Adrian Prantl [Tue, 10 Oct 2017 23:54:21 +0000 (23:54 +0000)]
Include getting generated struct offsets in CodegenABITypes
This change adds a new function, CodeGen::getFieldNumber, that
enables a user of clang's code generation to get the field number
in a generated LLVM IR struct that corresponds to a particular field
in a C struct.
It is important to expose this information in Clang's code generation
interface because there is no reasonable way for users of Clang's code
generation to get this information. In particular:
LLVM struct types do not include field names.
Clang adds a non-trivial amount of logic to the code generation of LLVM IR types for structs, in particular to handle padding and bit fields.
Richard Smith [Tue, 10 Oct 2017 22:33:17 +0000 (22:33 +0000)]
[Modules TS] Module ownership semantics for redeclarations.
When declaring an entity in the "purview" of a module, it's never a
redeclaration of an entity in the purview of a default module or in no module
("in the global module"). Don't consider those other declarations as possible
redeclaration targets if they're not visible, and reject any cases where we
pick a prior visible declaration that violates this rule.
This reinstates r315251 and r315256, reverted in r315309 and r315308
respectively, tweaked to avoid triggering a linkage calculation when declaring
implicit special members (this exposed our pre-existing issue with typedef
names for linkage changing the linkage of types whose linkage has already been
computed and cached in more cases). A testcase for that regression has been
added in r315366.
Support: Have directory_iterator::status() return FindFirstFileEx/FindNextFile results on Windows.
This allows clients to avoid an unnecessary fs::status() call on each
directory entry. Because the information returned by FindFirstFileEx
is a subset of the information returned by a regular status() call,
I needed to extract a base class from file_status that contains only
that information.
On my machine, this reduces the time required to enumerate a ThinLTO
cache directory containing 520k files from almost 4 minutes to less
than 2 seconds.
But when ppc intrinsic is called to compare vectors, the ppc intrinsic can return i32 even E->getType() is BoolTy, in this case EmitScalarConversion does nothing, an i32 type result is returned and causes crash later.
This patch detects this case and truncates the result before return.
Hans Wennborg [Tue, 10 Oct 2017 16:53:25 +0000 (16:53 +0000)]
For dllexport class templates, export specializations of member functions (PR34849) (take 2)
This is a re-commit of r315025, but making sure to only apply this to
specializations of class template member functions; i.e. not when the function
itself is a template.
Artem Dergachev [Tue, 10 Oct 2017 11:55:56 +0000 (11:55 +0000)]
[analyzer] MisusedMovedObject: Fix state-resetting a base-class sub-object.
If a method is resetting the state of an object that was moved from, it should
be safe to use this object again. However if the method was defined in a parent
class, but used in a child class, the reset didn't happen from the checker's
perspective.
Ivan A. Kosarev [Tue, 10 Oct 2017 09:39:32 +0000 (09:39 +0000)]
[CodeGen] Do not construct complete LValue base info in trivial cases
Besides obvious code simplification, avoiding explicit creation
of LValueBaseInfo objects makes it easier to make TBAA
information to be part of such objects.
This is part of D38126 reworked to be a separate patch to
simplify review.
Akira Hatanaka [Tue, 10 Oct 2017 01:24:33 +0000 (01:24 +0000)]
[Sema][ObjC] Preserve syntactic sugar when removing
ARCReclaimReturnedObject cast.
This is a follow-up to r314370.
Rather than throwing away the enclosing parentheses, this commit walks
down the expression until an ARCReclaimReturnedObject cast is found and
removes just the cast, preserving the syntactic sugar expressions
(parens and casts) that were visited up to that point.
Richard Smith [Tue, 10 Oct 2017 00:49:38 +0000 (00:49 +0000)]
[Modules TS] Avoid computing the linkage of the enclosing DeclContext for a declaration in the global module.
This works around a language issue where adding a typedef name for linkage
purposes changes the linkage of an already-defined class after it becomes
complete.
Richard Smith [Mon, 9 Oct 2017 23:42:09 +0000 (23:42 +0000)]
[Modules TS] Module ownership semantics for redeclarations.
