From ab3a338ba328efcfff39db750e2840794661d83d Mon Sep 17 00:00:00 2001
From: Mark Hoemmen <mhoemmen@nvidia.com>
Date: Wed, 10 Jul 2024 12:11:31 -0600
Subject: [PATCH 1/4] P2897R2: detectably_invalid

Update based on further discussion with David Sankel.

* Rename `is_checkably_valid` to `detectably_invalid`
    (with inverted return value)

* Ask LEWG to consider the alternative design that makes
    `is_sufficiently_aligned` a nonmember function in `<bit>`
    instead of a member function of `aligned_accessor`,
    while LWG review of R2 proceeds concurrently

* P2389R2 was voted into the Working Draft at St. Louis,
    so replace use of `dextents` in examples with `dims`.

* Minor formatting and non-wording fixes
---
 aligned_accessor/P2897R2.html        | 1282 +++++++++++++++-----------
 aligned_accessor/aligned_accessor.md |  491 +++++++---
 2 files changed, 1079 insertions(+), 694 deletions(-)

diff --git a/aligned_accessor/P2897R2.html b/aligned_accessor/P2897R2.html
index 7d9040a5..2b6c412a 100644
--- a/aligned_accessor/P2897R2.html
+++ b/aligned_accessor/P2897R2.html
@@ -4,7 +4,7 @@
   <meta charset="utf-8" />
   <meta name="generator" content="mpark/wg21" />
   <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes" />
-  <meta name="dcterms.date" content="2024-07-08" />
+  <meta name="dcterms.date" content="2024-07-10" />
   <title>aligned_accessor: An mdspan accessor expressing pointer overalignment</title>
   <style>
 code{white-space: pre-wrap;}
@@ -425,7 +425,7 @@ <h1 class="title" style="text-align:center"><code>aligned_accessor</code>:
   </tr>
   <tr>
     <td>Date: </td>
-    <td>2024-07-08</td>
+    <td>2024-07-10</td>
   </tr>
   <tr>
     <td style="vertical-align:top">Project: </td>
@@ -473,13 +473,21 @@ <h1 id="toctitle">Contents</h1>
 <li><a href="#is_sufficiently_aligned-is-specific-to-aligned_accessor" id="toc-is_sufficiently_aligned-is-specific-to-aligned_accessor"><span class="toc-section-number">5.6.1</span>
 <code>is_sufficiently_aligned</code> is specific to
 <code>aligned_accessor</code></a></li>
-<li><a href="#is_checkably_valid-generic-validity-check" id="toc-is_checkably_valid-generic-validity-check"><span class="toc-section-number">5.6.2</span> <code>is_checkably_valid</code>:
+<li><a href="#detectably_invalid-generic-validity-check" id="toc-detectably_invalid-generic-validity-check"><span class="toc-section-number">5.6.2</span> <code>detectably_invalid</code>:
 Generic validity check?</a></li>
-<li><a href="#arguments-against-and-for-is_checkably_valid" id="toc-arguments-against-and-for-is_checkably_valid"><span class="toc-section-number">5.6.3</span> Arguments against and for
-<code>is_checkably_valid</code></a></li>
+<li><a href="#arguments-against-and-for-detectably_invalid" id="toc-arguments-against-and-for-detectably_invalid"><span class="toc-section-number">5.6.3</span> Arguments against and for
+<code>detectably_invalid</code></a></li>
 <li><a href="#nonmember-customization-point-design" id="toc-nonmember-customization-point-design"><span class="toc-section-number">5.6.4</span> Nonmember customization point
 design</a></li>
-<li><a href="#conclusions" id="toc-conclusions"><span class="toc-section-number">5.6.5</span> Conclusions</a></li>
+<li><a href="#we-need-both-is_sufficiently_aligned-and-detectably_invalid" id="toc-we-need-both-is_sufficiently_aligned-and-detectably_invalid"><span class="toc-section-number">5.6.5</span> We need both
+<code>is_sufficiently_aligned</code> and
+<code>detectably_invalid</code></a></li>
+<li><a href="#alternative-nonmember-is_sufficiently_aligned" id="toc-alternative-nonmember-is_sufficiently_aligned"><span class="toc-section-number">5.6.6</span> Alternative: Nonmember
+<code>is_sufficiently_aligned</code></a></li>
+<li><a href="#do-accessors-need-to-check-anything-else" id="toc-do-accessors-need-to-check-anything-else"><span class="toc-section-number">5.6.7</span> Do accessors need to check
+anything else?</a></li>
+<li><a href="#naming-the-function" id="toc-naming-the-function"><span class="toc-section-number">5.6.8</span> Naming the function</a></li>
+<li><a href="#conclusions" id="toc-conclusions"><span class="toc-section-number">5.6.9</span> Conclusions</a></li>
 </ul></li>
 <li><a href="#explicit-conversions-as-the-model-for-precondition-asserting-conversions" id="toc-explicit-conversions-as-the-model-for-precondition-asserting-conversions"><span class="toc-section-number">5.7</span> Explicit conversions as the model
 for precondition-asserting conversions</a>
@@ -498,15 +506,18 @@ <h1 id="toctitle">Contents</h1>
 <li><a href="#implementation" id="toc-implementation"><span class="toc-section-number">6</span> Implementation</a></li>
 <li><a href="#example" id="toc-example"><span class="toc-section-number">7</span> Example</a></li>
 <li><a href="#references" id="toc-references"><span class="toc-section-number">8</span> References</a></li>
-<li><a href="#wording" id="toc-wording"><span class="toc-section-number">9</span> Wording</a>
+<li><a href="#acknowledgments" id="toc-acknowledgments"><span class="toc-section-number">9</span> Acknowledgments</a></li>
+<li><a href="#wording" id="toc-wording"><span class="toc-section-number">10</span> Wording</a>
 <ul>
-<li><a href="#add-subsection-mdspan.accessor.aligned-with-the-following" id="toc-add-subsection-mdspan.accessor.aligned-with-the-following"><span class="toc-section-number">9.1</span> Add subsection �
+<li><a href="#add-subsection-mdspan.accessor.aligned-with-the-following" id="toc-add-subsection-mdspan.accessor.aligned-with-the-following"><span class="toc-section-number">10.1</span> Add subsection �
 [mdspan.accessor.aligned] with the following</a></li>
-<li><a href="#members-mdspan.accessor.aligned.members" id="toc-members-mdspan.accessor.aligned.members"><span class="toc-section-number">9.2</span> Members
+<li><a href="#members-mdspan.accessor.aligned.members" id="toc-members-mdspan.accessor.aligned.members"><span class="toc-section-number">10.2</span> Members
 [mdspan.accessor.aligned.members]</a></li>
 </ul></li>
-<li><a href="#appendix-a-is_checkably_valid-nonmember-function-example" id="toc-appendix-a-is_checkably_valid-nonmember-function-example"><span class="toc-section-number">10</span> Appendix A:
-<code>is_checkably_valid</code> nonmember function example</a></li>
+<li><a href="#wording-for-alternative-nonmember-is_sufficiently_aligned-design" id="toc-wording-for-alternative-nonmember-is_sufficiently_aligned-design"><span class="toc-section-number">11</span> Wording for alternative nonmember
+<code>is_sufficiently_aligned</code> design</a></li>
+<li><a href="#appendix-a-detectably_invalid-nonmember-function-example" id="toc-appendix-a-detectably_invalid-nonmember-function-example"><span class="toc-section-number">12</span> Appendix A:
+<code>detectably_invalid</code> nonmember function example</a></li>
 </ul>
 </div>
 <h1 data-number="1" id="authors"><span class="header-section-number">1</span> Authors<a href="#authors" class="self-link"></a></h1>
@@ -565,20 +576,30 @@ <h1 data-number="2" id="revision-history"><span class="header-section-number">2<
 <li><p>Discuss optional suggestions from LEWG review of R1 on
 2024-06-28</p>
 <ul>
-<li><p>Add <code>is_checkably_valid</code> discussion</p></li>
-<li><p>Add <code>naughty_cast</code> discussion</p></li>
+<li><p>Add <code>explicit</code> converting constructor vs. named cast
+(“<code>naughty_cast</code>”) discussion</p></li>
+<li><p>Add <code>detectably_invalid</code> discussion</p></li>
+<li><p>Ask LEWG to consider the alternative design that makes
+<code>is_sufficiently_aligned</code> a nonmember function in
+<code>&lt;bit&gt;</code> instead of a member function of
+<code>aligned_accessor</code>, while LWG review of R2 proceeds
+concurrently</p></li>
 </ul></li>
+<li><p>P2389R2 was voted into the Working Draft at St. Louis, so replace
+use of <code>dextents</code> in examples with
+<code>dims</code>.</p></li>
 </ul></li>
 </ul>
 <h1 data-number="3" id="purpose-of-this-paper"><span class="header-section-number">3</span> Purpose of this paper<a href="#purpose-of-this-paper" class="self-link"></a></h1>
 <p>We propose adding <code>aligned_accessor</code> to the C++ Standard
-Library. This class template is an mdspan accessor policy that uses
-<code>assume_aligned</code> to decorate pointer access. We think it
-belongs in the Standard Library for two reasons. First, it would serve
-as a common vocabulary type for interfaces that take <code>mdspan</code>
-to declare their minimum alignment requirements. Second, it extends to
-<code>mdspan</code> accesses the optimizations that compilers can
-perform to pointers decorated with <code>assume_aligned</code>.</p>
+Library. This class template is an <code>mdspan</code> accessor policy
+that uses <code>assume_aligned</code> to decorate pointer access. We
+think it belongs in the Standard Library for two reasons. First, it
+would serve as a common vocabulary type for interfaces that take
+<code>mdspan</code> to declare their minimum alignment requirements.
+Second, it extends to <code>mdspan</code> accesses the optimizations
+that compilers can perform to pointers decorated with
+<code>assume_aligned</code>.</p>
 <p><code>aligned_accessor</code> is analogous to the various
 <code>atomic_accessor_*</code> templates proposed by P2689. Both that
 proposal and this one start with a Standard Library feature that
@@ -597,11 +618,15 @@ <h1 data-number="4" id="key-features"><span class="header-section-number">4</spa
 <code>default_accessor</code></p></li>
 <li><p><code>data_handle_type</code> is
 <code>ElementType*</code></p></li>
-<li><p>Constructor permits (implicit) conversion</p>
+<li><p>Permitted implicit conversions</p>
 <ul>
-<li><p>from nonconst to const <code>ElementType</code>, and</p></li>
-<li><p>from more overalignment to less overalignment</p></li>
+<li><p>from nonconst to const <code>ElementType</code>,</p></li>
+<li><p>from more overalignment to less overalignment, and</p></li>
+<li><p>from overalignment to no overalignment
+(<code>default_accessor</code>)</p></li>
 </ul></li>
+<li><p><code>explicit</code> converting constructor from
+<code>default_accessor</code> lets users assert overalignment</p></li>
 <li><p><code>is_sufficiently_aligned</code> checks pointer
 alignment</p></li>
 </ul>
@@ -619,10 +644,10 @@ <h1 data-number="4" id="key-features"><span class="header-section-number">4</spa
 these attributes should not apply to the result of <code>offset</code>,
 for the same reason that <code>offset_policy</code> is
 <code>default_accessor</code> and not <code>aligned_accessor</code>.</p>
-<p>The constructor is analogous to <code>default_accessor</code>’s
-constructor, in that it exists to permit conversion from nonconst
-<code>element_type</code> to const <code>element_type</code>. The
-constructor of <code>aligned_accessor</code> additionally permits
+<p>The converting constructor from <code>aligned_accessor</code> is
+analogous to <code>default_accessor</code>’s constructor, in that it
+exists to permit conversion from nonconst <code>element_type</code> to
+const <code>element_type</code>. It additionally permits implicit
 conversion from more overalignment to less overalignment – something
 that we expect users may need to do. For example, users may start with
 <code>aligned_accessor&lt;float, 128&gt;</code>, because their
@@ -635,10 +660,16 @@ <h1 data-number="4" id="key-features"><span class="header-section-number">4</spa
 This makes it easier for users to check preconditions, without needing
 to know how to cast a pointer to an integer of the correct size and
 signedness.</p>
+<p>Per discussion below, we would like LEWG to consider the alternative
+design that removes <code>is_sufficiently_aligned</code> from
+<code>aligned_accessor</code> and adds it to the
+<code>&lt;bit&gt;</code> header as a separate nonmember function. We
+think LWG review of <code>aligned_accessor</code> can proceed
+concurrently. We present wording for this design alternative below.</p>
 <h1 data-number="5" id="design-discussion"><span class="header-section-number">5</span> Design discussion<a href="#design-discussion" class="self-link"></a></h1>
 <h2 data-number="5.1" id="the-accessor-is-not-nestable"><span class="header-section-number">5.1</span> The accessor is not nestable<a href="#the-accessor-is-not-nestable" class="self-link"></a></h2>
 <p>We considered making <code>aligned_accessor</code> “wrap” any
-accessor type meeting the right requirements. For example,
+accessor type that meets the right requirements. For example,
 <code>aligned_accessor</code> could take the inner accessor as a
 template parameter, store an instance of it, and dispatch to its member
 functions. That would give users a way to apply multiple accessor
@@ -650,17 +681,15 @@ <h2 data-number="5.1" id="the-accessor-is-not-nestable"><span class="header-sect
 <code>data_handle_type</code> is a pointer type, but we could not check
 that their accessor’s <code>access</code> function actually dereferences
 the pointer. For instance, <code>access</code> might instead interpret
-the data handle as a file handle or a key into a distributed data
-store.</p>
+the pointer as a file handle or a key into a distributed data store.</p>
 <p>Second, even if the inner accessor’s <code>access</code> function
 actually did return the result of dereferencing the pointer, the outer
 <code>access</code> function might not be able to recover the effects of
 the inner <code>access</code> function, because <code>access</code>
-computes a <code>reference</code>, not a pointer. For example, our
-<code>aligned_accessor</code>’s access function can only apply
-<code>assume_aligned</code> to the pointer; knowledge of overalignment
-at compile time “lost” with the dereference at element
-<code>i</code>.</p>
+computes a <code>reference</code>, not a pointer. In order for
+<code>aligned_accessor</code>’s <code>access</code> function to get back
+that pointer, it would need to reach past the inner accessor’s public
+interface. That would defeat the purpose of generic nesting.</p>
 <p>Third, any way (not just this one) of nesting two generic accessors
 raises the question of order dependence. Even if it were possible to
 apply the effects of both the inner and outer accessors’
@@ -690,22 +719,28 @@ <h2 data-number="5.2" id="explicit-constructor-from-default_accessor"><span clas
 unaligned <code>mdspan</code>. Making it <code>explicit</code> would
 prevent implicit conversion. Thus, we have decided to add this
 <code>explicit</code> constructor in R1.</p>
-<p>Without the <code>explicit</code> constructor, users must do this to
-convert from nonaligned to aligned <code>mdspan</code>.</p>
+<p>Without the <code>explicit</code> constructor, users have two options
+for turning a nonaligned <code>mdspan</code> into an aligned
+<code>mdspan</code>. First, as in the following example, users could
+“take apart” the input nonaligned <code>mdspan</code> and use the pieces
+to construct an aligned <code>mdspan</code>, whose type they name
+completely.</p>
 <div class="sourceCode" id="cb1"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb1-1"><a href="#cb1-1" aria-hidden="true" tabindex="-1"></a><span class="dt">void</span> compute_with_aligned<span class="op">(</span></span>
-<span id="cb1-2"><a href="#cb1-2" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextent<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">2</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span> matrix<span class="op">)</span></span>
+<span id="cb1-2"><a href="#cb1-2" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">2</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span> matrix<span class="op">)</span></span>
 <span id="cb1-3"><a href="#cb1-3" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
 <span id="cb1-4"><a href="#cb1-4" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> std<span class="op">::</span><span class="dt">size_t</span> byte_alignment <span class="op">=</span> <span class="dv">4</span> <span class="op">*</span> <span class="kw">alignof</span><span class="op">(</span><span class="dt">float</span><span class="op">)</span>;</span>
-<span id="cb1-5"><a href="#cb1-5" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> aligned_matrix_t <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextent<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">2</span><span class="op">&gt;</span>,</span>
+<span id="cb1-5"><a href="#cb1-5" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> aligned_matrix_t <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">2</span><span class="op">&gt;</span>,</span>
 <span id="cb1-6"><a href="#cb1-6" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>layout_left, std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, byte_alignment<span class="op">&gt;&gt;</span>;</span>
 <span id="cb1-7"><a href="#cb1-7" aria-hidden="true" tabindex="-1"></a></span>
 <span id="cb1-8"><a href="#cb1-8" aria-hidden="true" tabindex="-1"></a>  aligned_matrix_t aligned_matrix<span class="op">{</span>matrix<span class="op">.</span>data_handle<span class="op">()</span>, matrix<span class="op">.</span>mapping<span class="op">()}</span>;</span>
 <span id="cb1-9"><a href="#cb1-9" aria-hidden="true" tabindex="-1"></a>  <span class="co">// ... use aligned_matrix ...</span></span>
 <span id="cb1-10"><a href="#cb1-10" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
-<p>Another option would be to use constructor template argument
-deduction (CTAD), as in the following.</p>
+<p>Second, as in the following example, users could construct an
+<code>aligned_accessor</code> explicitly and use constructor template
+argument deduction (CTAD) to construct the aligned <code>mdspan</code>
+from its pieces.</p>
 <div class="sourceCode" id="cb2"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb2-1"><a href="#cb2-1" aria-hidden="true" tabindex="-1"></a><span class="dt">void</span> compute_with_aligned<span class="op">(</span></span>
-<span id="cb2-2"><a href="#cb2-2" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextent<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">2</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span> matrix<span class="op">)</span></span>
+<span id="cb2-2"><a href="#cb2-2" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">2</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span> matrix<span class="op">)</span></span>
 <span id="cb2-3"><a href="#cb2-3" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
 <span id="cb2-4"><a href="#cb2-4" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> std<span class="op">::</span><span class="dt">size_t</span> byte_alignment <span class="op">=</span> <span class="dv">4</span> <span class="op">*</span> <span class="kw">alignof</span><span class="op">(</span><span class="dt">float</span><span class="op">)</span>;</span>
 <span id="cb2-5"><a href="#cb2-5" aria-hidden="true" tabindex="-1"></a></span>
@@ -713,19 +748,19 @@ <h2 data-number="5.2" id="explicit-constructor-from-default_accessor"><span clas
 <span id="cb2-7"><a href="#cb2-7" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, byte_alignment<span class="op">&gt;{}}</span>;</span>
 <span id="cb2-8"><a href="#cb2-8" aria-hidden="true" tabindex="-1"></a>  <span class="co">// ... use aligned_matrix ...</span></span>
 <span id="cb2-9"><a href="#cb2-9" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
-<p>However, <code>mdspan</code> users commonly define their own type
-aliases for <code>mdspan</code>, with application-specific names that
-make code more self-documenting. The <code>aligned_matrix_t</code>
-definition above is an an example. Such users would not normally rely on
-CTAD for conversion from a less specific to a more specific
-<code>mdspan</code>.</p>
+<p>The first approach would likely be more common. This is because
+<code>mdspan</code> users commonly define their own type aliases for
+<code>mdspan</code>, with application-specific names that make code more
+self-documenting. The <code>aligned_matrix_t</code> definition above is
+an an example.</p>
 <p>Adding an <code>explicit</code> constructor from
 <code>default_accessor</code> lets users get the same effect more
-concisely.</p>
-<div class="sourceCode" id="cb3"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb3-1"><a href="#cb3-1" aria-hidden="true" tabindex="-1"></a><span class="dt">void</span> compute_with_aligned<span class="op">(</span>std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextent<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">2</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span> matrix<span class="op">)</span></span>
+concisely, without needing to “take apart” the input
+<code>mdspan</code>.</p>
+<div class="sourceCode" id="cb3"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb3-1"><a href="#cb3-1" aria-hidden="true" tabindex="-1"></a><span class="dt">void</span> compute_with_aligned<span class="op">(</span>std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">2</span>, <span class="dt">int</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span> matrix<span class="op">)</span></span>
 <span id="cb3-2"><a href="#cb3-2" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
 <span id="cb3-3"><a href="#cb3-3" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> std<span class="op">::</span><span class="dt">size_t</span> byte_alignment <span class="op">=</span> <span class="dv">4</span> <span class="op">*</span> <span class="kw">alignof</span><span class="op">(</span><span class="dt">float</span><span class="op">)</span>;</span>
-<span id="cb3-4"><a href="#cb3-4" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> aligned_mdspan <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextent<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">2</span><span class="op">&gt;</span>,</span>
+<span id="cb3-4"><a href="#cb3-4" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> aligned_mdspan <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">2</span>, <span class="dt">int</span><span class="op">&gt;</span>,</span>
 <span id="cb3-5"><a href="#cb3-5" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>layout_left, std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, byte_alignment<span class="op">&gt;&gt;</span>;</span>
 <span id="cb3-6"><a href="#cb3-6" aria-hidden="true" tabindex="-1"></a></span>
 <span id="cb3-7"><a href="#cb3-7" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan aligned_matrix<span class="op">{</span>matrix<span class="op">}</span>;</span>
@@ -749,23 +784,23 @@ <h2 data-number="5.3" id="we-do-not-define-an-alias-for-aligned-mdspan"><span cl
 but we do not find it very useful, for the following reasons.</p>
 <ol type="1">
 <li><p>Users of <code>mdspan</code> commonly define their own type
-aliases whose names have meaningful names for their
-applications.</p></li>
+aliases whose names are meaningful for their applications.</p></li>
 <li><p>It would not save much typing.</p></li>
 </ol>
-<p>Examples may define aliases to make them more concise. One of them in
+<p>Examples may define aliases to make them more concise. One example in
 this proposal defines the following alias for an <code>mdspan</code> of
 <code>float</code> with alignment <code>byte_alignment</code>.</p>
 <div class="sourceCode" id="cb6"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb6-1"><a href="#cb6-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
-<span id="cb6-2"><a href="#cb6-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">1</span><span class="op">&gt;</span>,</span>
+<span id="cb6-2"><a href="#cb6-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">1</span>, <span class="dt">int</span><span class="op">&gt;</span>,</span>
 <span id="cb6-3"><a href="#cb6-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>layout_right, std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, byte_alignment<span class="op">&gt;&gt;</span>;</span></code></pre></div>
 <p>This lets the example use <code>aligned_mdspan&lt;32&gt;</code> and
 <code>aligned_mdspan&lt;16&gt;</code>.</p>
 <p>The above alias is specific to a particular example. A
 <em>general</em> version of alias would look like this.</p>
-<div class="sourceCode" id="cb7"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb7-1"><a href="#cb7-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="kw">class</span> Extents, <span class="kw">class</span> Layout, <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
-<span id="cb7-2"><a href="#cb7-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span>ElementType, Extents, Layout,</span>
-<span id="cb7-3"><a href="#cb7-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span>ElementType, byte_alignment<span class="op">&gt;&gt;</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb7"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb7-1"><a href="#cb7-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="kw">class</span> Extents, <span class="kw">class</span> Layout,</span>
+<span id="cb7-2"><a href="#cb7-2" aria-hidden="true" tabindex="-1"></a>  <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb7-3"><a href="#cb7-3" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span>ElementType, Extents, Layout,</span>
+<span id="cb7-4"><a href="#cb7-4" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span>ElementType, byte_alignment<span class="op">&gt;&gt;</span>;</span></code></pre></div>
 <p>This alias would save <em>some</em> typing. However, mdspan “power
 users” rarely type out all the template arguments. First, they can rely
 on CTAD to create <code>mdspan</code>s, and <code>auto</code> to return
@@ -775,18 +810,18 @@ <h2 data-number="5.3" id="we-do-not-define-an-alias-for-aligned-mdspan"><span cl
 users might define the following.</p>
 <div class="sourceCode" id="cb8"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb8-1"><a href="#cb8-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
 <span id="cb8-2"><a href="#cb8-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> vector_t <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span>ElementType,</span>
-<span id="cb8-3"><a href="#cb8-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">1</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span>;</span>
+<span id="cb8-3"><a href="#cb8-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span>;</span>
 <span id="cb8-4"><a href="#cb8-4" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
 <span id="cb8-5"><a href="#cb8-5" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> matrix_t <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span>ElementType,</span>
-<span id="cb8-6"><a href="#cb8-6" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">2</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span>;</span>
