The new driver uses an augmented linker wrapper to perform the device
linking phase, but to the user looks like a regular linker invocation.
Contrary to the old driver, the new driver contains all the information
necessary to produce a linked device image in the host object itself.
Currently, we infer the usage of the device linker by the user
specifying an offloading toolchain, e.g. (--offload-arch=...) or
(-fopenmp-targets=...), but this shouldn't be strictly necessary.
This patch introduces a new option `--offload-link` to tell
the driver to use the offloading linker instead. So a compilation flow
can now look like this,
```
clang foo.cu --offload-new-driver -fgpu-rdc --offload-arch=sm_70 -c
clang foo.o --offload-link -lcudart
```
I was considering if this could be merged into the `-fuse-ld` option,
but because the device linker wraps over the users linker it would
conflict with that. In the future it's possible to merge this into `lld`
completely or `gold` via a plugin and we would use this option to
enable the device linking feature. Let me know what you think for this.
Reviewed By: tra
Differential Revision: https://reviews.llvm.org/D126398
Const class members may be initialized with a defaulted default
constructor under the same conditions it would be allowed for a const
object elsewhere.
Differential Revision: https://reviews.llvm.org/D126170
Following the new flow for external object code emission,
provide flags to switch between integrated and external
backend similar to the integrated assembler options.
SPIR-V target is the only user of this functionality at
this point.
This patch also updated SPIR-V documentation to clarify
that integrated object code emission for SPIR-V is an
experimental feature.
Differential Revision: https://reviews.llvm.org/D125679
Since this didn't make it into the v14 release - anyone requesting the
v14 ABI shouldn't get this GCC-compatible change that isn't backwards
compatible with v14 Clang.
Differential Revision: https://reviews.llvm.org/D126334
For generic targets such as SPIR-V clang sets all OpenCL
extensions/features as supported by default. However
concrete targets are unlikely to support all extensions
features, which creates a problem when such generic SPIR-V
binary is compiled for a specific target later on.
To allow compile time diagnostics for unsupported features
this flag is now being exposed in the clang driver.
Differential Revision: https://reviews.llvm.org/D125243
Support for `__attribute__((no_builtin("foo")))` was added in https://reviews.llvm.org/D68028,
but builtins were still being used even when the attribute was placed on a function.
Reviewed By: hans
Differential Revision: https://reviews.llvm.org/D124701
WG14 DR423 (https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2148.htm#dr_423),
resolved during the C11 time frame, changed the way qualifiers are
handled on function return types and in cast expressions after it was
noted that these types are now directly observable via generic
selection expressions. In C, the function declarator is adjusted to
ignore all qualifiers (including _Atomic qualifiers).
Clang already handles the cast expression case correctly (by performing
the lvalue conversion, which drops the qualifiers as well), but with
these changes it will now also handle function declarations
appropriately.
Fixes#39595
Differential Revision: https://reviews.llvm.org/D125919
This brings clang/llvm into line with GCC. The Pass is still enabled for
the affected cores, but is now opt-in when using `-march=`.
I also took the opportunity to add release notes for this change.
Reviewed By: john.brawn
Differential Revision: https://reviews.llvm.org/D125775
This new CTU implementation is the natural extension of the normal single TU
analysis. The approach consists of two analysis phases. During the first phase,
we do a normal single TU analysis. During this phase, if we find a foreign
function (that could be inlined from another TU) then we don’t inline that
immediately, we rather mark that to be analysed later.
When the first phase is finished then we start the second phase, the CTU phase.
In this phase, we continue the analysis from that point (exploded node)
which had been enqueued during the first phase. We gradually extend the
exploded graph of the single TU analysis with the new node that was
created by the inlining of the foreign function.
We count the number of analysis steps of the first phase and we limit the
second (ctu) phase with this number.
This new implementation makes it convenient for the users to run the
single-TU and the CTU analysis in one go, they don't need to run the two
analysis separately. Thus, we name this new implementation as "onego" CTU.
Discussion:
https://discourse.llvm.org/t/rfc-much-faster-cross-translation-unit-ctu-analysis-implementation/61728
Differential Revision: https://reviews.llvm.org/D123773
The standard says:
The optional requires-clause ([temp.pre]) in an init-declarator or
member-declarator shall be present only if the declarator declares a
templated function ([dcl.fct]).
