Albert Gutowski [Tue, 4 Oct 2016 22:29:49 +0000 (22:29 +0000)]
Separate builtins for x84-64 and i386; implement __mulh and __umulh
Summary: We need x86-64-specific builtins if we want to implement some of the MS intrinsics - winnt.h contains definitions of some functions for i386, but not for x86-64 (for example _InterlockedOr64), which means that we cannot treat them as builtins for both i386 and x86-64, because then we have definitions of builtin functions in winnt.h on i386.
Hans Wennborg [Tue, 4 Oct 2016 21:00:57 +0000 (21:00 +0000)]
clang-cl: Expose the -flto option
We could hook up /GL as an alias for -flto, but that might be
confusing, as clang-cl in that mode would not be drop-in compatible
with cl.exe /GL, as it requires the linker to be lld.
Exposing -flto seems like a less confusing way to expose this
functionality.
Sanjay Patel [Tue, 4 Oct 2016 20:44:05 +0000 (20:44 +0000)]
[clang] make reciprocal estimate codegen a function attribute
The motivation for the change is that we can't have pseudo-global settings
for codegen living in TargetOptions because that doesn't work with LTO.
Ideally, these reciprocal attributes will be moved to the instruction-level
via FMF, metadata, or something else. But making them function attributes is
at least an improvement over the current state.
I'm committing this patch ahead of the related LLVM patch to avoid bot failures,
but if that patch needs to be reverted, then this should be reverted too.
Reid Kleckner [Tue, 4 Oct 2016 18:10:23 +0000 (18:10 +0000)]
Test what happens when tag lookup and redeclaration lookup disagree
Clang has a diagnostic for the what happens when an elaborated type
implicitly creates a tag declaration and the initial tag lookup fails,
but the redeclaration lookup succeeds and finds a non-tag type. However,
it wasn't tested, and looked like dead code. After much staring, we
discovered how to exercise it, and are now committing the test for
posterity.
In this example, the tag lookup will not find A, but then when we go to
insert a declaration of A at global scope, we discover the template
friend, which is not a tag type.
struct C {
template <typename> friend struct A;
};
struct B {
struct A *p;
};
Serge Pavlov [Tue, 4 Oct 2016 10:11:43 +0000 (10:11 +0000)]
Do not find friend function definitions inside non-instantiated class.
Previously if a file-level function was defined inside befriending
template class, it always was treated as defined. For instance, the code like:
```
int func(int x);
template<typename T> class C1 {
friend int func(int x) { return x; }
};
template<typename T> class C2 {
friend int func(int x) { return x; }
};
```
could not be compiled due to function redefinition, although not of the templates
is instantiated. Moreover, the body of friend function can contain use of template
parameters, attempt to get definition of such function outside any instantiation
causes compiler abnormal termination.
Other compilers (gcc, icc) follow viewpoint that the body of the function defined
in friend declaration becomes available when corresponding class is instantiated.
This patch implements this viewpoint in clang.
Definitions introduced by friend declarations in template classes are not added
to the redeclaration chain of corresponding function. Only when the template is
instantiated, instantiation of the function definition is placed to the chain.
The fix was made in collaboration with Richard Smith.
This change fixes PR8035, PR17923, PR22307 and PR25848.
Manuel Klimek [Tue, 4 Oct 2016 09:53:04 +0000 (09:53 +0000)]
Minor cleanups in clang-format.el.
- Enable lexical binding
- Formatting
- Enable file name completion for the clang-format-executable variable
- Add a missing docstring
- When available, use bufferpos-to-filepos and filepos-to-bufferpos. These functions given more precise mapping than byte-to-position and position-bytes.
- Rename arguments of clang-format-region to match the docstring
- Instead of binding local variables to nil and then assigning them, bind them directly to their values
- Make use of the fact that insert-file-contents returns the number of characters it inserted
- Use cl-destructuring-bind to make the code a bit shorter
- Use standard iteration (dolist) instead of mapc with a lambda, which is more common and shorter
- Remove a message that was most likely only present for debugging purposes
Manman Ren [Mon, 3 Oct 2016 21:26:46 +0000 (21:26 +0000)]
ObjectiveC: fix a seg fault when deserialing redeclaration of ObjCMethodDecl.
The deserialization of redeclartion can cause seg fault since getCanonicalDecl
of the redeclaration returns the lookup result on the ObjCContainerDecl,
which can be null if FindExternalVisibleDeclsByName is not done updating
the lookup results.
The fix is to return the redeclaration itself as the canonical decl. Note that
the handling for redeclaration of ObjCMethodDecl is not in line with other
redeclarables.
Justin Lebar [Mon, 3 Oct 2016 16:48:23 +0000 (16:48 +0000)]
[CUDA] Disallow overloading destructors.
Summary:
We'd attempted to allow this, but turns out we were doing a very bad
job. :)
Making this work properly would be a giant change in clang. For
example, we'd need to make CXXRecordDecl::getDestructor()
context-sensitive, because the destructor you end up with depends on
where you're calling it from.
For now (and hopefully for ever), just disallow overloading of
destructors in CUDA.