+<span id="cb8-6"><a href="#cb8-6" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">2</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left<span class="op">&gt;</span>;</span>
 <span id="cb8-7"><a href="#cb8-7" aria-hidden="true" tabindex="-1"></a></span>
 <span id="cb8-8"><a href="#cb8-8" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
 <span id="cb8-9"><a href="#cb8-9" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_vector_t <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span>ElementType,</span>
-<span id="cb8-10"><a href="#cb8-10" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">1</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left, </span>
+<span id="cb8-10"><a href="#cb8-10" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left, </span>
 <span id="cb8-11"><a href="#cb8-11" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span>ElementType, byte_alignment<span class="op">&gt;&gt;</span>;</span>
 <span id="cb8-12"><a href="#cb8-12" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
 <span id="cb8-13"><a href="#cb8-13" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_matrix_t <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span>ElementType,</span>
-<span id="cb8-14"><a href="#cb8-14" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">2</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left, </span>
+<span id="cb8-14"><a href="#cb8-14" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">2</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_left, </span>
 <span id="cb8-15"><a href="#cb8-15" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span>ElementType, byte_alignment<span class="op">&gt;&gt;</span>;</span></code></pre></div>
 <p>Such users may never type the characters “<code>mdspan</code>” again.
 For this reason, while we do not object to an
@@ -803,17 +838,18 @@ <h2 data-number="5.4" id="mdspan-construction-safety"><span class="header-sectio
 whether its accessor finds the caller’s data handle acceptable.</p>
 <p>This is true for any accessor type, not just for
 <code>aligned_accessor</code>. It is a design feature of
-<code>mdspan</code> that accessors can be stateless. Even if they have
-state, they do not need to be constructed with or store the data handle.
-As a result, an accessor might not see a data handle until
-<code>access</code> or <code>offset</code> is called. Both of those
-member functions are performance critical, so they cannot afford an
-extra branch on every call. Compare to <code>vector::operator[]</code>,
-which has preconditions but is not required to perform bounds checks.
-Using exceptions in the manner of <code>vector::at</code> could reduce
-performance and would also make <code>mdspan</code> unusable in a
-freestanding or no-exceptions context.</p>
-<p>Note that <code>aligned_accessor</code> does not introduce any
+<code>mdspan</code> that accessors can be stateless. Most of them have
+no state. Even if they have state, they do not need to be constructed
+with or store the data handle. As a result, an accessor might not see a
+data handle until <code>access</code> or <code>offset</code> is called.
+Both of those member functions are performance critical, so they cannot
+afford an extra branch on every call. Compare to
+<code>vector::operator[]</code>, which has preconditions but is not
+required to perform bounds checks. Using exceptions in the manner of
+<code>vector::at</code> could reduce performance and would also make
+<code>mdspan</code> unusable in a freestanding or no-exceptions
+context.</p>
+<p>Note that <code>aligned_accessor</code> does not introduce
 <em>additional</em> preconditions beyond those of the existing C++
 Standard Library feature <code>assume_aligned</code>. In the words of
 one LEWG reviewer, <code>aligned_accessor</code> is not any more
@@ -822,7 +858,7 @@ <h2 data-number="5.4" id="mdspan-construction-safety"><span class="header-sectio
 <p>Before submitting R0 of this paper, we considered an approach
 specific to <code>aligned_accessor</code>, that would wrap the pointer
 in a special data handle type. The data handle type would have an
-<code>explicit</code> constructor taking a raw pointer, with a
+<code>explicit</code> constructor that takes a raw pointer, with a
 precondition that the raw pointer have sufficient alignment. The
 constructor would be <code>explicit</code>, because it would have a
 precondition. This design would force the precondition back to
@@ -832,30 +868,38 @@ <h2 data-number="5.4" id="mdspan-construction-safety"><span class="header-sectio
 <span id="cb9-2"><a href="#cb9-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> acc_type <span class="op">=</span> aligned_accessor<span class="op">&lt;</span>element_type, byte_alignment<span class="op">&gt;</span>;</span>
 <span id="cb9-3"><a href="#cb9-3" aria-hidden="true" tabindex="-1"></a>mdspan x<span class="op">{</span>acc_type<span class="op">::</span>data_handle_type<span class="op">{</span>raw_pointer<span class="op">}</span>, mapping, acc_type<span class="op">{}}</span>;</span></code></pre></div>
 <p>We rejected this approach in favor of
-<code>is_sufficiently_aligned</code> for the following reasons. First,
-wrapping the pointer in a custom data handle class would make every
-<code>access</code> or <code>offset</code> call need to reach through
-the data handle’s interface, instead of just taking the raw pointer
-directly. The <code>access</code> function, and to some extent also
-<code>offset</code>, need to be as fast as possible. Their performance
-depends on compilers being able to optimize through function calls. The
-authors of <code>mdspan</code> carefully balanced generality with
-function call depth and other code complexity factors that may hinder
-compilers from optimizing. Performance of <code>aligned_accessor</code>
-matters as much or even more than performance of
-<code>default_accessor</code>, because <code>aligned_accessor</code>
-exists to communicate optimization potential. Second, the alignment
-precondition would still exist. Requiring the data handle type to throw
-an exception if the pointer is not sufficiently aligned would make
-<code>mdspan</code> unusable in a freestanding or no-exceptions context.
-Third, users should not have to pay for unneeded checks. The two
+<code>is_sufficiently_aligned</code> for the following reasons.</p>
+<ol type="1">
+<li><p>Wrapping the pointer in a custom data handle class would make
+every <code>access</code> or <code>offset</code> call need to reach
+through the data handle’s interface, instead of just taking the raw
+pointer directly. The <code>access</code> function, and to some extent
+also <code>offset</code>, need to be as fast as possible. Their
+performance depends on compilers being able to optimize through function
+calls. The authors of <code>mdspan</code> carefully balanced generality
+with function call depth and other code complexity factors that may
+hinder compilers from optimizing. Performance of
+<code>aligned_accessor</code> matters as much or even more than
+performance of <code>default_accessor</code>, because
+<code>aligned_accessor</code> exists to communicate optimization
+potential.</p></li>
+<li><p>The alignment precondition would still exist. Requiring the data
+handle type to throw an exception if the pointer is not sufficiently
+aligned would make <code>mdspan</code> unusable in a freestanding or
+no-exceptions context.</p></li>
+<li><p>Users should not have to pay for unneeded checks. The two
 examples in the wording express the two most common cases. If users get
 a pointer from a function like <code>aligned_alloc</code>, then they
 already know its alignment, because they asked for it. If users are
 computing alignment at run time to dispatch to a more optimized code
 path, then they know alignment before dispatch. In both cases, users
 already know the alignment before constructing the
-<code>mdspan</code>.</p>
+<code>mdspan</code>.</p></li>
+<li><p>The data handle is still a pointer, it’s just a pointer with a
+constraint on its values. Users would reasonably expect to be able to
+use the result of <code>data_handle()</code> with existing interfaces
+that expect a raw pointer.</p></li>
+</ol>
 <p>An LEWG poll on 2023-10-10, “[b]lock <code>aligned_accessor</code>
 progressing until we have a way of checking alignment requirements
 during <code>mdspan</code> construction,” resulted in no consensus.
@@ -899,10 +943,10 @@ <h2 data-number="5.4" id="mdspan-construction-safety"><span class="header-sectio
 <p>LEWG expressed an (unpolled) interest that we explore
 <code>mdspan</code> safety in subsequent work after the fall 2023 Kona
 WG21 meeting. LEWG asked us to explore safety in a way that is not
-specific to <code>aligned_accessor</code>. We consider the section below
-“Generalize <code>is_sufficiently_aligned</code> for all accessors?”,
-that discusses adding a member function <code>is_checkably_valid</code>
-to the accessor requirements, part of that exploration.</p>
+specific to <code>aligned_accessor</code>. Part of that exploration is
+in the section below “Generalize <code>is_sufficiently_aligned</code>
+for all accessors?”. We plan further exploration of this topic
+elsewhere.</p>
 <h2 data-number="5.5" id="is_sufficiently_aligned-is-not-constexpr"><span class="header-section-number">5.5</span>
 <code>is_sufficiently_aligned</code> is not <code>constexpr</code><a href="#is_sufficiently_aligned-is-not-constexpr" class="self-link"></a></h2>
 <p>LEWG reviewed R1 of this proposal at the June 2024 St. Louis WG21
@@ -938,30 +982,28 @@ <h3 data-number="5.6.1" id="is_sufficiently_aligned-is-specific-to-aligned_acces
 <code>mdspan</code> users, Standard C++ offers no way to check whether a
 pointer and a layout mapping’s <code>required_span_size()</code> form a
 valid range.</p>
-<h3 data-number="5.6.2" id="is_checkably_valid-generic-validity-check"><span class="header-section-number">5.6.2</span>
-<code>is_checkably_valid</code>: Generic validity check?<a href="#is_checkably_valid-generic-validity-check" class="self-link"></a></h3>
+<h3 data-number="5.6.2" id="detectably_invalid-generic-validity-check"><span class="header-section-number">5.6.2</span>
+<code>detectably_invalid</code>: Generic validity check?<a href="#detectably_invalid-generic-validity-check" class="self-link"></a></h3>
 <p>During the June 2024 St. Louis WG21 meeting, one LEWG reviewer
-pointed out that code that is generic on the accessor type currently has
-no way to check whether a given data handle is valid. Specifically,
-given a <code>size_t</code> <code>size</code> (e.g., the
-<code>required_span_size()</code> of a given layout mapping), there is
-no way to check whether <span class="math inline">[</span> 0,
-<code>size</code> <span class="math inline">)</span> forms an accessible
-range (see
+(please see Acknowledgments below) pointed out that code that is generic
+on the accessor type currently has no way to check whether a given data
+handle is valid. Specifically, given a <code>size_t</code>
+<code>size</code> (e.g., the <code>required_span_size()</code> of a
+given layout mapping), there is no way to check whether <span class="math inline">[</span> 0, <code>size</code> <span class="math inline">)</span> forms an accessible range (see
 <a href="https://eel.is/c++draft/views.multidim#mdspan.accessor.general-2">[mdspan.accessor.general]
 2</a>) of a given data handle and accessor. The reviewer suggested
 adding a new member function</p>
-<div class="sourceCode" id="cb10"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb10-1"><a href="#cb10-1" aria-hidden="true" tabindex="-1"></a><span class="dt">bool</span> is_checkably_valid<span class="op">(</span>data_handle_type handle, <span class="dt">size_t</span> size<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb10"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb10-1"><a href="#cb10-1" aria-hidden="true" tabindex="-1"></a><span class="dt">bool</span> detectably_invalid<span class="op">(</span>data_handle_type handle, <span class="dt">size_t</span> size<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span></code></pre></div>
 <p>to all <code>mdspan</code> accessors. This would return
-<code>false</code> if the implementation can show that <span class="math inline">[</span> 0, <code>size</code> <span class="math inline">)</span> is <em>not</em> an accessible range for
+<code>true</code> if the implementation can show that <span class="math inline">[</span> 0, <code>size</code> <span class="math inline">)</span> is <em>not</em> an accessible range for
 <code>handle</code> and the accessor, and <code>true</code> otherwise.
-The word “checkably” in the name would remind users that this is a “best
-effort” check. It might return <code>true</code> even if the handle is
-invalid or if <span class="math inline">[</span> 0, <code>size</code>
-<span class="math inline">)</span> is not an accessible range. Also, it
-might return different values on different implementations, depending on
-their ability to check e.g., pointer range validity. The function would
-have the following design features.</p>
+The word “detectably” in the name would remind users that this is a
+“best effort” check. It might return <code>false</code> even if the
+handle is invalid or if <span class="math inline">[</span> 0,
+<code>size</code> <span class="math inline">)</span> is not an
+accessible range. Also, it might return different values on different
+implementations, depending on their ability to check e.g., pointer range
+validity. The function would have the following design features.</p>
 <ul>
 <li><p>It must be a non-<code>static</code> member function, because in
 general, accessors may have state that determines validity of the data
@@ -979,21 +1021,22 @@ <h3 data-number="5.6.2" id="is_checkably_valid-generic-validity-check"><span cla
 <span id="cb11-4"><a href="#cb11-4" aria-hidden="true" tabindex="-1"></a>  LayoutMapping mapping,</span>
 <span id="cb11-5"><a href="#cb11-5" aria-hidden="true" tabindex="-1"></a>  Accessor accessor<span class="op">)</span></span>
 <span id="cb11-6"><a href="#cb11-6" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
-<span id="cb11-7"><a href="#cb11-7" aria-hidden="true" tabindex="-1"></a>  <span class="cf">if</span> <span class="op">(</span><span class="kw">not</span> accessor<span class="op">.</span>is_checkably_valid<span class="op">(</span>handle, mapping<span class="op">.</span>required_span_size<span class="op">()))</span> <span class="op">{</span></span>
+<span id="cb11-7"><a href="#cb11-7" aria-hidden="true" tabindex="-1"></a>  <span class="cf">if</span> <span class="op">(</span>accessor<span class="op">.</span>detectably_invalid<span class="op">(</span>handle, mapping<span class="op">.</span>required_span_size<span class="op">()))</span> <span class="op">{</span></span>
 <span id="cb11-8"><a href="#cb11-8" aria-hidden="true" tabindex="-1"></a>    <span class="cf">throw</span> std<span class="op">::</span>out_of_range<span class="op">(</span><span class="st">&quot;Invalid data handle and/or size&quot;</span><span class="op">)</span>;</span>
 <span id="cb11-9"><a href="#cb11-9" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
 <span id="cb11-10"><a href="#cb11-10" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> mdspan<span class="op">{</span>handle, mapping, accessor<span class="op">}</span>;</span>
 <span id="cb11-11"><a href="#cb11-11" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
-<h3 data-number="5.6.3" id="arguments-against-and-for-is_checkably_valid"><span class="header-section-number">5.6.3</span> Arguments against and for
-<code>is_checkably_valid</code><a href="#arguments-against-and-for-is_checkably_valid" class="self-link"></a></h3>
+<h3 data-number="5.6.3" id="arguments-against-and-for-detectably_invalid"><span class="header-section-number">5.6.3</span> Arguments against and for
+<code>detectably_invalid</code><a href="#arguments-against-and-for-detectably_invalid" class="self-link"></a></h3>
 <p>We didn’t include this feature in the original <code>mdspan</code>
-design because most data handle types have no way to check
-validity.<br />
-We didn’t want to give users the false impression that a validity check
-was doing anything meaningful. Standard C++ has no way to check a raw
-pointer <code>T*</code> and a size, though some implementations such as
-CHERI C++ ([Davis 2019] and [Watson 2020]) do have this feature. We
-designed <code>mdspan</code> accessors to be able to wrap libraries that
+design because most data handle types have no way to say with full
+accuracy whether a handle and size are valid. We didn’t want to give
+users the false impression that a validity check was doing anything
+meaningful. Standard C++ has no way to check a raw pointer
+<code>T*</code> and a size, though some implementations such as CHERI
+C++ ([Davis 2019] and [Watson 2020]) and run-time profiling and
+debugging systems such as Valgrind do have this feature. We designed
+<code>mdspan</code> accessors to be able to wrap libraries that
 implement a partitioned global address space (PGAS) programming model
 for accessing remote data over a network. (See
 <a href="https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2022/p0009r18.html#why-custom-accessors">P0009R18,
@@ -1007,77 +1050,170 @@ <h3 data-number="5.6.3" id="arguments-against-and-for-is_checkably_valid"><span
 “symmetric heap” (that is accessible to all participating parallel
 processes). Such libraries generally do not have validity checks for
 their handles.</p>
-<p>On the other hand, an <code>is_checkably_valid</code> function would
+<p>On the other hand, a <code>detectably_invalid</code> function would
 let happen any checks that <em>could</em> happen. For instance, a
 hypothetical “GPU device memory accessor” (not proposed for the C++
-Standard, but existing in projects like RAPIDS RAFT) might permit access
-to an allocation of GPU “device” memory from only GPU “device” code, not
-from ordinary “host” code. A common use case for GPU allocations is to
-allocate device memory in host code, then pass the pointer to device
-code for use there. Thus, it would be reasonable to create an
-<code>mdspan</code> in host code with that accessor. The accessor could
-use a CUDA run-time function like
+Standard, but existing in projects like
+<a href="https://github.com/rapidsai/raft">RAPIDS RAFT</a>) might permit
+access to an allocation of GPU “device” memory from only GPU “device”
+code, not from ordinary “host” code. A common use case for GPU
+allocations is to allocate device memory in host code, then pass the
+pointer to device code for use there. Thus, it would be reasonable to
+create an <code>mdspan</code> in host code with that accessor. The
+accessor could use a CUDA run-time function like
 <a href="https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__UNIFIED.html"><code>cudaPointerGetAttributes</code></a>
-to check if the pointer points to valid GPU memory.</p>
+to check if the pointer points to valid GPU memory. Even
+<code>default_accessor</code> could have a simple check like this.</p>
+<div class="sourceCode" id="cb12"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb12-1"><a href="#cb12-1" aria-hidden="true" tabindex="-1"></a><span class="dt">bool</span> detectably_invalid<span class="op">(</span>data_handle_type ptr, <span class="dt">size_t</span> size<span class="op">)</span> <span class="kw">const</span> <span class="op">{</span></span>
+<span id="cb12-2"><a href="#cb12-2" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> ptr <span class="op">==</span> <span class="kw">nullptr</span> <span class="op">&amp;&amp;</span> size <span class="op">!=</span> <span class="dv">0</span>;</span>
+<span id="cb12-3"><a href="#cb12-3" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
 <h3 data-number="5.6.4" id="nonmember-customization-point-design"><span class="header-section-number">5.6.4</span> Nonmember customization point
 design<a href="#nonmember-customization-point-design" class="self-link"></a></h3>
-<p>C++23 defines the generic interface of accessors through the
-“accessor policy requirements” in
-<a href="https://eel.is/c++draft/mdspan.accessor.reqmts">[mdspan.accessor.reqmts]</a>.
-Adding <code>is_checkably_valid</code> to these requirements would be a
-breaking change. Thus, generic code that wanted to call this function
-would need to fill in default behavior for existing C++23 or
-user-defined accessors. The following <code>is_checkably_valid</code>
-nonmember function shows one way users could do that. For the full
-source code and tests, please see this
-<a href="https://godbolt.org/z/14YPfMoq3">Compiler Explorer link</a> or
-Appendix A below.</p>
-<p>The point of this function is to show that even if C++26 adds
-<code>is_checkably_valid</code> to the accessor requirements and to all
-the Standard accessors, users could still work around the lack of this
-function in C++23 - compliant accessors. While it could make sense to
-standardize this function, users might like to fill in different default
-behavior, so we do not propose this here.</p>
-<div class="sourceCode" id="cb12"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb12-1"><a href="#cb12-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Accessor<span class="op">&gt;</span></span>
-<span id="cb12-2"><a href="#cb12-2" aria-hidden="true" tabindex="-1"></a><span class="kw">concept</span> has_is_checkably_valid <span class="op">=</span> <span class="kw">requires</span><span class="op">(</span>Accessor acc<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb12-3"><a href="#cb12-3" aria-hidden="true" tabindex="-1"></a>  <span class="co">// Making the existence of the type alias a separate constraint</span></span>
-<span id="cb12-4"><a href="#cb12-4" aria-hidden="true" tabindex="-1"></a>  <span class="co">// gives more user-friendly error messages.</span></span>
-<span id="cb12-5"><a href="#cb12-5" aria-hidden="true" tabindex="-1"></a>  <span class="kw">typename</span> Accessor<span class="op">::</span>data_handle_type;</span>
-<span id="cb12-6"><a href="#cb12-6" aria-hidden="true" tabindex="-1"></a>  <span class="op">{</span> std<span class="op">::</span>as_const<span class="op">(</span>acc<span class="op">).</span>is_checkably_valid<span class="op">(</span></span>
-<span id="cb12-7"><a href="#cb12-7" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>declval<span class="op">&lt;</span><span class="kw">typename</span> Accessor<span class="op">::</span>data_handle_type<span class="op">&gt;()</span>,</span>
-<span id="cb12-8"><a href="#cb12-8" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>declval<span class="op">&lt;</span>std<span class="op">::</span><span class="dt">size_t</span><span class="op">&gt;()</span></span>
-<span id="cb12-9"><a href="#cb12-9" aria-hidden="true" tabindex="-1"></a>    <span class="op">)</span> <span class="op">}</span> <span class="kw">noexcept</span> <span class="op">-&gt;</span> std<span class="op">::</span>convertible_to<span class="op">&lt;</span><span class="dt">bool</span><span class="op">&gt;</span>;</span>
-<span id="cb12-10"><a href="#cb12-10" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
-<span id="cb12-11"><a href="#cb12-11" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb12-12"><a href="#cb12-12" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Accessor<span class="op">&gt;</span></span>
-<span id="cb12-13"><a href="#cb12-13" aria-hidden="true" tabindex="-1"></a><span class="dt">bool</span> is_checkably_valid<span class="op">(</span>Accessor<span class="op">&amp;&amp;</span> accessor,</span>
-<span id="cb12-14"><a href="#cb12-14" aria-hidden="true" tabindex="-1"></a>  <span class="kw">typename</span> std<span class="op">::</span>remove_cvref_t<span class="op">&lt;</span>Accessor<span class="op">&gt;::</span>data_handle_type handle,</span>
-<span id="cb12-15"><a href="#cb12-15" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span></span>
-<span id="cb12-16"><a href="#cb12-16" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
-<span id="cb12-17"><a href="#cb12-17" aria-hidden="true" tabindex="-1"></a>  <span class="cf">if</span> <span class="kw">constexpr</span> <span class="op">(</span>has_is_checkably_valid<span class="op">&lt;</span>std<span class="op">::</span>remove_cvref_t<span class="op">&lt;</span>Accessor<span class="op">&gt;&gt;)</span> <span class="op">{</span></span>
-<span id="cb12-18"><a href="#cb12-18" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> std<span class="op">::</span>as_const<span class="op">(</span>accessor<span class="op">).</span>is_checkably_valid<span class="op">(</span>handle, size<span class="op">)</span>;</span>
-<span id="cb12-19"><a href="#cb12-19" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb12-20"><a href="#cb12-20" aria-hidden="true" tabindex="-1"></a>  <span class="cf">else</span> <span class="op">{</span></span>
-<span id="cb12-21"><a href="#cb12-21" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="kw">true</span>;</span>
-<span id="cb12-22"><a href="#cb12-22" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb12-23"><a href="#cb12-23" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
-<h3 data-number="5.6.5" id="conclusions"><span class="header-section-number">5.6.5</span> Conclusions<a href="#conclusions" class="self-link"></a></h3>
+<p>C++23 defines the generic interface of accessors through the accessor
+policy requirements
+<a href="https://eel.is/c++draft/mdspan.accessor.reqmts"><strong>[mdspan.accessor.reqmts]</strong></a>.
+Adding <code>detectably_invalid</code> to these requirements would be a
+breaking change to C++23. Thus, generic code that wanted to call this
+function would need to fill in default behavior for both Standard
+accessors defined in C++23, and user-defined accessors that comply with
+the C++23 accessor requirements. The following
+<code>detectably_invalid</code> nonmember function (not proposed in this
+paper) shows one way users could do that. Please see Appendix A below
+for the full source code of a demonstration, along with a Compiler
+Explorer link. This demonstration shows that breaking backwards
+compatibility with C++23 is unnecessary, because users can
+straightforwardly work around the lack of a
+<code>detectably_invalid</code> member function in C++23 - compliant
+accessors. Not standardizing this nonmember function work-around would
+also give users the freedom to fill in different default behavior. For
+example, some users may prefer to consider every (data handle, size)
+pair invalid unless proven otherwise, as a way to force use of custom
+accessors that have the ability to make accurate checks.</p>
+<div class="sourceCode" id="cb13"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb13-1"><a href="#cb13-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Accessor<span class="op">&gt;</span></span>
