Richard Smith [Fri, 3 Oct 2014 00:31:35 +0000 (00:31 +0000)]
Fix interaction of max_align_t and modules.
When building with modules enabled, we were defining max_align_t as a typedef
for a different anonymous struct type each time it was included, resulting in
an error if <stddef.h> is not covered by a module map and is included more than
once in the same modules-enabled compilation of C11 or C++11 code.
Patch to warn if 'override' is missing
for an overriding method if class has at least one
'override' specified on one of its methods.
Reviewed by Doug Gregor. rdar://18295240
(I have already checked in all llvm files with missing 'override'
methods and Bob Wilson has fixed a TableGen of FastISel so
no warnings are expected from build of llvm after this patch.
I have already verified this).
Hal Finkel [Thu, 2 Oct 2014 21:21:25 +0000 (21:21 +0000)]
Initial support for the align_value attribute
This adds support for the align_value attribute. This attribute is supported by
Intel's compiler (versions 14.0+), and several of my HPC users have requested
support in Clang. It specifies an alignment assumption on the values to which a
pointer points, and is used by numerical libraries to encourage efficient
generation of vector code.
Of course, we already have an aligned attribute that can specify enhanced
alignment for a type, so why is this additional attribute important? The
problem is that if you want to specify that an input array of T is, say,
64-byte aligned, you could try this:
typedef double aligned_double attribute((aligned(64)));
void foo(aligned_double *P) {
double x = P[0]; // This is fine.
double y = P[1]; // What alignment did those doubles have again?
}
the access here to P[1] causes problems. P was specified as a pointer to type
aligned_double, and any object of type aligned_double must be 64-byte aligned.
But if P[0] is 64-byte aligned, then P[1] cannot be, and this access causes
undefined behavior. Getting round this problem requires a lot of awkward
casting and hand-unrolling of loops, all of which is bad.
With the align_value attribute, we can accomplish what we'd like in a well
defined way:
typedef double *aligned_double_ptr attribute((align_value(64)));
void foo(aligned_double_ptr P) {
double x = P[0]; // This is fine.
double y = P[1]; // This is fine too.
}
This attribute does not create a new type (and so it not part of the type
system), and so will only "propagate" through templates, auto, etc. by
optimizer deduction after inlining. This seems consistent with Intel's
implementation (thanks to Alexey for confirming the various Intel-compiler
behaviors).
As a final note, I would have chosen to call this aligned_value, not
align_value, for better naming consistency with the aligned attribute, but I
think it would be more useful to users to adopt Intel's name.
in response to this change, support for sync_fetch_and_nand (and
sync_nand_and_fetch) was removed in r99522 in order to avoid miscompiling code
depending on the old semantics. However, at this point:
1. Many years have passed, and the amount of code relying on the old
semantics is likely smaller.
2. Through the work of many contributors, all LLVM backends have been updated
such that "atomicrmw nand" provides the newer GCC 4.4+ semantics (this process
was complete July of 2014 (added to the release notes in r212635).
3. The lack of this intrinsic is now a needless impediment to porting codes
from GCC to Clang (I've now seen several examples of this).
It is true, however, that we still set GNUC_MINOR to 2 (corresponding to GCC
4.2). To compensate for this, and to address the original concern regarding
code relying on the old semantics, I've added a warning that specifically
details the fact that the semantics have changed and that we provide the newer
semantics.
Jan Wen Voung [Thu, 2 Oct 2014 16:56:57 +0000 (16:56 +0000)]
[x32/NaCl] Check if method pointers straddle an eightbyte to classify Hi
Summary:
Currently, with struct my_struct { int x; method_ptr y; };
a call to foo(my_struct s) may end up dropping the last 4 bytes
of the method pointer for x86_64 NaCl and x32.
When checking Has64BitPointers, also check if the method pointer
straddles an eightbyte boundary and classify Hi as well as Lo if needed.
