Aaron Ballman [Mon, 11 May 2015 14:09:50 +0000 (14:09 +0000)]
Disable __has_cpp_attribute when not compiling in C++ mode. As this feature test macro only supports C++ style attributes, it doesn't apply to code compiled as C code, and can lead to diagnostics when given a scoped attribute.
Daniel Jasper [Mon, 11 May 2015 13:35:40 +0000 (13:35 +0000)]
clang-format: Improve column layout.
Specifically, calculate the deviation between the shortest and longest
element (which is used to prevent excessive whitespace) per column, not
overall. This automatically handles the corner cases of a single column
and a single row so that the actualy implementation becomes simpler.
Manuel Klimek [Mon, 11 May 2015 08:21:35 +0000 (08:21 +0000)]
Refactor the formatter of clang-format.
Pull various parts of the UnwrappedLineFormatter into their own
abstractions. NFC.
There are two things left for subsequent changes (to keep this
reasonably small)
- the UnwrappedLineFormatter now has a bad name
- the UnwrappedLineFormatter::format function is still too large
David Majnemer [Sun, 10 May 2015 21:48:08 +0000 (21:48 +0000)]
[MS ABI] Form member pointers from virtual funcs overriding vbases
We didn't supporting taking the address of virtual member functions
which overrode a method in a virtual base. We simply need to encode the
virtual base index in the member pointer.
This fixes PR23452.
N.B. There is no data member pointer side to this change because taking
the address of a virtual bases' data member gives you a member pointer
whose type is derived from the virtual bases' type, not the most derived
type.
David Majnemer [Sun, 10 May 2015 21:38:26 +0000 (21:38 +0000)]
[MS ABI] Update EH emission for MSVC 2015 compatibility
MSVC 2015 renamed the symbol found by name lookup for 'std::terminate'
so we cannot rely on using '?terminate@@YAXXZ'. Furthermore, it seems
that 2015 will be the first release of MSVC which permits inlining a
function which is noexcept into a function which isn't. This is
implemented by creating a cleanup for the invoker which jumps to
__std_terminate. Clang's implementation of this aspect of the MSVC
scheme is slightly less efficient in this respect because we use a
catch handler configured as a catch-all handler instead.
Craig Topper [Sun, 10 May 2015 18:40:12 +0000 (18:40 +0000)]
De-virtualize some const versions of getCanonicalDecl by redirecting to the non-const version. Most of the Decl hierarchy already did it this way this just makes the rest consistent.
Tobias Grosser [Fri, 8 May 2015 21:34:09 +0000 (21:34 +0000)]
Revert "Make emacs show when clang-format encountered a syntax error."
This reverts commit 236854, which caused clang-format to always print
'{ "IncompleteFormat": false }' at the top of an incompletely formatted file.
This output causes problems e.g. in Polly's automatic formatting checks. Daniel
tried to fix this in 236867, but this fix had to be reverted due to buildbot
failures. I revert this change as well for now as it is Friday night and
unlikely to be fixed immediately.
Renato Golin [Fri, 8 May 2015 21:04:50 +0000 (21:04 +0000)]
Using ARMTargetParser in Clang
This is a starting point for using the TargetParser in Clang, in a simple
enough part of the code that can be used without disrupting the crazy
platform support that we need to be compatible with other toolchains.
Also adding a few FIXME on obvious places that need replacing, but those
cases will indeed break a few of the platform assumptions, as arch/cpu names
change multiple times in the driver.
Finally, I'm changing the "neon-vfpv3" behaviour to match standard NEON, since
-mfpu=neon implies vfpv3 by default in both Clang and LLVM. That option
string is still supported as an alias to "neon".
Diego Novillo [Fri, 8 May 2015 20:59:56 +0000 (20:59 +0000)]
Fix BackendConsumer::EmitOptimizationMessage()
Patch from Geoff Berry <gberry@codeaurora.org>
Fix BackendConsumer::EmitOptimizationMessage() to check if the
DiagnosticInfoOptimizationBase object has a valid location before
calling getLocation() to avoid dereferencing a null pointer inside
getLocation() when no debug info is present.
Derek Schuff [Fri, 8 May 2015 16:47:21 +0000 (16:47 +0000)]
Do not emit thunks with available_externally linkage in comdats
Functions with available_externally linkage will not be emitted to object
files (they will just be undefined symbols), so it does not make sense to
put them in comdats.
Creates a second overload of maybeSetTrivialComdat that uses the GlobalObject
instead of the Decl, and uses that in several places that had the faulty
logic.
Daniel Jasper [Fri, 8 May 2015 13:51:14 +0000 (13:51 +0000)]
clang-format: Several improvements around formatting braced lists.
