David Majnemer [Thu, 6 Aug 2015 21:08:36 +0000 (21:08 +0000)]
[IR] Remove TerminateInst's "NameStr" argument
TerminateInst can't have a name because it doesn't produce a result. No
functionality change is intended, this is just a cleanup.
llvm-svn: 244276
David Majnemer [Thu, 6 Aug 2015 21:08:34 +0000 (21:08 +0000)]
PHIs don't need to be postprocessed
llvm-svn: 244275
David Majnemer [Thu, 6 Aug 2015 21:08:32 +0000 (21:08 +0000)]
Handle PHI nodes prefacing EH pads too
llvm-svn: 244274
David Majnemer [Thu, 6 Aug 2015 21:08:30 +0000 (21:08 +0000)]
handle phi nodes
llvm-svn: 244273
David Majnemer [Thu, 6 Aug 2015 21:07:55 +0000 (21:07 +0000)]
[WinEHPrepare] Add rudimentary support for the new EH instructions
Summary:
This adds somewhat basic preparation functionality including:
- Formation of funclets via coloring basic blocks.
- Cloning of polychromatic blocks to ensure that funclets have unique
program counters.
- Demotion of values used between different funclets.
- Some amount of cleanup once we have removed predecessors from basic
blocks.
- Verification that we are left with a CFG that makes some amount of
sense.
N.B. Arguments and numbering still need to be done.
Reviewers: rnk, JosephTremoulet
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11750
llvm-svn: 244272
Richard Smith [Thu, 6 Aug 2015 21:05:21 +0000 (21:05 +0000)]
[modules] Remove redundant lookups into non-primary DeclContexts. These were made unnecessary by r244192.
llvm-svn: 244271
Frederic Riss [Thu, 6 Aug 2015 21:05:06 +0000 (21:05 +0000)]
[dsymutil] Implement dSYM bundle creation
A dSYM bundle is a file hierarchy that looks slike this:
<bundle name>.dSYM/
Contents/
Info.plist
Resources/
DWARF/
<DWARF file(s)>
This is the default output mode of dsymutil.
llvm-svn: 244270
Frederic Riss [Thu, 6 Aug 2015 21:05:01 +0000 (21:05 +0000)]
[dsymutil] Add (unimplemented) --flat option
dsymutil should by default generate dSYM bundles which are filesystem
hierarchies containing the debug info and an additional Info.plist.
Currently llvm-dsymutil emits raw binaries containing the debug info.
This is what we call the 'flat mode'. Add a -f/-flat option that is
supposed to enable that flat mode, but don't wire it for now, only
pass it to the tests that will need it to stay functional once we
do bundle generation by default.
This basically makes this commit NFC and removes the noise from the
actual commit that adds support for bundle generation.
llvm-svn: 244269
Frederic Riss [Thu, 6 Aug 2015 21:04:55 +0000 (21:04 +0000)]
Thread premissions through sys::fs::create_director{y|ies}
llvm-svn: 244268
Reid Kleckner [Thu, 6 Aug 2015 21:01:32 +0000 (21:01 +0000)]
[WinEH] Add documentation motivating the new EH instructions
This adds documentation on how to use the new EH instructions added in
r243766.
Reviewers: majnemer, reames
Differential Revision: http://reviews.llvm.org/D11565
llvm-svn: 244267
David Majnemer [Thu, 6 Aug 2015 20:56:55 +0000 (20:56 +0000)]
[ItaniumCXXABI] Don't import RTTI data for classes with key functions
MinGW has some pretty strange behvaior around RTTI and
dllimport/dllexport:
- RTTI data is never imported
- RTTI data is only exported if the class has no key function.
llvm-svn: 244266
Sanjoy Das [Thu, 6 Aug 2015 20:43:41 +0000 (20:43 +0000)]
[IndVars] Fix PR24356.
Unsigned predicates increase or decrease agnostic of the signs of their
increments.
llvm-svn: 244265
Sanjoy Das [Thu, 6 Aug 2015 20:43:28 +0000 (20:43 +0000)]
[IndVars] Improved logging under DEBUG(); NFC.
Before this, we'd print the modified comparision in the "Simplified
comparison" case. That looked misleading.
llvm-svn: 244264
Renato Golin [Thu, 6 Aug 2015 20:34:51 +0000 (20:34 +0000)]
[ASAN/AArch64] Make sure aarch64 is not stable-runtime
llvm-svn: 244263
James Y Knight [Thu, 6 Aug 2015 20:26:32 +0000 (20:26 +0000)]
Convert a few classes over to use the new TrailingObjects helper.