When declaring an entity in the "purview" of a module, it's never a
redeclaration of an entity in the purview of a default module or in no module
("in the global module"). Don't consider those other declarations as possible
redeclaration targets if they're not visible, and reject any cases where we
pick a prior visible declaration that violates this rule.
Driver: hoist the `wchar_t` handling to the driver
Move the logic for determining the `wchar_t` type information into the
driver. Rather than passing the single bit of information of
`-fshort-wchar` indicate to the frontend the desired type of `wchar_t`
through a new `-cc1` option of `-fwchar-type` and indicate the
signedness through `-f{,no-}signed-wchar`. This replicates the current
logic which was spread throughout Basic into the
`RenderCharacterOptions`.
Most of the changes to the tests are to ensure that the frontend uses
the correct type. Add a new test set under `test/Driver/wchar_t.c` to
ensure that we calculate the proper types for the various cases.
Matt Arsenault [Fri, 6 Oct 2017 19:34:40 +0000 (19:34 +0000)]
OpenCL: Assume functions are convergent
This was done for CUDA functions in r261779, and for the same
reason this also needs to be done for OpenCL. An arbitrary
function could have a barrier() call in it, which in turn
requires the calling function to be convergent.
Zachary Turner [Fri, 6 Oct 2017 17:54:46 +0000 (17:54 +0000)]
[lit] Improve tool substitution in lit.
This addresses two sources of inconsistency in test configuration
files.
1. Substitution boundaries. Previously you would specify a
substitution, such as 'lli', and then additionally a set
of characters that should fail to match before and after
the tool. This was used, for example, so that matches that
are parts of full paths would not be replaced. But not all
tools did this, and those that did would often re-invent
the set of characters themselves, leading to inconsistency.
Now, every tool substitution defaults to using a sane set
of reasonable defaults and you have to explicitly opt out
of it. This actually fixed a few latent bugs that were
never being surfaced, but only on accident.
2. There was no standard way for the system to decide how to
locate a tool. Sometimes you have an explicit path, sometimes
we would search for it and build up a path ourselves, and
sometimes we would build up a full command line. Furthermore,
there was no standardized way to handle missing tools. Do we
warn, fail, ignore, etc? All of this is now encapsulated in
the ToolSubst class. You either specify an exact command to
run, or an instance of FindTool('<tool-name>') and everything
else just works. Furthermore, you can specify an action to
take if the tool cannot be resolved.
Alexey Bataev [Fri, 6 Oct 2017 17:00:28 +0000 (17:00 +0000)]
[OPENMP] Do not capture local static variables.
Previously we may erroneously try to capture locally declared static
variables, which will lead to crash for target-based constructs.
Patch fixes this problem.
Erich Keane [Fri, 6 Oct 2017 16:40:45 +0000 (16:40 +0000)]
Split X86::BI__builtin_cpu_init handling into own function[NFC]
The Cpu Init functionality is required for the target
attribute, so this patch simply splits it out into its own
function, exactly like CpuIs and CpuSupports.
Alexey Bataev [Fri, 6 Oct 2017 16:17:25 +0000 (16:17 +0000)]
[OPENMP] Capture references to global variables.
In C++11 variable to global variables are considered as constant
expressions and these variables are not captured in the outlined
regions. Patch allows capturing of such variables in the OpenMP regions.
Ivan A. Kosarev [Fri, 6 Oct 2017 08:17:48 +0000 (08:17 +0000)]
Refine generation of TBAA information in clang
This patch is an attempt to clarify and simplify generation and
propagation of TBAA information. The idea is to pack all values
that describe a memory access, namely, base type, access type and
offset, into a single structure. This is supposed to make further
changes, such as adding support for unions and array members,
easier to prepare and review.
DecorateInstructionWithTBAA() is no more responsible for
converting types to tags. These implicit conversions not only
complicate reading the code, but also suggest assigning scalar
access tags while we generally prefer full-size struct-path tags.
TBAAPathTag is replaced with TBAAAccessInfo; the latter is now
the type of the keys of the cache map that translates access
descriptors to metadata nodes.