This implements that limitation, and updates the tests to the best of my
ability to capture the intent of the original checks.
Differential Revision: https://reviews.llvm.org/D125711
function in promise_type
According to https://cplusplus.github.io/CWG/issues/2585.html, this
fixes https://github.com/llvm/llvm-project/issues/54881
Simply, the clang tried to found (do lookup and overload resolution. Is
there any better word to use than found?) allocation function in
promise_type and global scope. However, this is not consistent with the
standard. The standard behavior would be that the compiler shouldn't
lookup in global scope in case we lookup the allocation function name in
promise_type. In other words, the program is ill-formed if there is
incompatible allocation function in promise type.
Reviewed By: erichkeane
Differential Revision: https://reviews.llvm.org/D125517
When a preprocessor directive is unknown outside of a skipped
conditional block, we give an error diagnostic because we don't know
how to proceed with preprocessing. But when the directive is in a
skipped conditional block, we would not diagnose it on the theory that
the directive may be known to an implementation other than Clang.
Now, for unknown directives inside a skipped conditional block, we
diagnose the unknown directive as a warning if it is sufficiently
similar to a directive specific to preprocessor conditional blocks. For
example, we'll warn about `#esle` and suggest `#else` but we won't warn
about `#progma` because it's not a directive specific to preprocessor
conditional blocks.
Fixes#51598
Differential Revision: https://reviews.llvm.org/D124726
Fixes the `FIXME:` related to adding `forEachTemplateArgument` to the
core AST Matchers library.
Reviewed By: aaron.ballman
Differential Revision: http://reviews.llvm.org/D125383
With sufficiently tortured code, it's possible to cause a stack
overflow when parsing declarators. Thus, we now check for resource
exhaustion when recursively parsing declarators so that we can at least
warn the user we're about to crash before we actually crash.
Fixes#51642
Differential Revision: https://reviews.llvm.org/D124915
This adds an extension warning when using the preprocessor conditionals
in a language mode they're not officially supported in, and an opt-in
warning for compatibility with previous standards.
Fixes#55306
Differential Revision: https://reviews.llvm.org/D125178
D87451 added -mignore-xcoff-visibility for AIX targets and made it the default (which mimicked the behaviour of the XL 16.1 compiler on AIX).
However, ignoring hidden visibility has unwanted side effects and some libraries depend on visibility to hide non-ABI facing entities from user headers and
reserve the right to change these implementation details based on this (https://libcxx.llvm.org/DesignDocs/VisibilityMacros.html). This forces us to use
internal linkage fallbacks for these cases on AIX and creates an unwanted divergence in implementations on the plaform.
For these reasons, it's preferable to not add -mignore-xcoff-visibility by default, which is what this patch does.
Reviewed By: DiggerLin
Differential Revision: https://reviews.llvm.org/D125141
In order to do offloading compilation we need to embed files into the
host and create fatbainaries. Clang uses a special binary format to
bundle several files along with their metadata into a single binary
image. This is currently performed using the `-fembed-offload-binary`
option. However this is not very extensibile since it requires changing
the command flag every time we want to add something and makes optional
arguments difficult. This patch introduces a new tool called
`clang-offload-packager` that behaves similarly to CUDA's `fatbinary`.
This tool takes several input files with metadata and embeds it into a
single image that can then be embedded in the host.
Reviewed By: tra
Differential Revision: https://reviews.llvm.org/D125165
This adds the -Wgnu-line-marker diagnostic flag, grouped under -Wgnu,
to warn about use of the GNU linemarker preprocessor extension.
Fixes#55067
Differential Revision: https://reviews.llvm.org/D124534
CUDA/HIP programs use __noinline__ like a keyword e.g.
__noinline__ void foo() {} since __noinline__ is defined
as a macro __attribute__((noinline)) in CUDA/HIP runtime
header files.
However, gcc and clang supports __attribute__((__noinline__))
the same as __attribute__((noinline)). Some C++ libraries
use __attribute__((__noinline__)) in their header files.
