CXXDeductionGuideDecl inherits from FunctionDecl. For FunctionDecls,
the JSONVisitor includes a call to visit NamedDecl in order to provide
mangled names in the dump. This did not correctly exclude
CXXDeductionGuideDecl, which resulted in an assert being hit.
Prior to this, clang would always report
```
compile with '-ffixed-point' to enable fixed point types
```
whenever it sees `_Accum`, `_Fract`, or `_Sat` when fixed point
arithmetic is not enabled. This can break existing code that uses these
as variable names and doesn't use fixed point arithmetic like in some
microsoft headers
(https://github.com/llvm/llvm-project/pull/67750#issuecomment-1775264907).
Fixed point should not raise this error for these cases, so this removes
the error altogether and defaults to the usual error clang gives where
it can see these keywords as either unknown types or regular variables.
When emitting LLVM IR for gather loads/scatter stores, the predicate
parameter is cast to a type that depends on the loaded, resp. stored
type. That's correct for operation where we have a predicate per lane,
however it is not correct for quadword loads and stores (`LD1Q`, `ST1Q`)
where the predicate is per 128-bit chunk, independent from the ACLE
intrinsic type.
This can be universally handled by cast to the corresponding parameter
type of the intrinsic. The intrinsic itself should be defined in a way
that enforces relations between parameter types.
Recommits the changes from https://reviews.llvm.org/D148216.
Explicitly named globals are now matched literally, instead of emitting
a capture group for the name. This resolves#70047.
Metadata and annotations, on the other hand, are captured and matched
against by default, since their identifiers are not stable.
The reasons for revert (#63746) have been fixed:
The first issue, that of duplicated checkers, has already been resolved
in #70050.
This PR resolves the second issue listed in #63746, regarding the order
of named and unnamed globals. This is fixed by recording the index of
substrings containing global values, and sorting the checks according to
that index before emitting them. This results in global value checks
being emitted in the order they were seen instead of being grouped
separately.
Add clang builtins for the new tied wmma intrinsics.
These variations tie the destination
accumulator matrix to the input
accumulator matrix.
See https://github.com/llvm/llvm-project/pull/69903 for context.
When instantiation function, a call to Sema::resetFPOption was used to
set the FP options associated with AST node. However this function also
cleared FP pragma stack, and it is incorrect. Template instantiation
takes place on AST representation and semantic information like the FP
pragma stack should not affect it. This was a reason for miscompilation
in some cases.
To make the Sema interface more consistent, now `resetFPOptions` does
not clear FP pragma stack anymore. It is cleared in
`FpPragmaStackSaveRAII`, which is used in parsing only.
This change must fix https://github.com/llvm/llvm-project/issues/69717
(Problems with float_control pragma stack in Clang 17.x).
-fveclib= allows users to choose a vectorized libm so that loops
containing math functions are vectorized.
This is implemented as much as possible in the same way as in clang. The
driver test in veclib.f90 is copied from the clang test.
The data size is required for implementing the `memmove` optimization
for `std::copy`, `std::move` etc. correctly as well as replacing
`__compressed_pair` with `[[no_unique_address]]` in libc++. Since the
compiler already knows the data size, we can avoid some complexity by
exposing that information.
This reverts commit e8fe4de64ffb84924c41e54116a04570046eed74.
memcpy/memmove instrumentation for -fsanitize=alignment has been tested
on a huge code base. There were some cleanups but the number does not
justify a workaround.
This patch relands https://github.com/llvm/llvm-project/pull/70639
It was reverted because under certain conditions we triggered an
assertion
in `DIBuilder`. Specifically, in the original patch we called
`EmitGlobalVariable`
at the end of `CGDebugInfo::finalize`, after all the temporary `DIType`s
have
been uniqued. With limited debug-info such temporary nodes would be
created
more frequently, leaving us with non-uniqued nodes by the time we got to
`DIBuilder::finalize`; this violated its pre-condition and caused
assertions to trigger.
To fix this, the latest iteration of the patch moves
`EmitGlobalVariable` to the
beginning of `CGDebugInfo::finalize`. Now, when we create a temporary
`DIType` node as a result of emitting a variable definition, it will get
uniqued
in time. A test-case was added for this scenario.
We also now don't emit a linkage name for non-locationed constants since
LLDB doesn't make use of it anyway.