+<span id="cb13-2"><a href="#cb13-2" aria-hidden="true" tabindex="-1"></a><span class="kw">concept</span> has_detectably_invalid <span class="op">=</span> <span class="kw">requires</span><span class="op">(</span>Accessor acc<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb13-3"><a href="#cb13-3" aria-hidden="true" tabindex="-1"></a>  <span class="kw">typename</span> Accessor<span class="op">::</span>data_handle_type;</span>
+<span id="cb13-4"><a href="#cb13-4" aria-hidden="true" tabindex="-1"></a>  <span class="op">{</span> std<span class="op">::</span>as_const<span class="op">(</span>acc<span class="op">).</span>detectably_invalid<span class="op">(</span></span>
+<span id="cb13-5"><a href="#cb13-5" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>declval<span class="op">&lt;</span><span class="kw">typename</span> Accessor<span class="op">::</span>data_handle_type<span class="op">&gt;()</span>,</span>
+<span id="cb13-6"><a href="#cb13-6" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>declval<span class="op">&lt;</span>std<span class="op">::</span><span class="dt">size_t</span><span class="op">&gt;()</span></span>
+<span id="cb13-7"><a href="#cb13-7" aria-hidden="true" tabindex="-1"></a>    <span class="op">)</span> <span class="op">}</span> <span class="kw">noexcept</span> <span class="op">-&gt;</span> std<span class="op">::</span>same_as<span class="op">&lt;</span><span class="dt">bool</span><span class="op">&gt;</span>;</span>
+<span id="cb13-8"><a href="#cb13-8" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb13-9"><a href="#cb13-9" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb13-10"><a href="#cb13-10" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Accessor<span class="op">&gt;</span></span>
+<span id="cb13-11"><a href="#cb13-11" aria-hidden="true" tabindex="-1"></a><span class="dt">bool</span> detectably_invalid<span class="op">(</span>Accessor<span class="op">&amp;&amp;</span> accessor,</span>
+<span id="cb13-12"><a href="#cb13-12" aria-hidden="true" tabindex="-1"></a>  <span class="kw">typename</span> std<span class="op">::</span>remove_cvref_t<span class="op">&lt;</span>Accessor<span class="op">&gt;::</span>data_handle_type handle,</span>
+<span id="cb13-13"><a href="#cb13-13" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span></span>
+<span id="cb13-14"><a href="#cb13-14" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb13-15"><a href="#cb13-15" aria-hidden="true" tabindex="-1"></a>  <span class="cf">if</span> <span class="kw">constexpr</span> <span class="op">(</span>has_detectably_invalid<span class="op">&lt;</span>std<span class="op">::</span>remove_cvref_t<span class="op">&lt;</span>Accessor<span class="op">&gt;&gt;)</span> <span class="op">{</span></span>
+<span id="cb13-16"><a href="#cb13-16" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> std<span class="op">::</span>as_const<span class="op">(</span>accessor<span class="op">).</span>detectably_invalid<span class="op">(</span>handle, size<span class="op">)</span>;</span>
+<span id="cb13-17"><a href="#cb13-17" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb13-18"><a href="#cb13-18" aria-hidden="true" tabindex="-1"></a>  <span class="cf">else</span> <span class="op">{</span></span>
+<span id="cb13-19"><a href="#cb13-19" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="kw">false</span>;</span>
+<span id="cb13-20"><a href="#cb13-20" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb13-21"><a href="#cb13-21" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
+<h3 data-number="5.6.5" id="we-need-both-is_sufficiently_aligned-and-detectably_invalid"><span class="header-section-number">5.6.5</span> We need both
+<code>is_sufficiently_aligned</code> and
+<code>detectably_invalid</code><a href="#we-need-both-is_sufficiently_aligned-and-detectably_invalid" class="self-link"></a></h3>
+<p>The same LEWG reviewer who suggested <code>detectably_invalid</code>
+also said that it would make <code>is_sufficiently_aligned</code>
+unnecessary. We disagree for the following reasons.</p>
+<ol type="1">
+<li><p>Users will often want to check alignment separately from pointer
+range validity.</p></li>
+<li><p>Checking alignment may be much less expensive than checking
+pointer range validity.</p></li>
+</ol>
+<p>Regarding (1), we think the most common use case for
+<code>aligned_accessor</code>’s <code>explicit</code> converting
+constructor from <code>default_accessor</code> would be explicit
+construction of an <code>mdspan</code> with
+<code>aligned_accessor</code> from an <code>mdspan</code> with
+<code>default_accessor</code>. The latter exists, so the user has
+already asserted that the range formed by its data handle and
+<code>required_span_size()</code> is valid. Thus, the only thing the
+user would need to check would be whether the data handle is
+sufficiently aligned.</p>
+<h3 data-number="5.6.6" id="alternative-nonmember-is_sufficiently_aligned"><span class="header-section-number">5.6.6</span> Alternative: Nonmember
+<code>is_sufficiently_aligned</code><a href="#alternative-nonmember-is_sufficiently_aligned" class="self-link"></a></h3>
+<p>The reviewer responded to our argument above by suggesting that we
+remove <code>is_sufficiently_aligned</code> from
+<code>aligned_accessor</code> and make it a separate nonmember function.
+This is a reasonable alternative suggestion. We think that LWG review of
+<code>aligned_accessor</code> can proceed concurrently while LEWG
+considers this alternative.</p>
+<p>Into which header should this new function go? We think it belongs in
+<code>&lt;bit&gt;</code>, because it is fundamentally a bit arithmetic
+operation. There should be no reason why that could not be freestanding,
+like everything else in the <code>&lt;bit&gt;</code> header. Another
+option would be to put it in <code>&lt;memory&gt;</code> right after
+<code>assume_aligned</code>, and to mark it as freestanding (since
+<code>assume_aligned</code> is).</p>
+<h3 data-number="5.6.7" id="do-accessors-need-to-check-anything-else"><span class="header-section-number">5.6.7</span> Do accessors need to check
+anything else?<a href="#do-accessors-need-to-check-anything-else" class="self-link"></a></h3>
+<p>The only other thing an accessor’s user might want to check besides a
+(data handle, size) pair would be converting construction from another
+type of accessor. All <code>mdspan</code> components –
+<code>extents</code>, layout mappings, and accessors – implement
+conversions with preconditions via <code>explicit</code> constructors.
+(For more detail, please see the section below, “Explicit conversions as
+the model for precondition-asserting conversions.”) Accessors do
+<em>not</em> store their data handles, so the only reason to check
+whether converting construction is valid would be if the input or result
+accessor has separate run-time state. (Otherwise, the check could be a
+constraint or <code>static_assert</code>.) It’s rare for an accessor to
+need run-time state, so we don’t expect to need this feature in generic
+code. It would also be a separable addition from the feature of checking
+a data handle and size. Nevertheless, one could consider a design. We
+would favor just overloading <code>detectably_invalid</code> for
+accessors, as there would be no risk of ambiguity. Converting
+constructors only take one argument, so there would be no ambiguity
+between calling <code>detectably_invalid</code> with an accessor and
+calling it with a data handle and size.</p>
+<h3 data-number="5.6.8" id="naming-the-function"><span class="header-section-number">5.6.8</span> Naming the function<a href="#naming-the-function" class="self-link"></a></h3>
+<ol type="1">
+<li><p>The function describes a property: “this (data handle, size) pair
+is not known to be invalid.” It’s an adjective (like
+“<code>valid</code>” or “<code>is_valid</code>”), not a verb (like
+“check” as in “<code>check_valid</code>”).</p></li>
+<li><p>The function does not promise perfect accuracy. In the common
+case, it says whether it can <em>detect</em> whether the handle and size
+are _in_valid. Whether or not they are <em>valid</em> might be harder to
+say.</p></li>
+<li><p>As discussed above, users may also want to check converting
+constructors from other accessor types. However, there would be no risk
+of ambiguity between that and checking a data handle and size.
+Therefore, there’s no need for the function’s name to include the type
+of the thing being checked (e.g., “range”).</p></li>
+<li><p>Specifically, the function should not contain the word “pointer,”
+because a data handle is not necessarily a pointer. Even if
+<code>data_handle_type</code>​ is a pointer type, a data handle might not
+necessarily be a pointer to the elements in the Standard C++ sense. For
+example, it might be some opaque handle that a library represents as a
+type alias of <code>void*</code>​.</p></li>
+</ol>
+<p>These points together suggest the name
+<code>detectably_invalid</code>.</p>
+<h3 data-number="5.6.9" id="conclusions"><span class="header-section-number">5.6.9</span> Conclusions<a href="#conclusions" class="self-link"></a></h3>
 <ol type="1">
-<li><p>Adding <code>is_checkably_valid</code> to the accessor
+<li><p>Adding <code>detectably_invalid</code> to the accessor
 requirements and existing Standard accessors in C++26 would be a
-breaking change. Nevertheless, users could fill in reasonable behavior
-for C++23 accessors.</p></li>
+breaking change to C++23. Nevertheless, users could write code that
+fills in reasonable behavior for C++23 accessors.</p></li>
 <li><p>Few C++ implementations offer a way to check validity of a
 pointer range. Thus, users would experience
-<code>is_checkably_valid</code> as mostly not useful for the common case
+<code>detectably_invalid</code> as mostly not useful for the common case
 of <code>default_accessor</code> and other accessors that access a
 pointer range.</p></li>
 <li><p>Item (1) reduces the urgency of this suggestion for C++26. Item
 (2) reduces its potential to improve the <code>mdspan</code> user
 experience in a practical way. Therefore, we do not suggest adding
-<code>is_checkably_valid</code> to the accessor requirements in this
+<code>detectably_invalid</code> to the accessor requirements in this
 proposal. However, we do not discourage further work in separate
 proposals.</p></li>
+<li><p>We would like LEWG to consider the alternative design that
+removes <code>is_sufficiently_aligned</code> from
+<code>aligned_accessor</code> and adds it to the C++ Standard Library as
+a separate nonmember function. We think LWG review of
+<code>aligned_accessor</code> can proceed concurrently.</p></li>
 </ol>
 <h2 data-number="5.7" id="explicit-conversions-as-the-model-for-precondition-asserting-conversions"><span class="header-section-number">5.7</span> Explicit conversions as the
 model for precondition-asserting conversions<a href="#explicit-conversions-as-the-model-for-precondition-asserting-conversions" class="self-link"></a></h2>
@@ -1097,14 +1233,14 @@ <h3 data-number="5.7.1" id="example-conversion-to-aligned_accessor"><span class=
 <code>default_accessor</code>, even though users know that the
 <code>mdspan</code> is overaligned to <code>8 * sizeof(float)</code>
 bytes. The function’s parameter(s) don’t matter for this example.</p>
-<div class="sourceCode" id="cb13"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb13-1"><a href="#cb13-1" aria-hidden="true" tabindex="-1"></a>mdspan<span class="op">&lt;</span><span class="dt">float</span>, dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, layout_right, default_accessor<span class="op">&lt;</span><span class="dt">float</span><span class="op">&gt;&gt;</span></span>
-<span id="cb13-2"><a href="#cb13-2" aria-hidden="true" tabindex="-1"></a>  overaligned_view<span class="op">(</span>SomeParameters params<span class="op">)</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb14"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb14-1"><a href="#cb14-1" aria-hidden="true" tabindex="-1"></a>mdspan<span class="op">&lt;</span><span class="dt">float</span>, dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, layout_right, default_accessor<span class="op">&lt;</span><span class="dt">float</span><span class="op">&gt;&gt;</span></span>
+<span id="cb14-2"><a href="#cb14-2" aria-hidden="true" tabindex="-1"></a>  overaligned_view<span class="op">(</span>SomeParameters params<span class="op">)</span>;</span></code></pre></div>
 <p>Suppose also that users want to call some other function that they
 can’t change. This function takes an <code>mdspan</code> of
 <code>float</code> with <code>aligned_accessor&lt;float, 8&gt;</code>.
 Its return type doesn’t matter for this example.</p>
-<div class="sourceCode" id="cb14"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb14-1"><a href="#cb14-1" aria-hidden="true" tabindex="-1"></a>SomeReturnType use_overaligned_view<span class="op">(</span></span>
-<span id="cb14-2"><a href="#cb14-2" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">&lt;</span><span class="dt">float</span>, dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, layout_right, aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;)</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb15"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb15-1"><a href="#cb15-1" aria-hidden="true" tabindex="-1"></a>SomeReturnType use_overaligned_view<span class="op">(</span></span>
+<span id="cb15-2"><a href="#cb15-2" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">&lt;</span><span class="dt">float</span>, dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, layout_right, aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;)</span>;</span></code></pre></div>
 <h3 data-number="5.7.2" id="status-quo"><span class="header-section-number">5.7.2</span> Status quo<a href="#status-quo" class="self-link"></a></h3>
 <p>How do users call <code>use_overaligned_view</code> with the object
 returned from <code>overaligned_view</code>? The status quo offers two
@@ -1121,17 +1257,17 @@ <h3 data-number="5.7.2" id="status-quo"><span class="header-section-number">5.7.
 layout mapping, and accessor.)</p></li>
 </ol>
 <p>Way (1) looks like this.</p>
-<div class="sourceCode" id="cb15"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb15-1"><a href="#cb15-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> x <span class="op">=</span> overaligned_view<span class="op">(</span>params<span class="op">)</span>;</span>
-<span id="cb15-2"><a href="#cb15-2" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span></span>
-<span id="cb15-3"><a href="#cb15-3" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">&lt;</span><span class="dt">float</span>, dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, layout_right,</span>
-<span id="cb15-4"><a href="#cb15-4" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;(</span>x<span class="op">)</span></span>
-<span id="cb15-5"><a href="#cb15-5" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
-<p>Way (2) looks like this. Note use of CTAD.</p>
 <div class="sourceCode" id="cb16"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb16-1"><a href="#cb16-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> x <span class="op">=</span> overaligned_view<span class="op">(</span>params<span class="op">)</span>;</span>
 <span id="cb16-2"><a href="#cb16-2" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span></span>
-<span id="cb16-3"><a href="#cb16-3" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">{</span>x<span class="op">.</span>data_handle<span class="op">()</span>, x<span class="op">.</span>mapping<span class="op">()</span>,</span>
-<span id="cb16-4"><a href="#cb16-4" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;(</span>x<span class="op">.</span>accessor<span class="op">())}</span></span>
+<span id="cb16-3"><a href="#cb16-3" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">&lt;</span><span class="dt">float</span>, dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, layout_right,</span>
+<span id="cb16-4"><a href="#cb16-4" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;(</span>x<span class="op">)</span></span>
 <span id="cb16-5"><a href="#cb16-5" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
+<p>Way (2) looks like this. Note use of CTAD.</p>
+<div class="sourceCode" id="cb17"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb17-1"><a href="#cb17-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> x <span class="op">=</span> overaligned_view<span class="op">(</span>params<span class="op">)</span>;</span>
+<span id="cb17-2"><a href="#cb17-2" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span></span>
+<span id="cb17-3"><a href="#cb17-3" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">{</span>x<span class="op">.</span>data_handle<span class="op">()</span>, x<span class="op">.</span>mapping<span class="op">()</span>,</span>
+<span id="cb17-4"><a href="#cb17-4" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;(</span>x<span class="op">.</span>accessor<span class="op">())}</span></span>
+<span id="cb17-5"><a href="#cb17-5" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
 <p>Which way is less verbose depends on <code>mdspan</code>’s template
 arguments. Both ways, though, force the user to name the type
 <code>aligned_accessor&lt;float, 8&gt;</code> explicitly. Users know
@@ -1180,44 +1316,37 @@ <h3 data-number="5.7.4" id="alternative-explicit-cast-function-naughty_cast"><sp
 “<code>naughty_cast</code>.” For the above
 <code>use_overaligned_view</code> example, the <code>naughty_cast</code>
 analog of Way (2) would look like this.</p>
-<div class="sourceCode" id="cb17"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb17-1"><a href="#cb17-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> x <span class="op">=</span> overaligned_view<span class="op">(</span>params<span class="op">)</span>;</span>
-<span id="cb17-2"><a href="#cb17-2" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span></span>
-<span id="cb17-3"><a href="#cb17-3" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">{</span>x<span class="op">.</span>data_handle<span class="op">()</span>, x<span class="op">.</span>mapping<span class="op">()</span>,</span>
-<span id="cb17-4"><a href="#cb17-4" aria-hidden="true" tabindex="-1"></a>    naughty_cast<span class="op">&lt;</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;&gt;(</span>x<span class="op">.</span>accessor<span class="op">())}</span></span>
-<span id="cb17-5"><a href="#cb17-5" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb18"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb18-1"><a href="#cb18-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> x <span class="op">=</span> overaligned_view<span class="op">(</span>params<span class="op">)</span>;</span>
+<span id="cb18-2"><a href="#cb18-2" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span></span>
+<span id="cb18-3"><a href="#cb18-3" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">{</span>x<span class="op">.</span>data_handle<span class="op">()</span>, x<span class="op">.</span>mapping<span class="op">()</span>,</span>
+<span id="cb18-4"><a href="#cb18-4" aria-hidden="true" tabindex="-1"></a>    naughty_cast<span class="op">&lt;</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;&gt;(</span>x<span class="op">.</span>accessor<span class="op">())}</span></span>
+<span id="cb18-5"><a href="#cb18-5" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
 <p>One could imagine defining <code>naughty_cast</code> of
 <code>mdspan</code> by <code>naughty_cast</code> of its components. This
 would enable an analog of Way (1).</p>
-<div class="sourceCode" id="cb18"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb18-1"><a href="#cb18-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> x <span class="op">=</span> overaligned_view<span class="op">(</span>params<span class="op">)</span>;</span>
-<span id="cb18-2"><a href="#cb18-2" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span>naughty_cast<span class="op">&lt;</span></span>
-<span id="cb18-3"><a href="#cb18-3" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">&lt;</span><span class="dt">float</span>, dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, layout_right,</span>
-<span id="cb18-4"><a href="#cb18-4" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;&gt;(</span>x<span class="op">)</span></span>
-<span id="cb18-5"><a href="#cb18-5" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
-<p>Requiring <code>naughty_cast</code> for <code>mdspan</code>
-conversions with preconditions is a bit like requiring
-<code>static_cast</code> for narrowing arithmetic conversions. The
-advantage is that one can search a large code base for
-<code>static_cast</code>. The disadvantage is that the code base will
-become full of <code>static_cast</code>, thus making potentially
-incorrect conversions less searchable anyway.</p>
+<div class="sourceCode" id="cb19"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb19-1"><a href="#cb19-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> x <span class="op">=</span> overaligned_view<span class="op">(</span>params<span class="op">)</span>;</span>
+<span id="cb19-2"><a href="#cb19-2" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span>naughty_cast<span class="op">&lt;</span></span>
+<span id="cb19-3"><a href="#cb19-3" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">&lt;</span><span class="dt">float</span>, dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, layout_right,</span>
+<span id="cb19-4"><a href="#cb19-4" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;&gt;(</span>x<span class="op">)</span></span>
+<span id="cb19-5"><a href="#cb19-5" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
 <p>Another argument for <code>naughty_cast</code> besides searchability
 is to make conversions with preconditions “loud,” that is, easily seen
 in the code by human developers. However, the original Way (1) and Way
 (2) both are loud already in that they require a lot of extra code that
 spells out the result’s accessor type explicitly. The status quo’s
 difference in “volume” is implicit conversion</p>
-<div class="sourceCode" id="cb19"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb19-1"><a href="#cb19-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span>x<span class="op">)</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb20"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb20-1"><a href="#cb20-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span>x<span class="op">)</span>;</span></code></pre></div>
 <p>versus explicit construction.</p>
-<div class="sourceCode" id="cb20"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb20-1"><a href="#cb20-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span></span>
-<span id="cb20-2"><a href="#cb20-2" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">{</span>x<span class="op">.</span>data_handle<span class="op">()</span>, x<span class="op">.</span>mapping<span class="op">()</span>,</span>
-<span id="cb20-3"><a href="#cb20-3" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;(</span>x<span class="op">)})</span>;</span>
-<span id="cb20-4"><a href="#cb20-4" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
-<p>Adding <code>naughty_cast</code> to the latter doesn’t make it much
-louder.</p>
 <div class="sourceCode" id="cb21"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb21-1"><a href="#cb21-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span></span>
 <span id="cb21-2"><a href="#cb21-2" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">{</span>x<span class="op">.</span>data_handle<span class="op">()</span>, x<span class="op">.</span>mapping<span class="op">()</span>,</span>
-<span id="cb21-3"><a href="#cb21-3" aria-hidden="true" tabindex="-1"></a>    naughty_cast<span class="op">&lt;</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;(</span>x<span class="op">)})</span>;</span>
+<span id="cb21-3"><a href="#cb21-3" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;(</span>x<span class="op">)})</span>;</span>
 <span id="cb21-4"><a href="#cb21-4" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
+<p>Adding <code>naughty_cast</code> to the latter doesn’t make it much
+louder.</p>
+<div class="sourceCode" id="cb22"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb22-1"><a href="#cb22-1" aria-hidden="true" tabindex="-1"></a><span class="kw">auto</span> result <span class="op">=</span> use_overaligned_view<span class="op">(</span></span>
+<span id="cb22-2"><a href="#cb22-2" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">{</span>x<span class="op">.</span>data_handle<span class="op">()</span>, x<span class="op">.</span>mapping<span class="op">()</span>,</span>
+<span id="cb22-3"><a href="#cb22-3" aria-hidden="true" tabindex="-1"></a>    naughty_cast<span class="op">&lt;</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;(</span>x<span class="op">)})</span>;</span>
+<span id="cb22-4"><a href="#cb22-4" aria-hidden="true" tabindex="-1"></a><span class="op">)</span>;</span></code></pre></div>
 <p>There are other disadvantages to a <code>naughty_cast</code> design.
 The point of that design would be to remove or make
 non-<code>public</code> all the <code>explicit</code> constructors from
@@ -1227,27 +1356,27 @@ <h3 data-number="5.7.4" id="alternative-explicit-cast-function-naughty_cast"><sp
 one for the input type and one for the output type of the cast. The
 <code>naughty_caster</code> struct example below shows how one could do
 that.</p>
-<div class="sourceCode" id="cb22"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb22-1"><a href="#cb22-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Output, <span class="kw">class</span> Input<span class="op">&gt;</span></span>
-<span id="cb22-2"><a href="#cb22-2" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> naughty_caster <span class="op">{}</span>;</span>
-<span id="cb22-3"><a href="#cb22-3" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb22-4"><a href="#cb22-4" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Output, <span class="kw">class</span> Input<span class="op">&gt;</span></span>
-<span id="cb22-5"><a href="#cb22-5" aria-hidden="true" tabindex="-1"></a>Output naughty_cast<span class="op">(</span><span class="kw">const</span> Input<span class="op">&amp;</span> input<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb22-6"><a href="#cb22-6" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> naughty_caster<span class="op">&lt;</span>Output, Input<span class="op">&gt;::</span>cast<span class="op">(</span>input<span class="op">)</span>;</span>
