[Driver, CodeGen] rename options to disable an FP cast optimization
As suggested in the post-commit thread for rL331056, we should match these
clang options with the established vocabulary of the corresponding sanitizer
option. Also, the use of 'strict' is well-known for these kinds of knobs,
and we can improve the descriptive text in the docs.
So this intends to match the logic of D46135 but only change the words.
Matching LLVM commit to match this spelling of the attribute to follow shortly.
Emit error messages instead of compiler crashing when the target region
does not exist in the device code + fix crash when the location comes
from macros.
[Targets] Implement getConstraintRegister for ARM and AArch64
Summary:
The getConstraintRegister method is used by semantic checking of
inline assembly statements in order to diagnose conflicts between
clobber list and input/output lists. Currently ARM and AArch64 don't
override getConstraintRegister, so conflicts between registers
assigned to variables in asm labels and clobber lists are not
diagnosed. Such conflicts can cause assertion failures in the back end
and even miscompilations.
This patch implements getConstraintRegister for ARM and AArch64
targets. Since these targets don't have single-register constraints,
the implementation is trivial and just returns the register specified
in an asm label (if any).
Richard Smith [Mon, 30 Apr 2018 05:25:48 +0000 (05:25 +0000)]
PR37189 Fix incorrect end source location and spelling for a split '>>' token.
When a '>>' token is split into two '>' tokens (in C++11 onwards), or (as an
extension) when we do the same for other tokens starting with a '>', we can't
just use a location pointing to the first '>' as the location of the split
token, because that would result in our miscomputing the length and spelling
for the token. As a consequence, for example, a refactoring replacing 'A<X>'
with something else would sometimes replace one character too many, and
similarly diagnostics highlighting a template-id source range would highlight
one character too many.
Fix this by creating an expansion range covering the first character of the
'>>' token, whose spelling is '>'. For this to work, we generalize the
expansion range of a macro FileID to be either a token range (the common case)
or a character range (used in this new case).
Richard Smith [Sun, 29 Apr 2018 04:55:46 +0000 (04:55 +0000)]
PR37275 packed attribute should not apply to base classes
Clang incorrectly applied the packed attribute to base classes. Per GCC's
documentation and as can be observed from its behavior, packed only applies to
members, not base classes.
This change is conditioned behind -fclang-abi-compat so that an ABI break can
be avoided by users if desired.
Speculative. ClangMoveTests started failing on
http://lab.llvm.org:8011/builders/clang-x64-ninja-win7/builds/9958
after this change. I can't reproduce on my machine, let's see
if it was due to this change.
LLVM_ON_WIN32 is set exactly with MSVC and MinGW (but not Cygwin) in
HandleLLVMOptions.cmake, which is where _WIN32 defined too. Just use the
default macro instead of a reinvented one.
See thread "Replacing LLVM_ON_WIN32 with just _WIN32" on llvm-dev and cfe-dev.
No intended behavior change.
[Driver, CodeGen] add options to enable/disable an FP cast optimization
As discussed in the post-commit thread for:
rL330437 ( http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20180423/545906.html )
We need a way to opt-out of a float-to-int-to-float cast optimization because too much
existing code relies on the platform-specific undefined result of those casts when the
float-to-int overflows.
The LLVM changes associated with adding this function attribute are here:
rL330947
rL330950
rL330951
Also as suggested, I changed the LLVM doc to mention the specific sanitizer flag that
catches this problem:
rL330958
Oliver Stannard [Fri, 27 Apr 2018 14:03:32 +0000 (14:03 +0000)]
[ARM,AArch64] Add intrinsics for dot product instructions
The ACLE spec which describes these intrinsics hasn't been published yet, but
this is based on the final draft which will be published soon, and these have
already been implemented by GCC.
Oliver Stannard [Fri, 27 Apr 2018 13:56:02 +0000 (13:56 +0000)]
[ARM] Add __ARM_FEATURE_DOTPROD pre-defined macro
This adds a pre-defined macro to test if the compiler has support for the
v8.2-A dot rpoduct intrinsics in AArch32 mode.
The AAcrh64 equivalent has already been added by rL330229.
The ACLE spec which describes this macro hasn't been published yet, but this is
based on the final internal draft, and GCC has already implemented this.
[OpenCL] Add separate read_only and write_only pipe IR types
SPIR-V encodes the read_only and write_only access qualifiers of pipes,
so separate LLVM IR types are required to target SPIR-V. Other backends
may also find this useful.
