2013 01:17:10 +0000 (01:17 +0000)]
Add 'const' qualifiers to static const char* variables.
llvm-svn: 186371
Eli Friedman [Tue, 16 Jul 2013 00:21:28 +0000 (00:21 +0000)]
Fix alignment of class derived from empty class.
The record layout code didn't properly take into account that
an empty class at offset 0 can have an alignment greater than 1.
Patch by Andrea Di Biagio.
llvm-svn: 186370
Fariborz Jahanian [Tue, 16 Jul 2013 00:20:21 +0000 (00:20 +0000)]
ObjC migrator: build conforming interface
declaration (not yet used). wip.
llvm-svn: 186369
Rui Ueyama [Tue, 16 Jul 2013 00:03:51 +0000 (00:03 +0000)]
[PECOFF] Add default argument to addDir32NBReloc() as the default value 0 is reasonable.
llvm-svn: 186368
Eli Friedman [Tue, 16 Jul 2013 00:01:31 +0000 (00:01 +0000)]
Fix member refs with using decl + anonymous union.
Make sure we call BuildFieldReferenceExpr with the appropriate decl
when a member of an anonymous union is made public with a using decl.
Also, fix a crash on invalid field access into an anonymous union.
Fixes PR16630.
llvm-svn: 186367
Rafael Espindola [Mon, 15 Jul 2013 23:55:07 +0000 (23:55 +0000)]
Revert "Don't pass llvm::errs() all over the place. Diagnostics always go to stderr."
This reverts commit 185657. It will be used by unit tests.
llvm-svn: 186366
Rafael Espindola [Mon, 15 Jul 2013 23:51:47 +0000 (23:51 +0000)]
Add mingw32 to the XFAIL. I forgot about it when adding win32.
llvm-svn: 186365
Manman Ren [Mon, 15 Jul 2013 23:47:29 +0000 (23:47 +0000)]
PEI: Support for non-zero SPAdj at beginning of a basic block.
We can have a FrameSetup in one basic block and the matching FrameDestroy
in a different basic block when we have struct byval. In that case, SPAdj
is not zero at beginning of the basic block.
Modify PEI to correctly set SPAdj at beginning of each basic block using
DFS traversal. We used to assume SPAdj is 0 at beginning of each basic block.
PEI had an assert SPAdjCount || SPAdj == 0.
If we have a Destroy <n> followed by a Setup <m>, PEI will assert failure.
We can add an extra condition to make sure the pairs are matched:
The pairs start with a FrameSetup.
But since we are doing a much better job in the verifier, this patch removes
the check in PEI.
PR16393
llvm-svn: 186364
Sean Callanan [Mon, 15 Jul 2013 23:31:47 +0000 (23:31 +0000)]
Fixed a problem in IRForTarget where we would not
delete a constant after we replaced it with a
dynamically-computed value. Also ensured that we
replace all users of the constant if there are
multiple ones. Added a testcase.
<rdar://problem/
14379043>
llvm-svn: 186363
Enrico Granata [Mon, 15 Jul 2013 23:17:32 +0000 (23:17 +0000)]
Disabling the introspecting summary for __NSCFSet (essentially, for CF*SetRef)
llvm-svn: 186362
Ed Maste [Mon, 15 Jul 2013 22:59:08 +0000 (22:59 +0000)]
Remove unused RunLocker and related code
RunLocker was not used anywhere, and was the only instance of the
WriteLocker class. Remove both.
llvm-svn: 186361
Greg Clayton [Mon, 15 Jul 2013 22:54:20 +0000 (22:54 +0000)]
<rdar://problem/
13793059>
Added a setting to control timeout for kdp response packets. While I was at it, I also added a way to control the response timeout for gdb-remote packets.
KDP defaults to 5 seconds, and GDB defaults to 1 second. These were the default values that were in the code prior to adding these settings.
(lldb) settings set plugin.process.gdb-remote.packet-timeout 10
(lldb) settings set plugin.process.kdp-remote.packet-timeout 10
llvm-svn: 186360
Nadav Rotem [Mon, 15 Jul 2013 22:52:48 +0000 (22:52 +0000)]
PR16628: Fix a bug in the code that merges compares.