-<span id="cb22-7"><a href="#cb22-7" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
-<span id="cb22-8"><a href="#cb22-8" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb22-9"><a href="#cb22-9" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OutputElementType, <span class="dt">size_t</span> ByteAlignment,</span>
-<span id="cb22-10"><a href="#cb22-10" aria-hidden="true" tabindex="-1"></a>  <span class="kw">class</span> InputElementType<span class="op">&gt;</span></span>
-<span id="cb22-11"><a href="#cb22-11" aria-hidden="true" tabindex="-1"></a>  <span class="kw">requires</span> <span class="op">(</span>is_convertible_v<span class="op">&lt;</span>InputElementType<span class="op">(*)[]</span>,</span>
-<span id="cb22-12"><a href="#cb22-12" aria-hidden="true" tabindex="-1"></a>    OutputElementType<span class="op">(*)[]&gt;)</span> </span>
-<span id="cb22-13"><a href="#cb22-13" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> naughty_caster <span class="op">{</span></span>
-<span id="cb22-14"><a href="#cb22-14" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> output_type <span class="op">=</span></span>
-<span id="cb22-15"><a href="#cb22-15" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span>OutputElementType, ByteAlignment<span class="op">&gt;</span>;</span>
-<span id="cb22-16"><a href="#cb22-16" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> input_type <span class="op">=</span> default_accessor<span class="op">&lt;</span>InputElementType<span class="op">&gt;</span>;</span>
-<span id="cb22-17"><a href="#cb22-17" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb22-18"><a href="#cb22-18" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static</span> output_type cast<span class="op">(</span><span class="kw">const</span> input_type<span class="op">&amp;)</span> <span class="op">{</span></span>
-<span id="cb22-19"><a href="#cb22-19" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="op">{}</span>; </span>
-<span id="cb22-20"><a href="#cb22-20" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb22-21"><a href="#cb22-21" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb23"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb23-1"><a href="#cb23-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Output, <span class="kw">class</span> Input<span class="op">&gt;</span></span>
+<span id="cb23-2"><a href="#cb23-2" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> naughty_caster <span class="op">{}</span>;</span>
+<span id="cb23-3"><a href="#cb23-3" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb23-4"><a href="#cb23-4" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Output, <span class="kw">class</span> Input<span class="op">&gt;</span></span>
+<span id="cb23-5"><a href="#cb23-5" aria-hidden="true" tabindex="-1"></a>Output naughty_cast<span class="op">(</span><span class="kw">const</span> Input<span class="op">&amp;</span> input<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb23-6"><a href="#cb23-6" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> naughty_caster<span class="op">&lt;</span>Output, Input<span class="op">&gt;::</span>cast<span class="op">(</span>input<span class="op">)</span>;</span>
+<span id="cb23-7"><a href="#cb23-7" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
+<span id="cb23-8"><a href="#cb23-8" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb23-9"><a href="#cb23-9" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OutputElementType, <span class="dt">size_t</span> ByteAlignment,</span>
+<span id="cb23-10"><a href="#cb23-10" aria-hidden="true" tabindex="-1"></a>  <span class="kw">class</span> InputElementType<span class="op">&gt;</span></span>
+<span id="cb23-11"><a href="#cb23-11" aria-hidden="true" tabindex="-1"></a>  <span class="kw">requires</span> <span class="op">(</span>is_convertible_v<span class="op">&lt;</span>InputElementType<span class="op">(*)[]</span>,</span>
+<span id="cb23-12"><a href="#cb23-12" aria-hidden="true" tabindex="-1"></a>    OutputElementType<span class="op">(*)[]&gt;)</span> </span>
+<span id="cb23-13"><a href="#cb23-13" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> naughty_caster <span class="op">{</span></span>
+<span id="cb23-14"><a href="#cb23-14" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> output_type <span class="op">=</span></span>
+<span id="cb23-15"><a href="#cb23-15" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span>OutputElementType, ByteAlignment<span class="op">&gt;</span>;</span>
+<span id="cb23-16"><a href="#cb23-16" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> input_type <span class="op">=</span> default_accessor<span class="op">&lt;</span>InputElementType<span class="op">&gt;</span>;</span>
+<span id="cb23-17"><a href="#cb23-17" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb23-18"><a href="#cb23-18" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static</span> output_type cast<span class="op">(</span><span class="kw">const</span> input_type<span class="op">&amp;)</span> <span class="op">{</span></span>
+<span id="cb23-19"><a href="#cb23-19" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="op">{}</span>; </span>
+<span id="cb23-20"><a href="#cb23-20" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb23-21"><a href="#cb23-21" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span></code></pre></div>
 <p>This technique takes a lot of effort and code, when by far the common
 case is that <code>cast</code> has a trivial body. For any accessors
 with state, it would almost certainly call for breaks of encapsulation,
@@ -1283,58 +1412,58 @@ <h1 data-number="6" id="implementation"><span class="header-section-number">6</s
 <p>We have tested an implementation of this proposal with the <a href="https://github.com/kokkos/mdspan/">reference mdspan
 implementation</a>.</p>
 <h1 data-number="7" id="example"><span class="header-section-number">7</span> Example<a href="#example" class="self-link"></a></h1>
-<div class="sourceCode" id="cb23"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb23-1"><a href="#cb23-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
-<span id="cb23-2"><a href="#cb23-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span></span>
-<span id="cb23-3"><a href="#cb23-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">1</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_right, std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, byte_alignment<span class="op">&gt;&gt;</span>;</span>
-<span id="cb23-4"><a href="#cb23-4" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-5"><a href="#cb23-5" aria-hidden="true" tabindex="-1"></a><span class="co">// Interfaces that require 32-byte alignment,</span></span>
-<span id="cb23-6"><a href="#cb23-6" aria-hidden="true" tabindex="-1"></a><span class="co">// because they want to do 8-wide SIMD of float.</span></span>
-<span id="cb23-7"><a href="#cb23-7" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span> vectorized_axpy<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y, <span class="dt">float</span> alpha, aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> x<span class="op">)</span>;</span>
-<span id="cb23-8"><a href="#cb23-8" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">float</span> vectorized_norm<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y<span class="op">)</span>;</span>
-<span id="cb23-9"><a href="#cb23-9" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-10"><a href="#cb23-10" aria-hidden="true" tabindex="-1"></a><span class="co">// Interfaces that require 16-byte alignment,</span></span>
-<span id="cb23-11"><a href="#cb23-11" aria-hidden="true" tabindex="-1"></a><span class="co">// because they want to do 4-wide SIMD of float.</span></span>
-<span id="cb23-12"><a href="#cb23-12" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span> fill_x<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">16</span><span class="op">&gt;</span> x<span class="op">)</span>;</span>
-<span id="cb23-13"><a href="#cb23-13" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span> fill_y<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">16</span><span class="op">&gt;</span> y<span class="op">)</span>;</span>
-<span id="cb23-14"><a href="#cb23-14" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-15"><a href="#cb23-15" aria-hidden="true" tabindex="-1"></a><span class="co">// Helper functions for making overaligned array allocations.</span></span>
-<span id="cb23-16"><a href="#cb23-16" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-17"><a href="#cb23-17" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
-<span id="cb23-18"><a href="#cb23-18" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> delete_raw <span class="op">{</span></span>
-<span id="cb23-19"><a href="#cb23-19" aria-hidden="true" tabindex="-1"></a>  <span class="dt">void</span> <span class="kw">operator</span><span class="op">()(</span>ElementType<span class="op">*</span> p<span class="op">)</span> <span class="kw">const</span> <span class="op">{</span></span>
-<span id="cb23-20"><a href="#cb23-20" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>free<span class="op">(</span>p<span class="op">)</span>;</span>
-<span id="cb23-21"><a href="#cb23-21" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb23-22"><a href="#cb23-22" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
-<span id="cb23-23"><a href="#cb23-23" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-24"><a href="#cb23-24" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
-<span id="cb23-25"><a href="#cb23-25" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> allocation <span class="op">=</span> std<span class="op">::</span>unique_ptr<span class="op">&lt;</span>ElementType<span class="op">[]</span>, delete_raw<span class="op">&lt;</span>ElementType<span class="op">&gt;&gt;</span>;</span>
-<span id="cb23-26"><a href="#cb23-26" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-27"><a href="#cb23-27" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, std<span class="op">::</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
-<span id="cb23-28"><a href="#cb23-28" aria-hidden="true" tabindex="-1"></a>allocation<span class="op">&lt;</span>ElementType<span class="op">&gt;</span> allocate_raw<span class="op">(</span><span class="kw">const</span> std<span class="op">::</span><span class="dt">size_t</span> num_elements<span class="op">)</span></span>
-<span id="cb23-29"><a href="#cb23-29" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
-<span id="cb23-30"><a href="#cb23-30" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> std<span class="op">::</span><span class="dt">size_t</span> num_bytes <span class="op">=</span> num_elements <span class="op">*</span> <span class="kw">sizeof</span><span class="op">(</span>ElementType<span class="op">)</span>;</span>
-<span id="cb23-31"><a href="#cb23-31" aria-hidden="true" tabindex="-1"></a>  <span class="dt">void</span><span class="op">*</span> ptr <span class="op">=</span> std<span class="op">::</span>aligned_alloc<span class="op">(</span>byte_alignment, num_bytes<span class="op">)</span>;</span>
-<span id="cb23-32"><a href="#cb23-32" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> <span class="op">{</span>ptr, delete_raw<span class="op">&lt;</span>ElementType<span class="op">&gt;{}}</span>;</span>
-<span id="cb23-33"><a href="#cb23-33" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
-<span id="cb23-34"><a href="#cb23-34" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-35"><a href="#cb23-35" aria-hidden="true" tabindex="-1"></a><span class="dt">float</span> user_function<span class="op">(</span><span class="dt">size_t</span> num_elements, <span class="dt">float</span> alpha<span class="op">)</span></span>
-<span id="cb23-36"><a href="#cb23-36" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
-<span id="cb23-37"><a href="#cb23-37" aria-hidden="true" tabindex="-1"></a>  <span class="co">// Code using the above two interfaces needs to allocate to the max alignment.</span></span>
-<span id="cb23-38"><a href="#cb23-38" aria-hidden="true" tabindex="-1"></a>  <span class="co">// Users could also query aligned_accessor::byte_alignment</span></span>
-<span id="cb23-39"><a href="#cb23-39" aria-hidden="true" tabindex="-1"></a>  <span class="co">// for the various interfaces and take the max.</span></span>
-<span id="cb23-40"><a href="#cb23-40" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="dt">size_t</span> max_byte_alignment <span class="op">=</span> <span class="dv">32</span>;</span>
-<span id="cb23-41"><a href="#cb23-41" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> x_alloc <span class="op">=</span> allocate_raw<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
-<span id="cb23-42"><a href="#cb23-42" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> y_alloc <span class="op">=</span> allocate_raw<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
-<span id="cb23-43"><a href="#cb23-43" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-44"><a href="#cb23-44" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> x<span class="op">(</span>x_alloc<span class="op">.</span>get<span class="op">())</span>;</span>
-<span id="cb23-45"><a href="#cb23-45" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> y<span class="op">(</span>y_alloc<span class="op">.</span>get<span class="op">())</span>;</span>
-<span id="cb23-46"><a href="#cb23-46" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-47"><a href="#cb23-47" aria-hidden="true" tabindex="-1"></a>  fill_x<span class="op">(</span>x<span class="op">)</span>; <span class="co">// automatic conversion from 32-byte aligned to 16-byte aligned</span></span>
-<span id="cb23-48"><a href="#cb23-48" aria-hidden="true" tabindex="-1"></a>  fill_y<span class="op">(</span>y<span class="op">)</span>; <span class="co">// automatic conversion again</span></span>
-<span id="cb23-49"><a href="#cb23-49" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb23-50"><a href="#cb23-50" aria-hidden="true" tabindex="-1"></a>  vectorized_axpy<span class="op">(</span>y, alpha, x<span class="op">)</span>;</span>
-<span id="cb23-51"><a href="#cb23-51" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> vectorized_norm<span class="op">(</span>y<span class="op">)</span>;</span>
-<span id="cb23-52"><a href="#cb23-52" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
+<div class="sourceCode" id="cb24"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb24-1"><a href="#cb24-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb24-2"><a href="#cb24-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span></span>
+<span id="cb24-3"><a href="#cb24-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">1</span>, <span class="dt">int</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_right, std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, byte_alignment<span class="op">&gt;&gt;</span>;</span>
+<span id="cb24-4"><a href="#cb24-4" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-5"><a href="#cb24-5" aria-hidden="true" tabindex="-1"></a><span class="co">// Interfaces that require 32-byte alignment,</span></span>
+<span id="cb24-6"><a href="#cb24-6" aria-hidden="true" tabindex="-1"></a><span class="co">// because they want to do 8-wide SIMD of float.</span></span>
+<span id="cb24-7"><a href="#cb24-7" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span> vectorized_axpy<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y, <span class="dt">float</span> alpha, aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> x<span class="op">)</span>;</span>
+<span id="cb24-8"><a href="#cb24-8" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">float</span> vectorized_norm<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y<span class="op">)</span>;</span>
+<span id="cb24-9"><a href="#cb24-9" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-10"><a href="#cb24-10" aria-hidden="true" tabindex="-1"></a><span class="co">// Interfaces that require 16-byte alignment,</span></span>
+<span id="cb24-11"><a href="#cb24-11" aria-hidden="true" tabindex="-1"></a><span class="co">// because they want to do 4-wide SIMD of float.</span></span>
+<span id="cb24-12"><a href="#cb24-12" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span> fill_x<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">16</span><span class="op">&gt;</span> x<span class="op">)</span>;</span>
+<span id="cb24-13"><a href="#cb24-13" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span> fill_y<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">16</span><span class="op">&gt;</span> y<span class="op">)</span>;</span>
+<span id="cb24-14"><a href="#cb24-14" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-15"><a href="#cb24-15" aria-hidden="true" tabindex="-1"></a><span class="co">// Helper functions for making overaligned array allocations.</span></span>
+<span id="cb24-16"><a href="#cb24-16" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-17"><a href="#cb24-17" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
+<span id="cb24-18"><a href="#cb24-18" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> delete_raw <span class="op">{</span></span>
+<span id="cb24-19"><a href="#cb24-19" aria-hidden="true" tabindex="-1"></a>  <span class="dt">void</span> <span class="kw">operator</span><span class="op">()(</span>ElementType<span class="op">*</span> p<span class="op">)</span> <span class="kw">const</span> <span class="op">{</span></span>
+<span id="cb24-20"><a href="#cb24-20" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>free<span class="op">(</span>p<span class="op">)</span>;</span>
+<span id="cb24-21"><a href="#cb24-21" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb24-22"><a href="#cb24-22" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb24-23"><a href="#cb24-23" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-24"><a href="#cb24-24" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
+<span id="cb24-25"><a href="#cb24-25" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> allocation <span class="op">=</span> std<span class="op">::</span>unique_ptr<span class="op">&lt;</span>ElementType<span class="op">[]</span>, delete_raw<span class="op">&lt;</span>ElementType<span class="op">&gt;&gt;</span>;</span>
+<span id="cb24-26"><a href="#cb24-26" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-27"><a href="#cb24-27" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, std<span class="op">::</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb24-28"><a href="#cb24-28" aria-hidden="true" tabindex="-1"></a>allocation<span class="op">&lt;</span>ElementType<span class="op">&gt;</span> allocate_raw<span class="op">(</span><span class="kw">const</span> std<span class="op">::</span><span class="dt">size_t</span> num_elements<span class="op">)</span></span>
+<span id="cb24-29"><a href="#cb24-29" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb24-30"><a href="#cb24-30" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> std<span class="op">::</span><span class="dt">size_t</span> num_bytes <span class="op">=</span> num_elements <span class="op">*</span> <span class="kw">sizeof</span><span class="op">(</span>ElementType<span class="op">)</span>;</span>
+<span id="cb24-31"><a href="#cb24-31" aria-hidden="true" tabindex="-1"></a>  <span class="dt">void</span><span class="op">*</span> ptr <span class="op">=</span> std<span class="op">::</span>aligned_alloc<span class="op">(</span>byte_alignment, num_bytes<span class="op">)</span>;</span>
+<span id="cb24-32"><a href="#cb24-32" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> <span class="op">{</span>ptr, delete_raw<span class="op">&lt;</span>ElementType<span class="op">&gt;{}}</span>;</span>
+<span id="cb24-33"><a href="#cb24-33" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
+<span id="cb24-34"><a href="#cb24-34" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-35"><a href="#cb24-35" aria-hidden="true" tabindex="-1"></a><span class="dt">float</span> user_function<span class="op">(</span><span class="dt">size_t</span> num_elements, <span class="dt">float</span> alpha<span class="op">)</span></span>
+<span id="cb24-36"><a href="#cb24-36" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb24-37"><a href="#cb24-37" aria-hidden="true" tabindex="-1"></a>  <span class="co">// Code using the above two interfaces needs to allocate to the max alignment.</span></span>
+<span id="cb24-38"><a href="#cb24-38" aria-hidden="true" tabindex="-1"></a>  <span class="co">// Users could also query aligned_accessor::byte_alignment</span></span>
+<span id="cb24-39"><a href="#cb24-39" aria-hidden="true" tabindex="-1"></a>  <span class="co">// for the various interfaces and take the max.</span></span>
+<span id="cb24-40"><a href="#cb24-40" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="dt">size_t</span> max_byte_alignment <span class="op">=</span> <span class="dv">32</span>;</span>
+<span id="cb24-41"><a href="#cb24-41" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> x_alloc <span class="op">=</span> allocate_raw<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
+<span id="cb24-42"><a href="#cb24-42" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> y_alloc <span class="op">=</span> allocate_raw<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
+<span id="cb24-43"><a href="#cb24-43" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-44"><a href="#cb24-44" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> x<span class="op">(</span>x_alloc<span class="op">.</span>get<span class="op">())</span>;</span>
+<span id="cb24-45"><a href="#cb24-45" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> y<span class="op">(</span>y_alloc<span class="op">.</span>get<span class="op">())</span>;</span>
+<span id="cb24-46"><a href="#cb24-46" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-47"><a href="#cb24-47" aria-hidden="true" tabindex="-1"></a>  fill_x<span class="op">(</span>x<span class="op">)</span>; <span class="co">// automatic conversion from 32-byte aligned to 16-byte aligned</span></span>
+<span id="cb24-48"><a href="#cb24-48" aria-hidden="true" tabindex="-1"></a>  fill_y<span class="op">(</span>y<span class="op">)</span>; <span class="co">// automatic conversion again</span></span>
+<span id="cb24-49"><a href="#cb24-49" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb24-50"><a href="#cb24-50" aria-hidden="true" tabindex="-1"></a>  vectorized_axpy<span class="op">(</span>y, alpha, x<span class="op">)</span>;</span>
+<span id="cb24-51"><a href="#cb24-51" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> vectorized_norm<span class="op">(</span>y<span class="op">)</span>;</span>
+<span id="cb24-52"><a href="#cb24-52" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
 <h1 data-number="8" id="references"><span class="header-section-number">8</span> References<a href="#references" class="self-link"></a></h1>
 <ul>
 <li><p>Davis et al., “CheriABI: Enforcing Valid Pointer Provenance and
@@ -1347,14 +1476,19 @@ <h1 data-number="8" id="references"><span class="header-section-number">8</span>
 Available online [last accessed 2024-07-05]:
 <a href="https://doi.org/10.48456/tr-947">https://doi.org/10.48456/tr-947</a></p></li>
 </ul>
-<h1 data-number="9" id="wording"><span class="header-section-number">9</span> Wording<a href="#wording" class="self-link"></a></h1>
+<h1 data-number="9" id="acknowledgments"><span class="header-section-number">9</span> Acknowledgments<a href="#acknowledgments" class="self-link"></a></h1>
+<ul>
+<li>For <code>detectably_invalid</code>, credit (with permission) to
+David Sankel (Adobe), <code>dsankel@adobe.com</code></li>
+</ul>
+<h1 data-number="10" id="wording"><span class="header-section-number">10</span> Wording<a href="#wording" class="self-link"></a></h1>
 <blockquote>
 <p>Text in blockquotes is not proposed wording, but rather instructions
 for generating proposed wording. The � character is used to denote a
 placeholder section number which the editor shall determine.</p>
 <p>In <em>[version.syn]</em>, add</p>
 </blockquote>
-<div class="sourceCode" id="cb24"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb24-1"><a href="#cb24-1" aria-hidden="true" tabindex="-1"></a><span class="pp">#define __cpp_lib_aligned_accessor </span>YYYYMML<span class="pp"> </span><span class="co">// also in &lt;mdspan&gt;</span></span></code></pre></div>
+<div class="sourceCode" id="cb25"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb25-1"><a href="#cb25-1" aria-hidden="true" tabindex="-1"></a><span class="pp">#define __cpp_lib_aligned_accessor </span>YYYYMML<span class="pp"> </span><span class="co">// also in &lt;mdspan&gt;</span></span></code></pre></div>
 <blockquote>
 <p>Adjust the placeholder value as needed so as to denote this
 proposal’s date of adoption.</p>
@@ -1362,48 +1496,48 @@ <h1 data-number="9" id="wording"><span class="header-section-number">9</span> Wo
 <strong>[mdspan.syn]</strong>, after <code>class default_accessor</code>
 and before <code>class mdspan</code>, add the following.</p>
 </blockquote>
-<div class="sourceCode" id="cb25"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb25-1"><a href="#cb25-1" aria-hidden="true" tabindex="-1"></a><span class="co">// [mdspan.accessor.aligned], class template aligned_accessor</span></span>
-<span id="cb25-2"><a href="#cb25-2" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
-<span id="cb25-3"><a href="#cb25-3" aria-hidden="true" tabindex="-1"></a>  <span class="kw">class</span> aligned_accessor;</span></code></pre></div>
+<div class="sourceCode" id="cb26"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb26-1"><a href="#cb26-1" aria-hidden="true" tabindex="-1"></a><span class="co">// [mdspan.accessor.aligned], class template aligned_accessor</span></span>
+<span id="cb26-2"><a href="#cb26-2" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb26-3"><a href="#cb26-3" aria-hidden="true" tabindex="-1"></a>  <span class="kw">class</span> aligned_accessor;</span></code></pre></div>
 <blockquote>
 <p>At the end of <strong>[mdspan.accessor.default]</strong> and before
 <strong>[mdspan.mdspan]</strong>, add all the material that follows.</p>
 </blockquote>
-<h2 data-number="9.1" id="add-subsection-mdspan.accessor.aligned-with-the-following"><span class="header-section-number">9.1</span> Add subsection �
+<h2 data-number="10.1" id="add-subsection-mdspan.accessor.aligned-with-the-following"><span class="header-section-number">10.1</span> Add subsection �
 [mdspan.accessor.aligned] with the following<a href="#add-subsection-mdspan.accessor.aligned-with-the-following" class="self-link"></a></h2>
 <p><b> � Class template <code>aligned_accessor</code>
 [mdspan.accessor.aligned] </b></p>