These new types are `opencl.pipe_ro_t` and `opencl.pipe_wo_t`, which
replace `opencl.pipe_t`.
This replaces __get_pipe_num_packets(...) and __get_pipe_max_packets(...)
which took a read_only pipe with separate versions for read_only and
write_only pipes, namely:
Make MultiplexASTDeserializationListener part of the API [NFC]
Summary:
This patch moves the MultiplexASTDeserializationListener declaration into a public header.
We're currently using this multiplexer in the cling interpreter to attach another
ASTDeserializationListener during the execution (so, after the MultiplexConsumer is already
attached which prevents us from attaching more). So far we're doing this by patching clang
and making this class public, but it makes things easier if we make this instead just public in
upstream.
[CodeGen] Avoid destructing a callee-destructued struct type in a
function if a function delegates to another function.
Fix a bug introduced in r328731, which caused a struct with ObjC __weak
fields that was passed to a function to be destructed twice, once in the
callee function and once in another function the callee function
delegates to. To prevent this, keep track of the callee-destructed
structs passed to a function and disable their cleanups at the point of
the call to the delegated function.
This reapplies r331016, which was reverted in r331019 because it caused
an assertion to fail in EmitDelegateCallArg on a windows bot. I made
changes to EmitDelegateCallArg so that it doesn't try to deactivate
cleanups for structs that have trivial destructors (cleanups for those
structs are never pushed to the cleanup stack in EmitParmDecl).
[CodeGen] Avoid destructing a callee-destructued struct type in a
function if a function delegates to another function.
Fix a bug introduced in r328731, which caused a struct with ObjC __weak
fields that was passed to a function to be destructed twice, once in the
callee function and once in another function the callee function
delegates to. To prevent this, keep track of the callee-destructed
structs passed to a function and disable their cleanups at the point of
the call to the delegated function.
Richard Smith [Fri, 27 Apr 2018 02:00:13 +0000 (02:00 +0000)]
Parse A::template B as an identifier rather than as a template-id with no
template arguments.
This fixes some cases where we'd incorrectly accept "A::template B" when B is a
kind of template that requires template arguments (in particular, a variable
template or a concept).
Summary:
We need to know if ASTReader already has a DeserializationListner or
not, and this also helps to create a multiplexing deserialization
listener if there is one already attached.
[ConfigFiles] Update argument strings when merging argrument lists
Implementation of `InputArgList` assumes its field `ArgStrings` contains
strings for each argument exactly in the same order. This condition was
broken when arguments from config file and from invocation were merged.
This change fixes https://bugs.llvm.org/show_bug.cgi?id=37196 (Clang
config files can crash argument handling).
Richard Trieu [Wed, 25 Apr 2018 23:50:55 +0000 (23:50 +0000)]
Switch to Clang's isDigit function.
std::isdigit can be overloaded, causing the template deduction to fail. Use
Clang's isDigit function which to avoid this. Switch the other calls for
consistency.
[analyzer] Fix a crash on lifetime extension through aggregate initialization.
If 'A' is a C++ aggregate with a reference field of type 'C', in code like
A a = { C() };
C() is lifetime-extended by 'a'. The analyzer wasn't expecting this pattern and
crashing. Additionally, destructors aren't added in the CFG for this case,
so for now we shouldn't be inlining the constructor for C().
Ana Pazos [Wed, 25 Apr 2018 22:42:38 +0000 (22:42 +0000)]
[RISCV] More validations on the input value of -march=
Supporting additional rules for parsing ISA string.
- RISC-V ISA strings must be lowercase.
E.g.: rv32IMC is not supported, rv32imc is correct.
- Multi-letter extensions are to be separated by a single
underscore '_'. The extension prefix counts as a letter.
This means extensions that start with 's', 'sx' and 'sx'
are all multi-letter.
E.g.:
xasb is a single non-standard extension named 'xasb'
xa_sb are two extensions, the non-standard user level extension
'xa', and the supervisor level extension 'sb'.
- Standard user-level extensions are specified following
a canonical order, according to Table 22.1 in
RISC-V User-Level ISA V2.2.
- Non-standard user-level 'x' extensions,
standard supervisor-level 's' extensions and
non-standard supervisor-level 'sx' extensions
are also specified following a canonical order according
to Table 22.1 in RISC-V User-Level ISA V2.2:
'x' extensions, follwed by 's' extensions and then 'sx' extensions.
- Extensions might have a version number.
Underscores may be used to separate ISA subset components to
improve readability and to provide disambiguation.