Test Plan: test/CodeGenCXX/x86_64-arguments-nacl-x32.cpp
Summary: Commit r218863 broke this test case. This patch fixes it
by updating the expected output line. Should've been updated with
the original patch but for some reason it didn't fail during my
local make check.
David Blaikie [Wed, 1 Oct 2014 23:16:30 +0000 (23:16 +0000)]
Reduce the PR20399 test case.
I couldn't get something /really/ obvious, and I imagine Richard Smith
might be able to provide some text explaining the sequence of steps
that's demonstrated by these files - but at least it's a bit simpler
now.
Adrian Prantl [Wed, 1 Oct 2014 18:55:34 +0000 (18:55 +0000)]
Update CGDebugInfo to the updated API in LLVM.
Complex address expressions are no longer part of DIVariable, but
rather an extra argument to the debug intrinsics.
c++ error recovery. Build a valid AST when trying
to recover from parse error parsing the default
argument. Patch prevents crash after spewing 100s
of errors caused by someone who forgot to compile in c++11
mode. So no test. rdar://18508589
Adrian Prantl [Wed, 1 Oct 2014 17:55:09 +0000 (17:55 +0000)]
Update CGDebugInfo to the updated API in LLVM.
Complex address expressions are no longer part of DIVariable, but
rather an extra argument to the debug intrinsics.
Refactor Matcher<T> and DynTypedMatcher to reduce overhead of casts.
Summary:
This change introduces DynMatcherInterface and changes the internal
representation of DynTypedMatcher and Matcher<T> to use a generic
interface instead.
It removes unnecessary indirections and virtual function calls when
converting matchers by implicit and dynamic casts.
DynTypedMatcher now remembers the stricter type in the chain of casts
and checks it before calling into DynMatcherInterface.
This change improves our clang-tidy related benchmark by ~14%.
Also, it opens the door for more optimizations of this kind that are
coming in future changes.
As a side effect of removing these template instantiations, it also
speeds up compilation of Dynamic/Registry.cpp by ~17% and reduces the
number of
symbols generated by ~30%.
Oliver Stannard [Wed, 1 Oct 2014 09:03:02 +0000 (09:03 +0000)]
[ARM] Add support for Cortex-M7, FPv5-SP and FPv5-DP
The Cortex-M7 has 3 options for its FPU: none, FPv5-SP-D16 and
FPv5-DP-D16. FPv5 has the same instructions as FP-ARMv8, so it can be
modeled using the same target feature, and all double-precision
operations are already disabled by the fp-only-sp target features.
[OPENMP] Loop collapsing and codegen for 'omp simd' directive.
This patch implements collapsing of the loops (in particular, in
presense of clause 'collapse'). It calculates number of iterations N
and expressions nesessary to calculate the nested loops counters
values based on new iteration variable (that goes from 0 to N-1)
in Sema. It also adds Codegen for 'omp simd', which uses
(and tests) this feature.
Justin Bogner [Wed, 1 Oct 2014 03:33:52 +0000 (03:33 +0000)]
InstrProf: Avoid repeated linear searches in a hot path
When generating coverage regions, we were doing a linear search
through the existing regions in order to try to merge related ones.
Most of the time this would find what it was looking for in a small
number of steps and it wasn't a big deal, but in cases with many
regions and few mergeable ones this leads to an absurd compile time
regression.
This changes the coverage mapping logic to do a single sort and then
merge as we go, which is a bit simpler and about 100 times faster.
I've also added FIXMEs on a couple of behaviours that seem a little
suspect, while keeping them behaving as they were - I'll look into
these soon.
The test changes here are mostly tedious reorganization, because the
ordering of regions we output has become slightly (but not completely)
more consistent from the almost completely arbitrary ordering we got
before.
This struct has some members that are accessed directly and others
that need accessors, but it's all just public. This is confusing, so
I've changed it to a class and made more members private.