When CUDA/HIP programs include such header files,
clang will emit error about invalid attributes.
This patch fixes this issue by supporting __noinline__ as
a keyword, so that CUDA/HIP runtime could remove
the macro definition.
Reviewed by: Aaron Ballman, Artem Belevich
Differential Revision: https://reviews.llvm.org/D124866
The controlling expression of a _Generic selection expression undergoes
lvalue conversion, array conversion, and function conversion before
picking the association. This means that array types, function types,
and qualified types are all unreachable code if they're used as an
association. I've been caught by this twice in the past few months and
I figure that if a WG14 member can't seem to remember this rule, users
are also likely to struggle with it. So this adds an on-by-default
unreachable code diagnostic for generic selection expression
associations.
Note, we don't have to worry about function types as those are already
a constraint violation which generates an error.
Differential Revision: https://reviews.llvm.org/D125259
Ensures an -Wenum-conversion warning happens when one of the enums is
signed and the other is unsigned. Also adds a test file to verify these
warnings.
This warning would not happen since the -Wsign-conversion would make a
diagnostic then return, never allowing the -Wenum-conversion checks.
For example:
C
enum PE { P = -1 };
enum NE { N };
enum NE conv(enum PE E) { return E; }
Before this would only create a diagnostic with -Wsign-conversion and
never on -Wenum-conversion. Now it will create a diagnostic for both
-Wsign-conversion and -Wenum-conversion.
I could change it to just warn on -Wenum-conversion as that was what I
initially did. Seeing PR35200 (or GitHub Issue 316268), I let both
diagnostics check so that the sign conversion could generate a warning.
Like regular assignment, compound assignment operators can be assumed to
write to their left-hand side operand. So we strengthen the requirements
there. (Previously only the default read access had been required.)
Just like operator->, operator->* can also be assumed to dereference the
left-hand side argument, so we require read access to the pointee. This
will generate new warnings if the left-hand side has a pt_guarded_by
attribute. This overload is rarely used, but it was trivial to add, so
why not. (Supporting the builtin operator requires changes to the TIL.)
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D124966
This includes a fix for the libc++ issue I ran across with friend
declarations not properly being identified as overloads.
This reverts commit 45c07db31c.
Similar to the existing bitwise reduction builtins, this lowers to a llvm.vector.reduce.mul intrinsic call.
For other reductions, we've tried to share builtins for float/integer vectors, but the fmul reduction intrinsic also take a starting value argument and can either do unordered or serialized, but not reduction-trees as specified for the builtins. However we address fmul support this shouldn't affect the integer case.
Differential Revision: https://reviews.llvm.org/D117829
This reverts commit f6dff93641.
This diagnostic is also in the -Wcomment group, which is in the -Wall
group, so the diagnostic is enabled in a wider context than GCC does.
That turns out to be disruptive for the Linux kernel builds still using
-std=gnu89 because the kernel requires C source files to start with //
comments: https://kernel.org/doc/html/v5.18-rc5/process/license-rules.html#license-identifier-syntax
Compared to the old implementation:
* In C++, we only recurse into aggregate classes.
* Unnamed bit-fields are not printed.
* Constant evaluation is supported.
* Proper conversion is done when passing arguments through `...`.
* Additional arguments are supported and are injected prior to the
format string; this directly supports use with `fprintf`, for example.
* An arbitrary callable can be passed rather than only a function
pointer. In particular, in C++, a function template or overload set is
acceptable.
* All text generated by Clang is printed via `%s` rather than directly;
this avoids issues where Clang's pretty-printing output might itself
contain a `%` character.
* Fields of types that we don't know how to print are printed with a
`"*%p"` format and passed by address to the print function.
* No return value is produced.
Reviewed By: aaron.ballman, erichkeane, yihanaa
Differential Revision: https://reviews.llvm.org/D124221
GCC warns with a pedantic warning when -std=gnu89, but Clang would only
diagnose in -std=c89 mode. Clang now matches the GCC behavior in both
modes.