This initial commit serves as an example -- the remainder of the
classes using pointer arithmetic for trailing objects will be
converted in subsequent changes.
Differential Revision: http://reviews.llvm.org/D11298
llvm-svn: 244262
James Y Knight [Thu, 6 Aug 2015 20:26:18 +0000 (20:26 +0000)]
Remove unused 'Owned' flag from TemplateArgumentList.
llvm-svn: 244261
Pete Cooper [Thu, 6 Aug 2015 20:22:46 +0000 (20:22 +0000)]
Convert a bunch of loops to foreach. NFC.
After r244074, we now have a successors() method to iterate over
all the successors of a TerminatorInst. This commit changes a bunch
of eligible loops to use it.
llvm-svn: 244260
Rui Ueyama [Thu, 6 Aug 2015 20:05:27 +0000 (20:05 +0000)]
Update comments.
llvm-svn: 244259
Rui Ueyama [Thu, 6 Aug 2015 19:57:21 +0000 (19:57 +0000)]
COFF: Update a comment.
llvm-svn: 244258
Tom Stellard [Thu, 6 Aug 2015 19:43:02 +0000 (19:43 +0000)]
AMDGPU/SI: Add Fiji support
Patch by: Alex Deucher
llvm-svn: 244255
Tom Stellard [Thu, 6 Aug 2015 19:28:38 +0000 (19:28 +0000)]
AMDGPU/SI: Add support for 32-bit immediate SMRD offsets on CI
Reviewers: arsenm
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11604
llvm-svn: 244254
Tom Stellard [Thu, 6 Aug 2015 19:28:30 +0000 (19:28 +0000)]
AMDGPU/SI: Use ComplexPatterns for SMRD addressing modes
Summary: This allows us to consolidate several of the TableGen patterns.
Reviewers: arsenm
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11602
llvm-svn: 244253
David Blaikie [Thu, 6 Aug 2015 19:23:33 +0000 (19:23 +0000)]
Fix memory leaks by avoiding extra manual dynamic allocation
Improvement to r244212.
llvm-svn: 244252
Nico Weber [Thu, 6 Aug 2015 19:21:25 +0000 (19:21 +0000)]
Fix nested CrashRecoveryContexts with LLVM_ENABLE_THREADS=OFF, allow them.
libclang uses a CrashRecoveryContext, and building a module does too. If a
module gets built through libclang, nested CrashRecoveryContexts are used. They
work fine with threads as things are stored in ThreadLocal variables, but in
LLVM_ENABLE_THREADS=OFF builds the two recovery contexts would write to the
same globals.
To fix, keep active CrashRecoveryContextImpls in a list and have the global
point to the innermost one, and do something similar for
tlIsRecoveringFromCrash.
Necessary (but not sufficient) for PR11974 and PR20325
http://reviews.llvm.org/D11770
llvm-svn: 244251
Kostya Serebryany [Thu, 6 Aug 2015 19:19:55 +0000 (19:19 +0000)]
[libFuzzer] move the mutators to public interface so that custom mutators may reuse these functions directly
llvm-svn: 244250
Rafael Espindola [Thu, 6 Aug 2015 19:12:03 +0000 (19:12 +0000)]
On freebsd it is possible to open a directory. Test with another error.
llvm-svn: 244249
Nico Rieck [Thu, 6 Aug 2015 19:10:45 +0000 (19:10 +0000)]
Rename inst_range() to instructions() for consistency. NFC
llvm-svn: 244248
Kit Barton [Thu, 6 Aug 2015 19:01:57 +0000 (19:01 +0000)]
Fix possible infinite loop in shrink wrapping when searching for save/restore
points.
There is an infinite loop that can occur in Shrink Wrapping while searching
for the Save/Restore points.
Part of this search checks whether the save/restore points are located in
different loop nests and if so, uses the (post) dominator trees to find the
immediate (post) dominator blocks. However, if the current block does not have
any immediate (post) dominators then this search will result in an infinite
loop. This can occur in code containing an infinite loop.
The modification checks whether the immediate (post) dominator is different from
the current save/restore block. If it is not, then the search terminates and the
current location is not considered as a valid save/restore point for shrink wrapping.