 <p><b> �.1 Overview [mdspan.accessor.aligned.overview] </b></p>
-<div class="sourceCode" id="cb26"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb26-1"><a href="#cb26-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> the_byte_alignment<span class="op">&gt;</span></span>
-<span id="cb26-2"><a href="#cb26-2" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> aligned_accessor <span class="op">{</span></span>
-<span id="cb26-3"><a href="#cb26-3" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> offset_policy <span class="op">=</span> default_accessor<span class="op">&lt;</span>ElementType<span class="op">&gt;</span>;</span>
-<span id="cb26-4"><a href="#cb26-4" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> element_type <span class="op">=</span> ElementType;</span>
-<span id="cb26-5"><a href="#cb26-5" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> reference <span class="op">=</span> ElementType<span class="op">&amp;</span>;</span>
-<span id="cb26-6"><a href="#cb26-6" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> ElementType<span class="op">*</span>;</span>
-<span id="cb26-7"><a href="#cb26-7" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb26-8"><a href="#cb26-8" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static</span> <span class="kw">constexpr</span> <span class="dt">size_t</span> byte_alignment <span class="op">=</span> the_byte_alignment;</span>
-<span id="cb26-9"><a href="#cb26-9" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb26-10"><a href="#cb26-10" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> aligned_accessor<span class="op">()</span> <span class="kw">noexcept</span> <span class="op">=</span> <span class="cf">default</span>;</span>
-<span id="cb26-11"><a href="#cb26-11" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb26-12"><a href="#cb26-12" aria-hidden="true" tabindex="-1"></a>  <span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OtherElementType, <span class="dt">size_t</span> other_byte_alignment<span class="op">&gt;</span></span>
-<span id="cb26-13"><a href="#cb26-13" aria-hidden="true" tabindex="-1"></a>    <span class="kw">constexpr</span> aligned_accessor<span class="op">(</span></span>
-<span id="cb26-14"><a href="#cb26-14" aria-hidden="true" tabindex="-1"></a>      aligned_accessor<span class="op">&lt;</span>OtherElementType, other_byte_alignment<span class="op">&gt;)</span> <span class="kw">noexcept</span>;</span>
-<span id="cb26-15"><a href="#cb26-15" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb26-16"><a href="#cb26-16" aria-hidden="true" tabindex="-1"></a>  <span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OtherElementType<span class="op">&gt;</span></span>
-<span id="cb26-17"><a href="#cb26-17" aria-hidden="true" tabindex="-1"></a>    <span class="kw">explicit</span> <span class="kw">constexpr</span> aligned_accessor<span class="op">(</span></span>
-<span id="cb26-18"><a href="#cb26-18" aria-hidden="true" tabindex="-1"></a>      default_accessor<span class="op">&lt;</span>OtherElementType<span class="op">&gt;)</span> <span class="kw">noexcept</span>;</span>
-<span id="cb26-19"><a href="#cb26-19" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb26-20"><a href="#cb26-20" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="kw">operator</span> default_accessor<span class="op">&lt;</span>element_type<span class="op">&gt;()</span> <span class="kw">const</span> <span class="op">{</span></span>
-<span id="cb26-21"><a href="#cb26-21" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="op">{}</span>;</span>
-<span id="cb26-22"><a href="#cb26-22" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb26-23"><a href="#cb26-23" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb26-24"><a href="#cb26-24" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> reference access<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span>
-<span id="cb26-25"><a href="#cb26-25" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb26-26"><a href="#cb26-26" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="kw">typename</span> offset_policy<span class="op">::</span>data_handle_type</span>
-<span id="cb26-27"><a href="#cb26-27" aria-hidden="true" tabindex="-1"></a>    offset<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span>
-<span id="cb26-28"><a href="#cb26-28" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb26-29"><a href="#cb26-29" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static</span> <span class="dt">bool</span> is_sufficiently_aligned<span class="op">(</span>data_handle_type p<span class="op">)</span>;</span>
-<span id="cb26-30"><a href="#cb26-30" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb27"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb27-1"><a href="#cb27-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> the_byte_alignment<span class="op">&gt;</span></span>
+<span id="cb27-2"><a href="#cb27-2" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> aligned_accessor <span class="op">{</span></span>
+<span id="cb27-3"><a href="#cb27-3" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> offset_policy <span class="op">=</span> default_accessor<span class="op">&lt;</span>ElementType<span class="op">&gt;</span>;</span>
+<span id="cb27-4"><a href="#cb27-4" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> element_type <span class="op">=</span> ElementType;</span>
+<span id="cb27-5"><a href="#cb27-5" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> reference <span class="op">=</span> ElementType<span class="op">&amp;</span>;</span>
+<span id="cb27-6"><a href="#cb27-6" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> ElementType<span class="op">*</span>;</span>
+<span id="cb27-7"><a href="#cb27-7" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb27-8"><a href="#cb27-8" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static</span> <span class="kw">constexpr</span> <span class="dt">size_t</span> byte_alignment <span class="op">=</span> the_byte_alignment;</span>
+<span id="cb27-9"><a href="#cb27-9" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb27-10"><a href="#cb27-10" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> aligned_accessor<span class="op">()</span> <span class="kw">noexcept</span> <span class="op">=</span> <span class="cf">default</span>;</span>
+<span id="cb27-11"><a href="#cb27-11" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb27-12"><a href="#cb27-12" aria-hidden="true" tabindex="-1"></a>  <span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OtherElementType, <span class="dt">size_t</span> other_byte_alignment<span class="op">&gt;</span></span>
+<span id="cb27-13"><a href="#cb27-13" aria-hidden="true" tabindex="-1"></a>    <span class="kw">constexpr</span> aligned_accessor<span class="op">(</span></span>
+<span id="cb27-14"><a href="#cb27-14" aria-hidden="true" tabindex="-1"></a>      aligned_accessor<span class="op">&lt;</span>OtherElementType, other_byte_alignment<span class="op">&gt;)</span> <span class="kw">noexcept</span>;</span>
+<span id="cb27-15"><a href="#cb27-15" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb27-16"><a href="#cb27-16" aria-hidden="true" tabindex="-1"></a>  <span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OtherElementType<span class="op">&gt;</span></span>
+<span id="cb27-17"><a href="#cb27-17" aria-hidden="true" tabindex="-1"></a>    <span class="kw">explicit</span> <span class="kw">constexpr</span> aligned_accessor<span class="op">(</span></span>
+<span id="cb27-18"><a href="#cb27-18" aria-hidden="true" tabindex="-1"></a>      default_accessor<span class="op">&lt;</span>OtherElementType<span class="op">&gt;)</span> <span class="kw">noexcept</span>;</span>
+<span id="cb27-19"><a href="#cb27-19" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb27-20"><a href="#cb27-20" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="kw">operator</span> default_accessor<span class="op">&lt;</span>element_type<span class="op">&gt;()</span> <span class="kw">const</span> <span class="op">{</span></span>
+<span id="cb27-21"><a href="#cb27-21" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="op">{}</span>;</span>
+<span id="cb27-22"><a href="#cb27-22" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb27-23"><a href="#cb27-23" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb27-24"><a href="#cb27-24" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> reference access<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span>
+<span id="cb27-25"><a href="#cb27-25" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb27-26"><a href="#cb27-26" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="kw">typename</span> offset_policy<span class="op">::</span>data_handle_type</span>
+<span id="cb27-27"><a href="#cb27-27" aria-hidden="true" tabindex="-1"></a>    offset<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span>
+<span id="cb27-28"><a href="#cb27-28" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb27-29"><a href="#cb27-29" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static</span> <span class="dt">bool</span> is_sufficiently_aligned<span class="op">(</span>data_handle_type p<span class="op">)</span>;</span>
+<span id="cb27-30"><a href="#cb27-30" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span></code></pre></div>
 <p><span class="marginalizedparent"><a class="marginalized">1</a></span>
 <em>Mandates</em>:</p>
 <ul>
@@ -1426,11 +1560,11 @@ <h2 data-number="9.1" id="add-subsection-mdspan.accessor.aligned-with-the-follow
 <span class="math inline">[0, <em>n</em>)</span> is an accessible range
 for an object <code>p</code> of type <code>data_handle_type</code> and
 an object of type <code>aligned_accessor</code> if and only if <span class="math inline">[</span><code>p</code>, <code>p</code> + <span class="math inline"><em>n</em>)</span> is a valid range.</p>
-<h2 data-number="9.2" id="members-mdspan.accessor.aligned.members"><span class="header-section-number">9.2</span> Members
+<h2 data-number="10.2" id="members-mdspan.accessor.aligned.members"><span class="header-section-number">10.2</span> Members
 [mdspan.accessor.aligned.members]<a href="#members-mdspan.accessor.aligned.members" class="self-link"></a></h2>
-<div class="sourceCode" id="cb27"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb27-1"><a href="#cb27-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OtherElementType, <span class="dt">size_t</span> other_byte_alignment<span class="op">&gt;</span></span>
-<span id="cb27-2"><a href="#cb27-2" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> aligned_accessor<span class="op">(</span></span>
-<span id="cb27-3"><a href="#cb27-3" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span>OtherElementType, other_byte_alignment<span class="op">&gt;)</span> <span class="kw">noexcept</span> <span class="op">{}</span></span></code></pre></div>
+<div class="sourceCode" id="cb28"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb28-1"><a href="#cb28-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OtherElementType, <span class="dt">size_t</span> other_byte_alignment<span class="op">&gt;</span></span>
+<span id="cb28-2"><a href="#cb28-2" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> aligned_accessor<span class="op">(</span></span>
+<span id="cb28-3"><a href="#cb28-3" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span>OtherElementType, other_byte_alignment<span class="op">&gt;)</span> <span class="kw">noexcept</span> <span class="op">{}</span></span></code></pre></div>
 <p><span class="marginalizedparent"><a class="marginalized">1</a></span>
 <em>Constraints</em>:
 <code>is_convertible_v&lt;OtherElementType(*)[], element_type(*)[]&gt;</code>
@@ -1439,14 +1573,14 @@ <h2 data-number="9.2" id="members-mdspan.accessor.aligned.members"><span class="
 <em>Mandates</em>:
 <code>gcd(other_byte_alignment, byte_alignment) == byte_alignment</code>
 is <code>true</code>.</p>
-<div class="sourceCode" id="cb28"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb28-1"><a href="#cb28-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OtherElementType<span class="op">&gt;</span></span>
-<span id="cb28-2"><a href="#cb28-2" aria-hidden="true" tabindex="-1"></a>  <span class="kw">explicit</span> <span class="kw">constexpr</span> aligned_accessor<span class="op">(</span></span>
-<span id="cb28-3"><a href="#cb28-3" aria-hidden="true" tabindex="-1"></a>    default_accessor<span class="op">&lt;</span>OtherElementType<span class="op">&gt;)</span> <span class="kw">noexcept</span> <span class="op">{}</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb29"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb29-1"><a href="#cb29-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OtherElementType<span class="op">&gt;</span></span>
+<span id="cb29-2"><a href="#cb29-2" aria-hidden="true" tabindex="-1"></a>  <span class="kw">explicit</span> <span class="kw">constexpr</span> aligned_accessor<span class="op">(</span></span>
+<span id="cb29-3"><a href="#cb29-3" aria-hidden="true" tabindex="-1"></a>    default_accessor<span class="op">&lt;</span>OtherElementType<span class="op">&gt;)</span> <span class="kw">noexcept</span> <span class="op">{}</span>;</span></code></pre></div>
 <p><span class="marginalizedparent"><a class="marginalized">3</a></span>
 <em>Constraints</em>:
 <code>is_convertible_v&lt;OtherElementType(*)[], element_type(*)[]&gt;</code>
 is <code>true</code>.</p>
-<div class="sourceCode" id="cb29"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb29-1"><a href="#cb29-1" aria-hidden="true" tabindex="-1"></a><span class="kw">constexpr</span> reference access<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb30"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb30-1"><a href="#cb30-1" aria-hidden="true" tabindex="-1"></a><span class="kw">constexpr</span> reference access<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span></code></pre></div>
 <p><span class="marginalizedparent"><a class="marginalized">4</a></span>
 <em>Preconditions</em>: <code>p</code> points to an object
 <code>X</code> of a type similar (<strong>[conv.qual]</strong>) to
@@ -1455,8 +1589,8 @@ <h2 data-number="9.2" id="members-mdspan.accessor.aligned.members"><span class="
 <p><span class="marginalizedparent"><a class="marginalized">5</a></span>
 <em>Effects</em>: Equivalent to:
 <code>return assume_aligned&lt;byte_alignment&gt;(p)[i];</code></p>
-<div class="sourceCode" id="cb30"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb30-1"><a href="#cb30-1" aria-hidden="true" tabindex="-1"></a><span class="kw">constexpr</span> <span class="kw">typename</span> offset_policy<span class="op">::</span>data_handle_type</span>
-<span id="cb30-2"><a href="#cb30-2" aria-hidden="true" tabindex="-1"></a>  offset<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb31"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb31-1"><a href="#cb31-1" aria-hidden="true" tabindex="-1"></a><span class="kw">constexpr</span> <span class="kw">typename</span> offset_policy<span class="op">::</span>data_handle_type</span>
+<span id="cb31-2"><a href="#cb31-2" aria-hidden="true" tabindex="-1"></a>  offset<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span>;</span></code></pre></div>
 <p><span class="marginalizedparent"><a class="marginalized">6</a></span>
 <em>Preconditions</em>: <code>p</code> points to an object
 <code>X</code> of a type similar (<strong>[conv.qual]</strong>) to
@@ -1464,7 +1598,7 @@ <h2 data-number="9.2" id="members-mdspan.accessor.aligned.members"><span class="
 <code>byte_alignment</code> (<strong>[basic.align]</strong>).</p>
 <p><span class="marginalizedparent"><a class="marginalized">7</a></span>
 <em>Effects</em>: Equivalent to: <code>return p + i;</code></p>
-<div class="sourceCode" id="cb31"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb31-1"><a href="#cb31-1" aria-hidden="true" tabindex="-1"></a><span class="kw">static</span> <span class="dt">bool</span> is_sufficiently_aligned<span class="op">(</span>data_handle_type p<span class="op">)</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb32"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb32-1"><a href="#cb32-1" aria-hidden="true" tabindex="-1"></a><span class="kw">static</span> <span class="dt">bool</span> is_sufficiently_aligned<span class="op">(</span>data_handle_type p<span class="op">)</span>;</span></code></pre></div>
 <p><span class="marginalizedparent"><a class="marginalized">8</a></span>
 <em>Preconditions</em>: <code>p</code> points to an object
 <code>X</code> of a type similar (<strong>[conv.qual]</strong>) to
@@ -1485,236 +1619,286 @@ <h2 data-number="9.2" id="members-mdspan.accessor.aligned.members"><span class="
 less optimized function
 <code>compute_without_requiring_overalignment</code> that has no
 overalignment requirement.</p>
-<div class="sourceCode" id="cb32"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb32-1"><a href="#cb32-1" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span></span>
-<span id="cb32-2"><a href="#cb32-2" aria-hidden="true" tabindex="-1"></a>compute_using_fourfold_overalignment<span class="op">(</span></span>
-<span id="cb32-3"><a href="#cb32-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">size_t</span>, <span class="dv">1</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_right,</span>
-<span id="cb32-4"><a href="#cb32-4" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">4</span> <span class="op">*</span> <span class="kw">alignof</span><span class="op">(</span><span class="dt">float</span><span class="op">)&gt;&gt;</span> x<span class="op">)</span>;</span>
-<span id="cb32-5"><a href="#cb32-5" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb32-6"><a href="#cb32-6" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span></span>
-<span id="cb32-7"><a href="#cb32-7" aria-hidden="true" tabindex="-1"></a>compute_without_requiring_overalignment<span class="op">(</span></span>
-<span id="cb32-8"><a href="#cb32-8" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">size_t</span>, <span class="dv">1</span><span class="op">&gt;&gt;</span> x<span class="op">)</span>;</span>
-<span id="cb32-9"><a href="#cb32-9" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb32-10"><a href="#cb32-10" aria-hidden="true" tabindex="-1"></a><span class="dt">void</span> compute<span class="op">(</span>std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">size_t</span>, <span class="dv">1</span><span class="op">&gt;&gt;</span> x<span class="op">)</span></span>
-<span id="cb32-11"><a href="#cb32-11" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
-<span id="cb32-12"><a href="#cb32-12" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> accessor <span class="op">=</span> aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">4</span> <span class="op">*</span> <span class="kw">sizeof</span><span class="op">(</span><span class="dt">float</span><span class="op">)&gt;{}</span>;</span>
-<span id="cb32-13"><a href="#cb32-13" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> x_handle <span class="op">=</span> x<span class="op">.</span>data_handle<span class="op">()</span>;</span>
-<span id="cb32-14"><a href="#cb32-14" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb32-15"><a href="#cb32-15" aria-hidden="true" tabindex="-1"></a>  <span class="cf">if</span> <span class="op">(</span>accessor<span class="op">.</span>is_sufficiently_aligned<span class="op">(</span>x_handle<span class="op">))</span> <span class="op">{</span></span>
-<span id="cb32-16"><a href="#cb32-16" aria-hidden="true" tabindex="-1"></a>    compute_using_fourfold_overalignment<span class="op">(</span>mdspan<span class="op">{</span>x_handle, x<span class="op">.</span>mapping<span class="op">()</span>, accessor<span class="op">})</span>;</span>
-<span id="cb32-17"><a href="#cb32-17" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb32-18"><a href="#cb32-18" aria-hidden="true" tabindex="-1"></a>  <span class="cf">else</span> <span class="op">{</span></span>
-<span id="cb32-19"><a href="#cb32-19" aria-hidden="true" tabindex="-1"></a>    compute_without_requiring_overalignment<span class="op">(</span>x<span class="op">)</span>;</span>
-<span id="cb32-20"><a href="#cb32-20" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb32-21"><a href="#cb32-21" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
+<div class="sourceCode" id="cb33"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb33-1"><a href="#cb33-1" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span></span>
+<span id="cb33-2"><a href="#cb33-2" aria-hidden="true" tabindex="-1"></a>compute_using_fourfold_overalignment<span class="op">(</span></span>
+<span id="cb33-3"><a href="#cb33-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_right,</span>
+<span id="cb33-4"><a href="#cb33-4" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">4</span> <span class="op">*</span> <span class="kw">alignof</span><span class="op">(</span><span class="dt">float</span><span class="op">)&gt;&gt;</span> x<span class="op">)</span>;</span>
+<span id="cb33-5"><a href="#cb33-5" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb33-6"><a href="#cb33-6" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span></span>
+<span id="cb33-7"><a href="#cb33-7" aria-hidden="true" tabindex="-1"></a>compute_without_requiring_overalignment<span class="op">(</span></span>
+<span id="cb33-8"><a href="#cb33-8" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;&gt;</span> x<span class="op">)</span>;</span>
+<span id="cb33-9"><a href="#cb33-9" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb33-10"><a href="#cb33-10" aria-hidden="true" tabindex="-1"></a><span class="dt">void</span> compute<span class="op">(</span>std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;&gt;</span> x<span class="op">)</span></span>
+<span id="cb33-11"><a href="#cb33-11" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb33-12"><a href="#cb33-12" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> accessor <span class="op">=</span> aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">4</span> <span class="op">*</span> <span class="kw">sizeof</span><span class="op">(</span><span class="dt">float</span><span class="op">)&gt;{}</span>;</span>
+<span id="cb33-13"><a href="#cb33-13" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> x_handle <span class="op">=</span> x<span class="op">.</span>data_handle<span class="op">()</span>;</span>
+<span id="cb33-14"><a href="#cb33-14" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb33-15"><a href="#cb33-15" aria-hidden="true" tabindex="-1"></a>  <span class="cf">if</span> <span class="op">(</span>accessor<span class="op">.</span>is_sufficiently_aligned<span class="op">(</span>x_handle<span class="op">))</span> <span class="op">{</span></span>
+<span id="cb33-16"><a href="#cb33-16" aria-hidden="true" tabindex="-1"></a>    compute_using_fourfold_overalignment<span class="op">(</span>mdspan<span class="op">{</span>x_handle, x<span class="op">.</span>mapping<span class="op">()</span>, accessor<span class="op">})</span>;</span>
+<span id="cb33-17"><a href="#cb33-17" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb33-18"><a href="#cb33-18" aria-hidden="true" tabindex="-1"></a>  <span class="cf">else</span> <span class="op">{</span></span>
+<span id="cb33-19"><a href="#cb33-19" aria-hidden="true" tabindex="-1"></a>    compute_without_requiring_overalignment<span class="op">(</span>x<span class="op">)</span>;</span>
+<span id="cb33-20"><a href="#cb33-20" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb33-21"><a href="#cb33-21" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
 <p>–<em>end example</em>]</p>
 <p>[<em>Example:</em> The following example shows how users can fulfill
 the preconditions of <code>aligned_accessor</code> by using existing C++
 Standard Library functionality to create overaligned allocations. First,
 the <code>allocate_overaligned</code> helper function uses
 <code>aligned_alloc</code> to create an overaligned allocation.</p>
-<div class="sourceCode" id="cb33"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb33-1"><a href="#cb33-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
-<span id="cb33-2"><a href="#cb33-2" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> delete_with_free <span class="op">{</span></span>
-<span id="cb33-3"><a href="#cb33-3" aria-hidden="true" tabindex="-1"></a>  <span class="dt">void</span> <span class="kw">operator</span><span class="op">()(</span>ElementType<span class="op">*</span> p<span class="op">)</span> <span class="kw">const</span> <span class="op">{</span></span>
-<span id="cb33-4"><a href="#cb33-4" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>free<span class="op">(</span>p<span class="op">)</span>;</span>
-<span id="cb33-5"><a href="#cb33-5" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb33-6"><a href="#cb33-6" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
-<span id="cb33-7"><a href="#cb33-7" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb33-8"><a href="#cb33-8" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
-<span id="cb33-9"><a href="#cb33-9" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> allocation <span class="op">=</span> std<span class="op">::</span>unique_ptr<span class="op">&lt;</span>ElementType<span class="op">[]</span>, delete_with_free<span class="op">&lt;</span>ElementType<span class="op">&gt;&gt;</span>;</span>
-<span id="cb33-10"><a href="#cb33-10" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb33-11"><a href="#cb33-11" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
-<span id="cb33-12"><a href="#cb33-12" aria-hidden="true" tabindex="-1"></a>allocation<span class="op">&lt;</span>ElementType<span class="op">&gt;</span> allocate_overaligned<span class="op">(</span><span class="kw">const</span> <span class="dt">size_t</span> num_elements<span class="op">)</span></span>
-<span id="cb33-13"><a href="#cb33-13" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
-<span id="cb33-14"><a href="#cb33-14" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> <span class="dt">size_t</span> num_bytes <span class="op">=</span> num_elements <span class="op">*</span> <span class="kw">sizeof</span><span class="op">(</span>ElementType<span class="op">)</span>;</span>
-<span id="cb33-15"><a href="#cb33-15" aria-hidden="true" tabindex="-1"></a>  <span class="dt">void</span><span class="op">*</span> ptr <span class="op">=</span> std<span class="op">::</span>aligned_alloc<span class="op">(</span>byte_alignment, num_bytes<span class="op">)</span>;</span>
-<span id="cb33-16"><a href="#cb33-16" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> <span class="op">{</span>ptr, delete_with_free<span class="op">&lt;</span>ElementType<span class="op">&gt;{}}</span>;</span>
