Reland with debug traces to try to understand a bug that only happens on
one CI configuration
===
This introduces a way detect the libstdc++ version,
use that to enable workarounds.
The version is cached.
This should make it easier in the future to find and remove
these hacks.
I did not find the need for enabling a hack between or after
specific versions, so it's left as a future exercise.
We can extend this fature to other libraries as the need arise.
===
At the time `__reference_constructs_from_temporary` got implemented,
`__reference_binds_to_temporary` was mentioned as deprecated in
`LanguageExtensions.rst`, but no deprecation warning was emitted. This
PR adds the previously missing warning.
This PR reverts a change made in #126846.
#126846 introduced an ``-Wimplicit-int-conversion`` diagnosis for
```c++
int8_t x = something;
x = -x; // warning here
```
This is technically correct since -x could have a width of 9, but this
pattern is common in codebases.
Reverting this change would also introduce the false positive I fixed in
#126846:
```c++
bool b(signed char c) {
return -c >= 128; // -c can be 128
}
```
This false positive is uncommon, so I think it makes sense to revert the
change.
With the current behavior the following example yields a linker error:
"multiple definition of `foo.default'"
// Translation Unit 1
__attribute__((target_clones("dotprod, sve"))) int foo(void) { return 1; }
// Translation Unit 2
int foo(void) { return 0; }
__attribute__((target_version("dotprod"))) int foo(void);
__attribute__((target_version("sve"))) int foo(void);
int bar(void) { return foo(); }
That is because foo.default is generated twice. As a user I don't find
this particularly intuitive. If I wanted the default to be generated in
TU1 I'd rather write target_clones("dotprod, sve", "default")
explicitly.
When changing the code I noticed that the RISC-V target defers the
resolver emission when encountering a target_version definition. This
seems accidental since it only makes sense for AArch64, where we only
emit a resolver once we've processed the entire TU, and only if the
default version is present. I've changed this so that RISC-V immediately
emmits the resolver. I adjusted the codegen tests since the functions
now appear in a different order.
Implements https://github.com/ARM-software/acle/pull/377
The previous approach broke code generation for the MS ABI due to an
unintended code path during constraint substitution. This time we
address the issue by inspecting the evaluation contexts and thereby
avoiding that code path.
This reapplies 96eced624 (#102857).
Microsoft helpfully defines `THIS` to `void` in two different platform
SDK headers, at least one of which is reachable via <Windows.h>. We have
a user who ran into a build because of `THIS` unfortunate macro name
collision.
Rename the members to better match our naming conventions.
Fixes#142186
This introduces a way detect the libstdc++ version, use that to enable
workarounds.
The version is cached.
This should make it easier in the future to find and remove these hacks.
I did not find the need for enabling a hack between or after specific
versions, so it's left as a future exercise.
We can extend this fature to other libraries as the need arise.
The 'capture' modifier is an OpenACC 3.3NEXT (AKA 3.4) feature, which
permits a new kind of identifying the memory location of variables in a
data clause. However, it is only valid on data, combined, or compute
constructs.
This patch implements all of the proper restrictions.
Fixes#141397
Element-wise math builtins (e.g.
__builtin_elementwise_max/__builtin_elementwise_pow etc.) fail when
their arguments have different qualifications, but should be
well-formed. The fix is to use hasSameUnqualifiedType to check if the
arguments match.
This fixes a regression introduced in
https://github.com/llvm/llvm-project/pull/130537,
which was reported here:
https://github.com/llvm/llvm-project/pull/132401#issuecomment-2839690599
When deducing member pointers class, ignore top level qualifiers on the
argument side, since the class portion of a member pointer is a
nested-name-specifier, and these just nominate an entity.
Qualifiers on the parameter side are fine since deduction allows the
parameter side to be more qualified, and these qualifiers won't be part
of the result.
Since this regression was never released, there are no release notes.
This patch fixes:
clang/lib/Sema/SemaSPIRV.cpp:132:3: error: default label in switch
which covers all enumeration values
[-Werror,-Wcovered-switch-default]
The patch introduce __builtin_spirv_generic_cast_to_ptr_explicit which
is lowered to the llvm.spv.generic.cast.to.ptr.explicit intrinsic.
The SPIR-V builtins are now split into 3 differents file:
BuiltinsSPIRVCore.td,
BuiltinsSPIRVVK.td for Vulkan specific builtins, BuiltinsSPIRVCL.td for
OpenCL specific builtins
and BuiltinsSPIRVCommon.td for common ones.