Richard Smith [Tue, 30 Sep 2014 23:10:19 +0000 (23:10 +0000)]
Enable both C and C++ modules with -fmodules, by switching -fcxx-modules to
being on by default. -fno-cxx-modules can still be used to enable C modules but
not C++ modules, but C++ modules is not significantly less stable than C
modules any more.
Also remove some of the scare words from the modules documentation. We're
certainly not going to remove modules support (though we might change the
interface), and it works well enough to bootstrap and build lots of
non-trivial code.
Note that this does not represent a commitment to the current interface nor
implementation, and we still intend to follow whatever direction the C and C++
committees take regarding modules support.
Ben Langmuir [Tue, 30 Sep 2014 20:00:18 +0000 (20:00 +0000)]
Avoid a crash after loading an #undef'd macro in code completion
In code-completion, don't assume there is a MacroInfo for everything,
since we aren't serializing the def corresponding to a later #undef in
the same module. Also setup the HadMacro bit correctly for undefs to
avoid an assertion failure.
Eli Bendersky [Tue, 30 Sep 2014 17:38:34 +0000 (17:38 +0000)]
CUDA: mark the target of implicit intrinsics properly
r218624 implemented target inference for implicit special members. However,
other entities can be implicit - for example intrinsics. These can not have
inference running on them, so they should be marked host device as before. This
is the safest and most flexible setting, since by construction these functions
don't invoke anything, and we'd like them to be invokable from both host and
device code. LLVM's intrinsics definitions (where these intrinsics come from in
the case of CUDA/NVPTX) have no notion of target, so both host and device
intrinsics can be supported this way.
David Majnemer [Tue, 30 Sep 2014 06:45:43 +0000 (06:45 +0000)]
MS ABI: Correct layout for empty records
Empty records do not always have size equivalent to their alignment.
They only do so when their alignment is at least as large as the minimum
empty struct size: 1 byte in C++ and 4 bytes in C.
Revert r218616, "Refactor Matcher<T> and DynTypedMatcher to reduce overhead of casts."
MSC17, aka VS2012, cannot compile it.
clang/include/clang/ASTMatchers/ASTMatchersInternal.h(387) : error C4519: default template arguments are only allowed on a class template
clang/include/clang/ASTMatchers/ASTMatchersInternal.h(443) : see reference to class template instantiation 'clang::ast_matchers::internal::Matcher<T>' being compiled
Hans Wennborg [Mon, 29 Sep 2014 23:06:57 +0000 (23:06 +0000)]
Don't trap when passing non-POD arguments to variadic functions in MS-compatibility mode
Clang warns (treated as error by default, but still ignored in system headers)
when passing non-POD arguments to variadic functions, and generates a trap
instruction to crash the program if that code is ever run.
Unfortunately, MSVC happily generates code for such calls without a warning,
and there is code in system headers that use it.
This makes Clang not insert the trap instruction when in -fms-compatibility
mode, while still generating the warning/error message.
Use ClangToLLVMArgsMapping in CodeGenTypes::GetFunctionType(). NFC.
This is the last piece of CGCall code that had implicit assumptions about
the order in which Clang arguments are translated to LLVM ones (positions
of inalloca argument, sret, this, padding arguments etc.) Now all of
this data is encapsulated in ClangToLLVMArgsMapping. If this information
would be required somewhere else, this class can be moved to a separate
header or pulled into CGFunctionInfo.
Eli Bendersky [Mon, 29 Sep 2014 20:38:29 +0000 (20:38 +0000)]
CUDA: Fix incorrect target inference for implicit members.
As PR20495 demonstrates, Clang currenlty infers the CUDA target (host/device,
etc) for implicit members (constructors, etc.) incorrectly. This causes errors
and even assertions in Clang when compiling code (assertions in C++11 mode where
implicit move constructors are added into the mix).
Fix the problem by inferring the target from the methods the implicit member
should call (depending on its base classes and fields).