-<span id="cb33-17"><a href="#cb33-17" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
+<div class="sourceCode" id="cb34"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb34-1"><a href="#cb34-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
+<span id="cb34-2"><a href="#cb34-2" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> delete_with_free <span class="op">{</span></span>
+<span id="cb34-3"><a href="#cb34-3" aria-hidden="true" tabindex="-1"></a>  <span class="dt">void</span> <span class="kw">operator</span><span class="op">()(</span>ElementType<span class="op">*</span> p<span class="op">)</span> <span class="kw">const</span> <span class="op">{</span></span>
+<span id="cb34-4"><a href="#cb34-4" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>free<span class="op">(</span>p<span class="op">)</span>;</span>
+<span id="cb34-5"><a href="#cb34-5" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb34-6"><a href="#cb34-6" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb34-7"><a href="#cb34-7" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb34-8"><a href="#cb34-8" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
+<span id="cb34-9"><a href="#cb34-9" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> allocation <span class="op">=</span> std<span class="op">::</span>unique_ptr<span class="op">&lt;</span>ElementType<span class="op">[]</span>, delete_with_free<span class="op">&lt;</span>ElementType<span class="op">&gt;&gt;</span>;</span>
+<span id="cb34-10"><a href="#cb34-10" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb34-11"><a href="#cb34-11" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb34-12"><a href="#cb34-12" aria-hidden="true" tabindex="-1"></a>allocation<span class="op">&lt;</span>ElementType<span class="op">&gt;</span> allocate_overaligned<span class="op">(</span><span class="kw">const</span> <span class="dt">size_t</span> num_elements<span class="op">)</span></span>
+<span id="cb34-13"><a href="#cb34-13" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb34-14"><a href="#cb34-14" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> <span class="dt">size_t</span> num_bytes <span class="op">=</span> num_elements <span class="op">*</span> <span class="kw">sizeof</span><span class="op">(</span>ElementType<span class="op">)</span>;</span>
+<span id="cb34-15"><a href="#cb34-15" aria-hidden="true" tabindex="-1"></a>  <span class="dt">void</span><span class="op">*</span> ptr <span class="op">=</span> std<span class="op">::</span>aligned_alloc<span class="op">(</span>byte_alignment, num_bytes<span class="op">)</span>;</span>
+<span id="cb34-16"><a href="#cb34-16" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> <span class="op">{</span>ptr, delete_with_free<span class="op">&lt;</span>ElementType<span class="op">&gt;{}}</span>;</span>
+<span id="cb34-17"><a href="#cb34-17" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
 <p>Second, this example presumes that some third-party library provides
 functions requiring arrays of <code>float</code> to have 32-byte
 alignment.</p>
-<div class="sourceCode" id="cb34"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb34-1"><a href="#cb34-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
-<span id="cb34-2"><a href="#cb34-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dextents<span class="op">&lt;</span><span class="dt">int</span>, <span class="dv">1</span><span class="op">&gt;</span>,</span>
-<span id="cb34-3"><a href="#cb34-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>layout_right, std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, byte_alignment<span class="op">&gt;&gt;</span>;</span>
-<span id="cb34-4"><a href="#cb34-4" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb34-5"><a href="#cb34-5" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span> vectorized_axpy<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y, <span class="dt">float</span> alpha, aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> x<span class="op">)</span>;</span>
-<span id="cb34-6"><a href="#cb34-6" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">float</span> vectorized_norm<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y<span class="op">)</span>;</span></code></pre></div>
+<div class="sourceCode" id="cb35"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb35-1"><a href="#cb35-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb35-2"><a href="#cb35-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span> std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">1</span>, <span class="dt">int</span><span class="op">&gt;</span>,</span>
+<span id="cb35-3"><a href="#cb35-3" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>layout_right, std<span class="op">::</span>aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, byte_alignment<span class="op">&gt;&gt;</span>;</span>
+<span id="cb35-4"><a href="#cb35-4" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb35-5"><a href="#cb35-5" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">void</span> vectorized_axpy<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y, <span class="dt">float</span> alpha, aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> x<span class="op">)</span>;</span>
+<span id="cb35-6"><a href="#cb35-6" aria-hidden="true" tabindex="-1"></a><span class="kw">extern</span> <span class="dt">float</span> vectorized_norm<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y<span class="op">)</span>;</span></code></pre></div>
 <p>Third and finally, the user’s function <code>user_function</code>
 would begin by allocating “raw” overaligned arrays with
 <code>allocate_overaligned</code>. It would then create aligned
 <code>mdspan</code> with them, and pass the resulting
 <code>mdspan</code> into the third-party library’s functions.</p>
-<div class="sourceCode" id="cb35"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb35-1"><a href="#cb35-1" aria-hidden="true" tabindex="-1"></a><span class="dt">float</span> user_function<span class="op">(</span><span class="dt">size_t</span> num_elements, <span class="dt">float</span> alpha<span class="op">)</span></span>
-<span id="cb35-2"><a href="#cb35-2" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
-<span id="cb35-3"><a href="#cb35-3" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="dt">size_t</span> max_byte_alignment <span class="op">=</span> <span class="dv">32</span>;</span>
-<span id="cb35-4"><a href="#cb35-4" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> x_alloc <span class="op">=</span> allocate_overaligned<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
-<span id="cb35-5"><a href="#cb35-5" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> y_alloc <span class="op">=</span> allocate_overaligned<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
-<span id="cb35-6"><a href="#cb35-6" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb35-7"><a href="#cb35-7" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> x<span class="op">(</span>x_alloc<span class="op">.</span>get<span class="op">())</span>;</span>
-<span id="cb35-8"><a href="#cb35-8" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> y<span class="op">(</span>y_alloc<span class="op">.</span>get<span class="op">())</span>;</span>
-<span id="cb35-9"><a href="#cb35-9" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb35-10"><a href="#cb35-10" aria-hidden="true" tabindex="-1"></a>  <span class="co">// ... fill the elements of x and y ...</span></span>
-<span id="cb35-11"><a href="#cb35-11" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb35-12"><a href="#cb35-12" aria-hidden="true" tabindex="-1"></a>  vectorized_axpy<span class="op">(</span>y, alpha, x<span class="op">)</span>;</span>
-<span id="cb35-13"><a href="#cb35-13" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> vectorized_norm<span class="op">(</span>y<span class="op">)</span>;</span>
-<span id="cb35-14"><a href="#cb35-14" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
+<div class="sourceCode" id="cb36"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb36-1"><a href="#cb36-1" aria-hidden="true" tabindex="-1"></a><span class="dt">float</span> user_function<span class="op">(</span><span class="dt">size_t</span> num_elements, <span class="dt">float</span> alpha<span class="op">)</span></span>
+<span id="cb36-2"><a href="#cb36-2" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb36-3"><a href="#cb36-3" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="dt">size_t</span> max_byte_alignment <span class="op">=</span> <span class="dv">32</span>;</span>
+<span id="cb36-4"><a href="#cb36-4" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> x_alloc <span class="op">=</span> allocate_overaligned<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
+<span id="cb36-5"><a href="#cb36-5" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> y_alloc <span class="op">=</span> allocate_overaligned<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
+<span id="cb36-6"><a href="#cb36-6" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb36-7"><a href="#cb36-7" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> x<span class="op">(</span>x_alloc<span class="op">.</span>get<span class="op">())</span>;</span>
+<span id="cb36-8"><a href="#cb36-8" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> y<span class="op">(</span>y_alloc<span class="op">.</span>get<span class="op">())</span>;</span>
+<span id="cb36-9"><a href="#cb36-9" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb36-10"><a href="#cb36-10" aria-hidden="true" tabindex="-1"></a>  <span class="co">// ... fill the elements of x and y ...</span></span>
+<span id="cb36-11"><a href="#cb36-11" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb36-12"><a href="#cb36-12" aria-hidden="true" tabindex="-1"></a>  vectorized_axpy<span class="op">(</span>y, alpha, x<span class="op">)</span>;</span>
+<span id="cb36-13"><a href="#cb36-13" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> vectorized_norm<span class="op">(</span>y<span class="op">)</span>;</span>
+<span id="cb36-14"><a href="#cb36-14" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
 <p>–<em>end example</em>]</p>
-<h1 data-number="10" id="appendix-a-is_checkably_valid-nonmember-function-example"><span class="header-section-number">10</span> Appendix A:
-<code>is_checkably_valid</code> nonmember function example<a href="#appendix-a-is_checkably_valid-nonmember-function-example" class="self-link"></a></h1>
+<h1 data-number="11" id="wording-for-alternative-nonmember-is_sufficiently_aligned-design"><span class="header-section-number">11</span> Wording for alternative
+nonmember <code>is_sufficiently_aligned</code> design<a href="#wording-for-alternative-nonmember-is_sufficiently_aligned-design" class="self-link"></a></h1>
+<blockquote>
+<p>Text in blockquotes is not proposed wording, but rather instructions
+for generating proposed wording. The instructions here are expressed as
+a “diff” atop P2897R2, that is, as if the wording proposed in P2897R2
+had already been applied to the Working Draft. The � character is used
+to denote a placeholder section number which the editor shall
+determine.</p>
+</blockquote>
+<blockquote>
+<p>From the <code>aligned_accessor</code> synopsis in
+<strong>[mdspan.accessor.aligned.overview]</strong>, remove the member
+function
+<code>static bool is_sufficiently_aligned(data_handle_type p);</code>.</p>
+</blockquote>
+<blockquote>
+<p>From the list of members of <code>aligned_accessor</code> in
+<strong>[mdspan.accessor.aligned.members]</strong>, remove
+<code>static bool is_sufficiently_aligned(data_handle_type p);</code>
+from after line 7 (“<em>Effects</em>: Equivalent to:
+<code>return p + i;</code>”), and remove line 8
+(“<em>Preconditions</em>: …”) and line 9 (“<em>Returns</em>: …”).
+(Retain the <em>Example</em> using <code>is_sufficiently_aligned</code>
+that immediately follows the deleted lines.)</p>
+</blockquote>
+<blockquote>
+<p>To the Header <code>&lt;bit&gt;</code> synopsis
+<strong>[bit.syn]</strong>, after <code>enum class endian</code> and
+before the <code>}</code> that closes <code>namespace std</code>, add
+the following.</p>
+</blockquote>
+<div class="sourceCode" id="cb37"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb37-1"><a href="#cb37-1" aria-hidden="true" tabindex="-1"></a><span class="co">// [bit.aligned], is_sufficiently_aligned</span></span>
+<span id="cb37-2"><a href="#cb37-2" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb37-3"><a href="#cb37-3" aria-hidden="true" tabindex="-1"></a>  <span class="dt">bool</span> is_sufficiently_aligned<span class="op">(</span>ElementType<span class="op">*)</span>;</span></code></pre></div>
+<blockquote>
+<p>After <strong>[bit.endian]</strong>, add a new section
+<strong>[bit.aligned]</strong> and put in it all the material that
+follows.</p>
+</blockquote>
+<div class="sourceCode" id="cb38"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb38-1"><a href="#cb38-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, <span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb38-2"><a href="#cb38-2" aria-hidden="true" tabindex="-1"></a><span class="dt">bool</span> is_sufficiently_aligned<span class="op">(</span>ElementType<span class="op">*</span> p<span class="op">)</span>;</span></code></pre></div>
+<p><span class="marginalizedparent"><a class="marginalized">1</a></span>
+<em>Preconditions</em>: <code>p</code> points to an object
+<code>X</code> of a type similar (<strong>[conv.qual]</strong>) to
+<code>ElementType</code>.</p>
+<p><span class="marginalizedparent"><a class="marginalized">2</a></span>
+<em>Returns</em>: <code>true</code> if <code>X</code> has alignment at
+least <code>byte_alignment</code>, else <code>false</code>.</p>
+<h1 data-number="12" id="appendix-a-detectably_invalid-nonmember-function-example"><span class="header-section-number">12</span> Appendix A:
+<code>detectably_invalid</code> nonmember function example<a href="#appendix-a-detectably_invalid-nonmember-function-example" class="self-link"></a></h1>
 <p>This section is nonnormative. This is the full source code with tests
-for the <code>is_checkably_valid</code> nonmember function example
+for the <code>detectably_invalid</code> nonmember function example
 above. Please see this
-<a href="https://godbolt.org/z/hKh3vK11f">Compiler Explorer
-link</a>.</p>
-<div class="sourceCode" id="cb36"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb36-1"><a href="#cb36-1" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;cassert&gt;</span></span>
-<span id="cb36-2"><a href="#cb36-2" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;concepts&gt;</span></span>
-<span id="cb36-3"><a href="#cb36-3" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;cstdint&gt;</span></span>
-<span id="cb36-4"><a href="#cb36-4" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;iostream&gt;</span></span>
-<span id="cb36-5"><a href="#cb36-5" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;stdexcept&gt;</span></span>
-<span id="cb36-6"><a href="#cb36-6" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;type_traits&gt;</span></span>
-<span id="cb36-7"><a href="#cb36-7" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;utility&gt;</span></span>
-<span id="cb36-8"><a href="#cb36-8" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-9"><a href="#cb36-9" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Accessor<span class="op">&gt;</span></span>
-<span id="cb36-10"><a href="#cb36-10" aria-hidden="true" tabindex="-1"></a><span class="kw">concept</span> has_is_checkably_valid <span class="op">=</span> <span class="kw">requires</span><span class="op">(</span>Accessor acc<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb36-11"><a href="#cb36-11" aria-hidden="true" tabindex="-1"></a>  <span class="kw">typename</span> Accessor<span class="op">::</span>data_handle_type;</span>
-<span id="cb36-12"><a href="#cb36-12" aria-hidden="true" tabindex="-1"></a>  <span class="op">{</span> std<span class="op">::</span>as_const<span class="op">(</span>acc<span class="op">).</span>is_checkably_valid<span class="op">(</span></span>
-<span id="cb36-13"><a href="#cb36-13" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>declval<span class="op">&lt;</span><span class="kw">typename</span> Accessor<span class="op">::</span>data_handle_type<span class="op">&gt;()</span>,</span>
-<span id="cb36-14"><a href="#cb36-14" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>declval<span class="op">&lt;</span>std<span class="op">::</span><span class="dt">size_t</span><span class="op">&gt;()</span></span>
-<span id="cb36-15"><a href="#cb36-15" aria-hidden="true" tabindex="-1"></a>    <span class="op">)</span> <span class="op">}</span> <span class="kw">noexcept</span> <span class="op">-&gt;</span> std<span class="op">::</span>convertible_to<span class="op">&lt;</span><span class="dt">bool</span><span class="op">&gt;</span>;</span>
-<span id="cb36-16"><a href="#cb36-16" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
-<span id="cb36-17"><a href="#cb36-17" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-18"><a href="#cb36-18" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Accessor<span class="op">&gt;</span></span>
-<span id="cb36-19"><a href="#cb36-19" aria-hidden="true" tabindex="-1"></a><span class="dt">bool</span> is_checkably_valid<span class="op">(</span>Accessor<span class="op">&amp;&amp;</span> accessor,</span>
-<span id="cb36-20"><a href="#cb36-20" aria-hidden="true" tabindex="-1"></a>  <span class="kw">typename</span> std<span class="op">::</span>remove_cvref_t<span class="op">&lt;</span>Accessor<span class="op">&gt;::</span>data_handle_type handle,</span>
-<span id="cb36-21"><a href="#cb36-21" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span></span>
-<span id="cb36-22"><a href="#cb36-22" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
-<span id="cb36-23"><a href="#cb36-23" aria-hidden="true" tabindex="-1"></a>  <span class="cf">if</span> <span class="kw">constexpr</span> <span class="op">(</span>has_is_checkably_valid<span class="op">&lt;</span>std<span class="op">::</span>remove_cvref_t<span class="op">&lt;</span>Accessor<span class="op">&gt;&gt;)</span> <span class="op">{</span></span>
-<span id="cb36-24"><a href="#cb36-24" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> std<span class="op">::</span>as_const<span class="op">(</span>accessor<span class="op">).</span>is_checkably_valid<span class="op">(</span>handle, size<span class="op">)</span>;</span>
-<span id="cb36-25"><a href="#cb36-25" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-26"><a href="#cb36-26" aria-hidden="true" tabindex="-1"></a>  <span class="cf">else</span> <span class="op">{</span></span>
-<span id="cb36-27"><a href="#cb36-27" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="kw">true</span>;</span>
-<span id="cb36-28"><a href="#cb36-28" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-29"><a href="#cb36-29" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
-<span id="cb36-30"><a href="#cb36-30" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-31"><a href="#cb36-31" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> A <span class="op">{</span></span>
-<span id="cb36-32"><a href="#cb36-32" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> <span class="dt">float</span><span class="op">*</span>;</span>
-<span id="cb36-33"><a href="#cb36-33" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-34"><a href="#cb36-34" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static</span> <span class="dt">bool</span> is_checkably_valid<span class="op">(</span>data_handle_type ptr, std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span> <span class="kw">noexcept</span> <span class="op">{</span></span>
-<span id="cb36-35"><a href="#cb36-35" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> ptr <span class="op">!=</span> <span class="kw">nullptr</span> <span class="op">||</span> size <span class="op">==</span> <span class="dv">0</span>;</span>
-<span id="cb36-36"><a href="#cb36-36" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-37"><a href="#cb36-37" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
-<span id="cb36-38"><a href="#cb36-38" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-39"><a href="#cb36-39" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> B <span class="op">{</span></span>
-<span id="cb36-40"><a href="#cb36-40" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> <span class="dt">float</span><span class="op">*</span>;</span>
-<span id="cb36-41"><a href="#cb36-41" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
-<span id="cb36-42"><a href="#cb36-42" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-43"><a href="#cb36-43" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> C <span class="op">{</span></span>
-<span id="cb36-44"><a href="#cb36-44" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> <span class="dt">float</span><span class="op">*</span>;</span>
-<span id="cb36-45"><a href="#cb36-45" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-46"><a href="#cb36-46" aria-hidden="true" tabindex="-1"></a>  <span class="co">// This is nonconst, so it&#39;s not actually called.</span></span>
-<span id="cb36-47"><a href="#cb36-47" aria-hidden="true" tabindex="-1"></a>  <span class="dt">bool</span> is_checkably_valid<span class="op">(</span>data_handle_type ptr, std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb36-48"><a href="#cb36-48" aria-hidden="true" tabindex="-1"></a>    <span class="cf">throw</span> std<span class="op">::</span>runtime_error<span class="op">(</span><span class="st">&quot;C::is_checkably_valid: uh oh&quot;</span><span class="op">)</span>;</span>
-<span id="cb36-49"><a href="#cb36-49" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-50"><a href="#cb36-50" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
-<span id="cb36-51"><a href="#cb36-51" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-52"><a href="#cb36-52" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> D <span class="op">{</span></span>
-<span id="cb36-53"><a href="#cb36-53" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> <span class="dt">float</span><span class="op">*</span>;</span>
-<span id="cb36-54"><a href="#cb36-54" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-55"><a href="#cb36-55" aria-hidden="true" tabindex="-1"></a>  <span class="co">// This is const but not noexcept, so it&#39;s not actually called.</span></span>
-<span id="cb36-56"><a href="#cb36-56" aria-hidden="true" tabindex="-1"></a>  <span class="dt">bool</span> is_checkably_valid<span class="op">(</span>data_handle_type ptr, std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span> <span class="kw">const</span> <span class="op">{</span></span>
-<span id="cb36-57"><a href="#cb36-57" aria-hidden="true" tabindex="-1"></a>    <span class="cf">throw</span> std<span class="op">::</span>runtime_error<span class="op">(</span><span class="st">&quot;D::is_checkably_valid: uh oh&quot;</span><span class="op">)</span>;</span>
-<span id="cb36-58"><a href="#cb36-58" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-59"><a href="#cb36-59" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
-<span id="cb36-60"><a href="#cb36-60" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-61"><a href="#cb36-61" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-62"><a href="#cb36-62" aria-hidden="true" tabindex="-1"></a><span class="dt">int</span> main<span class="op">()</span></span>
-<span id="cb36-63"><a href="#cb36-63" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
-<span id="cb36-64"><a href="#cb36-64" aria-hidden="true" tabindex="-1"></a>  <span class="dt">float</span><span class="op">*</span> ptr <span class="op">=</span> <span class="kw">nullptr</span>;</span>
-<span id="cb36-65"><a href="#cb36-65" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-66"><a href="#cb36-66" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>A<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-67"><a href="#cb36-67" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> is_checkably_valid<span class="op">(</span>A<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-68"><a href="#cb36-68" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-69"><a href="#cb36-69" aria-hidden="true" tabindex="-1"></a>  A a<span class="op">{}</span>;</span>
-<span id="cb36-70"><a href="#cb36-70" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>a, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-71"><a href="#cb36-71" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> is_checkably_valid<span class="op">(</span>a, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-72"><a href="#cb36-72" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-73"><a href="#cb36-73" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> A a_c<span class="op">{}</span>;</span>
-<span id="cb36-74"><a href="#cb36-74" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>a_c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-75"><a href="#cb36-75" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> is_checkably_valid<span class="op">(</span>a_c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-76"><a href="#cb36-76" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-77"><a href="#cb36-77" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>B<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-78"><a href="#cb36-78" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>B<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-79"><a href="#cb36-79" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-80"><a href="#cb36-80" aria-hidden="true" tabindex="-1"></a>  <span class="co">// B doesn&#39;t know how to check pointer validity.</span></span>
-<span id="cb36-81"><a href="#cb36-81" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-82"><a href="#cb36-82" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>B<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-83"><a href="#cb36-83" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>B<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-84"><a href="#cb36-84" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-85"><a href="#cb36-85" aria-hidden="true" tabindex="-1"></a>  B b<span class="op">{}</span>;</span>
-<span id="cb36-86"><a href="#cb36-86" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>b, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-87"><a href="#cb36-87" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>b, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-88"><a href="#cb36-88" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-89"><a href="#cb36-89" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> B b_c<span class="op">{}</span>;</span>