E.g.: rv32i2_m3_a1_f2_d2
- Version numbers are divided into major and minor numbers,
separated by a 'p'. If the minor version is 0, then 'p0' can
be omitted.
- Additional checks for dependent extensions and invalid
extensions combinations.
E.g.:
'e' requires rv32
'e' can't be combined with 'f' nor 'd'
'q' requires rv64
- TODO items have also been marked with comments in the code.
[analyzer] Enable analysis of WebKit "unified sources".
Normally the analyzer begins path-sensitive analysis from functions within
the main file, even though the path is allowed to go through any functions
within the translation unit.
When a recent version of WebKit is compiled, the "unified sources" technique
is used, that assumes #including multiple code files into a single main file.
Such file would have no functions defined in it, so the analyzer wouldn't be
able to find any entry points for path-sensitive analysis.
This patch pattern-matches unified file names that are similar to those
used by WebKit and allows the analyzer to find entry points in the included
code files. A more aggressive/generic approach is being planned as well.
Martin Storsjo [Wed, 25 Apr 2018 21:23:59 +0000 (21:23 +0000)]
[Driver] Fix implicit config files from prefixed symlinks
If -no-canonical-prefixes isn't used, the clang executable name used
is the one of the actual executable, not the name of the symlink that
the user invoked.
In these cases, the target prefix was overridden based on the clang
executable name. (On the other hand the implicit -target option
that such a symlink adds, is added as an actual command line parameter
in tools/driver/driver.cop, before resolving the symlink and finding
the actual clang executable.
Use the original ClangNameParts (set from argv[0] in
tools/driver/driver.cpp) if it seems to be initialized propery.
All existing tests of this feature used -no-canonical-prefixes
(possibly because it also makes the driver look in the directory
of the symlink instead of the directory of the executable); add
another one that uses --config-user-dir= to specify the directory
instead. (For actual users of such symlinks, outisde of the test
suite, the directory is probably the same for both.)
This makes this feature work more like what the documentation
describes.
Eli Friedman [Wed, 25 Apr 2018 19:14:05 +0000 (19:14 +0000)]
[TargetInfo] Sort target features before passing them to the backend
Passing the features in random order will lead to unpredictable results
when some of the features are related (like the architecture-version
features on ARM).
It might be possible to fix this particular case in the ARM target code,
to avoid adding overlapping target features. But we should probably be
sorting in any case: the behavior shouldn't depend on StringMap's
hashing algorithm.
Peter Szecsi [Wed, 25 Apr 2018 17:28:03 +0000 (17:28 +0000)]
[ASTImporter] FriendDecl importing improvements
There are only a few cases of importing a frienddecl which is currently supported.
This patch aims to improve the friend import process.
Set FriendObjectKind in case of decls, insert friend into the friend chain
correctly, checks structurally equivalent in a more advanced manner.
Test cases added as well.
[CodeComplete] Fix completion in the middle of ident in ctor lists.
Summary:
The example that was broken before (^ designates completion points):
class Foo {
Foo() : fie^ld^() {} // no completions were provided here.
int field;
};
To fix it we don't cut off lexing after an identifier followed by code
completion token is lexed. Instead we skip the rest of identifier and
continue lexing.
This is consistent with behavior of completion when completion token is
right before the identifier.
[c++2a] [concepts] Add rudimentary parsing support for template concept declarations
This patch is a tweak of changyu's patch: https://reviews.llvm.org/D40381. It differs in that the recognition of the 'concept' token is moved into the machinery that recognizes declaration-specifiers - this allows us to leverage the attribute handling machinery more seamlessly.
See the test file to get a sense of the basic parsing that this patch supports.
There is much more work to be done before concepts are usable...
Yaxun Liu [Wed, 25 Apr 2018 01:10:37 +0000 (01:10 +0000)]
[HIP] Add hip input kind and codegen for kernel launching
HIP is a language similar to CUDA (https://github.com/ROCm-Developer-Tools/HIP/blob/master/docs/markdown/hip_kernel_language.md ).
The language syntax is very similar, which allows a hip program to be compiled as a CUDA program by Clang. The main difference
is the host API. HIP has a set of vendor neutral host API which can be implemented on different platforms. Currently there is open source
implementation of HIP runtime on amdgpu target (https://github.com/ROCm-Developer-Tools/HIP).
This patch adds support of input kind and language standard hip.
When hip file is compiled, both LangOpts.CUDA and LangOpts.HIP is turned on. This allows compilation of hip program as CUDA
in most cases and only special handling of hip program is needed LangOpts.HIP is checked.