Fixes#18427
We had a think-o that would allow a user to declare a scoped
enumeration in C language modes "as a C++11 extension". This is a
think-o because there's no way for the user to spell the name of the
enumerators; C does not have '::' for a fully-qualified name. See
commit d0d87b5972 for details on why this
is unintentional for C.
Fixes#42372
If the operand to `sizeof` is an expression of VLA type, the operand is
still evaluated, so we should not issue a diagnostic about ignoring the
side effects in this case, as they're not actually ignored.
Fixes#48010
The flag was added when the C++20 draft did not allow for concept
caching. The final C++20 standard permits the caching, so flag is
redundant. See http://wg21.link/p2104r0.
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D125014
There are many more instances of this pattern, but I chose to limit this change to .rst files (docs), anything in libcxx/include, and string literals. These have the highest chance of being seen by end users.
Reviewed By: #libc, Mordante, martong, ldionne
Differential Revision: https://reviews.llvm.org/D124708
We'd nondeterministically assert (and later crash) when calculating the size or
alignment of a __bf16 type when the type isn't supported on a target because of
reading uninitialized values. Now we check whether the type is supported first.
Fixes#50171
When constant evaluating the initializer for an object of vector type,
we would call APInt::trunc() but truncate to the same bit-width the
object already had, which would cause an assertion. Instead, use
APInt::truncOrSelf() so that we no longer assert in this situation.
Fix#50216
We were failing to check if the controlling expression is dependent or
not when testing whether it has side effects. This would trigger an
assertion. Instead, if the controlling expression is dependent, we
suppress the check and diagnostic.
This fixes Issue 50227.
C89 allowed a type specifier to be elided with the resulting type being
int, aka implicit int behavior. This feature was subsequently removed
in C99 without a deprecation period, so implementations continued to
support the feature. Now, as with implicit function declarations, is a
good time to reevaluate the need for this support.
This patch allows -Wimplicit-int to issue warnings in C89 mode (off by
default), defaults the warning to an error in C99 through C17, and
disables support for the feature entirely in C2x. It also removes a
warning about missing declaration specifiers that really was just an
implicit int warning in disguise and other minor related cleanups.
This reverts commit a97899108e.
The patch caused some problems with the libc++ `__range_adaptor_closure`
that I haven't been able to figure out the cause of, so I am reverting
while I figure out whether this is a solvable problem/issue with the
CFE, or libc++ depending on an older 'incorrect' behavior.
The Itanium C++ ABI says prefixes are substitutable. For most prefixes
we already handle this: the manglePrefix(const DeclContext *, bool) and
manglePrefix(QualType) overloads explicitly handles substitutions or
defer to functions that handle substitutions on their behalf. The
manglePrefix(NestedNameSpecifier *) overload, however, is different and
handles some cases implicitly, but not all. The Identifier case was not
handled; this change adds handling for it, as well as a test case.
Reviewed By: erichkeane
Differential Revision: https://reviews.llvm.org/D122663
This patch turns on support for CR bit accesses for Power8 and above. The reason
why CR bits are turned on as the default for Power8 and above is that because
later architectures make use of builtins and instructions that require CR bit
accesses (such as the use of setbc in the vector string isolate predicate
and bcd builtins on Power10).
This patch also adds the clang portion to allow for turning on CR bits in the
front end if the user so desires to.
Differential Revision: https://reviews.llvm.org/D124060
This reverts commit 0c31da4838.
I've solved the issue with the PointerUnion by making the
`FunctionTemplateDecl` pointer be a NamedDecl, that could be a
`FunctionDecl` or `FunctionTemplateDecl` depending. This is enforced
with an assert.
This reverts commit 4b6c2cd647.
The patch caused numerous ARM 32 bit build failures, since we added a
5th item to the PointerUnion, and went over the 2-bits available in the
32 bit pointers.
As reported here: https://github.com/llvm/llvm-project/issues/44178
Concepts are not supposed to be instantiated until they are checked, so
this patch implements that and goes through significant amounts of work
to make sure we properly re-instantiate the concepts correctly.
Differential Revision: https://reviews.llvm.org/D119544
Summary:
Some previous patches introduced the `--offload-new-driver` flag, which
is a generic way to enable the new driver, and the `--offload-host-only`
and `--offload-device-only` flags which allow users to compile for one
side, making it easier to inspect intermediate code for offloading
compilations. This patch just documents them in the command line
reference.
