As a follow-up to #121013 (which optimized `ranges::copy`) and #121026
(which optimized `ranges::copy_backward`), this PR enhances the
performance of `std::ranges::{move, move_backward}` for
`vector<bool>::iterator`, addressing a subtask outlined in issue #64038.
The optimizations bring performance improvements analogous to those
achieved for the `{copy, copy_backward}` algorithms: up to 2000x for
aligned moves and 60x for unaligned moves. Moreover, comprehensive
tests covering up to 4 storage words (256 bytes) with odd and even bit
sizes are provided, which validate the proposed optimizations in this
patch.
This is to modify a list of libcxx tests written under the assumption
that iterators for std::array, std::string_view, and std::string are
pointers. The motivation for this PR is to make the tests more universal
and potentially being used to test other C++ standard library
implementations, for example
[microsoft/STL](https://github.com/microsoft/STL).
I can confirm that this patch makes a number of tests compatible with
microsoft STL:
`Failed : 204 (2.12%)` -> `Failed : 136 (1.42%)`
, and does not break any tests on `libcxx`.
This is not a complete list of such incompatibilities, but I am hoping
this will start a discussion about whether we are open to accepting such
changes.
As reported in https://reviews.llvm.org/D151953#4472195, the std::move
algorithm (and various other functions that relied on it) stopped working
after starting to use `__constexpr_memmove` in its implementation. This
patch fixes the underlying issue in `__constexpr_memmove` and adds tests
for various related algorithms and functions that were not exercising
trivial move-only types.
Differential Revision: https://reviews.llvm.org/D154613
Some tests in our test suite are unbelievably slow on GCC due to the
use of the always_inline attribute. See [1] for more details.
This patch introduces the GCC-ALWAYS_INLINE-FIXME lit feature to
disable tests that are plagued by that issue. At the same time, it
moves several existing tests from ad-hoc `UNSUPPORTED: gcc-12` markup
to the new GCC-ALWAYS_INLINE-FIXME feature, and marks the slowest tests
reported by the CI as `UNSUPPORTED: GCC-ALWAYS_INLINE-FIXME`.
[1]: https://discourse.llvm.org/t/rfc-stop-supporting-extern-instantiations-with-gcc/71277/1
Differential Revision: https://reviews.llvm.org/D152736
This has multiple benefits:
- The optimizations are also performed for the `ranges::` versions of the algorithms
- Code duplication is reduced
- it is simpler to add this optimization for other segmented iterators,
like `ranges::join_view::iterator`
- Algorithm code is removed from `<deque>`
Reviewed By: ldionne, huixie90, #libc
Spies: mstorsjo, sstefan1, EricWF, libcxx-commits, mgorny
Differential Revision: https://reviews.llvm.org/D132505
This reverts commit a6e1080b87db8fbe0e1afadd96af5a3c0bd5e279.
Fix the conditions when the `memmove` optimization can be applied and refactor them out into a reusable type trait, fix and significantly expand the tests.
Differential Revision: https://reviews.llvm.org/D139235
Instead of using `reverse_iterator`, share the optimization between the 4 algorithms. The key observation here that `memmove` applies to both `copy` and `move` identically, and to their `_backward` versions very similarly. All algorithms now follow the same pattern along the lines of:
```
if constexpr (can_memmove<InIter, OutIter>) {
memmove(first, last, out);
} else {
naive_implementation(first, last, out);
}
```
A follow-up will delete `unconstrained_reverse_iterator`.
This patch removes duplication and divergence between `std::copy`, `std::move` and `std::move_backward`. It also improves testing:
- the test for whether the optimization is used only applied to `std::copy` and, more importantly, was essentially a no-op because it would still pass if the optimization was not used;
- there were no tests to make sure the optimization is not used when the effect would be visible.
Differential Revision: https://reviews.llvm.org/D130695
When we ship LLVM 16, <ranges> won't be considered experimental anymore.
We might as well do this sooner rather than later.
Differential Revision: https://reviews.llvm.org/D132151
The return value for both of these algorithms is specified as
```
`{last, result - N}` for the overloads in namespace `ranges`.
```
But the current implementation instead returns `{first, result - N}`.
Also add both algorithms to the relevant "robust" tests.
Differential Revision: https://reviews.llvm.org/D130968
It is meant to be used in ranges algorithm tests.
It is much simplified version of C++23's tuple + zip_view.
Using std::swap would cause compilation failure and using `std::move` would not create the correct rvalue proxy which would result in copies.
Differential Revision: https://reviews.llvm.org/D129099
This patch also adds a new optimization to `std::move`. It unwraps three `reverse_iterator`s if the wrapped iterator is a `contiguous_iterator` and the iterated type is trivially_movable. This allows us to simplify `ranges::move_backward` to a forward to `std::move` without any pessimization.
Reviewed By: var-const, #libc
Spies: libcxx-commits, mgorny
Differential Revision: https://reviews.llvm.org/D126616