Phabricator: http://reviews.llvm.org/D11607
llvm-svn: 244247
Peter Collingbourne [Thu, 6 Aug 2015 19:00:42 +0000 (19:00 +0000)]
LibDriver: Replace references to lld-link2 with lld-link.
llvm-svn: 244246
Argyrios Kyrtzidis [Thu, 6 Aug 2015 18:46:36 +0000 (18:46 +0000)]
[Frontend] Fix crash when serializing diagnostics with really long text.
rdar://
21896690
llvm-svn: 244245
Quentin Colombet [Thu, 6 Aug 2015 18:44:34 +0000 (18:44 +0000)]
[Reassociation] Fix miscompile for va_arg arguments.
iisUnmovableInstruction() had a list of instructions hardcoded which are
considered unmovable. The list lacked (at least) an entry for the va_arg
and cmpxchg instructions.
Fix this by introducing a new Instruction::mayBeMemoryDependent()
instead of maintaining another instruction list.
Patch by Matthias Braun <matze@braunis.de>.
Differential Revision: http://reviews.llvm.org/D11577
rdar://problem/
22118647
llvm-svn: 244244
Peter Collingbourne [Thu, 6 Aug 2015 18:37:54 +0000 (18:37 +0000)]
test: Use lld-link instead of lld-link2 as the LTO linker on Windows.
The old COFF linker has been removed and lld-link now refers to the new linker.
llvm-svn: 244243
David Blaikie [Thu, 6 Aug 2015 18:29:32 +0000 (18:29 +0000)]
Fix memory ownership in the NeonEmitter by using values instead of pointers (smart or otherwise)
Improvement to the memory leak fix in 244196.
Address validity is required for the Intrinsic objects, but since the
collections only ever grow (no elements are removed), deque provides
sufficient guarantees (that the objects will never be reallocated/moved
around) for this use case.
llvm-svn: 244241
Alex Lorenz [Thu, 6 Aug 2015 18:26:36 +0000 (18:26 +0000)]
MIR Parser: Report an error when parsing duplicate memory operand flags.
llvm-svn: 244240
Cong Hou [Thu, 6 Aug 2015 18:17:29 +0000 (18:17 +0000)]
Revert r244154 which causes some build failure. See https://llvm.org/bugs/show_bug.cgi?id=24377.
llvm-svn: 244239
Duncan P. N. Exon Smith [Thu, 6 Aug 2015 18:15:25 +0000 (18:15 +0000)]
Update docs for accessing !dbg attachments
llvm-svn: 244238
Renato Golin [Thu, 6 Aug 2015 18:15:17 +0000 (18:15 +0000)]
[ARM/Unwind] Fix wrong usage of write-back on register save
The register save routine in libunwind was using write-back addressing
mode to r0 for thumb, when that was not only different from the ARM
version and more importantly the register restore, but also saving the
wrong address.
Patch by Manuel Freiberger.
Fixes PR24331.
llvm-svn: 244237
Kit Barton [Thu, 6 Aug 2015 18:02:53 +0000 (18:02 +0000)]
This patch changes the interface to enable the shrink wrapping optimization.
It adds a new constructor, which takes a std::function predicate function that
is run at the beginning of shrink wrapping to determine whether the optimization
should run on the given machine function. The std::function can be overridden by
each target, allowing target-specific decisions to be made on each machine
function.
This is necessary for PowerPC, as the decision to run shrink wrapping is
partially based on the ABI. Futhermore, this operates nicely with the GCC iFunc
capability, which allows option overrides on a per-function basis.
Phabricator: http://reviews.llvm.org/D11421
llvm-svn: 244235
Kostya Serebryany [Thu, 6 Aug 2015 17:52:54 +0000 (17:52 +0000)]
[sanitizer] 2-nd attempt. Add the flag handle_sigfpe that is default true to handle SIGFPE crashes same as SIGSEV crashes, patch by Karl Skomski. This time the test is enabled only on x86-64 (it broke on ARM)
llvm-svn: 244234
Chad Rosier [Thu, 6 Aug 2015 17:37:18 +0000 (17:37 +0000)]
[AArch64] Use a static function and other minor cleanup for readability. NFC.
llvm-svn: 244233
Reid Kleckner [Thu, 6 Aug 2015 17:17:44 +0000 (17:17 +0000)]
[cmake] Handle Dragonfly BSD like FreeBSD
Fixes build break reported in PR24358.