By default -fsanitize=address already compiles with this check, why not use it.
For compatibly it can be disabled with env ASAN_OPTIONS=detect_stack_use_after_return=0.
Reviewed By: eugenis, kda, #sanitizers, hans
Differential Revision: https://reviews.llvm.org/D124057
This reverts commit b0bc93da92.
Changes: `s/_WIN32/_WIN64/g` in clang/test/SemaCXX/attr-trivial-abi.cpp.
The calling convention is specific to 64-bit windows. It's even in the name: `CCK_MicrosoftWin64`.
After this, the test passes with both `-triple i686-pc-win32` and `-triple x86_64-pc-win32`. Phew!
Reviewed By: gribozavr2
Differential Revision: https://reviews.llvm.org/D123059
Temporarily revert the option to fix
https://bugs.chromium.org/p/chromium/issues/detail?id=1319387
This reverts option default and documentation of the commit. Test
changes are not being reverted as they are improvement and break
reliance on option defaults.
Additional memory usage is a problem on mobile devices with low memory.
Even heavy thread desktop programs may need some FakeStack tunning.
This reverts commit 4b4437c084.
Introduced by 23a5090c6, this style option marker indicated
'clang-format 9', though its respective option was available in
an earlier release.
Differential Revision: https://reviews.llvm.org/D123299
This change adds an option to detect all null dereferences for
non-default address spaces, except for address spaces 256, 257 and 258.
Those address spaces are special since null dereferences are not errors.
All address spaces can be considered (except for 256, 257, and 258) by
using -analyzer-config
core.NullDereference:DetectAllNullDereferences=true. This option is
false by default, retaining the original behavior.
A LIT test was enhanced to cover this case, and the rst documentation
was updated to describe this behavior.
Reviewed By: steakhal
Differential Revision: https://reviews.llvm.org/D122841
When doing overload resolution, we have to check that candidates' parameter types are equal before trying to find a better candidate through checking which candidate is more constrained.
This revision adds this missing check and makes us diagnose those cases as ambiguous calls when the types are not equal.
Fixes GitHub issue https://github.com/llvm/llvm-project/issues/53640
Reviewed By: erichkeane
Differential Revision: https://reviews.llvm.org/D123182
By default -fsanitize=address already compiles with this check,
why not use it.
For compatibly it can be disabled with env ASAN_OPTIONS=detect_stack_use_after_return=0.
Reviewed By: eugenis, kda, #sanitizers, hans
Differential Revision: https://reviews.llvm.org/D124057
These command-line flags are alternates to providing the -x
c++-*-header indicators that we are building a header unit.
Act on fmodule-header= for headers on the c/l:
If we have x.hh -fmodule-header, then we should treat that header
as a header unit input (equivalent to -xc++-header-unit-header x.hh).
Likewise, for fmodule-header={user,system} the source should be now
recognised as a header unit input (since this can affect the job list
that we need).
It's not practical to recognise a header without any suffix so
-fmodule-header=system foo isn't going to happen. Although
-fmodule-header=system foo.hh will work OK. However we can make it
work if the user indicates that the item without a suffix is a valid
header. (so -fmodule-header=system -xc++-header vector)
Differential Revision: https://reviews.llvm.org/D121589
This is extended to all `std::` functions that take a reference to a
value and return a reference (or pointer) to that same value: `move`,
`forward`, `move_if_noexcept`, `as_const`, `addressof`, and the
libstdc++-specific function `__addressof`.
We still require these functions to be declared before they can be used,
but don't instantiate their definitions unless their addresses are
taken. Instead, code generation, constant evaluation, and static
analysis are given direct knowledge of their effect.
This change aims to reduce various costs associated with these functions
-- per-instantiation memory costs, compile time and memory costs due to
creating out-of-line copies and inlining them, code size at -O0, and so
on -- so that they are not substantially more expensive than a cast.
Most of these improvements are very small, but I measured a 3% decrease
in -O0 object file size for a simple C++ source file using the standard
library after this change.
