The benchmark currently uses makeCartesianProductBenchmark, which
doesn't make a ton of sense, since minmax_element always goes through
every element one by one. The runtime doesn't depend on the values
of the elements.
Fixes#120758
This patch implements the forwarding to frozen C++03 headers as
discussed in
https://discourse.llvm.org/t/rfc-freezing-c-03-headers-in-libc. In the
RFC, we initially proposed selecting the right headers from the Clang
driver, however consensus seemed to steer towards handling this in the
library itself. This patch implements that direction.
At a high level, the changes basically amount to making each public
header look like this:
```
// inside <vector>
#ifdef _LIBCPP_CXX03_LANG
# include <__cxx03/vector>
#else
// normal <vector> content
#endif
```
In most cases, public headers are simple umbrella headers so there isn't
much code in the #else branch. In other cases, the #else branch contains
the actual implementation of the header.
That benchmark isn't really useful, since it doesn't benchmark anything
from libc++ (besides `operator new`). The implementation of the
benchmark also has serious problems like the fact that it allocates an
unknown amount of memory without deallocating it.
This PR enhances the test coverage for std::vector::assign by adding new
tests for several important test cases that were previously missing, as
shown in the following table:
| test cases | forward_iterator | input_iterator |
|-----------------------------------|------------------|----------------|
| new_size > capacity() | Yes | Yes |
| size() < new_size <= capacity() | No | No |
| new_size <= size() | No | No |
Similarly, no tests have previously covered `assign(InputIterator, InputIterator)`
and `assign(size_type, const value_type&)` for `vector<bool>`.
With this patch applied, all missing tests are covered.
This change has a long history. It was first attempted naively in
https://reviews.llvm.org/D131425, which didn't work because we broke the
ability for code to include e.g. <stdio.h> multiple times and get
different definitions based on the pre-defined macros.
However, in #86843 we managed to simplify <stddef.h> by including the
underlying system header outside of any include guards, which worked.
This patch applies the same simplification we did to <stddef.h> to the
other headers that currently mention __need_FOO macros explicitly.
This commit reverts c3276a96d9 and 1901da32, which added a test to
ensure that type traits are derived from integral_constant. While that
is a fine test to add, the commit didn't go through a PR and as a result
it looks like some of our CI has been broken by it.
This should be an uncontroversial change, but let's re-land it via a PR
to get our usual CI coverage.
This disentangles the code which previously had a mix of many #ifdefs, a
non-versioned namespace and a versioned namespace. It also makes it
clearer which parts of <new> are implemented on Windows by including <new.h>.
This allows running the test suite against the native Standard Library
on most systems, and against libstdc++ installed at a custom location.
Of course, these configurations don't run 100% clean at the moment. In
particular, running against the native stdlib is almost guaranteed not
to work out-of-the-box, since the test suite generally contains tests
for things that have been implemented on tip-of-trunk but not released
to most major platforms yet. However, having an easy way to run the test
suite against that library is still both useful and interesting.
This patch fixes a const-qualification on the return type of a method
of `limited_allocator`, which is widely used for testing allocator-aware
containers.
I'm exploring marking microsoft/STL's std::expected as [[nodiscard]],
which affects all functions returning std::expected, including its
own monadic member functions.
As usual, libc++'s test suite contains calls to these member functions
to make sure they compile, but it's discarding the returns. I'm adding
void casts to silence the [[nodiscard]] warnings without altering
what the test is covering.
This patch documents the underlying API for implementing atomics on a
platform.
This doesn't change the operations that std::atomic is based on, but it
reorganizes the C11 / GCC implementation split to make it clearer what's
the base support layer and what's not.
This patch refactors the tests around aligned allocation and sized
deallocation to avoid relying on passing the -fsized-deallocation or
-faligned-allocation flags by default. Since both of these features are
enabled by default in >= C++14 mode, it now makes sense to make that
assumption in the test suite.
A notable exception is MinGW and some older compilers, where sized
deallocation is still not enabled by default. We treat that as a "bug"
in the test suite and we work around it by explicitly adding
-fsized-deallocation, but only under those configurations.