Patch by John Marino.
llvm-svn: 244232
Reid Kleckner [Thu, 6 Aug 2015 16:57:49 +0000 (16:57 +0000)]
docs: Fix typo in Phabricator.rst
Summary: Fix a small typo in Phabricator.rst: s/your/you.
Reviewers: rnk
Subscribers: rnk, llvm-commits
Differential Revision: http://reviews.llvm.org/D11585
llvm-svn: 244231
Alex Lorenz [Thu, 6 Aug 2015 16:55:53 +0000 (16:55 +0000)]
MIR Serialization: Serialize the 'invariant' machine memory operand flag.
llvm-svn: 244230
Richard Diamond [Thu, 6 Aug 2015 16:55:03 +0000 (16:55 +0000)]
Fix an alignment error in `llvm::expandAtomicRMWToCmpXchg` without breaking the build where X86 isn't enabled.
Summary: Divide the primitive size in bits by eight so the initial load's alignment is in bytes as expected. Tested with the included unit test.
Reviewers: rengolin, jfb
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11804
llvm-svn: 244229
Alex Lorenz [Thu, 6 Aug 2015 16:49:30 +0000 (16:49 +0000)]
MIR Serialization: Serialize the 'non-temporal' machine memory operand flag.
llvm-svn: 244228
Rui Ueyama [Thu, 6 Aug 2015 16:47:46 +0000 (16:47 +0000)]
COFF: Make test commands shorter. NFC.
llvm-svn: 244227
Rui Ueyama [Thu, 6 Aug 2015 16:19:35 +0000 (16:19 +0000)]
COFF: Remove the old COFF linker and make link an alias to link2.
It's time to remove old COFF linker because the new one is now complete.
llvm-svn: 244226
Chad Rosier [Thu, 6 Aug 2015 15:50:12 +0000 (15:50 +0000)]
[AArch64] Improve the readability of the ld/st optimization pass. NFC.
llvm-svn: 244222
Douglas Katzman [Thu, 6 Aug 2015 15:44:12 +0000 (15:44 +0000)]
[SPARC] Don't compare arch name as a string, use the enum instead.
Fixes PR22695
llvm-svn: 244221
Rafael Espindola [Thu, 6 Aug 2015 15:33:21 +0000 (15:33 +0000)]
Delete broken code.
This was using a hard coded string table and getting it for every symbol.
The symbol name was already available.
llvm-svn: 244220
Rafael Espindola [Thu, 6 Aug 2015 15:33:19 +0000 (15:33 +0000)]
bar
llvm-svn: 244219
Rafael Espindola [Thu, 6 Aug 2015 15:18:59 +0000 (15:18 +0000)]
Bring test back but don't check the filesytem dependent error.
llvm-svn: 244218
Jonathan Peyton [Thu, 6 Aug 2015 15:16:54 +0000 (15:16 +0000)]
Fix link error on Windows when LIBOMP_USE_DEBUGGER is off
Two symbols for the external debugger support were incorrectly exported when LIBOMP_USE_DEBUGGER=off.
Differential Revision: http://reviews.llvm.org/D11763
llvm-svn: 244217
Rafael Espindola [Thu, 6 Aug 2015 15:08:23 +0000 (15:08 +0000)]
Move the error handling functions to Error.h. NFC.
llvm-svn: 244216
Rafael Espindola [Thu, 6 Aug 2015 14:58:50 +0000 (14:58 +0000)]
Port the error functions from ELF to COFF.
This has a few advantages
* Less C++ code (about 300 lines less).
* Less machine code (about 14 KB of text on a linux x86_64 build).
* It is more debugger friendly. Just set a breakpoint on the exit function and
you get the complete lld stack trace of when the error was found.
* It is a more robust API. The errors are handled early and we don't get a
std::error_code hot potato being passed around.
