Alexey Bataev [Tue, 22 Dec 2015 12:21:47 +0000 (12:21 +0000)]
[OPENMP 4.5] Parsing/sema for 'depend(sink:vec)' clause in 'ordered' directive.
OpenMP 4.5 adds 'depend(sink:vec)' in 'ordered' directive for doacross loop synchronization. Patch adds parsing and semantic analysis for this clause.
David Majnemer [Mon, 21 Dec 2015 22:09:34 +0000 (22:09 +0000)]
[clang-cl] Add support for /Brepro
The /Brepro flag controls whether or not the compiler should embed
timestamps into the object file. Object files which do not embed
timestamps are not suitable for incremental linking but are suitable for
hermetic build systems and staged self-hosts of clang.
A normal clang spelling of this flag has been added,
-mincremental-linker-compatible.
Vedant Kumar [Mon, 21 Dec 2015 20:21:15 +0000 (20:21 +0000)]
Reapply "[CodeGen] Fix assignments of inline layouts into the byref structure"
When using blocks, a byref structure is created to represent the
closure. The "byref.layout" field of this structure is an i8*. However,
some 'inline' layouts are represented as i64's, not i8*'s.
Prior to r246985 we cast the i64 'inline' layout to an i8* before
assigning it into the byref structure. This patch brings the cast back
and adds a regression test.
The original version of this patch was too invasive. This version only adds the
cast to BuildByrefLayout.
Vedant Kumar [Mon, 21 Dec 2015 19:30:37 +0000 (19:30 +0000)]
[CodeGen] Fix assignments of inline layouts into the byref structure
When using blocks, a byref structure is created to represent the
closure. The "byref.layout" field of this structure is an i8*. However,
some 'inline' layouts are represented as i64's, not i8*'s.
Prior to r246985 we cast the i64 'inline' layout to an i8* before
assigning it into the byref structure. This patch brings the cast back
and adds a regression test.
Daniel Jasper [Mon, 21 Dec 2015 12:14:17 +0000 (12:14 +0000)]
clang-format: Extend detection of the "main" #include to use the filename
Before, the first (non-system) header in a file was considered to be
the main include. This is conservative as it makes clang-format change
the #include order less often. Instead implement some basic usage of
the filename itself. With this patch, clang-format considers every
header to be a main include if the header file's basename is a prefix
to the filename the #include is in.
Paul Robinson [Sat, 19 Dec 2015 19:41:48 +0000 (19:41 +0000)]
Driver part of debugger tuning.
Adds driver options named -glldb and -gsce to mean -g plus tuning for
lldb and SCE debuggers respectively; the existing -ggdb option does
the same for gdb. Existing options -ggdb0, -ggdb1 etc. unpack into
-ggdb -g<N>. (There will not be -glldb<N> or -gsce<N> options.) The
tuning gets a target-specific default in the driver, and is passed
into cc1 with the new -debugger-tuning option.
As fallout, fixes where '-gsplit-dwarf -g0' would ignore the -g0 part
on Linux.
After configuration the following additional targets will be generated:
stage2-instrumented:
Builds a stage1 x86 compiler, runtime, and required tools (llvm-config, llvm-profdata) then uses that compiler to build an instrumented stage2 compiler.
stage2-instrumented-generate-profdata:
Depends on "stage2-instrumented" and will use the instrumented compiler to generate profdata based on the training files in <clang>/utils/perf-training
stage2:
Depends on "stage2-instrumented-generate-profdata" and will use the stage1 compiler with the stage2 profdata to build a PGO-optimized compiler.
stage2-check-llvm:
Depends on stage2 and runs check-llvm using the stage3 compiler.
stage2-check-clang:
Depends on stage2 and runs check-clang using the stage3 compiler.
stage2-check-all:
Depends on stage2 and runs check-all using the stage3 compiler.
stage2-test-suite:
Depends on stage2 and runs the test-suite using the stage3 compiler (requires in-tree test-suite).
