[PATCH v1 0/3] Implement aligned_accessor [P2897R7].

[Tomasz Kaminski]

I have created a separate bugzilla for it:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120994

On Tue, Jul 8, 2025 at 9:58 AM Luc Grosheintz <luc.grosheintz@gmail.com>
wrote:

> How would you like to track this on bugzilla:
>
>    1. As part of submdspan:
>    https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110352
>    https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2023/p2630r4.html
>
>    2. As a new issue, because it's a different paper:
>    https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2024/p2897r7.html
>
> On 7/3/25 12:33, Luc Grosheintz wrote:
> > This patch series implements the aligned_accessor paper P2897R7 in three
> > parts:
> >
> >    - Implement `is_sufficiently_aligned` which is part of <memory>.
> >    - Prepare the accessor tests for reuse.
> >    - Implement aligned_accessor.
> >
> > A couple of remarks:
> >
> >    - The paper P2897R7 and spec N5008 don't specify that the aligment
> >    for is_sufficiently_aligned must be a power of two.
> >
> >    - The reasoning for why is_sufficiently_aligned isn't constexpr is
> >    nicely described in the paper.
> >
> >    - Use of `class` in is_sufficiently_aligned is for consistency within
> >    that file.
> >
> >    - The tests create new unsupported tests and expected failures. The
> >    testsuite doesn't have all that many of those; so there's likely a
> >    strategy to avoid this. However, I don't know how.
> >
> >    - These changes are independent of mdspan, but due to the precise
> >    location of the code it might conflict with the mdspan patch series.
> >
> >    - I skipped updating `cxxapi-data.csv` for is_sufficiently_aligned
> >    due to consistency with the rest of the mdspan patches, i.e. there
> >    will be a bulk update of the file later.
> >
> >    - Each commit was tested with/without PCH and with/without
> >    _GLIBCXX_DEBUG filtered by 20_util/is_sufficiently_aligned
> >    and 23_containers/mdspan.
> >    The last commit was tested fully with/without PCH. All tests on
> >    x86_64-linux.
> >
> > As always I'm happy to reorganize into different commits, if the
> > grouping doesn't make sense.
> >
> > Luc Grosheintz (3):
> >    libstdc++: Implement is_sufficiently_aligned.
> >    libstdc++: Prepare test code for default_accessor for reuse.
> >    libstdc++: Implement aligned_accessor from mdspan.
> >
> >   libstdc++-v3/include/bits/align.h             |  16 ++
> >   libstdc++-v3/include/bits/version.def         |  18 ++
> >   libstdc++-v3/include/bits/version.h           |  20 +++
> >   libstdc++-v3/include/std/mdspan               |  72 ++++++++
> >   libstdc++-v3/include/std/memory               |   1 +
> >   libstdc++-v3/src/c++23/std.cc.in              |   4 +-
> >   .../20_util/is_sufficiently_aligned/1.cc      |  31 ++++
> >   .../20_util/is_sufficiently_aligned/2.cc      |   7 +
> >   .../23_containers/mdspan/accessors/aligned.cc |  43 +++++
> >   .../mdspan/accessors/aligned_ftm.cc           |   6 +
> >   .../mdspan/accessors/aligned_neg.cc           |  33 ++++
> >   .../accessors/debug/aligned_access_neg.cc     |  23 +++
> >   .../accessors/debug/aligned_offset_neg.cc     |  23 +++
> >   .../23_containers/mdspan/accessors/default.cc |  99 -----------
> >   .../23_containers/mdspan/accessors/generic.cc | 168 ++++++++++++++++++
> >   15 files changed, 464 insertions(+), 100 deletions(-)
> >   create mode 100644
> libstdc++-v3/testsuite/20_util/is_sufficiently_aligned/1.cc
> >   create mode 100644
> libstdc++-v3/testsuite/20_util/is_sufficiently_aligned/2.cc
> >   create mode 100644
> libstdc++-v3/testsuite/23_containers/mdspan/accessors/aligned.cc
> >   create mode 100644
> libstdc++-v3/testsuite/23_containers/mdspan/accessors/aligned_ftm.cc
> >   create mode 100644
> libstdc++-v3/testsuite/23_containers/mdspan/accessors/aligned_neg.cc
> >   create mode 100644
> libstdc++-v3/testsuite/23_containers/mdspan/accessors/debug/aligned_access_neg.cc
> >   create mode 100644
> libstdc++-v3/testsuite/23_containers/mdspan/accessors/debug/aligned_offset_neg.cc
> >   delete mode 100644
> libstdc++-v3/testsuite/23_containers/mdspan/accessors/default.cc
> >   create mode 100644
> libstdc++-v3/testsuite/23_containers/mdspan/accessors/generic.cc
> >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://gcc.gnu.org/pipermail/libstdc++/attachments/20250708/a5f27d92/attachment.htm>