We now automatically infer the `const` and `nothrow` attributes on these
now-builtin functions, in particular meaning that we get a warning for
an unused call to one of these functions.
In C++20 onwards, we disallow taking the addresses of these functions,
per the C++20 "addressable function" rule. In earlier language modes, a
compatibility warning is produced but the address can still be taken.
The same infrastructure is extended to the existing MSVC builtin
`__GetExceptionInfo`, which is now only recognized in namespace `std`
like it always should have been.
This is a re-commit of
fc30901096,
a571f82a50,
64c045e25b, and
de6ddaeef3,
and reverts aa643f455a.
This change also includes a workaround for users using libc++ 3.1 and
earlier (!!), as apparently happens on AIX, where std::move sometimes
returns by value.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D123345
Revert "Fixup D123950 to address revert of D123345"
This reverts commit aa643f455a.
This reverts commit 69dd89fdcb.
This reverts commit 04000c2f92.
The current states breaks libstdc++ usage (https://reviews.llvm.org/D119136#3455423).
The fixup has been reverted as it caused other valid code to be disallowed.
I think we should start from the clean state by reverting all relevant commits.
WG14 has elected to remove support for K&R C functions in C2x. The
feature was introduced into C89 already deprecated, so after this long
of a deprecation period, the committee has made an empty parameter list
mean the same thing in C as it means in C++: the function accepts no
arguments exactly as if the function were written with (void) as the
parameter list.
This patch implements WG14 N2841 No function declarators without
prototypes (http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2841.htm)
and WG14 N2432 Remove support for function definitions with identifier
lists (http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2432.pdf).
It also adds The -fno-knr-functions command line option to opt into
this behavior in other language modes.
Differential Revision: https://reviews.llvm.org/D123955
C89 had a questionable feature where the compiler would implicitly
declare a function that the user called but was never previously
declared. The resulting function would be globally declared as
extern int func(); -- a function without a prototype which accepts zero
or more arguments.
C99 removed support for this questionable feature due to severe
security concerns. However, there was no deprecation period; C89 had
the feature, C99 didn't. So Clang (and GCC) both supported the
functionality as an extension in C99 and later modes.
C2x no longer supports that function signature as it now requires all
functions to have a prototype, and given the known security issues with
the feature, continuing to support it as an extension is not tenable.
This patch changes the diagnostic behavior for the
-Wimplicit-function-declaration warning group depending on the language
mode in effect. We continue to warn by default in C89 mode (due to the
feature being dangerous to use). However, because this feature will not
be supported in C2x mode, we've diagnosed it as being invalid for so
long, the security concerns with the feature, and the trivial
workaround for users (declare the function), we now default the
extension warning to an error in C99-C17 mode. This still gives users
an easy workaround if they are extensively using the extension in those
modes (they can disable the warning or use -Wno-error to downgrade the
error), but the new diagnostic makes it more clear that this feature is
not supported and should be avoided. In C2x mode, we no longer allow an
implicit function to be defined and treat the situation the same as any
other lookup failure.
Differential Revision: https://reviews.llvm.org/D122983
Reimplements MisExpect diagnostics from D66324 to reconstruct its
original checking methodology only using MD_prof branch_weights
metadata.
New checks rely on 2 invariants:
1) For frontend instrumentation, MD_prof branch_weights will always be
populated before llvm.expect intrinsics are lowered.
2) for IR and sample profiling, llvm.expect intrinsics will always be
lowered before branch_weights are populated from the IR profiles.
These invariants allow the checking to assume how the existing branch
weights are populated depending on the profiling method used, and emit
the correct diagnostics. If these invariants are ever invalidated, the
MisExpect related checks would need to be updated, potentially by
re-introducing MD_misexpect metadata, and ensuring it always will be
transformed the same way as branch_weights in other optimization passes.
Frontend based profiling is now enabled without using LLVM Args, by
introducing a new CodeGen option, and checking if the -Wmisexpect flag
has been passed on the command line.
Reviewed By: tejohnson
Differential Revision: https://reviews.llvm.org/D115907
Summary:
This patch removes the OpenMP sections in the release notes. These will
be filled once the release is close and implementations are finalized.