Fixed a bug with writing to a wrong map in
getTBAABaseTypeMetadata() (former getTBAAStructTypeInfo()).
We now check for valid base access types every time we
dereference a field. The original code only checks the top-level
base type. See isValidBaseType() / isTBAAPathStruct() calls.
Some entities have been renamed to sound more adequate and less
confusing/misleading in presence of path-aware TBAA information.
Now we do not lookup twice for the same cache entry in
getAccessTagInfo().
Akira Hatanaka [Fri, 6 Oct 2017 07:12:46 +0000 (07:12 +0000)]
[CodeGen] Emit a helper function for __builtin_os_log_format to reduce
code size.
Currently clang expands a call to __builtin_os_log_format into a long
sequence of instructions at the call site, causing code size to
increase in some cases.
This commit attempts to reduce code size by emitting a helper function
that can be shared by calls to __builtin_os_log_format with similar
formats and arguments. The helper function has linkonce_odr linkage to
enable the linker to merge identical functions across translation units.
Attribute 'noinline' is attached to the helper function at -Oz so that
the inliner doesn't inline functions that can potentially be merged.
This commit also fixes a bug where the generated IR writes past the end
of the buffer when "%m" is the last specifier appearing in the format
string passed to __builtin_os_log_format.
Summary:
* Rename -shared-libasan to -shared-libsan, keeping the old name as alias.
* Add -static-libsan for targets that default to shared.
* Remove an Android special case. It is now possible (but untested) to use static compiler-rt libraries there.
* Support libclang_rt.ubsan_standalone as a shared library.
Unlike GCC, this change applies -shared-libsan / -static-libsan to all sanitizers.
I don't see a point in multiple flags like -shared-libubsan, considering that most sanitizers
are not compatible with each other, and each link has basically a single shared/static choice.
Richard Smith [Thu, 5 Oct 2017 19:35:51 +0000 (19:35 +0000)]
Fix two-phase name lookup for non-dependent overloaded operators.
If we resolve an overloaded operator call to a specific function during
template definition, don't perform ADL during template instantiation.
Doing so finds overloads that we're not supposed to find.
Alexey Bataev [Thu, 5 Oct 2017 17:51:39 +0000 (17:51 +0000)]
[OPENMP] Fix mapping|privatization of implicitly captured variables.
If the `defaultmap(tofrom:scalar)` clause is specified, the scalars must
be mapped with 'tofrom' modifiers, otherwise they must be captured as
firstprivates.
Ivan A. Kosarev [Thu, 5 Oct 2017 11:08:17 +0000 (11:08 +0000)]
[CodeGen] Unify generation of scalar and struct-path TBAA tags
This patch makes it possible to produce access tags in a uniform
manner regardless whether the resulting tag will be a scalar or a
struct-path one. getAccessTagInfo() now takes care of the actual
translation of access descriptors to tags and can handle all
kinds of accesses. Facilities that specific to scalar accesses
are eliminated.
Some more details:
* DecorateInstructionWithTBAA() is not responsible for conversion
of types to access tags anymore. Instead, it takes an access
descriptor (TBAAAccessInfo) and generates corresponding access
tag from it.
* getTBAAInfoForVTablePtr() reworked to
getTBAAVTablePtrAccessInfo() that now returns the
virtual-pointer access descriptor and not the virtual-point
type metadata.
* Added function getTBAAMayAliasAccessInfo() that returns the
descriptor for may-alias accesses.
* getTBAAStructTagInfo() renamed to getTBAAAccessTagInfo() as now
it is the only way to generate access tag by a given access
descriptor. It is capable of producing both scalar and
struct-path tags, depending on options and availability of the
base access type. getTBAAScalarTagInfo() and its cache
ScalarTagMetadataCache are eliminated.
* Now that we do not need to care about whether the resulting
access tag should be a scalar or struct-path one,
getTBAAStructTypeInfo() is renamed to getBaseTypeInfo().
* Added function getTBAAAccessInfo() that constructs access
descriptor by a given QualType access type.