Compares return i1 but they compare different types.
llvm-svn: 186359
Jason Molenda [Mon, 15 Jul 2013 22:40:17 +0000 (22:40 +0000)]
Fix typeo in diagnose-unwind.py.
llvm-svn: 186358
Hal Finkel [Mon, 15 Jul 2013 22:29:40 +0000 (22:29 +0000)]
PPC: Refactoring to support subtarget feature changing
This change mirrors the changes that were made to the X86 and ARM targets to
support subtarget feature changing. As indicated in r182899, the mechanism is
still undergoing revision, and so as with the X86 and ARM targets, there is no
test case yet (there is no effective functionality change).
llvm-svn: 186357
David Blaikie [Mon, 15 Jul 2013 22:28:45 +0000 (22:28 +0000)]
Further simplify test case from r186119/r186035.
llvm-svn: 186356
Rui Ueyama [Mon, 15 Jul 2013 22:20:10 +0000 (22:20 +0000)]
[PECOFF][Writer] Follow up patch for r186336.
- Make a const reference instead of copying an object
- Fix a comment
llvm-svn: 186355
Rafael Espindola [Mon, 15 Jul 2013 22:16:53 +0000 (22:16 +0000)]
XFAIL on windows too and document the XFAILs.
llvm-svn: 186354
Rui Ueyama [Mon, 15 Jul 2013 22:15:32 +0000 (22:15 +0000)]
[PECOFF][Writer] s/dyn_cast/cast/ where we don't expect null return value.
llvm-svn: 186353
Rui Ueyama [Mon, 15 Jul 2013 21:35:15 +0000 (21:35 +0000)]
[PECOFF][Writer] Refactoring: Move private methods below public ones.
llvm-svn: 186352
David Blaikie [Mon, 15 Jul 2013 21:33:26 +0000 (21:33 +0000)]
Test case for r18266 (serialization support for TagDecl:IsCompleteDefinitionRequired)
llvm-svn: 186351
Manman Ren [Mon, 15 Jul 2013 21:26:31 +0000 (21:26 +0000)]
Machine Verifier: verify FrameSetup and FrameDestroy
1> on every path through the CFG, a FrameSetup <n> is always followed by a
FrameDestroy <n> and a FrameDestroy is always followed by a FrameSetup.
2> stack adjustments are identical on all CFG edges to a merge point.
3> frame is destroyed at end of a return block.
PR16393
llvm-svn: 186350
Fariborz Jahanian [Mon, 15 Jul 2013 21:22:08 +0000 (21:22 +0000)]
ObjC migrator: finding conforming protocol
candidates for each class. wip.
llvm-svn: 186349
Kaelyn Uhrain [Mon, 15 Jul 2013 21:18:33 +0000 (21:18 +0000)]
Fix a typo in the diagnostic note added in r186342.
llvm-svn: 186348
Greg Clayton [Mon, 15 Jul 2013 21:10:17 +0000 (21:10 +0000)]
Fix issues with GCC debugging. GCC emits DWARF in unique ways that LLDB wasn't handling. This fix will fix cases where classes are forward declared using DW_TAG_structure_type and then actually defined using DW_TAG_class_type. LLDB, when it finds a forward declaration, would try and find and parse the complete type. It does this by:
1 - looking up the type basename in the type index
2 - iterate through all matches and look for decl contexts (namespace/class hierarchy) that match
The issue was the decl context matching wasn't watching for DW_TAG_class_type/DW_TAG_structure_type mismatches, and it wasn't also getting the name for DIE's that didn't have a DW_AT_name, but did have a DW_AT_specification that had a name.
llvm-svn: 186347
Rui Ueyama [Mon, 15 Jul 2013 21:00:53 +0000 (21:00 +0000)]
[PECOFF] Fix comment.
llvm-svn: 186346
Rafael Espindola [Mon, 15 Jul 2013 20:52:01 +0000 (20:52 +0000)]
Remove an extra is_directory call.
