move padded layouts and dims from Experimental to the standard namespace

[nmm0]

Closes #404

[Copilot]

Pull request overview

Moves layout_left_padded / layout_right_padded and dims out of the experimental/proposed namespace into the standard namespace to align with their C++26 standardization (closes #404).

Changes:

• Update padded layout definitions and related traits/usages to live in MDSPAN_IMPL_STANDARD_NAMESPACE / Kokkos:: rather than ...::Experimental::.
• Update dims usage to MDSPAN_IMPL_STANDARD_NAMESPACE::dims and adjust the experimental header accordingly.
• Refactor tests, compilation tests, and benchmarks to use the new namespace locations.

Reviewed changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
tests/test_submdspan_static_slice.cpp	Updates expected padded layout types to Kokkos::layout_*_padded.
tests/test_submdspan.cpp	Switches padded layout aliases to Kokkos::layout_*_padded; minor whitespace cleanup.
tests/test_layout_padded_right.cpp	Updates padded layout and detail trait references from experimental/proposed to standard namespace.
tests/test_layout_padded_left.cpp	Updates padded layout and detail trait references from experimental/proposed to standard namespace.
tests/test_dims.cpp	Updates dims usage to MDSPAN_IMPL_STANDARD_NAMESPACE::dims.
include/experimental/__p2642_bits/layout_padded_fwd.hpp	Removes proposed-namespace wrapping so padded layouts are declared in the standard namespace.
include/experimental/__p2642_bits/layout_padded.hpp	Removes proposed-namespace wrapping and updates internal references accordingly.
include/experimental/__p2630_bits/submdspan_mapping.hpp	Updates submdspan mapping code to refer to layout_*_padded in the standard namespace.
include/experimental/__p2389_bits/dims.hpp	Moves dims alias to the standard namespace (and removes proposed namespace wrapper).
include/experimental/__p0009_bits/layout_stride.hpp	Updates trait checks for padded mappings to reference standard detail::is_layout_*_padded_mapping.
include/experimental/__p0009_bits/layout_right.hpp	Updates converting ctor constraints/checks to use standard padded-mapping traits and checks.
include/experimental/__p0009_bits/layout_left.hpp	Updates converting ctor constraints/checks to use standard padded-mapping traits and checks.
compilation_tests/ctest_layout_convertible.cpp	Updates compilation-time checks to use Kokkos::layout_*_padded.
benchmarks/sum/sum_3d_common.hpp	Removes unused proposed-namespace alias (now that padded/dims moved).
benchmarks/fill.hpp	Removes unused proposed-namespace alias (now that padded/dims moved).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[mhoemmen]

Should we consider also regenerating the single-header file for Compiler Explorer as part of this PR? It would be excellent for the single-header version to have the same namespaces as the installable version. Thanks! : - )

[crtrott]

Should we consider also regenerating the single-header file for Compiler Explorer as part of this PR? It would be excellent for the single-header version to have the same namespaces as the installable version. Thanks! : - )

Let's do that in a separate PR, and maybe after your big guy is through. We also are trying to actually create a release, likely coinciding with Kokkos 5.1.0. That opens an interesting question though: currently Kokkos leadership is leaning towards a recommendation (not enforcement) that we have a unified versioning. The idea being that it will be less confusing inside the Kokkos ecosystem about what belongs together and is tested together. I.e. some of us are just exasperated by stuff like the Visual Studio / MSVC versioning, or even the CUDA versioning where it's kinda hard to decide what is what. Any strong feelings on that? If we go with that recommendation for mdspan, we would release as mdspan 5.1 ...

[mhoemmen]

@mhoemmen wrote:

Should we consider also regenerating the single-header file for Compiler Explorer as part of this PR?

@crtrott wrote:

Let's do that in a separate PR, and maybe after your big guy is through.

That's fine! PR #408 actually fixes the dims namespace issue (implementing acceptance of P2389 in C++26) as a drive-by, because it was easier to use dims to write tests. It also adds a test for dims.

@crtrott wrote:

... currently Kokkos leadership is leaning towards a recommendation (not enforcement) that we have a unified versioning.... Any strong feelings on that? If we go with that recommendation for mdspan, we would release as mdspan 5.1 ...

I think that's a reasonable idea. I'm talking it over with colleagues. Would you plan the same thing for the reference std::linalg implementation as well?