Richard Smith [Fri, 18 Dec 2015 22:40:25 +0000 (22:40 +0000)]
Split RequireCompleteType into a function that actually requires that the type
is complete (with an error produced if not) and a function that merely queries
whether the type is complete. Either way we'll trigger instantiation if
necessary, but only the former will diagnose and recover from missing module
imports.
The intent of this change is to prevent a class of bugs where code would call
RequireCompleteType(..., 0) and then ignore the result. With modules, we must
check the return value and use it to determine whether the definition of the
type is visible.
This also fixes a debug info quality issue: calls to isCompleteType do not
trigger the emission of debug information for a type in limited-debug-info
mode. This allows us to avoid emitting debug information for type definitions
in more cases where we believe it is safe to do so.
Richard Smith [Fri, 18 Dec 2015 22:19:11 +0000 (22:19 +0000)]
[modules] Don't try to use the definition of a class if
RequireCompleteType(..., 0) says we're not permitted to do so. The definition
might not be visible, even though we know what it is.
Richard Smith [Fri, 18 Dec 2015 21:45:41 +0000 (21:45 +0000)]
Wire a SourceLocation into IsDerivedFrom and move the RequireCompleteType call
for the derived class into it. This is mostly just a cleanup, but could in
principle be a bugfix if there is some codepath that reaches here and didn't
previously require a complete type (I couldn't find any such codepath, though).
Bob Wilson [Fri, 18 Dec 2015 20:37:54 +0000 (20:37 +0000)]
PIC should not be enabled by default on Darwin with -static.
r245667 changed -static so that it doesn't override an explicit -fPIC
option, but -static should still change the default for Darwin for -fno-PIC.
This matches longstanding GCC and Clang behavior on Darwin and changing it
would be disruptive, with no significant benefit.
http://reviews.llvm.org/D15455
rdar://problem/23811045
Alexey Bataev [Fri, 18 Dec 2015 10:24:53 +0000 (10:24 +0000)]
[OPENMP] Remove explicit call for implicit barrier
#pragma omp parallel needs an implicit barrier that is currently done by an explicit call to __kmpc_barrier. However, the runtime already ensures a barrier in __kmpc_fork_call which currently leads to two barriers per region per thread.
Differential Revision: http://reviews.llvm.org/D15561
Alexey Bataev [Fri, 18 Dec 2015 07:58:25 +0000 (07:58 +0000)]
[OPENMP] Fix for http://llvm.org/PR25878: Error compiling an OpenMP program
OpenMP codegen tried to emit the code for its constructs even if it was detected as a dead-code. Added checks to ensure that the code is emitted if the code is not dead.
Alexey Bataev [Fri, 18 Dec 2015 05:05:56 +0000 (05:05 +0000)]
[OPENMP 4.5] Parsing/sema analysis for 'depend(source)' clause in 'ordered' directive.
OpenMP 4.5 adds 'depend(source)' clause for 'ordered' directive to support cross-iteration dependence. Patch adds parsing and semantic analysis for this construct.
Douglas Gregor [Fri, 18 Dec 2015 00:52:31 +0000 (00:52 +0000)]
ObjC properties: consider ownership of properties from protocols when synthesizing.
When determining whether ownership was explicitly written for a
property when it is being synthesized, also consider that the original
property might have come from a protocol. Fixes rdar://problem/23931441.
Artem Belevich [Thu, 17 Dec 2015 22:25:22 +0000 (22:25 +0000)]
[CUDA] runtime wrapper header tweaks
* Pull in host-only implementations of few CUDA-specific math functions.
* #nclude <cmath> early to prevent its inclusion from CUDA headers after
they've messed with __THROW macro.
Easwaran Raman [Thu, 17 Dec 2015 19:14:27 +0000 (19:14 +0000)]
Attach maximum function count to Module when using PGO mode.
This sets the maximum entry count among all functions in the program to the module using module flags. This allows the optimizer to use this information.
