Secondary cache entries are now released to the OS from least recent to
most recent entries. This helps to avoid unnecessary scans of the cache
since entries ready to be released (specifically, entries that are
considered old relative to the configurable release interval) will
always be at the tail of the list of committed entries by the LRU
ordering. For this same reason, the `OldestTime` variable is no longer
needed to indicate when releases are necessary so it has been removed.
The plan was to make `eh-assembly.s` contain both the legacy and the new
tests, but the new tests require `--no-type-check` because the type
checker for the new EH is in progress. In case this drags on further
than expected, this renames the current file to `-legacy.s` in order to
follow the current naming scheme in `test/CodeGen/WebAssembly`.
After landing this first, `eh-assembly-new.s` in #108668 will be renamed
to `eh-assembly.s`.
This is an implementation of `ctime` and includes `ctime_r`.
According to documentation, `ctime` and `ctime_r` are defined as the
following:
```c
char *ctime(const time_t *timep);
char *ctime_r(const time_t *restrict timep, char buf[restrict 26]);
```
closes#86567
There is possibility of
static_tls_begin is set and static_tls_end is not yet
The test reproduces the case.
Stack trace looks like this:
* `MsanThread::Init`
* `SetThreadStackAndTls`
* `GetThreadStackAndTls`
* `GetThreadStackTopAndBottom`
* `pthread_getattr_np`
* `realloc`
* `__sanitizer_malloc_hook`
* TLS access
* `___interceptor___tls_get_addr`
* `DTLS_on_tls_get_addr`
The issue is that `SetThreadStackAndTls` implementation
stores `tls_begin` before `GetThreadStackTopAndBottom`,
and `tls_end` after. So we have partially initialized
state in `DTLS_on_tls_get_addr`.
In the process of adding scalarization support for DirectX target
intrinsics I found that intrinsics that weren't marked with `IntrNoMem`
did not get removed by
`RecursivelyDeleteTriviallyDeadInstructionsPermissive`. So this change
is to make it more clear that our intrinsics don't have side effects.
I only added `IntrNoMem` to the intrinics in `IntrinsicsDirectX.td` I
was involved with. There a potentially a few other cases that might
warrant this attribute, but will need input on the others.
These are no longer experimental after
051054e6f74303bc880221e88671745f363964cc. I left the section because we
will be adding intrinsics for Zvkgs and Zvbc32e.
Unlike scalar, where AArch64 prefers expanding scmp/ucmp with select,
under Neon we can use the arithmetic expansion to generate fewer
instructions. Notably it also prevents the scalarization of vselect
during vector-legalization.
Fusion of reshapes by collapsing patterns were restricted to single
result operations, but the implementation supports multi result ops.
This PR removes the restriction, since it is not necessary.
This replaces the previous Code Owners section of our developer policy
with a new section for Maintainers. It also updates most of the places
we mention "code owner" in the documentation (it does not update the
files named `Code Owners.rst` or similar because those should be updated
when the subprojects add their `Maintainers.rst` file).
The wording was taken from what was proposed in the RFC (including all
suggested amendments from folks on the thread).
Please see the RFC for more details:
https://discourse.llvm.org/t/rfc-proposing-changes-to-the-community-code-ownership-policy/80714/
The "clustered" nodes for buildvector nodes must be postponed in
accordance with the global flag, otherwise it may cause crash because of
the dependency between phi nodes.
`gnu_get_libc_version` unlike `confstr` is not
intercepted. We should be able to use this
function earier.
Looks like we use `confstr` staring from
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60038
but there is no specific reason to refer it over
`gnu_get_libc_version`.
Add new entry points for more complex data transfer involving
descriptors. These functions will be called when converting
`cuf.data_transfer` operations.
If the operand node for truncs is not created during construction, but
one of the previous ones is reused instead, need to correctly identify
its index, to correctly emit the code.
Fixes https://github.com/llvm/llvm-project/issues/108700
Current MalloChecker logic suppresses FP caused by refcounting only for
C++ destructors. The same pattern occurs a lot in C in objects with
intrusive refcounting. See #104229 for code example.
To extend current logic to C, suspect all release functions as candidate
for suppression.
Closes: #104229
This ABI break only affects fancy pointer which have a different value
representation when pointing to a base of T instead of T itself. This
seems like a rather small set of fancy pointers, which themselves
already represent a very small niche. This patch swaps a pointer to T
with a pointer to base of T in a few library-internal types.