Add a method to calculate the number of arguments given QualType
expnads to. Use this method in ClangToLLVMArgMapping calculation.
This number may be cached in CodeGenTypes for efficiency, if needed.
Samuel Benzaquen [Mon, 29 Sep 2014 18:43:20 +0000 (18:43 +0000)]
Refactor Matcher<T> and DynTypedMatcher to reduce overhead of casts.
Summary:
This change introduces DynMatcherInterface and changes the internal
representation of DynTypedMatcher and Matcher<T> to use a generic
interface instead.
It removes unnecessary indirections and virtual function calls when
converting matchers by implicit and dynamic casts.
DynTypedMatcher now remembers the stricter type in the chain of casts
and checks it before calling into DynMatcherInterface.
This change improves our clang-tidy related benchmark by ~14%.
Also, it opens the door for more optimizations of this kind that are
coming in future changes.
As a side effect of removing these template instantiations, it also
speeds up compilation of Dynamic/Registry.cpp by ~17% and reduces the number of
symbols generated by ~30%.
Hoist the logic which determines the way QualType is expanded
into a separate method. Remove a bunch of copy-paste and simplify
getTypesFromArgs() / ExpandTypeFromArgs() / ExpandTypeToArgs() methods.
David Majnemer [Sun, 28 Sep 2014 06:39:30 +0000 (06:39 +0000)]
CodeGen: Don't crash when initializing pointer-to-member fields in bases
Clang uses two types to talk about a C++ class, the
NonVirtualBaseLLVMType and the LLVMType. Previously, we would allow one
of these to be packed and the other not.
This is problematic. If both don't agree on a common subset of fields,
then routines like getLLVMFieldNo will point to the wrong field. Solve
this by copying the 'packed'-ness of the complete type to the
non-virtual subobject. For this to work, we need to take into account
the non-virtual subobject's size and alignment when we are computing the
layout of the complete object.
Ben Langmuir [Fri, 26 Sep 2014 22:42:23 +0000 (22:42 +0000)]
Ensure that all module build failures get diagnosed
Otherwise we can end up silently skipping an import. If we happen to be
building another module at the time, we may build a mysteriously broken
module and not know why it seems to be missing symbols.
Don't link in sanitizer runtimes if -nostdlib/-nodefaultlibs is provided.
It makes no sense to link in sanitizer runtimes in this case: the user
probably doesn't want to see any system/toolchain libs in his link if he
provides these flags, and the link will most likely fail anyway - as sanitizer
runtimes depend on libpthread, libdl, libc etc.
Also, see discussion in https://code.google.com/p/address-sanitizer/issues/detail?id=344
Ben Langmuir [Fri, 26 Sep 2014 15:27:29 +0000 (15:27 +0000)]
Fix an assertion failure trying to emit a trivial destructor in ObjC++
If a base class declares a destructor, we will add the implicit
destructor for the subclass in
ActOnFields -> AddImplicitlyDeclaredMembersToClass
But in Objective C++, we did not compute whether we have a trivial
destructor until after that in
CXXRecordDecl::completeDefinition()
This was leading to a mismatch between the class, which thought it had
no trivial destructor, and the CXXDestructorDecl, which considered
itself trivial. It turns out the reason we delayed setting this until
completeDefinition() was for a warning that has since been removed as
part of -Warc-abi, so we just do it eagerly now.
Alexander Musman [Fri, 26 Sep 2014 06:28:25 +0000 (06:28 +0000)]
Small fix for bug 18635.
(clang crashed in CodeGen in llvm::Module::getNamedValue on
thread_local std::unique_ptr<int>).
Differential Revision: http://reviews.llvm.org/D5353
Hal Finkel [Fri, 26 Sep 2014 05:04:30 +0000 (05:04 +0000)]
Support the assume_aligned function attribute
In addition to __builtin_assume_aligned, GCC also supports an assume_aligned
attribute which specifies the alignment (and optional offset) of a function's
return value. Here we implement support for the assume_aligned attribute by making
use of the @llvm.assume intrinsic.