-<span id="cb36-90"><a href="#cb36-90" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>b_c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-91"><a href="#cb36-91" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>b_c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-92"><a href="#cb36-92" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-93"><a href="#cb36-93" aria-hidden="true" tabindex="-1"></a>  <span class="co">// If users make is_checkably_valid nonconst or not noexcept,</span></span>
-<span id="cb36-94"><a href="#cb36-94" aria-hidden="true" tabindex="-1"></a>  <span class="co">// the nonmember function falls back to a default implementation.</span></span>
-<span id="cb36-95"><a href="#cb36-95" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-96"><a href="#cb36-96" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
-<span id="cb36-97"><a href="#cb36-97" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>C<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-98"><a href="#cb36-98" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>C<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-99"><a href="#cb36-99" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-100"><a href="#cb36-100" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb36-101"><a href="#cb36-101" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;C{} threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
-<span id="cb36-102"><a href="#cb36-102" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-103"><a href="#cb36-103" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-104"><a href="#cb36-104" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
-<span id="cb36-105"><a href="#cb36-105" aria-hidden="true" tabindex="-1"></a>    <span class="kw">const</span> C c_c<span class="op">{}</span>;</span>
-<span id="cb36-106"><a href="#cb36-106" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>c_c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-107"><a href="#cb36-107" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>c_c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-108"><a href="#cb36-108" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-109"><a href="#cb36-109" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb36-110"><a href="#cb36-110" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;const C threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
-<span id="cb36-111"><a href="#cb36-111" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-112"><a href="#cb36-112" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-113"><a href="#cb36-113" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
-<span id="cb36-114"><a href="#cb36-114" aria-hidden="true" tabindex="-1"></a>    C c<span class="op">{}</span>;</span>
-<span id="cb36-115"><a href="#cb36-115" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-116"><a href="#cb36-116" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-117"><a href="#cb36-117" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-118"><a href="#cb36-118" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb36-119"><a href="#cb36-119" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;nonconst C threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
-<span id="cb36-120"><a href="#cb36-120" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-121"><a href="#cb36-121" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-122"><a href="#cb36-122" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
-<span id="cb36-123"><a href="#cb36-123" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>D<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-124"><a href="#cb36-124" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>D<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-125"><a href="#cb36-125" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-126"><a href="#cb36-126" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb36-127"><a href="#cb36-127" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;D{} threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
-<span id="cb36-128"><a href="#cb36-128" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-129"><a href="#cb36-129" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-130"><a href="#cb36-130" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
-<span id="cb36-131"><a href="#cb36-131" aria-hidden="true" tabindex="-1"></a>    <span class="kw">const</span> D d_c<span class="op">{}</span>;</span>
-<span id="cb36-132"><a href="#cb36-132" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>d_c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-133"><a href="#cb36-133" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>d_c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-134"><a href="#cb36-134" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-135"><a href="#cb36-135" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb36-136"><a href="#cb36-136" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;const D threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
-<span id="cb36-137"><a href="#cb36-137" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-138"><a href="#cb36-138" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-139"><a href="#cb36-139" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
-<span id="cb36-140"><a href="#cb36-140" aria-hidden="true" tabindex="-1"></a>    D d<span class="op">{}</span>;</span>
-<span id="cb36-141"><a href="#cb36-141" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>d, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
-<span id="cb36-142"><a href="#cb36-142" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span>is_checkably_valid<span class="op">(</span>d, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
-<span id="cb36-143"><a href="#cb36-143" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-144"><a href="#cb36-144" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
-<span id="cb36-145"><a href="#cb36-145" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;nonconst D threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
-<span id="cb36-146"><a href="#cb36-146" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
-<span id="cb36-147"><a href="#cb36-147" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-148"><a href="#cb36-148" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;Made it to the end</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
-<span id="cb36-149"><a href="#cb36-149" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> <span class="dv">0</span>;</span>
-<span id="cb36-150"><a href="#cb36-150" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
+<a href="https://godbolt.org/z/4P9MGbhdj">Compiler Explorer link</a> for
+a test with five different compilers: GCC 14.1, Clang 18.1.0, MSVC
+v19.40 (VS17.10), and nvc++ 24.5.</p>
+<div class="sourceCode" id="cb39"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb39-1"><a href="#cb39-1" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;cassert&gt;</span></span>
+<span id="cb39-2"><a href="#cb39-2" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;concepts&gt;</span></span>
+<span id="cb39-3"><a href="#cb39-3" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;cstdint&gt;</span></span>
+<span id="cb39-4"><a href="#cb39-4" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;iostream&gt;</span></span>
+<span id="cb39-5"><a href="#cb39-5" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;stdexcept&gt;</span></span>
+<span id="cb39-6"><a href="#cb39-6" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;type_traits&gt;</span></span>
+<span id="cb39-7"><a href="#cb39-7" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;utility&gt;</span></span>
+<span id="cb39-8"><a href="#cb39-8" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-9"><a href="#cb39-9" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Accessor<span class="op">&gt;</span></span>
+<span id="cb39-10"><a href="#cb39-10" aria-hidden="true" tabindex="-1"></a><span class="kw">concept</span> has_detectably_invalid <span class="op">=</span> <span class="kw">requires</span><span class="op">(</span>Accessor acc<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb39-11"><a href="#cb39-11" aria-hidden="true" tabindex="-1"></a>  <span class="kw">typename</span> Accessor<span class="op">::</span>data_handle_type;</span>
+<span id="cb39-12"><a href="#cb39-12" aria-hidden="true" tabindex="-1"></a>  <span class="op">{</span> std<span class="op">::</span>as_const<span class="op">(</span>acc<span class="op">).</span>detectably_invalid<span class="op">(</span></span>
+<span id="cb39-13"><a href="#cb39-13" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>declval<span class="op">&lt;</span><span class="kw">typename</span> Accessor<span class="op">::</span>data_handle_type<span class="op">&gt;()</span>,</span>
+<span id="cb39-14"><a href="#cb39-14" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>declval<span class="op">&lt;</span>std<span class="op">::</span><span class="dt">size_t</span><span class="op">&gt;()</span></span>
+<span id="cb39-15"><a href="#cb39-15" aria-hidden="true" tabindex="-1"></a>    <span class="op">)</span> <span class="op">}</span> <span class="kw">noexcept</span> <span class="op">-&gt;</span> std<span class="op">::</span>convertible_to<span class="op">&lt;</span><span class="dt">bool</span><span class="op">&gt;</span>;</span>
+<span id="cb39-16"><a href="#cb39-16" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb39-17"><a href="#cb39-17" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-18"><a href="#cb39-18" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> Accessor<span class="op">&gt;</span></span>
+<span id="cb39-19"><a href="#cb39-19" aria-hidden="true" tabindex="-1"></a><span class="dt">bool</span> detectably_invalid<span class="op">(</span>Accessor<span class="op">&amp;&amp;</span> accessor,</span>
+<span id="cb39-20"><a href="#cb39-20" aria-hidden="true" tabindex="-1"></a>  <span class="kw">typename</span> std<span class="op">::</span>remove_cvref_t<span class="op">&lt;</span>Accessor<span class="op">&gt;::</span>data_handle_type handle,</span>
+<span id="cb39-21"><a href="#cb39-21" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span></span>
+<span id="cb39-22"><a href="#cb39-22" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb39-23"><a href="#cb39-23" aria-hidden="true" tabindex="-1"></a>  <span class="cf">if</span> <span class="kw">constexpr</span> <span class="op">(</span>has_detectably_invalid<span class="op">&lt;</span>std<span class="op">::</span>remove_cvref_t<span class="op">&lt;</span>Accessor<span class="op">&gt;&gt;)</span> <span class="op">{</span></span>
+<span id="cb39-24"><a href="#cb39-24" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> std<span class="op">::</span>as_const<span class="op">(</span>accessor<span class="op">).</span>detectably_invalid<span class="op">(</span>handle, size<span class="op">)</span>;</span>
+<span id="cb39-25"><a href="#cb39-25" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-26"><a href="#cb39-26" aria-hidden="true" tabindex="-1"></a>  <span class="cf">else</span> <span class="op">{</span></span>
+<span id="cb39-27"><a href="#cb39-27" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="kw">false</span>;</span>
+<span id="cb39-28"><a href="#cb39-28" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-29"><a href="#cb39-29" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
+<span id="cb39-30"><a href="#cb39-30" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-31"><a href="#cb39-31" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> A <span class="op">{</span></span>
+<span id="cb39-32"><a href="#cb39-32" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> <span class="dt">float</span><span class="op">*</span>;</span>
+<span id="cb39-33"><a href="#cb39-33" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-34"><a href="#cb39-34" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static</span> <span class="dt">bool</span> detectably_invalid<span class="op">(</span>data_handle_type ptr, std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span> <span class="kw">noexcept</span> <span class="op">{</span></span>
+<span id="cb39-35"><a href="#cb39-35" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> ptr <span class="op">==</span> <span class="kw">nullptr</span> <span class="op">&amp;&amp;</span> size <span class="op">!=</span> <span class="dv">0</span>;</span>
+<span id="cb39-36"><a href="#cb39-36" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-37"><a href="#cb39-37" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb39-38"><a href="#cb39-38" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-39"><a href="#cb39-39" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> B <span class="op">{</span></span>
+<span id="cb39-40"><a href="#cb39-40" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> <span class="dt">float</span><span class="op">*</span>;</span>
+<span id="cb39-41"><a href="#cb39-41" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb39-42"><a href="#cb39-42" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-43"><a href="#cb39-43" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> C <span class="op">{</span></span>
+<span id="cb39-44"><a href="#cb39-44" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> <span class="dt">float</span><span class="op">*</span>;</span>
+<span id="cb39-45"><a href="#cb39-45" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-46"><a href="#cb39-46" aria-hidden="true" tabindex="-1"></a>  <span class="co">// This is nonconst, so it&#39;s not actually called.</span></span>
+<span id="cb39-47"><a href="#cb39-47" aria-hidden="true" tabindex="-1"></a>  <span class="dt">bool</span> detectably_invalid<span class="op">(</span>data_handle_type ptr, std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb39-48"><a href="#cb39-48" aria-hidden="true" tabindex="-1"></a>    <span class="cf">throw</span> std<span class="op">::</span>runtime_error<span class="op">(</span><span class="st">&quot;C::detectably_invalid: uh oh&quot;</span><span class="op">)</span>;</span>
+<span id="cb39-49"><a href="#cb39-49" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-50"><a href="#cb39-50" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb39-51"><a href="#cb39-51" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-52"><a href="#cb39-52" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> D <span class="op">{</span></span>
+<span id="cb39-53"><a href="#cb39-53" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> <span class="dt">float</span><span class="op">*</span>;</span>
+<span id="cb39-54"><a href="#cb39-54" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-55"><a href="#cb39-55" aria-hidden="true" tabindex="-1"></a>  <span class="co">// This is const but not noexcept, so it&#39;s not actually called.</span></span>
+<span id="cb39-56"><a href="#cb39-56" aria-hidden="true" tabindex="-1"></a>  <span class="dt">bool</span> detectably_invalid<span class="op">(</span>data_handle_type ptr, std<span class="op">::</span><span class="dt">size_t</span> size<span class="op">)</span> <span class="kw">const</span> <span class="op">{</span></span>
+<span id="cb39-57"><a href="#cb39-57" aria-hidden="true" tabindex="-1"></a>    <span class="cf">throw</span> std<span class="op">::</span>runtime_error<span class="op">(</span><span class="st">&quot;D::detectably_invalid: uh oh&quot;</span><span class="op">)</span>;</span>
+<span id="cb39-58"><a href="#cb39-58" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-59"><a href="#cb39-59" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb39-60"><a href="#cb39-60" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-61"><a href="#cb39-61" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-62"><a href="#cb39-62" aria-hidden="true" tabindex="-1"></a><span class="dt">int</span> main<span class="op">()</span></span>
+<span id="cb39-63"><a href="#cb39-63" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb39-64"><a href="#cb39-64" aria-hidden="true" tabindex="-1"></a>  <span class="dt">float</span><span class="op">*</span> ptr <span class="op">=</span> <span class="kw">nullptr</span>;</span>
+<span id="cb39-65"><a href="#cb39-65" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-66"><a href="#cb39-66" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>A<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-67"><a href="#cb39-67" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>detectably_invalid<span class="op">(</span>A<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-68"><a href="#cb39-68" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-69"><a href="#cb39-69" aria-hidden="true" tabindex="-1"></a>  A a<span class="op">{}</span>;</span>
+<span id="cb39-70"><a href="#cb39-70" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>a, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-71"><a href="#cb39-71" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>detectably_invalid<span class="op">(</span>a, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-72"><a href="#cb39-72" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-73"><a href="#cb39-73" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> A a_c<span class="op">{}</span>;</span>
+<span id="cb39-74"><a href="#cb39-74" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>a_c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-75"><a href="#cb39-75" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>detectably_invalid<span class="op">(</span>a_c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-76"><a href="#cb39-76" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-77"><a href="#cb39-77" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>B<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-78"><a href="#cb39-78" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>B<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-79"><a href="#cb39-79" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-80"><a href="#cb39-80" aria-hidden="true" tabindex="-1"></a>  <span class="co">// B doesn&#39;t know how to check pointer validity.</span></span>
+<span id="cb39-81"><a href="#cb39-81" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-82"><a href="#cb39-82" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>B<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-83"><a href="#cb39-83" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>B<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-84"><a href="#cb39-84" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-85"><a href="#cb39-85" aria-hidden="true" tabindex="-1"></a>  B b<span class="op">{}</span>;</span>
+<span id="cb39-86"><a href="#cb39-86" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>b, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-87"><a href="#cb39-87" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>b, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-88"><a href="#cb39-88" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-89"><a href="#cb39-89" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> B b_c<span class="op">{}</span>;</span>
+<span id="cb39-90"><a href="#cb39-90" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>b_c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-91"><a href="#cb39-91" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>b_c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-92"><a href="#cb39-92" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-93"><a href="#cb39-93" aria-hidden="true" tabindex="-1"></a>  <span class="co">// If users make detectably_invalid nonconst or not noexcept,</span></span>
+<span id="cb39-94"><a href="#cb39-94" aria-hidden="true" tabindex="-1"></a>  <span class="co">// the nonmember function falls back to a default implementation.</span></span>
+<span id="cb39-95"><a href="#cb39-95" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-96"><a href="#cb39-96" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
+<span id="cb39-97"><a href="#cb39-97" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>C<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-98"><a href="#cb39-98" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>C<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-99"><a href="#cb39-99" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-100"><a href="#cb39-100" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb39-101"><a href="#cb39-101" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;C{} threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
+<span id="cb39-102"><a href="#cb39-102" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-103"><a href="#cb39-103" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-104"><a href="#cb39-104" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
+<span id="cb39-105"><a href="#cb39-105" aria-hidden="true" tabindex="-1"></a>    <span class="kw">const</span> C c_c<span class="op">{}</span>;</span>
+<span id="cb39-106"><a href="#cb39-106" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>c_c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-107"><a href="#cb39-107" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>c_c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-108"><a href="#cb39-108" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-109"><a href="#cb39-109" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb39-110"><a href="#cb39-110" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;const C threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
+<span id="cb39-111"><a href="#cb39-111" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-112"><a href="#cb39-112" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-113"><a href="#cb39-113" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
+<span id="cb39-114"><a href="#cb39-114" aria-hidden="true" tabindex="-1"></a>    C c<span class="op">{}</span>;</span>
+<span id="cb39-115"><a href="#cb39-115" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-116"><a href="#cb39-116" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-117"><a href="#cb39-117" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-118"><a href="#cb39-118" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb39-119"><a href="#cb39-119" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;nonconst C threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
+<span id="cb39-120"><a href="#cb39-120" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-121"><a href="#cb39-121" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-122"><a href="#cb39-122" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
+<span id="cb39-123"><a href="#cb39-123" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>D<span class="op">{}</span>, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-124"><a href="#cb39-124" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>D<span class="op">{}</span>, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-125"><a href="#cb39-125" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-126"><a href="#cb39-126" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb39-127"><a href="#cb39-127" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;D{} threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
+<span id="cb39-128"><a href="#cb39-128" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-129"><a href="#cb39-129" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-130"><a href="#cb39-130" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
+<span id="cb39-131"><a href="#cb39-131" aria-hidden="true" tabindex="-1"></a>    <span class="kw">const</span> D d_c<span class="op">{}</span>;</span>
+<span id="cb39-132"><a href="#cb39-132" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>d_c, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-133"><a href="#cb39-133" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>d_c, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-134"><a href="#cb39-134" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-135"><a href="#cb39-135" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb39-136"><a href="#cb39-136" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;const D threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
+<span id="cb39-137"><a href="#cb39-137" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-138"><a href="#cb39-138" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-139"><a href="#cb39-139" aria-hidden="true" tabindex="-1"></a>  <span class="cf">try</span> <span class="op">{</span></span>
+<span id="cb39-140"><a href="#cb39-140" aria-hidden="true" tabindex="-1"></a>    D d<span class="op">{}</span>;</span>
+<span id="cb39-141"><a href="#cb39-141" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>d, ptr, <span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb39-142"><a href="#cb39-142" aria-hidden="true" tabindex="-1"></a>    <span class="ot">assert</span><span class="op">(</span><span class="kw">not</span> detectably_invalid<span class="op">(</span>d, ptr, <span class="dv">1</span><span class="op">))</span>;</span>
+<span id="cb39-143"><a href="#cb39-143" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-144"><a href="#cb39-144" aria-hidden="true" tabindex="-1"></a>  <span class="cf">catch</span> <span class="op">(</span><span class="kw">const</span> std<span class="op">::</span>runtime_error<span class="op">&amp;</span> e<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb39-145"><a href="#cb39-145" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;nonconst D threw runtime_error: &quot;</span> <span class="op">&lt;&lt;</span> e<span class="op">.</span>what<span class="op">()</span> <span class="op">&lt;&lt;</span> <span class="st">&quot;</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
+<span id="cb39-146"><a href="#cb39-146" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb39-147"><a href="#cb39-147" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb39-148"><a href="#cb39-148" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;Made it to the end</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
+<span id="cb39-149"><a href="#cb39-149" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> <span class="dv">0</span>;</span>
+<span id="cb39-150"><a href="#cb39-150" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
 </div>
 </div>
 </body>
diff --git a/aligned_accessor/aligned_accessor.md b/aligned_accessor/aligned_accessor.md
index 4b03b57b..10fbbbb7 100644
--- a/aligned_accessor/aligned_accessor.md
+++ b/aligned_accessor/aligned_accessor.md
@@ -68,14 +68,23 @@ toc: true
     