* In most cases the error function in a better position to print diagnostics
(it has more context).
llvm-svn: 244215
Rafael Espindola [Thu, 6 Aug 2015 14:51:02 +0000 (14:51 +0000)]
Reuse variable. NFC.
llvm-svn: 244214
Aaron Ballman [Thu, 6 Aug 2015 14:07:29 +0000 (14:07 +0000)]
Silencing a -Wtype-limits warning; NFC.
llvm-svn: 244213
Yaron Keren [Thu, 6 Aug 2015 13:18:59 +0000 (13:18 +0000)]
Plug memory leaks in AsmWriterEmitter::EmitPrintAliasInstruction.
llvm-svn: 244212
Chad Rosier [Thu, 6 Aug 2015 12:49:40 +0000 (12:49 +0000)]
Fix minor typos. NFC.
llvm-svn: 244211
Renato Golin [Thu, 6 Aug 2015 12:42:46 +0000 (12:42 +0000)]
Revert "[sanitizer] Add the flag handle_sigfpe that is default true to handle SIGFPE crashes same as SIGSEV crashes, patch by Karl Skomski"
This reverts commit r244136, it was breaking the ARM bots for too long. We should investigate it offline.
llvm-svn: 244210
Alexey Bataev [Thu, 6 Aug 2015 12:30:57 +0000 (12:30 +0000)]
[OPENMP 4.1] Allow references in init expression for loop-based constructs.
OpenMP 4.1 allows to use variables with reference types in private clauses and, therefore, in init expressions of the cannonical loop forms.
llvm-svn: 244209
Aaron Ballman [Thu, 6 Aug 2015 12:15:47 +0000 (12:15 +0000)]
Correcting and adding tests for r244206.
llvm-svn: 244208
Michael Kuperstein [Thu, 6 Aug 2015 11:57:15 +0000 (11:57 +0000)]
Mark calls in thunk functions as tail-call optimization candidates
When a thunk is generated with a call to the original adjusted function,
the thunk appears in the debugger call stack. We want the backend to perform
tail-call optimization on the call, to make it invisible to the debugger.
This fixes PR24235
Patch by: amjad.aboud@intel.com
Differential Revision: http://reviews.llvm.org/D11476
llvm-svn: 244207
Aaron Ballman [Thu, 6 Aug 2015 11:56:57 +0000 (11:56 +0000)]
Add the "quit" command as a way to terminate clang-query interactive sessions.
llvm-svn: 244206
Renato Golin [Thu, 6 Aug 2015 10:37:59 +0000 (10:37 +0000)]
Revert "Divide the primitive size in bits by eight so the initial load's alignment is in bytes as expected. Tested with the included unit test."
This reverts commit r244155, as it was breaking the buildbots for too long.
Should be reapplied with proper fix.
llvm-svn: 244205
NAKAMURA Takumi [Thu, 6 Aug 2015 09:49:17 +0000 (09:49 +0000)]
llvm/lib/IR/AttributeImpl.h: Move comment block not to cover typedef, introduced in r244164. [-Wdocumentation]
llvm-svn: 244204
Michael Liao [Thu, 6 Aug 2015 09:06:20 +0000 (09:06 +0000)]
Removing tailing whitespaces
llvm-svn: 244203
Michael Kuperstein [Thu, 6 Aug 2015 08:45:34 +0000 (08:45 +0000)]
[X86] Improve EmitLoweredSelect for contiguous CMOV pseudo instructions.
This change improves EmitLoweredSelect() so that multiple contiguous CMOV pseudo
instructions with the same (or exactly opposite) conditions get lowered using a single
new basic-block. This eliminates unnecessary extra basic-blocks (and CFG merge points)
when contiguous CMOVs are being lowered.
Patch by: kevin.b.smith@intel.com
Differential Revision: http://reviews.llvm.org/D11428
llvm-svn: 244202
Michael Kuperstein [Thu, 6 Aug 2015 08:24:38 +0000 (08:24 +0000)]
[X86] Add support for _MM_ALIGN16
Differential Revision: http://reviews.llvm.org/D11753
llvm-svn: 244201
Chandler Carruth [Thu, 6 Aug 2015 08:17:06 +0000 (08:17 +0000)]
[PM/AA] Clean up and homogenize comments throughout basic-aa.
llvm-svn: 244200
Yaron Keren [Thu, 6 Aug 2015 07:59:26 +0000 (07:59 +0000)]
Fix Visual C++ error C2248:
'llvm::TrailingObjects<`anonymous-namespace'::Class1,short,llvm::NoTrailingTypeArg>::additionalSizeToAlloc' :
cannot access protected member declared in class
'llvm::TrailingObjects<`anonymous-namespace'::Class1,short,llvm::NoTrailingTypeArg>'
I'm not sure how this compiles with gcc.