This is part of D37826 reworked to be a separate patch to
simplify review.
Ivan A. Kosarev [Thu, 5 Oct 2017 10:47:51 +0000 (10:47 +0000)]
[CodeGen] Unify generation of scalar and struct-path TBAA tags
This patch makes it possible to produce access tags in a uniform
manner regardless whether the resulting tag will be a scalar or a
struct-path one. getAccessTagInfo() now takes care of the actual
translation of access descriptors to tags and can handle all
kinds of accesses. Facilities that specific to scalar accesses
are eliminated.
Some more details:
* DecorateInstructionWithTBAA() is not responsible for conversion
of types to access tags anymore. Instead, it takes an access
descriptor (TBAAAccessInfo) and generates corresponding access
tag from it.
* getTBAAInfoForVTablePtr() reworked to
getTBAAVTablePtrAccessInfo() that now returns the
virtual-pointer access descriptor and not the virtual-point
type metadata.
* Added function getTBAAMayAliasAccessInfo() that returns the
descriptor for may-alias accesses.
* getTBAAStructTagInfo() renamed to getTBAAAccessTagInfo() as now
it is the only way to generate access tag by a given access
descriptor. It is capable of producing both scalar and
struct-path tags, depending on options and availability of the
base access type. getTBAAScalarTagInfo() and its cache
ScalarTagMetadataCache are eliminated.
* Now that we do not need to care about whether the resulting
access tag should be a scalar or struct-path one,
getTBAAStructTypeInfo() is renamed to getBaseTypeInfo().
* Added function getTBAAAccessInfo() that constructs access
descriptor by a given QualType access type.
This is part of D37826 reworked to be a separate patch to
simplify review.
Artem Dergachev [Thu, 5 Oct 2017 08:43:32 +0000 (08:43 +0000)]
[analyzer] Fix leak false positives on stuff put in C++/ObjC initializer lists.
The analyzer now realizes that C++ std::initializer_list objects and
Objective-C boxed structure/array/dictionary expressions can potentially
maintain a reference to the objects that were put into them. This avoids
false memory leak posivites and a few other issues.
This is a conservative behavior; for now, we do not model what actually happens
to the objects after being passed into such initializer lists.
Richard Smith [Thu, 5 Oct 2017 00:43:38 +0000 (00:43 +0000)]
Remove PendingBody mechanism for function and ObjC method deserialization.
In its place, track on the canonical function declaration whether there is a
declaration with a body (and if so, which one). This brings function definition
handling in line with what we do in all other contexts, and is necessary to
allow us to merge declarations within multiple definitions of the same function
(eg, PR33924).
Yaxun Liu [Wed, 4 Oct 2017 20:32:17 +0000 (20:32 +0000)]
[OpenCL] Clean up and add missing fields for block struct
Currently block is translated to a structure equivalent to
struct Block {
void *isa;
int flags;
int reserved;
void *invoke;
void *descriptor;
};
Except invoke, which is the pointer to the block invoke function,
all other fields are useless for OpenCL, which clutter the IR and
also waste memory since the block struct is passed to the block
invoke function as argument.
On the other hand, the size and alignment of the block struct is
not stored in the struct, which causes difficulty to implement
__enqueue_kernel as library function, since the library function
needs to know the size and alignment of the argument which needs
to be passed to the kernel.
This patch removes the useless fields from the block struct and adds
size and align fields. The equivalent block struct will become
struct Block {
int size;
int align;
generic void *invoke;
/* custom fields */
};
It also changes the pointer to the invoke function to be
a generic pointer since the address space of a function
may not be private on certain targets.
Artem Dergachev [Wed, 4 Oct 2017 15:59:40 +0000 (15:59 +0000)]
[analyzer] Fix autodetection of binding types.
In ProgramState::getSVal(Location, Type) API which dereferences a pointer value,
when the optional Type parameter is not supplied and the Location is not typed,
type should have been guessed on a best-effort basis by inspecting the Location
more deeply. However, this never worked; the auto-detected type was instead
a pointer type to the correct type.