Vedant Kumar [Mon, 3 Oct 2016 15:29:22 +0000 (15:29 +0000)]
[ARC] Ignore qualifiers in copy-restore expressions
When ARC is enabled, an ObjCIndirectCopyRestoreExpr models the passing
of a function argument s.t:
* The argument is copied into a temporary,
* The temporary is passed into the function, and
* After the function call completes, the temporary is move-assigned
back to the original location of the argument.
The argument type and the parameter type must agree "except possibly in
qualification". This commit weakens an assertion in EmitCallArg() to
actually reflect that.
Yaxun Liu [Mon, 3 Oct 2016 14:41:50 +0000 (14:41 +0000)]
[OpenCL] Fix bug in __builtin_astype causing invalid LLVM cast instructions
__builtin_astype is used to cast OpenCL opaque types to other types, as such, it needs to be able to handle casting from and to pointer types correctly.
Current it cannot handle 1) casting between pointers of different addr spaces 2) casting between pointer type and non-pointer types.
Alex Lorenz [Mon, 3 Oct 2016 12:22:17 +0000 (12:22 +0000)]
Fix PR 28885: Fix AST Printer output for the inherited constructor using
declarations.
This commit ensures that the correct record type is printed out for the
using declarations that represent C++ inherited constructors.
It fixes a regression introduced in r274049 which changed the name that's
stored in the using declarations that correspond to inherited constructors.
Alex Lorenz [Mon, 3 Oct 2016 12:12:03 +0000 (12:12 +0000)]
Fix PR 28885: Fix AST Printer output for the inherited constructor using
declarations.
This commit ensures that the correct record type is printed out for the
using declarations that represent C++ inherited constructors.
It fixes a regression introduced in r274049 which changed the name that's
stored in the using declarations that correspond to inherited constructors.
Artem Dergachev [Mon, 3 Oct 2016 08:11:50 +0000 (08:11 +0000)]
[analyzer] Improve CloneChecker diagnostics
Highlight code clones referenced by the warning message with the help of
the extra notes feature recently introduced in r283092.
Change warning text to more clang-ish. Remove suggestions from the copy-paste
error checker diagnostics, because currently our suggestions are strictly 50%
wrong (we do not know which of the two code clones contains the error), and
for that reason we should not sound as if we're actually suggesting this.
Hopefully a better solution would bring them back.
Make sure the suspicious clone pair structure always mentions
the correct variable for the second clone.
Artem Dergachev [Mon, 3 Oct 2016 08:03:51 +0000 (08:03 +0000)]
[analyzer] Add extra notes to ObjCDeallocChecker
The report is now highlighting instance variables and properties
referenced by the warning message with the help of the
extra notes feature recently introduced in r283092.
Artem Dergachev [Mon, 3 Oct 2016 07:58:26 +0000 (07:58 +0000)]
[analyzer] Extend bug reports with extra notes
These diagnostics are separate from the path-sensitive engine's path notes,
and can be added manually on top of path-sensitive or path-insensitive reports.
The new note diagnostics would appear as note:-diagnostic on console and
as blue bubbles in scan-build. In plist files they currently do not appear,
because format needs to be discussed with plist file users.
The analyzer option "-analyzer-config notes-as-events=true" would convert
notes to normal path notes, and put them at the beginning of the path.
This is a temporary hack to show the new notes in plist files.
A few checkers would be updated in subsequent commits,
including tests for this new feature.
Michal Gorny [Sun, 2 Oct 2016 19:28:57 +0000 (19:28 +0000)]
[cmake] Install 'clang-cpp' symlink
Install the 'clang-cpp' symlink used to spawn the preprocessor. The code
handling this suffix is already included in Driver. FreeBSD is already
creating such a symlink in ports, and a similar one was requested
by Gentoo/FreeBSD team. The goal is to handle software that takes a C
preprocessor via a variable but does not handle passing options
correctly (i.e. 'clang -E' does not work).
Hal Finkel [Sun, 2 Oct 2016 02:10:45 +0000 (02:10 +0000)]
[PowerPC] Enable soft-float for PPC64, and +soft-float -> -hard-float
Enable soft-float support on PPC64, as the backend now supports it. Also, the
backend now uses -hard-float instead of +soft-float, so set the target features
accordingly.
Craig Topper [Sat, 1 Oct 2016 21:03:50 +0000 (21:03 +0000)]
[AVX-512] Check rounding mode for builtins that take one. Rounding mode should be either _MM_FROUND_CUR_DIRECTION or a 2-bit rounding mode ORed with _MM_FROUND_NO_EXC.
Craig Topper [Sat, 1 Oct 2016 21:03:46 +0000 (21:03 +0000)]
[AVX-512] Add _MM_FROUND_NO_EXC to test cases that pass a rounding mode intrinsics. This is preparation for a follow up commit that will check validity of rounding mode argument.
Richard Trieu [Sat, 1 Oct 2016 00:15:24 +0000 (00:15 +0000)]
Fix crash when emitting error.
With templated classes, is possible to not be able to determine is a member
function is a special member function before the class is instantiated. Only
these special member functions can be defaulted. In some cases, knowing
whether a function is a special member function can't be determined until
instantiation, so an uninstantiated function could possibly be defaulted too.