In particular:
* If the difference between the longest and shortest element, we copped
out of column format completely. Now, we instead allow to arrange
these in a single column, essentially enforcing a one-per-line format.
* Allow column layout even if there are braced lists. Especially, if
there are many short lists, this can be beneficial. The bad case,
where there is a long nested init list is usually caught as we now
limit the length difference of the longest and shortest element.
John Brawn [Fri, 8 May 2015 12:52:18 +0000 (12:52 +0000)]
[ARM] Give an error on invalid -march values
llvm::Triple::getARMCPUForArch now returns nullptr for invalid -march
values, instead of silently translating it to arm7tdmi. Use this to
give an error message, which is consistent with how gcc behaves.
Alexey Bataev [Fri, 8 May 2015 11:47:16 +0000 (11:47 +0000)]
[OPENMP] Fixed atomic construct with non-integer expressions.
Do not emit 'atomicrmw' instruction for simple atomic constructs with non-integer expressions.
Extra features are:
* -fsanitize-coverage=indirect-calls - coverage for indirect calls
* -fsanitize-coverage=trace-bb - tracing for basic blocks
* -fsanitize-coverage=trace-cmp - tracing for cmp instructions
* -fsanitize-coverage=8bit-counters - frequency counters
Levels and features can be combined in comma-separated list, and
can be disabled by subsequent -fno-sanitize-coverage= flags, e.g.:
-fsanitize-coverage=bb,trace-bb,8bit-counters -fno-sanitize-coverage=trace-bb
is equivalient to:
-fsanitize-coverage=bb,8bit-counters
Original semantics of -fsanitize-coverage flag is preserved:
* -fsanitize-coverage=0 disables the coverage
* -fsanitize-coverage=1 is a synonym for -fsanitize-coverage=func
* -fsanitize-coverage=2 is a synonym for -fsanitize-coverage=bb
* -fsanitize-coverage=3 is a synonym for -fsanitize-coverage=edge
* -fsanitize-coverage=4 is a synonym for -fsanitize-coverage=edge,indirect-calls
Driver tries to diagnose invalid flag usage, in particular:
* At most one level (func,bb,edge) must be specified.
* "trace-bb" and "8bit-counters" features require some level to be specified.
Artem Belevich [Thu, 7 May 2015 19:34:16 +0000 (19:34 +0000)]
[cuda] Include GPU binary into host object file and generate init/deinit code.
- added -fcuda-include-gpubinary option to incorporate results of
device-side compilation into host-side one.
- generate code to register GPU binaries and associated kernels
with CUDA runtime and clean-up on exit.
- added test case for init/deinit code generation.
James Dennett [Thu, 7 May 2015 18:48:18 +0000 (18:48 +0000)]
Replace the broken LambdaCapture::isInitCapture API.
A LambdaCapture does not have sufficient information
to correctly determine whether it is an init-capture or not.
Doing so requires knowledge held in the LambdaExpr itself.
It the case of a nested capture of an init-capture it is not
sufficient to check (as LambdaCapture::isInitCapture did)
whether the associated VarDecl was from an init-capture.
This patch moves isInitCapture to LambdaExpr and updates
Capture->isInitCapture() to Lambda->isInitCapture(Capture).
Alexey Bataev [Thu, 7 May 2015 06:28:46 +0000 (06:28 +0000)]
Fix for http://llvm.org/PR23392: magick/feature.c from ImageMagick-6.9.1-2 ICEs.
Fix for codegen of static variables declared inside of captured statements. Captured statements are actually a transparent DeclContexts, so we have to skip them when trying to get a mangled name for statics.
Differential Revision: http://reviews.llvm.org/D9522
David Majnemer [Thu, 7 May 2015 06:15:46 +0000 (06:15 +0000)]
[MS ABI] Implement thread-safe initialization using the MSVC 2015 ABI
The MSVC 2015 ABI utilizes a rather straightforward adaptation of the
algorithm found in the appendix of N2382. While we are here, implement
support for emitting cleanups if an exception is thrown while we are
intitializing a static local variable.
Alexey Bataev [Thu, 7 May 2015 04:25:17 +0000 (04:25 +0000)]
[OPENMP] Generate !llvm.mem.loop_parallel_access metadata for loops with dynamic/guided scheduling.
Inner bodies of OpenMP worksharing loop-based constructs with dynamic or guided scheduling are allowed to be marked with !llvm.mem.parallel_loop_access metadata for better optimization. Worksharing constructs with static scheduling cannot be marked this way (according to OpenMP standard "A data dependence between the same logical iterations in two such loops is guaranteed").
Constructs with auto and runtime scheduling are also not marked because automatically chosen scheduling may be static also.