Aren't protecteded members accessible only with protected or public inheritance?
llvm-svn: 244199
Chandler Carruth [Thu, 6 Aug 2015 07:57:58 +0000 (07:57 +0000)]
[PM/AA] Run clang-format over all of basic-aa before making more
substantive edits.
llvm-svn: 244198
Chandler Carruth [Thu, 6 Aug 2015 07:33:15 +0000 (07:33 +0000)]
[PM/AA] Hoist the interface for BasicAA into a header file.
This is the first mechanical step in preparation for making this and all
the other alias analysis passes available to the new pass manager. I'm
factoring out all the totally boring changes I can so I'm moving code
around here with no other changes. I've even minimized the formatting
churn.
I'll reformat and freshen comments on the interface now that its located
in the right place so that the substantive changes don't triger this.
llvm-svn: 244197
Yaron Keren [Thu, 6 Aug 2015 07:28:36 +0000 (07:28 +0000)]
Plug a memory leak in NeonEmitter: Intrinsics allocated were never released.
llvm-svn: 244196
Peter Collingbourne [Thu, 6 Aug 2015 05:26:35 +0000 (05:26 +0000)]
COFF: Assign the correct symbol type to internal functions.
The COFFSymbolRef::isFunctionDefinition() function tests for several conditions
that are not related to whether a symbol is a function, but rather whether
the symbol meets the requirements for a function definition auxiliary record,
which excludes certain symbols such as internal functions and undefined
references. The test we need to determine the symbol type is much simpler:
we only need to compare the complex type against IMAGE_SYM_DTYPE_FUNCTION.
llvm-svn: 244195
Richard Smith [Thu, 6 Aug 2015 05:13:41 +0000 (05:13 +0000)]
Update lldb's ExternalASTSources to match Clang r244161.
llvm-svn: 244194
Alex Denisov [Thu, 6 Aug 2015 04:51:14 +0000 (04:51 +0000)]
[ObjC] Circular containers: add support of subclasses
llvm-svn: 244193
Richard Smith [Thu, 6 Aug 2015 04:23:48 +0000 (04:23 +0000)]
[modules] Defer setting up the lookup table for a DeclContext until we can
determine the primary context, rather than sometimes registering the lookup
table on the wrong context.
This exposed a couple of bugs:
* the odr violation check didn't deal properly with mergeable declarations
if the declaration retained by name lookup wasn't in the canonical
definition of the class
* the (broken) RewriteDecl mechanism would emit two name lookup tables for
the same DeclContext into the same module file (one as part of the
rewritten declaration and one as a visible update for the old declaration)
These are both fixed too.
llvm-svn: 244192
Rui Ueyama [Thu, 6 Aug 2015 03:45:27 +0000 (03:45 +0000)]
COFF: ARM: Sort .pdata section correctly.
On ARM, exception handler entries in .pdata section are 8 byte long.
llvm-svn: 244191
Jason Molenda [Thu, 6 Aug 2015 03:27:10 +0000 (03:27 +0000)]
Second part of indicating when the user is stopped in optimized code.
The first part was in r243508 -- the extent of the UI changes in that
patchset was to add "[opt]" to the frame-format when a stack frame was
built with optimized code.
In this change, when a stack frame built with optimization is selected,
a message will be printed to the async output channel --
opt1.c was compiled with optimization - stepping may behave oddly; variables may not be available.
The warning will be only be printed once per source file in a debug session.
These warnings may be disabled by
settings set target.process.optimization-warnings false
Internally, a new Process::PrintWarning() method has been added for
warnings that we want to print only once to the user. It takes a type
of warning (currently only eWarningsOptimization) and an object
pointer (CompileUnit*) - the warning will only be printed once for a
given object pointer value.
This is a bit of a prototype of this change - I think we will be
tweaking it more in the future. But I wanted to land this and see
how it goes. Advanced users will find these warnings unnecessary
noise and will quickly disable them - but anyone who maintains a
debugger knows that debugging optimized code, without realizing it,
is a constant source of confusion and frustation for more typical
debugger users.
I imagine there will be more of these "warn once per whatever" style
warnings that we will want to add in the future and we'll need to
come up with a better way for enabling/disabling them. But I'm not
srue what form that warning settings should take and I didn't want
to code up something that we regret later, so for now I just added
another process setting for this one warning.
<rdar://problem/
19281172>
llvm-svn: 244190
Chandler Carruth [Thu, 6 Aug 2015 02:05:46 +0000 (02:05 +0000)]
[PM/AA] Simplify the AliasAnalysis interface by removing a wrapper
around a DataLayout interface in favor of directly querying DataLayout.