Add a case to the error diagnostic when the function marked with a default is
not known to be a special member function.
Richard Smith [Fri, 30 Sep 2016 23:16:08 +0000 (23:16 +0000)]
Fix bogus "inline namespace cannot be reopened as non-inline" diagnostic to
just warn that the second declaration is missing the 'inline' keyword. This is
valid, and we shouldn't be suggesting otherwise.
CodeGen: inherit DLLExport attribute in Windows Itanium
When emitting the fundamental type information constants, inherit the
DLLExportAttr from `__fundamental_type_info`. We would previously not
honor the `__declspec(dllexport)` on the type information.
Richard Smith [Fri, 30 Sep 2016 22:41:36 +0000 (22:41 +0000)]
P0035R4: add predefined __STDCPP_DEFAULT_NEW_ALIGNMENT__ macro. By default, we
assume that ::operator new provides no more alignment than is necessary for any
primitive type, except when we're on a GNU OS, where glibc's malloc guarantees
to provide 64-bit alignment on 32-bit systems and 128-bit alignment on 64-bit
systems. This can be controlled by the command-line -fnew-alignment flag.
double c = 3.0;
float64x2_t v = {0.0, 1.0};
c += vget_low_f64(v);
This restores one more valid behavior pre r266366, and is a incremental
follow up from work committed in r274646.
While here, make the check more strict, add FIXMEs, clean up variable
names to match what they can actually be and update testcases to reflect
that. We now reject:
[CUDA] Remove incorrect comment in CUDASetLambdaAttrs.
I'd said that nvcc doesn't allow you to add __host__ or __device__
attributes on lambdas in all circumstances, but I believe this was user
error on my part. I can't reproduce these warnings/errors if I pass
--expt-extended-lambda to nvcc.
[CUDA] Emit a warning if a CUDA host/device/global attribute is placed after '(...)'.
Summary:
This is probably the sane place for the attribute to go, but nvcc
specifically rejects it. Other GNU-style attributes are allowed in this
position (although judging from the warning it emits for
host/device/global, those attributes are applied to the lambda's
anonymous struct, not to the function itself).
It would be nice to have a FixIt message here, but doing so, or even
just getting the correct range for the attribute, including its '((' and
'))'s, is apparently Hard.
Martin Storsjo [Fri, 30 Sep 2016 19:13:46 +0000 (19:13 +0000)]
[MS] Implement __iso_volatile loads/stores as builtins
These are supposed to produce the same as normal volatile
pointer loads/stores. When -volatile:ms is specified,
normal volatile pointers are forced to have atomic semantics
(as is the default on x86 in MSVC mode). In that case,
these builtins should still produce non-atomic volatile
loads/stores without acquire/release semantics, which
the new test verifies.
These are only available on ARM (and on AArch64,
although clang doesn't support AArch64/Windows yet).
This implements what is missing for PR30394, making it possible
to compile C++ for ARM in MSVC mode with MSVC headers.
Samuel Antao [Fri, 30 Sep 2016 15:34:19 +0000 (15:34 +0000)]
[CUDA][OpenMP] Add a generic offload action builder
Summary:
This patch proposes a new class to generate and record action dependences related with offloading. The builder provides three main functionalities:
- Add device dependences to host actions.
- Add host dependence to device actions.
- Register device top-level actions.
The constructor of the builder detect the programming models that should be supported, and generates a specialized builder for each. If a new programming model is to be added in the future, only a new specialized builder has to be implemented.
When the specialized builder is generated, it produces programming-model-specific diagnostics.
A CUDA specialized builder is proposed in the patch that mostly consists of the partition of the current `buildCudaAction` by the three different functionalities.
Richard Smith [Thu, 29 Sep 2016 21:30:12 +0000 (21:30 +0000)]
Switch to a different workaround for unimplementability of P0145R3 in MS ABIs.
Instead of ignoring the evaluation order rule, ignore the "destroy parameters
in reverse construction order" rule for the small number of problematic cases.
This only causes incorrect behavior in the rare case where both parameters to
an overloaded operator <<, >>, ->*, &&, ||, or comma are of class type with
non-trivial destructor, and the program is depending on those parameters being
destroyed in reverse construction order.
We could do a little better here by reversing the order of parameter
destruction for those functions (and reversing the argument evaluation order
for all direct calls, not just those with operator syntax), but that is not a
complete solution to the problem, as the same situation can be reached by an
indirect function call.
[sanitizer-coverage/libFuzzer] make the guards for trace-pc 32-bit; create one array of guards per function, instead of one guard per BB. reorganize the code so that trace-pc-guard does not create unneeded globals
Jonas Hahnfeld [Thu, 29 Sep 2016 07:43:08 +0000 (07:43 +0000)]
XFAIL Driver/darwin-stdlib.cpp if CLANG_DEFAULT_CXX_STDLIB is set
Until someone rewrites the stdlib logic for Darwin so that we don't need
to pass down the -stdlib argument to cc1.
(see https://llvm.org/bugs/show_bug.cgi?id=30548)
projects / clang / log