The patch also introduces a new header defining its SPIR-V friendly
equivalent (__spirv_GenericCastToPtrExplicit_ToGlobal,
__spirv_GenericCastToPtrExplicit_ToLocal and
__spirv_GenericCastToPtrExplicit_ToPrivate). The functions are declared
as aliases to the new builtin allowing C-like languages to have a
definition to rely on as well as gaining proper front-end diagnostics.
The motivation for the header is to provide a stable binding for
applications or library (such as SYCL) and allows non SPIR-V targets to
provide an implementation (via libclc or similar to how it is done for
gpuintrin.h).
This adds support under LoongArch for the target("..") attributes.
The supported formats are:
- "arch=<arch>" strings, that specify the architecture features for a
function as per the -march=arch option.
- "tune=<cpu>" strings, that specify the tune-cpu cpu for a function as
per -mtune.
- "<feature>", "no-<feature>" enabled/disables the specific feature.
The bulk of the changes are in `CallExpr`
We cache Begin/End source locs in the trailing objects, in the space
left by making the offset to the trailing objects static.
We also set a flag to indicate that we are calling an explicit object
member function, further reducing the cost of getBeginLoc.
Fixes#140876
Summary:
When there are overloaded C++ operators in the global namespace the AST
node for these is not a `BinaryExpr` but a `CXXOperatorCallExpr`. Modify
the uses to handle this case, basically just treating it as a binary
expression with two arguments.
Fixes https://github.com/llvm/llvm-project/issues/141085
032ad59 taught the instantiator to expand template argument packs for
rewrite. However we might already be in a pack expansion when we
synthesizing the CTAD guide, so we reset the ArgPackSubstIndex to ensure
it doesn't get confused.
No release note because this is a regression in Clang 21.
Fixes https://github.com/llvm/llvm-project/issues/141425
---------
Co-authored-by: Erich Keane <ekeane@nvidia.com>
Resolves#138939
When enabling `--fno-exceptions` flag, discarded statements containing
`try/catch/throw` in an independent context can be avoided from being
rejected.
We did not handle the case where a variable could be initialized by a
CXXParenListInitExpr.
---------
Co-authored-by: Shafik Yaghmour <shafik.yaghmour@intel.com>
Adds resource name argument to resource class constructors and to builtin functions that initialize resource handles
`__builtin_hlsl_resource_handlefrombinding` and `__builtin_hlsl_resource_handlefromimplicitbinding`.
Part 1/4 of https://github.com/llvm/llvm-project/issues/105059
I think the intent of df18ee96206 was to substitute only those non-packs
into a pack expansion type (e.g. `T` in `T::pack...`), so let's hold off
pack expansions explicitly, in case there are calls coming from a
substitution of pack expansion.
Fixes https://github.com/llvm/llvm-project/issues/53609
Introduce the `reentrant_capability` attribute, which may be specified
alongside the `capability(..)` attribute to denote that the defined
capability type is reentrant. Marking a capability as reentrant means
that acquiring the same capability multiple times is safe, and does not
produce warnings on attempted re-acquisition.
The most significant changes required are plumbing to propagate if the
attribute is present to a CapabilityExpr, and introducing
ReentrancyDepth to the LockableFactEntry class.
These are identified by misc-include-cleaner. I've filtered out those
that break builds. Also, I'm staying away from llvm-config.h,
config.h, and Compiler.h, which likely cause platform- or
compiler-specific build failures.
`static_assert(std::is_xx_v<MyType>);` is a common pattern to check that
a type meets a requirement.
This patch produces diagnostics notes when such assertion fails. The
first type trait for which we provide detailed explanation is
std::is_trivially_relocatable.
We employ the same mechanisn when a type trait appears an an unsatisfied
atomic constraint.
I plan to also support `std::is_trivially_replaceable` in a follow up
PR, and hopefully, over time we can support more type traits.
We check the accessibility of constructors when initializing a default
argument whose type is not an aggregate.
Make sure the check is performed within the correct DeclContext.
Otherwise, it will be delayed until the end of the declaration, at which
point the context is mismatched.
Fixes#62444
Fixes https://github.com/llvm/llvm-project/issues/83608
Since arrays decay into pointers, no need to check them for arguments.
This commit reverts part of the changes from the commit "[OpenCL] Check
for invalid kernel arguments in array types"
3b238ed6626983beb238b95eada4172184fb2d29.