This patch also adds support of kernel launching of HIP program using HIP host API.
When -x hip is not specified, there is no behaviour change for CUDA.
Patch by Greg Rodgers.
Revised and lit test added by Yaxun Liu.
[analyzer] Add support for the note diagnostic pieces to plist output format.
Note diagnostic pieces are an additional way of highlighting code sections to
the user. They aren't part of the normal path diagnostic sequence. They can
also be attached to path-insensitive reports.
Notes are already supported by the text output and scan-build.
Expanding our machine-readable plist output format to be able to represent notes
opens up the possibility for various analyzer GUIs to pick them up.
Joel E. Denny [Tue, 24 Apr 2018 14:50:23 +0000 (14:50 +0000)]
[Attr] Print enum attributes at correct position
For example, given:
void fn() {
enum __attribute__((deprecated)) T *p;
}
-ast-print produced:
void fn() {
enum T __attribute__((deprecated(""))) *p;
}
-ast-print on that produced:
void fn() {
enum T *p __attribute__((deprecated("")));
}
The attribute is on enum T in the first case, but it's on p in the
other cases.
Details:
Within enum declarations, enum attributes were always printed after
the tag and any member list. When no member list was present but the
enum was a type specifier in a variable declaration, the attribute
then applied to the variable not the enum, changing the semantics.
This patch fixes that by always printing attributes between the enum's
keyword and tag, as clang already does for structs, unions, and
classes.
Commit 08c258670c ("[OpenCL] Generic address space has been added in
OpenCL v2.0.", 2014-11-26) did not mark the thread_local keyword as a
storage class specifier, whereas it did mark _Thread_local as such.
According to the C++14 spec s7.1.1, thread_local is a storage class
specifier, so mark it as such.
I will add a test for this in a follow-up commit that adds keyword
restrictions to the OpenCL C++ language mode.
[CodeComplete] Fix completion at the end of keywords
Summary:
Make completion behave consistently no matter if it is run at the
start, in the middle or at the end of an identifier that happens to
be a keyword or a macro name. Since completion is often ran on
incomplete identifiers, they may turn into keywords by accident.
For example, we should produce same results for all of these
completion points:
// ^ is completion point.
^class
cla^ss
class^
Previously clang produced different results for the last case (as if
the completion point was after a space: `class ^`).
This change also updates some offsets in tests that (unintentionally?)
relied on the old behavior.
Aleksei Sidorin [Tue, 24 Apr 2018 10:11:53 +0000 (10:11 +0000)]
[ASTImporter] Allow testing of import sequences; fix import of typedefs for anonymous decls
This patch introduces the ability to test an arbitrary sequence of imports
between a given set of virtual source files. This should finally allow
us to write simple tests and fix annoying issues inside ASTImporter
that cause failures in CSA CTU. This is done by refactoring
ASTImporterTest functions and introducing `testImportSequence` facility.
As a side effect, `testImport` facility was generalized a bit more. It
should now allow import of non-decl AST nodes; however, there is still no
test using this ability.
As a "test for test", there is also a fix for import anonymous TagDecls
referred by typedef. Before this patch, the setting of typedef for anonymous
structure was delayed; however, this approach misses the corner case if
an enum constant is imported directly. In this patch, typedefs for
anonymous declarations are imported right after the anonymous declaration
is imported, without any delay.
Thanks to Adam Balogh for suggestions included into this patch.
Martin Storsjo [Tue, 24 Apr 2018 08:50:11 +0000 (08:50 +0000)]
[GCC] Match a GCC version with a patch suffix without a third version component
Previously it would only accept a string as a GCC version if it had
either two components and no suffix, or three components with an
optional suffix.
Debian and ubuntu provided mingw compilers have lib/gcc/target entries
like "5.3-posix" and "5.3-win32". This doesn't try to make any specific
preference between them (other than lexical sorting of the suffix).
Erik Verbruggen [Tue, 24 Apr 2018 08:39:46 +0000 (08:39 +0000)]
[libclang] Only mark CXCursors for explicit attributes with a type
All attributes have a source range associated with it. However, implicit
attributes are added by the compiler, and not added because the user
wrote something in the input. So no token type should be set to
CXCursor_*Attr.
The problem was visible when a class gets marked by e.g.
MSInheritanceAttr, which has the full CXXRecordDecl's range as its
own range. The effect of marking that range as CXCursor_UnexposedAttr
was that all cursors for the record decl, including all child decls,
would become CXCursor_UnexposedAttr.
projects / clang / log