Artem Belevich [Thu, 17 Dec 2015 18:12:36 +0000 (18:12 +0000)]
[CUDA] Make vtable construction aware of host/device side of CUDA compilation.
C++ emits vtables for classes that have key function present in the
current TU. While we compile CUDA the fact that key function was found
in this TU does not mean that we are going to generate code for it. E.g.
vtable for a class with host-only methods should not (and can not) be
generated on device side, because we'll never generate code for them
during device-side compilation.
This patch adds an extra CUDA-specific check during key method computation
and filters out potential key methods that are not suitable for this side
of CUDA compilation.
When we codegen vtable, entries for unsuitable methods are set to null.
Alexey Bataev [Thu, 17 Dec 2015 06:55:08 +0000 (06:55 +0000)]
[OPENMP] Fix for http://llvm.org/PR25142: openmp: Assertion failed: DD && "queried property of class with no definition", file AST/DeclCXX.h
Added processing for template specialization during data-sharing attributes analysis
Devin Coughlin [Thu, 17 Dec 2015 00:28:33 +0000 (00:28 +0000)]
[analyzer] Better detect when C++ object was constructed into existing region.
When the analyzer evaluates a CXXConstructExpr, it looks ahead in the CFG for
the current block to detect what region the object should be constructed into.
If the constructor was directly constructed into a local variable or field
region then there is no need to explicitly bind the constructed value to
the local or field when analyzing the DeclStmt or CXXCtorInitializer that
called the constructor.
Unfortunately, there were situations in which the CXXConstructExpr was
constructed into a temporary region but when evaluating the corresponding
DeclStmt or CXXCtorInitializer the analyzer assumed the object was constructed
into the local or field. This led to spurious warnings about uninitialized
values (PR25777).
To avoid these false positives, this commit factors out the logic for
determining when a CXXConstructExpr will be directly constructed into existing
storage, adds the inverse logic to detect when the corresponding later bind can
be safely skipped, and adds assertions to make sure these two checks are in
sync.
[PS4] Fix the unit test to be compatible with clang driver. NFC
".exe" extension is inherently checked by llvm::fs::can_execute()
This patch fixes the linker extension in clang driver and updates the
unit test to accommodate the the check string on windows.
Chris Bieneman [Wed, 16 Dec 2015 20:17:07 +0000 (20:17 +0000)]
[CMake] Name the bootstrap stages stage[0-9]
When you start chaining bootstrap stages the CMake-generated targets get unwieldy. This change supports naming the bootstrap targets and creating wrapper targets in the top-level build file.
Without this patch the default target generated for a second stage build was "bootstrap" with it the target is "stage2".
Also with some CMake goop setting CLANG_BOOTSTRAP_TARGETS, you can expose third stage targets as "stage3" instead of "bootstrap-bootstrap"
Diego Novillo [Wed, 16 Dec 2015 19:52:05 +0000 (19:52 +0000)]
Add -fsyntax-only to fix failure in read-only directories.
Internally, this test is executed in a read-only directory, which causes
it to fail because the driver tries to generate a file unnecessarily.
Adding -fsyntax-only fixes the issue (thanks to Artem Belevich for
figuring out the root cause).
Some tests are missing the {{(.exe)?}} suffix on the exectables
which the FileCheck is grepping for. This will ensure, the lit tests
are clean on windows
Artem Belevich [Wed, 16 Dec 2015 18:51:59 +0000 (18:51 +0000)]
[CUDA] renamed cuda_runtime.h wrapper to __cuda_runtime.h
Currently it's easy to break CUDA compilation by passing
"-isystem /path/to/cuda/include" to compiler which leads to
compiler including real cuda_runtime.h from there instead
of the wrapper we need.
Renaming the wrapper ensures that we can include the wrapper
regardless of user-specified include paths and files.