I checked that opening a directory on windows does fail, so this saves a "stat".
llvm-svn: 186345
Marshall Clow [Mon, 15 Jul 2013 20:46:11 +0000 (20:46 +0000)]
Make std::forward and std::move (and std::move_if_noexcept) constexpr in C++14
llvm-svn: 186344
Hal Finkel [Mon, 15 Jul 2013 20:22:58 +0000 (20:22 +0000)]
Fix register subclass handling in PPCInstrInfo::insertSelect
PPCInstrInfo::insertSelect and PPCInstrInfo::canInsertSelect were computing the
common subclass of the true and false inputs, and then selecting either the
32-bit or the 64-bit isel variant based on the result of calling
PPC::GPRCRegClass.hasSubClassEq(RC) and PPC::G8RCRegClass.hasSubClassEq(RC)
(where RC is the common subclass). Unfortunately, this is not quite right: if
we have something like this:
%vreg8<def> = SELECT_CC_I8 %vreg4<kill>, %vreg7<kill>, %vreg6<kill>, 76;
G8RC_and_G8RC_NOX0:%vreg8 CRRC:%vreg4 G8RC_NOX0:%vreg7,%vreg6
then the common subclass of G8RC_and_G8RC_NOX0 and G8RC_NOX0 is G8RC_NOX0, and
G8RC_NOX0 is not a subclass of G8RC (because it also contains the ZERO8
pseudo-register). As a result, we also need to check the common subclass
against GPRC_NOR0 and G8RC_NOX0 explicitly.
This had not been a problem for clients of insertSelect that called
canInsertSelect first (because it had a compensating mistake), but insertSelect
is also used by the PPC pseudo-instruction expander, and this error was causing
a problem in that context.
This problem was found by csmith.
llvm-svn: 186343
Kaelyn Uhrain [Mon, 15 Jul 2013 19:54:54 +0000 (19:54 +0000)]
Move the "->" to "." fixit from r186128 into a separate note since
recovery is not attempted with the fixit. Also move the associated test
case from FixIt/fixit.cpp to SemaCXX/member-expr.cpp since the fixit is
no longer automatically applied.
llvm-svn: 186342
Reid Kleckner [Mon, 15 Jul 2013 19:41:21 +0000 (19:41 +0000)]
[mc-coff] Resolve aliases when emitting COFF relocations
This is consistent with the ELF object writer.
Add some COFF tests that relocate against an alias.
Reviewers: espindola
Differential Revision: http://llvm-reviews.chandlerc.com/D1079
llvm-svn: 186341
Samuel Benzaquen [Mon, 15 Jul 2013 19:25:06 +0000 (19:25 +0000)]
Add support for type traversal matchers.
Summary:
Fixup the type traversal macros/matchers to specify the supported types.
Make the marshallers a little more generic to support any variadic function.
Update the doc script.
Reviewers: klimek
CC: cfe-commits, revane
Differential Revision: http://llvm-reviews.chandlerc.com/D1023
llvm-svn: 186340
Tom Stellard [Mon, 15 Jul 2013 19:00:09 +0000 (19:00 +0000)]
R600/SI: Add support for 64-bit loads
https://bugs.freedesktop.org/show_bug.cgi?id=65873
llvm-svn: 186339
Hal Finkel [Mon, 15 Jul 2013 18:57:05 +0000 (18:57 +0000)]
Remove invalid assert in DAGTypeLegalizer::RemapValue
There is a comment at the top of DAGTypeLegalizer::PerformExpensiveChecks
which, in part, says:
// Note that these invariants may not hold momentarily when processing a node:
// the node being processed may be put in a map before being marked Processed.
Unfortunately, this assert would be valid only if the above-mentioned invariant
held unconditionally. This was causing llc to assert when, in fact,
everything was fine.
Thanks to Richard Sandiford for investigating this issue!
Fixes PR16562.
llvm-svn: 186338
Sean Callanan [Mon, 15 Jul 2013 18:43:36 +0000 (18:43 +0000)]
Actually use the return value we get back when
creating a persistent variable, rather than making
a (potentially expensive) lookup by name.
<rdar://problem/
14337653>
llvm-svn: 186337
Rui Ueyama [Mon, 15 Jul 2013 18:43:01 +0000 (18:43 +0000)]
[PECOFF][Writer] Emit .reloc section.