This wrapper specifically helped handle the case where this no
DataLayout, but LLVM now requires it simplifynig all of this. I've
updated callers to directly query DataLayout. This in turn exposed
a bunch of places where we should have DataLayout readily available but
don't which I've fixed. This then in turn exposed that we were passing
DataLayout around in a bunch of arguments rather than making it readily
available so I've also fixed that.
No functionality changed.
llvm-svn: 244189
Kostya Serebryany [Thu, 6 Aug 2015 01:29:13 +0000 (01:29 +0000)]
[libFuzzer] add one more mutation strategy: byte shuffling
llvm-svn: 244188
Argyrios Kyrtzidis [Thu, 6 Aug 2015 01:06:55 +0000 (01:06 +0000)]
[test] Another attempt to fix the test.
llvm-svn: 244187
Eric Christopher [Thu, 6 Aug 2015 01:01:12 +0000 (01:01 +0000)]
Rename the non-coding style conformant functions in namespace Builtins
to match the rest of their brethren and reformat the bits that need it.
llvm-svn: 244186
Alex Lorenz [Thu, 6 Aug 2015 00:44:07 +0000 (00:44 +0000)]
MIR Serialization: Initial serialization of the machine operand target flags.
This commit implements the initial serialization of the machine operand target
flags. It extends the 'TargetInstrInfo' class to add two new methods that help
to provide text based serialization for the target flags.
This commit can serialize only the X86 target flags, and the target flags for
the other targets will be serialized in the follow-up commits.
Reviewers: Duncan P. N. Exon Smith
llvm-svn: 244185
Davide Italiano [Thu, 6 Aug 2015 00:18:52 +0000 (00:18 +0000)]
[llvm-objdump] Add missing call to exit(1).
Reported by: Rafael Espindola.
llvm-svn: 244184
Rafael Espindola [Thu, 6 Aug 2015 00:00:50 +0000 (00:00 +0000)]
For completeness, add a 64 bit big endian test.
llvm-svn: 244183
Frederic Riss [Wed, 5 Aug 2015 23:53:38 +0000 (23:53 +0000)]
Revert "Make sure all temporary files get created under %T."
This reverts commit r244163. The workaround shouldn't be necessary
after r244172, and moreover the commit was slightly buggy as it
dis a simple mkdir without removing the directory first, which could
cause 'File exists' errors.
llvm-svn: 244182
Duncan P. N. Exon Smith [Wed, 5 Aug 2015 23:52:42 +0000 (23:52 +0000)]
ValueMapper: Rotate distinct node remapping algorithm
Rotate the algorithm for remapping distinct nodes in order to simplify
how uniquing cycles get resolved. This removes some of the recursion,
and, most importantly, exposes all uniquing cycles at the top-level.
Besides being a little more efficient -- temporary MDNodes won't live as
long -- the clearer logic should help protect against bugs like those
fixed in r243961 and r243976.
What are uniquing cycles? Why do they present challenges when remapping
metadata?
!0 = !{!1}
!1 = !{!0}
!0 and !1 form a simple uniquing cycle. When remapping from one
metadata graph to another, every uniquing cycle gets "duplicated"
through a dance:
!0-temp = !{!1?} ; map(!0): clone !0, VM[!0] = !0-temp
!1-temp = !{!0?} ; ..map(!1): clone !1, VM[!1] = !1-temp
!1-temp = !{!0-temp} ; ..map(!1): remap !1's operands
!2 = !{!0-temp} ; ..map(!1): uniquify: !1-temp => !2
!0-temp = !{!2} ; map(!0): remap !0's operands
!3 = !{!2} ; map(!0): uniquify: !0-temp => !3
; Result
!2 = !{!3}
!3 = !{!2}
(In the two "uniquify" steps above, the operands of !X-temp are compared
to the operands of !X. If they're the same, then !X-temp gets RAUW'ed
to !X; if they're different, then !X-temp is promoted to a new unique
node. The latter case always hits in for uniquing cycles, so we
duplicate all the nodes involved.)
Why is this a problem? Uniquable Metadata nodes that have temporary
node as transitive operands keep RAUW support until the temporary nodes
get finalized. With non-cycles, this happens automatically: when a
uniquable node's count of unresolved operands drops to zero, it
immediately sheds its own RAUW support (possibly triggering the same in
any node that references it). However, uniquing cycles create a
reference cycle, and uniqued nodes that transitively reference a
uniquing cycle are "stuck" in an unresolved state until someone calls
`MDNode::resolveCycles()` on a node in the unresolved subgraph.