We build a NestedNameSpecifier that records the CXXRecordDecl in which
__super appeared. Name lookup is performed in all base classes of the
recorded CXXRecordDecl. Use of __super is allowed only inside class and
member function scope.
David Majnemer [Thu, 25 Sep 2014 19:43:56 +0000 (19:43 +0000)]
MS ABI: Avoid hashing back reference keys in mangler
This patch replaces the back reference StringMap from the MS mangler
with a SmallVector of strings. My previous patches reduced the number of
hashes involved in back reference lookups, this one removes them
completely. The back reference map contains at most 10 entries, which
are likely to be of varying sizes and different initial subsequences,
and which can easily became huge (due to templates and namespaces).
The solution presented is the simplest possible one. Nevertheless, it's
enough to reduce compilation times for a particular test case from 11.1s
to 9s, versus 8.58s for the Itanium ABI. Possible further improvements
include using a sorted vector (carefully to not introduce an extra
comparison), storing the string contents in a common arena, and/or keep
the string storage in the context for reuse.
Richard Smith [Wed, 24 Sep 2014 23:55:00 +0000 (23:55 +0000)]
Fix handling of preincrement on bit-fields. This gives a bit-field in C++, but
we were failing to find that bit-field when performing integer promotions. This
brings us closer to following the standard, and closer to GCC.
In C, this change is technically a regression: we get bit-field promotions
completely wrong in C, promoting cases that are categorically not bit-field
designators. This change makes us do so slightly more consistently, though.
The attribute documentation now conforms to Aaron Ballman's renaming of the
thread safety attributes, as well as the new paper that is due to be published
in the conference on Source Code Analysis and Manipulation (SCAM 2014) later
this week. In addition, recent changes to the analysis, such as checking
of references and negative capabilities, are now documented.
Downgrade error about stdcall decls with no prototype to a warning
Fixes PR21027. The MIDL compiler produces code that does this.
If we wanted to improve the warning, I think we could do this:
void __stdcall f(); // Don't warn without -Wstrict-prototypes.
void g() {
f(); // Might warn, the user probably meant for f to take no args.
f(1, 2, 3); // Warn, we have no idea what args f takes.
f(1); // Error, this is insane, one of these calls is broken.
}
David Blaikie [Wed, 24 Sep 2014 17:01:27 +0000 (17:01 +0000)]
Include debug info for types referenced only via explicit cast expressions.
Most of the debug info emission is powered essentially from function
definitions - if we emit the definition of a function, we emit the types
of its parameters, the members of those types, and so on and so forth.
For types that aren't referenced even indirectly due to this - because
they only appear in temporary expressions, not in any named variable, we
need to explicitly emit/add them as is done here. This is not the only
case of such code, and we might want to consider handling "void
func(void*); ... func(new T());" (currently debug info for T is not
emitted) at some point, though GCC doesn't. There's a much broader
solution to these issues, but it's a lot of work for possibly marginal
gain (but might help us improve the default -fno-standalone-debug
behavior to be even more aggressive in some places). See the original
review thread for more details.
Patch by jyoti allur (jyoti.yalamanchili@gmail.com)!
David Majnemer [Wed, 24 Sep 2014 11:04:09 +0000 (11:04 +0000)]
Sema: Inherit the flexible array property from struct fields
A record which contains a flexible array member is itself a flexible
array member. A struct which contains such a record should also
consider itself to be a flexible array member.
Follow-up to r218292: Add more REVERTIBLE_TYPE_TRAITs.
r218292 reverted r197496 because it broke things. In addition to breaking
things, r197496 also made all traits starting with __is_ revertible.
Reinstantiate that part of r197496 because code out there (e.g. libc++) depends
on this behavior. Fixes PR21045.