Fix a typo in the header generation message.
Before:
Generating header from
/home/vscode/llvm-project/llvm/../libc/newhdrgen/yaml/ctype.yaml and
/home/vscode/llvm-project/libc/include/ctype.h.def
After:
Generating header ctype.h from
/home/vscode/llvm-project/llvm/../libc/newhdrgen/yaml/ctype.yaml and
/home/vscode/llvm-project/libc/include/ctype.h.def
Works towards P0619R4 / #99985.
The use of `std::get_temporary_buffer` and `std::return_temporary_buffer`
are replaced with `unique_ptr`-based RAII buffer holder.
Escape hatches:
- `_LIBCPP_ENABLE_CXX20_REMOVED_TEMPORARY_BUFFER` restores
`std::get_temporary_buffer` and `std::return_temporary_buffer`.
Drive-by changes:
- In `<syncstream>`, states that `get_temporary_buffer` is now removed,
because `<syncstream>` is added in C++20.
When building FortranRuntime out of tree on aarch64, the list of include
directories for the files comming from FortranFloat128MathILib
has to contain flang/runtime. I did this via
INTERFACE_INCLUDE_DIRECTORIES property.
Resolves: #70930 (and probably latest comments from
https://github.com/clangd/clangd/issues/251)
by fixing racing for the shared `DiagStorage` value which caused messing
with args inside the storage and then formatting the following message
with `getArgSInt(1)` == 2:
```
def err_module_odr_violation_function : Error<
"%q0 has different definitions in different modules; "
"%select{definition in module '%2'|defined here}1 "
"first difference is "
```
which causes `HandleSelectModifier` to go beyond the `ArgumentLen` so
the recursive call to `FormatDiagnostic` was made with `DiagStr` >
`DiagEnd` that leads to infinite `while (DiagStr != DiagEnd)`.
**The Main Idea:**
Reuse the existing `DiagStorageAllocator` logic to make all
`DiagnosticBuilder`s having independent states.
Also, encapsulating the rest of state (e.g. ID and Loc) into
`DiagnosticBuilder`.
**TODO (if it will be requested by reviewer):**
- [x] add a test (I have no idea how to turn a whole bunch of my
proprietary code which leads `clangd` to OOM into a small public
example.. probably I must try using
[this](https://github.com/llvm/llvm-project/issues/70930#issuecomment-2209872975)
instead)
- [x] [`Diag.CurDiagID !=
diag::fatal_too_many_errors`](https://github.com/llvm/llvm-project/pull/108187#pullrequestreview-2296395489)
- [ ] ? get rid of `DiagStorageAllocator` at all and make
`DiagnosticBuilder` having they own `DiagnosticStorage` coz it seems
pretty small so should fit the stack for short-living
`DiagnosticBuilder` instances
The test code with ignore_tkr(tk) on character dummy passed by
fir.boxchar<> was crashing the compiler in [an
assert](2afe678f0a/flang/lib/Optimizer/Dialect/FIRType.cpp (L632))
in `changeElementType`.
It makes little sense to call changeElementType on a fir.boxchar since
this type is lossy (the shape is not part of it). Just skip it in the
code dealing with ignore(tk) when hitting this case
Allows unloading and reloading of ignored libraries. We don't attempt to
reuse or free memory of unloaded library. So TSan will assert if an
ignored library is reloaded 128 times.
Co-authored-by: Vitaly Buka <vitalybuka@gmail.com>
This patch removes from the list of allowed clauses for a handful of
compound constructs those that are specifically disallowed by the OpenMP
spec. In particular, the following restrictions are followed:
- (regarding combined constructs) If _directive-name-A_ is `target`, the
`copyin` clause must not be specified.
- (regarding composite constructs) If _directive-name-A_ is
`distribute`, the `ordered` clause must not be specified.
These restrictions are listed in the OpenMP Specification version 5.2,
sections 17.4 and 17.5. Since it's a similar case as PR #90754, I'm
adding people involved in that decision as reviewers here.
Change macroName() to accept a StringRef to avoid extra string copy.
Simplify ASTNode comparison function.
Use equal_range() instead of calling lower_bound() and upper_bound()
separately for std::multimap.
No need to use std::make_pair.
In the GSL analysis, we don't track the `this` object if the conversion
is not from gsl::owner to gsl pointer, we want to be conservative here
to avoid triggering false positives.
Fixes#108272