Original commit message:
"""
When an LLDB user asks for the value of a static data member, LLDB
starts
by searching the Names accelerator table for the corresponding variable
definition DIE. For static data members with out-of-class definitions
that
works fine, because those get represented as global variables with a
location
and making them eligible to be added to the Names table. However,
in-class
definitions won’t get indexed because we usually don't emit global
variables
for them. So in DWARF we end up with a single `DW_TAG_member` that
usually holds the constant initializer. But we don't get a corresponding
CU-level `DW_TAG_variable` like we do for out-of-class definitions.
To make it more convenient for debuggers to get to the value of inline
static data
members, this patch makes sure we emit definitions for static variables
with
constant initializers the same way we do for other static variables.
This also aligns
Clang closer to GCC, which produces CU-level definitions for inline
statics and also
emits these into `.debug_pubnames`.
The implementation keeps track of newly created static data members.
Then in `CGDebugInfo::finalize`, we emit a global `DW_TAG_variable` with
a
`DW_AT_const_value` for any of those declarations that didn't end up
with a
definition in the `DeclCache`.
The newly emitted `DW_TAG_variable` will look as follows:
```
0x0000007b: DW_TAG_structure_type
DW_AT_calling_convention (DW_CC_pass_by_value)
DW_AT_name ("Foo")
...
0x0000008d: DW_TAG_member
DW_AT_name ("i")
DW_AT_type (0x00000062 "const int")
DW_AT_external (true)
DW_AT_declaration (true)
DW_AT_const_value (4)
Newly added
vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
0x0000009a: DW_TAG_variable
DW_AT_specification (0x0000008d "i")
DW_AT_const_value (4)
DW_AT_linkage_name ("_ZN2t2IiE1iIfEE")
```
This patch also drops the `DW_AT_const_value` off of the declaration
since we
now always have it on the definition. This ensures that the
`DWARFParallelLinker`
can type-merge class with static members where we couldn't attach the
constant
on the declaration in some CUs.
"""
Dependent changes:
* https://github.com/llvm/llvm-project/pull/71800
Closes https://github.com/llvm/llvm-project/issues/61460.
We have FunctionParmPackExpr that serves as the unexpanded expression
but from which the visitor collects none, which may lead to assertion
failure during the template instantiation.
This is a rather short list that does not include the list
of resolvded CWG issues, which is rather long (P3046).
Some of the links are currently invalid as they have yet to be
publicly published.
With opaque pointers, `CreatePointerBitCastOrAddrSpaceCast` can be replaced with `CreateAddrSpaceCast`.
Replace or remove uses of `CreatePointerBitCastOrAddrSpaceCast`.
Opaque pointer cleanup effort.
Emission of `mustprogress` attribute previously occurred only within
`EmitFunctionBody`, after generating the function body. Other routines
for function body creation may lack the attribute, potentially leading
to suboptimal optimizations later in the pipeline. Attribute emission
is now anticipated prior to generating the function body.
Fixes: https://github.com/llvm/llvm-project/issues/69833.
This is a hacky fix to get the bots green, because just reverting the
original commit causes more failures. This should be replaced with a
proper fix by the author.
This reverts commit 8ee07a4be7f7d8654ecf25e7ce0a680975649544.
The revert is breaking AMDGPU backend tests (which I didn't have
enabled), and I don't want to risk breakages over the weekend, so just
revert for now.
This reverts commit 3bb7ecc5a515d2ddd2708257096e14e3fe3b839d.
The revert breaks AMDGPU backend tests (which I didn't have enabled
locally), and I don't want to risk more breakages before the weekend, so
just restore things as they were.
This reverts commit cd7ba9f3d090afb5d3b15b0dcf379d15d1e11e33.
This is causing test failures on builds without assertions:
/data/users/smeenai/llvm-project/clang/test/Analysis/builtin_signbit.cpp:14:21: error: CHECK-BE32-NEXT: expected string not found in input
// CHECK-BE32-NEXT: entry:
^
Bots are also failing, e.g. https://lab.llvm.org/buildbot/#/builders/67/builds/13266/steps/7/logs/stdio
This reverts commit 82f68a992b9f89036042d57a5f6345cb2925b2c1.
cd7ba9f3d090afb5d3b15b0dcf379d15d1e11e33 needs to be reverted to fix
test failures on builds without assertions, and this one needs to be
reverted first for that.