Distinct nodes should help here (and mostly do): since they aren't
uniqued anywhere, they are guaranteed not to be RAUW'ed. They
effectively form a barrier between uniqued nodes, breaking some uniquing
cycles, and shielding uniqued nodes from uniquing cycles.
Unfortunately, with this barrier in place, the unresolved subgraph(s)
can be disjoint from the top-level node. The mapping algorithm needs to
find at least one representative from each disjoint subgraph. But which
nodes are *stuck*, and which will get resolved automatically? And which
nodes are in the unresolved subgraph? The old logic was conservative.
This commit rotates the logic for distinct nodes, so that we have access
to unresolved nodes at the top-level call to `llvm::MapMetadata()`.
Each time we return to the top-level, we know that all temporaries have
been RAUW'ed away. Here, it's safe (and necessary) to call
`resolveCycles()` immediately on unresolved operands.
This should also perform better than the old algorithm. The recursion
stack is shorter, temporary nodes don't live as long, and there are
fewer tracking references to unresolved nodes. As the debug info graph
introduces more 'distinct' nodes, remapping should incrementally get
cheaper and cheaper.
Aside from possible performance improvements (and reduced cruft in the
`LLVMContext`), there should be no functionality change here.
llvm-svn: 244181
Rui Ueyama [Wed, 5 Aug 2015 23:51:50 +0000 (23:51 +0000)]
COFF, ELF2: Pass output file path implicitly using Config global variable.
Various parameters are passed implicitly using Config global variable
already. Output file path is no different from others, so there was no
special reason to handle that differnetly.
This patch changes the signature of writeResult(SymbolTable *, StringRef)
to writeResult(SymbolTable *).
llvm-svn: 244180
Justin Bogner [Wed, 5 Aug 2015 23:49:44 +0000 (23:49 +0000)]
Driver: Strip -header-include-file and -diagnostic-log-file from crash repro scripts
These two arguments tend to refer to a local path that won't exist
when we try to reproduce a bug. Strip them.
llvm-svn: 244179
Eric Christopher [Wed, 5 Aug 2015 23:48:05 +0000 (23:48 +0000)]
Rename DescriptionString -> DataLayoutString as it matches the actual
use of the string.
llvm-svn: 244178
Kostya Serebryany [Wed, 5 Aug 2015 23:44:42 +0000 (23:44 +0000)]
[libFuzzer] avoid build warnings in non-assert build (useful warning in this case)
llvm-svn: 244177
Argyrios Kyrtzidis [Wed, 5 Aug 2015 23:44:06 +0000 (23:44 +0000)]
[test] Fix tests in non-darwin bots.
llvm-svn: 244176
Rui Ueyama [Wed, 5 Aug 2015 23:43:53 +0000 (23:43 +0000)]
COFF: Simplify Writer interface by hiding Writer class.
llvm-svn: 244175
Wei Mi [Wed, 5 Aug 2015 23:40:30 +0000 (23:40 +0000)]
Add a stat to show how often the limit to decompose GEPs in BasicAA is reached.
Differential Revision: http://reviews.llvm.org/D9689
llvm-svn: 244174
Rafael Espindola [Wed, 5 Aug 2015 23:40:20 +0000 (23:40 +0000)]
Add a big endian testcase.
llvm-svn: 244173
Frederic Riss [Wed, 5 Aug 2015 23:33:50 +0000 (23:33 +0000)]
[dsymutil] Do not create temporary files in -no-output mode.
The files were never written to and then deleted, but they were created
nonetheless. To prevent that, create a wrapper around the 2 variants of
createUniqueFile and use the one that only does an access(Exists) call
to check for name unicity in -no-output mode.
llvm-svn: 244172
Frederic Riss [Wed, 5 Aug 2015 23:33:45 +0000 (23:33 +0000)]
[dsymutil] Update source used to generate test binary.
Forgot to include that in the last commit.
llvm-svn: 244171
Diego Novillo [Wed, 5 Aug 2015 23:27:40 +0000 (23:27 +0000)]
Add -fno-coverage-mapping flag.
This new flag allows the user to disable a previous instance of
-fcoverage-mapping, if needed.
llvm-svn: 244170