The existing exceptions tests for `vector<T>` have several issues: some
tests did not throw exceptions at all, making them not useful for
exception-safety testing, and some tests did not throw exceptions at the
intended points, failing to serve their expected purpose. This PR fixes
those tests for vector's constructors. Morever, this PR extracted common
classes and utilities into a separate header file, and renamed those
classes using more descriptive names.
While reference comparators are a terrible idea and it's not entirely
clear whether they are supported, fixing the unintended ABI break is
straightforward so we should do it as a first step.
Fixes#118559
The increasing_allocator<T> class, originally introduced to test shrink_to_fit()
for std::vector, std::vector<bool>, and std::basic_string, has duplicated
definitions across several test files. Given the potential utility of this
class for capacity-related tests in various sequence containers, this patch
refactors the definition of increasing_allocator<T> into a single, reusable
location.
This PR optimizes the input iterator overload of `assign(_InputIterator,
_InputIterator)` in `std::vector<_Tp, _Allocator>` by directly assigning
to already initialized memory, rather than first destroying existing
elements and then constructing new ones. By eliminating unnecessary
destruction and construction, the proposed algorithm enhances the
performance by up to 2x for trivial element types (e.g.,
`std::vector<int>`), up to 2.6x for non-trivial element types like
`std::vector<std::string>`, and up to 3.4x for more complex non-trivial
types (e.g., `std::vector<std::vector<int>>`).
### Google Benchmarks
Benchmark tests (`libcxx/test/benchmarks/vector_operations.bench.cpp`)
were conducted for the `assign()` implementations before and after this
patch. The tests focused on trivial element types like
`std::vector<int>`, and non-trivial element types such as
`std::vector<std::string>` and `std::vector<std::vector<int>>`.
#### Before
```
-------------------------------------------------------------------------------------------------
Benchmark Time CPU Iterations
-------------------------------------------------------------------------------------------------
BM_AssignInputIterIter/vector_int/1024/1024 1157 ns 1169 ns 608188
BM_AssignInputIterIter<32>/vector_string/1024/1024 14559 ns 14710 ns 47277
BM_AssignInputIterIter<32>/vector_vector_int/1024/1024 26846 ns 27129 ns 25925
```
#### After
```
-------------------------------------------------------------------------------------------------
Benchmark Time CPU Iterations
-------------------------------------------------------------------------------------------------
BM_AssignInputIterIter/vector_int/1024/1024 561 ns 566 ns 1242251
BM_AssignInputIterIter<32>/vector_string/1024/1024 5604 ns 5664 ns 128365
BM_AssignInputIterIter<32>/vector_vector_int/1024/1024 7927 ns 8012 ns 88579
```
This PR simplifies the implementation of std::vector's move constructor
with an alternative allocator by invoking __init_with_size() instead of
calling assign(), which ultimately calls __assign_with_size(). The
advantage of using __init_with_size() lies in its internal use of
an exception guard, which simplifies the code. Furthermore, from a
semantic standpoint, it is more intuitive for a constructor to call
an initialization function than an assignment function.
The Android clang-r536225 compiler identifies as Clang 19, but it is
based on commit fc57f88f007497a4ead0ec8607ac66e1847b02d6, which predates
the official LLVM 19.0.0 release.
Some tests need fixes:
* The sized delete tests fail because clang-r536225 leaves sized
deallocation off by default.
* std::array<T[0]> is true when this Android Clang version is used with
a trunk libc++, but we expect it to be false in the test. In practice,
Clang and libc++ usually come from the same commit on Android.
This addresses the issue uncovered by #115361. Previously, we weren't
building benchmarks in many cases due to the following block:
e58949632e/libcxx/CMakeLists.txt (L162-L172)
We need to passthrough the necessary variables into the benchmarks
subbuild and use correct syntax.
In particular, test everything with both a normal and a min_allocator,
add tests for a few corner cases and add tests with types that are
trivially relocatable. Also add tests that count the number of
assignments performed by vector::erase, since that is mandated by the
Standard.
This patch is a preparation for optimizing vector::erase.