Emit .reloc section. This is the first step to support DLL creation. The
executable doesn't need .reloc section, but the DLL does.
Reviewers: Bigcheese
CC: llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1126
llvm-svn: 186336
Howard Hinnant [Mon, 15 Jul 2013 18:21:11 +0000 (18:21 +0000)]
Bill Fisher: This patch fixes an ill-formed comparison when parsing control escapes, e.g. "\cA\ca". The code will now throw an error_escape exception for invalid control sequences like "\c:" or "\c".
I've added the test cases to bad_escape.pass.cpp.
llvm-svn: 186335
Howard Hinnant [Mon, 15 Jul 2013 18:09:11 +0000 (18:09 +0000)]
A few fixes to tests for Windows port.
llvm-svn: 186334
Stephen Lin [Mon, 15 Jul 2013 17:55:02 +0000 (17:55 +0000)]
Remove trailing whitespace
llvm-svn: 186333
Chandler Carruth [Mon, 15 Jul 2013 17:36:21 +0000 (17:36 +0000)]
Revert r186316 while I track down an ASan failure and an assert from
a bot.
This reverts the commit which introduced a new implementation of the
fancy SROA pass designed to reduce its overhead. I'll skip the huge
commit log here, refer to r186316 if you're looking for how this all
works and why it works that way.
llvm-svn: 186332
Chandler Carruth [Mon, 15 Jul 2013 17:27:42 +0000 (17:27 +0000)]
Re-revert r86040, which was un-reverted in r186199.
This breaks the build of basic patterns with repeated friend
declarations. See the added test case in SemaCXX/friend.cpp or the test
case reported to the original commit log.
Original commit log:
If we friend a declaration twice, that should not make it visible to
name lookup in the surrounding context. Slightly rework how we handle
friend declarations to inherit the visibility of the prior
declaration, rather than setting a friend declaration to be visible
whenever there was a prior declaration.
llvm-svn: 186331
Aaron Ballman [Mon, 15 Jul 2013 16:53:32 +0000 (16:53 +0000)]
Teaching llvm-tblgen to not emit a switch statement when there are no case statements.
llvm-svn: 186330
Reid Kleckner [Mon, 15 Jul 2013 16:40:52 +0000 (16:40 +0000)]
Revert "[Option] Store arg strings in a set backed by a BumpPtrAllocator"
This broke clang's crash-report.c test, and I haven't been able to
figure it out yet.
This reverts commit r186319.
llvm-svn: 186329
Timur Iskhodzhanov [Mon, 15 Jul 2013 16:11:39 +0000 (16:11 +0000)]
Fix check_lint warnings in sanitizers' runtime libraries
llvm-svn: 186328
Ariel J. Bernal [Mon, 15 Jul 2013 15:37:05 +0000 (15:37 +0000)]
Fix UseAuto not transforming iterator when non-fully qualifiers are used and
using inline namespaces is specified.
UseAuto used to fail to transform iterators when using inline namespaces and
non-fully qualified types, relying on a using directive previously declared.
- This fix uses the already define isFromStdNamespace matcher.
- Fixed tests and added a new test using inline namespaces.
- Added CustomMatchers to reuse common matchers among transforms.
llvm-svn: 186327
Tom Stellard [Mon, 15 Jul 2013 15:20:05 +0000 (15:20 +0000)]
Add integer-gentype.inc: Missing file from r185839
llvm-svn: 186326
Ashok Thirumurthi [Mon, 15 Jul 2013 15:05:33 +0000 (15:05 +0000)]
Adds a passing test for step-over and step-out when the inferior is about to call longjmp (non-local goto) in a single-threaded inferior.
llvm-svn: 186325
Daniel Jasper [Mon, 15 Jul 2013 15:04:42 +0000 (15:04 +0000)]
clang-format: Improve c-style cast detection.
Before:
#define x ((int) - 1)
#define p(q) ((int *) & q)
After:
#define x ((int)-1)
#define p(q) ((int *)&q)
llvm-svn: 186324
Marshall Clow [Mon, 15 Jul 2013 14:57:19 +0000 (14:57 +0000)]
Add macro _LIBCPP_CONSTEXPR_AFTER_CXX11 for functions that have been marked constexpr post C++11
llvm-svn: 186323
Daniel Jasper [Mon, 15 Jul 2013 14:33:14 +0000 (14:33 +0000)]
Improvement of change r186320.