Chris Bieneman [Wed, 16 Dec 2015 18:42:14 +0000 (18:42 +0000)]
[CMake] ExternalProject for compiler-rt needs to depend on llvm-config and clang
The add_dependencies call on compiler-rt-configure adds llvm-config and clang to the phony target, but not to the actual configure custom command. We need the dependency bound to the custom command so that it can't be re-ordered by Ninja.
Specifically, it is sometimes necessary to keep certain #includes as
the first #include, even before the main #include for a .cc file.
Switching the category to be signed instead of unsigned isn't ideal,
but it seems as good of an option as any and is fully backwards
compatible.
Akira Hatanaka [Wed, 16 Dec 2015 06:25:38 +0000 (06:25 +0000)]
[Objective-c] Fix a crash that occurs when ObjCTypeParamList::back() is
called on an empty list.
This commit makes Parser::parseObjCTypeParamListOrProtocolRefs return
nullptr if it sees an invalid type parameter (e.g., __kindof) in the
type parameter list.
Chris Bieneman [Wed, 16 Dec 2015 01:02:44 +0000 (01:02 +0000)]
[CMake] Add support for generating profdata for clang from training files
Summary:
This patch adds support for using LIT to drive generating PGO profile data for clang.
This first pass implementation should work on Linux and Unix based platforms. If you build clang using CMake with LLVM_BUILD_INSTRUMENTED=On the CMake build generates a generate-profdata target that will use the just-built clang to build any test files (see hello_world.cpp as an example). Each test compile will generate profraw files for each clang process. After all tests have run CMake will merge the profraw files using llvm-profdata.
Future opportunities for extension:
* Support for Build->Profile->Build bootstrapping
* Support for linker order file generation using a similar mechanism and the same training data
* Support for Windows
Evgeniy Stepanov [Tue, 15 Dec 2015 23:00:20 +0000 (23:00 +0000)]
Cross-DSO control flow integrity (Clang part).
Clang-side cross-DSO CFI.
* Adds a command line flag -f[no-]sanitize-cfi-cross-dso.
* Links a runtime library when enabled.
* Emits __cfi_slowpath calls is bitset test fails.
* Emits extra hash-based bitsets for external CFI checks.
* Sets a module flag to enable __cfi_check generation during LTO.
[Microsoft][C++] Clang doesn't support a use of "this" pointer inside inline asm
Clang doesn’t support a use of “this” pointer inside inline asm.
When I tried to compile a class or a struct (see example) with an inline asm that contains "this" pointer.
Clang returns with an error.
This patch fixes that.
error: expected unqualified-id
For example:
'''
struct A {
void f() {
__asm mov eax, this
// error: expected unqualified-id
}
};
'''
Differential Revision: http://reviews.llvm.org/D15115
Alexey Bataev [Tue, 15 Dec 2015 10:55:09 +0000 (10:55 +0000)]
[OPENMP 4.5] Codegen for 'hint' clause of 'critical' directive
OpenMP 4.5 defines 'hint' clause for 'critical' directive. Patch adds codegen for this clause.
NAKAMURA Takumi [Tue, 15 Dec 2015 09:37:01 +0000 (09:37 +0000)]
clang/test/Analysis/padding_c.c: Suppress a test incompatible to i686-linux.
error: 'warning' diagnostics expected but not seen:
File clang/test/Analysis/padding_c.c Line 194 (directive at clang/test/Analysis/padding_c.c:193): Excessive padding in 'struct DefaultAttrAlign'
1 error generated.
Benjamin Kramer [Tue, 15 Dec 2015 09:30:31 +0000 (09:30 +0000)]
[libclang] Add a flag to create the precompiled preamble on the first parse.
Summary:
The current default is to create the preamble on the first reparse, aka
second parse. This is useful for clients that do not want to block when
opening a file because serializing the preamble takes a bit of time.
However, this makes the reparse much more expensive and that may be on the
critical path as it's the first interaction a user has with the source code.
YouCompleteMe currently optimizes for the first code interaction by parsing
the file twice when loaded. That's just unnecessarily slow and this flag
helps to avoid that.