If we have more than one reduction variable we need to be consistent
wrt. indexing. In 3de645efe30b83ba1b6d7e500486c4f441a17a61 we broke this
as the buffer type was reduced to a singleton but the index computation
was not adjusted to account for that offset. This fixes it by
interleaving the reduction variables properly in a array-of-struct
style. We can revert it back to struct-of-array in a follow up if turns
out to be a problem. I doubt it since half the accesses should benefit
from the locallity this layout offers and only the other half were
consecutive before.
Summary:
This patch reworks how we handle global constructors in OpenMP.
Previously, we emitted individual kernels that were all registered and
called individually. In order to provide more generic support, this
patch moves all handling of this to the target backend and the runtime
plugin. This has the benefit of supporting the GNU extensions for
constructors an destructors, removing a class of failures related to
shared library destruction order, and allows targets other than OpenMP
to use the same support without needing to change the frontend.
This is primarily done by calling kernels that the backend emits to
iterate a list of ctor / dtor functions. For x64, this is automatic and
we get it for free with the standard `dlopen` handling. For AMDGPU, we
emit `amdgcn.device.init` and `amdgcn.device.fini` functions which
handle everything atuomatically and simply need to be called. For NVPTX,
a patch https://github.com/llvm/llvm-project/pull/71549 provides the
kernels to call, but the runtime needs to set up the array manually by
pulling out all the known constructor / destructor functions.
One concession that this patch requires is the change that for GPU
targets in OpenMP offloading we will use `llvm.global_dtors` instead of
using `atexit`. This is because `atexit` is a separate runtime function
that does not mesh well with the handling we're trying to do here. This
should be equivalent in all cases except for cases where we would need
to destruct manually such as:
```
struct S { ~S() { foo(); } };
void foo() {
static S s;
}
```
However this is broken in many other ways on the GPU, so it is not
regressing any support, simply increasing the scope of what we can
handle.
This changes the handling of ctors / dtors. This patch now outputs a
information message regarding the deprecation if the old format is used.
This will be completely removed in a later release.
Depends on: https://github.com/llvm/llvm-project/pull/71549
Sometimes bpf developer might want to develop different codes
based on particular cpu versioins. For example, cpu v1/v2/v3
branch target is 16bit while cpu v4 branch target is 32bit,
thus cpu v4 allows more aggressive loop unrolling than cpu v1/v2/v3
(see [1] for a kernel selftest failure due to this).
We would like to maintain aggressive loop unrolling for cpu v4
while limit loop unrolling for earlier cpu versions.
Another example, signed divide also only available with cpu v4.
Actually, adding cpu specific macros are fairly common
in llvm. For example, x86 has maco like 'i486', '__pentium_mmx__', etc.
AArch64 has '__ARM_NEON', '__ARM_FEATURE_SVE', etc.
This patch added __BPF_CPU_VERSION__ macro. Current possible values
are 0/1/2/3/4. The following are the -mcpu=... to __BPF_CPU_VERSION__
mapping:
```
cpu __BPF_CPU_VERSION__
no -mcpu=<...> 1
-mcpu=v1 1
-mcpu=v2 2
-mcpu=v3 3
-mcpu=v4 4
-mcpu=generic 1
-mcpu=probe 0
```
This patch also added some macros for developers to identify some cpu
insn features:
```
feature macro enabled in which cpu
__BPF_FEATURE_JMP_EXT >= v2
__BPF_FEATURE_JMP32 >= v3
__BPF_FEATURE_ALU32 >= v3
__BPF_FEATURE_LDSX >= v4
__BPF_FEATURE_MOVSX >= v4
__BPF_FEATURE_BSWAP >= v4
__BPF_FEATURE_SDIV_SMOD >= v4
__BPF_FEATURE_GOTOL >= v4
__BPF_FEATURE_ST >= v4
```
[1]
https://lore.kernel.org/bpf/3e3a8a30-dde0-43a1-981e-2274962780ef@linux.dev/
Currently flang's runtime libraries are only built for the specific CRT
that LLVM itself was built against. This patch adds the cmake logic for
building a separate runtime for each CRT configuration and adds a flag
for selecting a CRT configuration to link against.
This is a follow-up to febf5c97bba7910e796041c9518fce01f31ae826 and
another instance of #64121.
The added test only fails if Clang is built with libc++ and a enabled
debug check for strict weak ordering.