Fixed a test that by now passed for the wrong reason.
Before:
llvm::outs() << "
aaaaaaaaaaaaaaaaaaa: " <<
aaaaaaaaaaaaa(
aaaaaaaaaaaaaaaaaaaaaaaaaaaa);
After:
llvm::outs() << "
aaaaaaaaaaaaaaaaaaa: "
<<
aaaaaaaaaaaaa(
aaaaaaaaaaaaaaaaaaaaaaaaaaaa);
Also reformatted Format.cpp with the latest changes (1 formatting fix
and 1 layout change of a <<-chain).
llvm-svn: 186322
Job Noorman [Mon, 15 Jul 2013 14:25:26 +0000 (14:25 +0000)]
Test commit to see if write access works.
llvm-svn: 186321
Daniel Jasper [Mon, 15 Jul 2013 14:12:30 +0000 (14:12 +0000)]
Reid Kleckner [Mon, 15 Jul 2013 13:46:24 +0000 (13:46 +0000)]
[Option] Store arg strings in a set backed by a BumpPtrAllocator
No functionality change.
This is preparing to move response file parsing into lib/Option so it
can be shared between clang and lld. This change isn't just a
micro-optimization. Clang's driver uses a std::set<std::string> to
unique arguments while parsing response files, so this matches that.
llvm-svn: 186319
Aaron Ballman [Mon, 15 Jul 2013 13:41:33 +0000 (13:41 +0000)]
Replacing an empty switch with its moral equivalent. No functional changes intended.
llvm-svn: 186318
Rafael Espindola [Mon, 15 Jul 2013 12:18:30 +0000 (12:18 +0000)]
XFAIL this on freebsd to bring the bot back.
Joerg Sonnenberger tells me one can open a directory in freebsd. I will try
to centralize our calls to open so that we can handle O_BINARY in one place,
and will then handle this there too.
llvm-svn: 186317
Chandler Carruth [Mon, 15 Jul 2013 10:30:19 +0000 (10:30 +0000)]
Reimplement SROA yet again. Same fundamental principle, but a totally
different core implementation strategy.
Previously, SROA would build a relatively elaborate partitioning of an
alloca, associate uses with each partition, and then rewrite the uses of
each partition in an attempt to break apart the alloca into chunks that
could be promoted. This was very wasteful in terms of memory and compile
time because regardless of how complex the alloca or how much we're able
to do in breaking it up, all of the datastructure work to analyze the
partitioning was done up front.
The new implementation attempts to form partitions of the alloca lazily
and on the fly, rewriting the uses that make up that partition as it
goes. This has a few significant effects:
1) Much simpler data structures are used throughout.
2) No more double walk of the recursive use graph of the alloca, only
walk it once.
3) No more complex algorithms for associating a particular use with
a particular partition.
4) PHI and Select speculation is simplified and happens lazily.
5) More precise information is available about a specific use of the
alloca, removing the need for some side datastructures.
Ultimately, I think this is a much better implementation. It removes
about 300 lines of code, but arguably removes more like 500 considering
that some code grew in the process of being factored apart and cleaned
up for this all to work.
I've re-used as much of the old implementation as possible, which
includes the lion's share of code in the form of the rewriting logic.
The interesting new logic centers around how the uses of a partition are
sorted, and split into actual partitions.
Each instruction using a pointer derived from the alloca gets
a 'Partition' entry. This name is totally wrong, but I'll do a rename in
a follow-up commit as there is already enough churn here. The entry
describes the offset range accessed and the nature of the access. Once
we have all of these entries we sort them in a very specific way:
increasing order of begin offset, followed by whether they are
splittable uses (memcpy, etc), followed by the end offset or whatever.
Sorting by splittability is important as it simplifies the collection of
uses into a partition.
Once we have these uses sorted, we walk from the beginning to the end
building up a range of uses that form a partition of the alloca.
Overlapping unsplittable uses are merged into a single partition while
splittable uses are broken apart and carried from one partition to the
next. A partition is also introduced to bridge splittable uses between
the unsplittable regions when necessary.