Fixed the issue and added various test cases to demonstrate which parts of the
analyzer were affected (uninitialized pointer argument checker, C++ trivial copy
modeling, Google test API modeling checker).
Additionally, autodetected void types are automatically replaced with char,
in order to simplify checker APIs. Which means that if the location is a void
pointer, getSVal() would read the first byte through this pointer
and return its symbolic value.
Jonas Hahnfeld [Wed, 4 Oct 2017 13:32:59 +0000 (13:32 +0000)]
[OpenMP] Fix passing of -m arguments correctly
The recent fix in D38258 was wrong: getAuxTriple() only returns
non-null values for the CUDA toolchain. That is why the now added
test for PPC and X86 failed.
Fix assertion failure in thread safety analysis (PR34800).
Summary:
Fix an assertion failure (http://llvm.org/PR34800) and clean up unused code relevant to the fixed logic.
A bit of context: when `SExprBuilder::translateMemberExpr` is called on a member expression that involves a conversion operator, for example, `til::Project` constructor can't just call `getName()` on it, since the name is not a simple identifier. In order to handle this case I've introduced an optional string to print the member name to. I discovered that the other two `til::Project` constructors are not used, so it was better to delete them instead of ensuring they work correctly with the new logic.
Richard Smith [Wed, 4 Oct 2017 01:49:22 +0000 (01:49 +0000)]
PR34822: Fix a collection of related bugs with our handling of C89 implicit function declarations.
We were injecting the function into the wrong semantic context, resulting in it
failing to be registered as a global for redeclaration lookup. As a
consequence, we accepted invalid code since r310616.
Fixing that resulted in the "out-of-scope declaration" diagnostic firing a lot
more often. It turned out that warning codepath was non-conforming, because it
did not cause us to inject the implicitly-declared function into the enclosing
block scope. We now only warn if the type of the out-of-scope declaration
doesn't match the type of an implicitly-declared function; in all other cases,
we produce the normal warning for an implicitly-declared function.
Richard Smith [Tue, 3 Oct 2017 20:36:00 +0000 (20:36 +0000)]
Suppress -Wmissing-braces warning when aggregate-initializing a struct with a single field that is itself an aggregate.
In C++, such initialization of std::array<T, N> types is guaranteed to work by
the standard, is completely idiomatic, and the "suggested" alternative from
Clang was technically invalid.
Alexey Bataev [Tue, 3 Oct 2017 20:00:00 +0000 (20:00 +0000)]
[OPENMP] Allow use of declare target directive inside struct
declaration.
Patch allows using of the `#pragma omp declare target`| `#pragma omp end
declare target` directives inside the structures if we need to mark as
declare target only some static members.
Yaxun Liu [Tue, 3 Oct 2017 14:34:29 +0000 (14:34 +0000)]
[OpenCL] Fix checking of vector type casting
Currently clang allows the following code
int a;
int b = (const int) a;
However it does not the following code
int4 a;
int4 b = (const int4) a;
This is because Clang compares the qualified types instead of unqualified types for vector type casting, which causes the inconsistency.
Ivan A. Kosarev [Tue, 3 Oct 2017 11:31:42 +0000 (11:31 +0000)]
[CodeGen] Fix propagation of TBAA info for atomic accesses
This patch fixes clang to propagate complete TBAA information for
atomic accesses and not just the final access types. Prepared
against D38456 and requires it to be committed first.
This is part of D37826 reworked to be a separate patch to
simplify review.
With this patch we implement a concept of TBAA access descriptors
that are capable of representing both scalar and struct-path
accesses in a generic way.
This is part of D37826 reworked to be a separate patch to
simplify review.
Vedant Kumar [Tue, 3 Oct 2017 01:27:26 +0000 (01:27 +0000)]
[ubsan] Skip alignment checks which are folded away
Don't emit alignment checks which the IR constant folder throws away.
I've tested this out on X86FastISel.cpp. While this doesn't decrease
end-to-end compile-time significantly, it results in 122 fewer type
checks (1% reduction) overall, without adding any real complexity.