Differential Revision: http://reviews.llvm.org/D9518
Alexey Bataev [Thu, 7 May 2015 03:54:03 +0000 (03:54 +0000)]
[OPENMP] Fixed codegen for 'reduction' clause.
Fixed codegen for reduction operations min, max, && and ||. Codegen for them is quite similar and I was confused by this similarity.
Also added a call to kmpc_end_reduce() in atomic part of reduction codegen (call to kmpc_end_reduce_nowait() is not required).
Differential Revision: http://reviews.llvm.org/D9513
Kaelyn Takata [Thu, 7 May 2015 00:11:02 +0000 (00:11 +0000)]
When performing delayed typo correction in a for-range loop's variable
declaration, ensure the loop variable is properly marked as invalid when
it is an "auto" variable.
All callers should be passing `CXXConstructorDecl` or
`CXXDestructorDecl` here, so use `cast<>` instead of `dyn_cast<>` when
setting up the `GlobalDecl`.
Note that this might now violate the column limit and we probably need an
alternative way of indenting these then. However, that is still strictly better
than the messy formatting that clang-format did before.
Daniel Jasper [Wed, 6 May 2015 11:16:43 +0000 (11:16 +0000)]
clang-format: Fix another assertion discovered by the fuzzer.
In the process, fix an old todo that I don't really know how to write
tests for. The problem is that Clang's lexer creates very strange token
sequences for these. However, the new approach seems generally better
and easier to read so I am submitting it nonetheless.
Ahmed Bougacha [Wed, 6 May 2015 02:08:27 +0000 (02:08 +0000)]
Revert "[analyzer] scan-build: support spaces in compiler path and arguments."
This reverts commit r236423 and its followup r236533, as indiscriminate
quoting makes for too much quoting (and clang doesn't like both '"-c"'
and -D"FOO=bar").
Justin Bogner [Tue, 5 May 2015 21:46:14 +0000 (21:46 +0000)]
InstrProf: Don't start or end coverage regions inside of system macros
It doesn't make much sense to try to show coverage inside system
macros, and source locations in builtins confuses the coverage
mapping. Just avoid doing this.
Fixes an assert that fired when a __block storage specifier starts a
region.
Ulrich Weigand [Tue, 5 May 2015 19:36:42 +0000 (19:36 +0000)]
[SystemZ] Add support for z13 low-level vector builtins
This adds low-level builtins to allow access to all of the z13 vector
instructions. Note that instructions whose semantics can be described
by standard C (including clang extensions) do not get any builtins.
For each instructions whose semantics *cannot* (fully) be described, we
define a builtin named __builtin_s390_<insn> that directly maps to this
instruction. These are intended to be compatible with GCC.
For instructions that also set the condition code, the builtin will take
an extra argument of type "int *" at the end. The integer pointed to by
this argument will be set to the post-instruction CC value.
For many instructions, the low-level builtin is mapped to the corresponding
LLVM IR intrinsic. However, a number of instructions can be represented
in standard LLVM IR without requiring use of a target intrinsic.
Some instructions require immediate integer operands within a certain
range. Those are verified at the Sema level.
Ulrich Weigand [Tue, 5 May 2015 19:35:52 +0000 (19:35 +0000)]
[SystemZ] Add support for z13 and its vector facility
This patch adds support for the z13 architecture type. For compatibility
with GCC, a pair of options -mvx / -mno-vx can be used to selectively
enable/disable use of the vector facility.
When the vector facility is present, we default to the new vector ABI.
This is characterized by two major differences:
- Vector types are passed/returned in vector registers
(except for unnamed arguments of a variable-argument list function).
- Vector types are at most 8-byte aligned.
The reason for the choice of 8-byte vector alignment is that the hardware
is able to efficiently load vectors at 8-byte alignment, and the ABI only
guarantees 8-byte alignment of the stack pointer, so requiring any higher
alignment for vectors would require dynamic stack re-alignment code.
However, for compatibility with old code that may use vector types, when
*not* using the vector facility, the old alignment rules (vector types
are naturally aligned) remain in use.
These alignment rules are not only implemented at the C language level,
but also at the LLVM IR level. This is done by selecting a different
DataLayout string depending on whether the vector ABI is in effect or not.
Kaelyn Takata [Tue, 5 May 2015 19:17:03 +0000 (19:17 +0000)]
Allow TransformTypos to ignore corrections to a specified VarDecl.
This is needed to prevent a TypoExpr from being corrected to a variable
when the TypoExpr is a subexpression of that variable's initializer.
Also exclude more keywords from the correction candidate pool when the
subsequent token is .* or ->* since keywords like "new" or "return"
aren't valid on the left side of those operators.
projects / clang / log