I've looked at the performance PRs fairly closely. PR15471 no longer
will even load (the module is invalid). Not sure what is up there.
PR15412 improves by between 5% and 10%, however it is nearly impossible
to know what is holding it up as SROA (the entire pass) takes less time
than reading the IR for that test case. The analysis takes the same time
as running mem2reg on the final allocas. I suspect (without much
evidence) that the new implementation will scale much better however,
and it is just the small nature of the test cases that makes the changes
small and noisy. Either way, it is still simpler and cleaner I think.
llvm-svn: 186316
Alexey Samsonov [Mon, 15 Jul 2013 08:43:35 +0000 (08:43 +0000)]
DebugInfo: Factor out parsing compile unit DIEs to a separate function. Improve code style and comments.
No functionality change.
llvm-svn: 186315
Craig Topper [Mon, 15 Jul 2013 08:24:27 +0000 (08:24 +0000)]
Add 'static' and 'const' qualifiers to some arrays of strings.
llvm-svn: 186314
Craig Topper [Mon, 15 Jul 2013 08:02:13 +0000 (08:02 +0000)]
Add 'const' qualifier to some arrays.
llvm-svn: 186312
Craig Topper [Mon, 15 Jul 2013 07:22:00 +0000 (07:22 +0000)]
Make some arrays 'static const'
llvm-svn: 186311
Craig Topper [Mon, 15 Jul 2013 07:15:05 +0000 (07:15 +0000)]
Add include to hopefully fix windows build.
llvm-svn: 186310
Craig Topper [Mon, 15 Jul 2013 07:02:45 +0000 (07:02 +0000)]
Add const qualifier to some static arrays.
llvm-svn: 186309
Craig Topper [Mon, 15 Jul 2013 06:54:12 +0000 (06:54 +0000)]
Add 'static' keyword to some const arrays for consistency.
llvm-svn: 186308
Craig Topper [Mon, 15 Jul 2013 06:39:13 +0000 (06:39 +0000)]
Make some arrays 'static const'
llvm-svn: 186307
Serge Pavlov [Mon, 15 Jul 2013 06:14:07 +0000 (06:14 +0000)]
Fix to PR12262 - assertion when substituting explicit template arguments
does not substitute a sizeof-pack expression.
The solution is proposed by Richard Smith.
Differential Revision: http://llvm-reviews.chandlerc.com/D869
llvm-svn: 186306
Craig Topper [Mon, 15 Jul 2013 05:44:18 +0000 (05:44 +0000)]
Revert 186302 to fix build bots.
llvm-svn: 186305
Jason Molenda [Mon, 15 Jul 2013 04:47:33 +0000 (04:47 +0000)]
Bump version number to 310.99.0.
llvm-svn: 186304
Craig Topper [Mon, 15 Jul 2013 04:37:54 +0000 (04:37 +0000)]
Revert part of 186302 to fix buildbots.
llvm-svn: 186303
Craig Topper [Mon, 15 Jul 2013 04:28:48 +0000 (04:28 +0000)]
Use llvm::array_lengthof to replace sizeof(array)/sizeof(array[0]).
llvm-svn: 186302
Craig Topper [Mon, 15 Jul 2013 04:27:47 +0000 (04:27 +0000)]
Use llvm::array_lengthof to replace sizeof(array)/sizeof(array[0]).
llvm-svn: 186301
Craig Topper [Mon, 15 Jul 2013 03:38:40 +0000 (03:38 +0000)]
Use llvm::array_lengthof to replace sizeof(array)/sizeof(array[0]).
llvm-svn: 186300
Jason Molenda [Mon, 15 Jul 2013 03:25:21 +0000 (03:25 +0000)]
Commit Timothee Besset's patch to update Host/linux/Host.cpp to keep up
with the changes in r186211.
llvm-svn: 186299
NAKAMURA Takumi [Mon, 15 Jul 2013 03:04:13 +0000 (03:04 +0000)]
Mark llvm/test/Object/extract.ll as XFAIL:mingw32, for now.
FIXME: Investigate Win32's TimeValue stuff!
llvm-svn: 186298