     * Discuss optional suggestions from LEWG review of R1 on 2024-06-28
 
-        * Add `is_checkably_valid` discussion
+        * Add `explicit` converting constructor vs.
+            named cast ("`naughty_cast`") discussion
 
-        * Add `naughty_cast` discussion
+        * Add `detectably_invalid` discussion
+
+        * Ask LEWG to consider the alternative design that makes
+            `is_sufficiently_aligned` a nonmember function in `<bit>`
+            instead of a member function of `aligned_accessor`,
+            while LWG review of R2 proceeds concurrently
+
+    * P2389R2 was voted into the Working Draft at St. Louis,
+        so replace use of `dextents` in examples with `dims`.
 
 # Purpose of this paper
 
 We propose adding `aligned_accessor` to the C++ Standard Library.
-This class template is an mdspan accessor policy
+This class template is an `mdspan` accessor policy
 that uses `assume_aligned` to decorate pointer access.
 We think it belongs in the Standard Library for two reasons.
 First, it would serve as a common vocabulary type
@@ -104,11 +113,16 @@ reach their maximum value.
 
 * `data_handle_type` is `ElementType*`
 
-* Constructor permits (implicit) conversion
+* Permitted implicit conversions
+
+    * from nonconst to const `ElementType`,
 
-    * from nonconst to const `ElementType`, and
+    * from more overalignment to less overalignment, and
 
-    * from more overalignment to less overalignment
+    * from overalignment to no overalignment (`default_accessor`)
+
+* `explicit` converting constructor from `default_accessor`
+    lets users assert overalignment
 
 * `is_sufficiently_aligned` checks pointer alignment
 
@@ -127,11 +141,12 @@ However, these attributes should not apply
 to the result of `offset`, for the same reason that
 `offset_policy` is `default_accessor` and not `aligned_accessor`.
 
-The constructor is analogous to `default_accessor`'s constructor,
+The converting constructor from `aligned_accessor`
+is analogous to `default_accessor`'s constructor,
 in that it exists to permit conversion
 from nonconst `element_type` to const `element_type`.
-The constructor of `aligned_accessor` additionally permits
-conversion from more overalignment to less overalignment --
+It additionally permits implicit conversion
+from more overalignment to less overalignment --
 something that we expect users may need to do.
 For example, users may start with `aligned_accessor<float, 128>`,
 because their allocation function promises 128-byte alignment.
@@ -145,11 +160,19 @@ This makes it easier for users to check preconditions,
 without needing to know how to cast a pointer to an integer
 of the correct size and signedness.
 
+Per discussion below,
+we would like LEWG to consider the alternative design that
+removes `is_sufficiently_aligned` from `aligned_accessor` and
+adds it to the `<bit>` header as a separate nonmember function.
+We think LWG review of `aligned_accessor` can proceed concurrently.
+We present wording for this design alternative below.
+
 # Design discussion
 
 ## The accessor is not nestable
 
-We considered making `aligned_accessor` "wrap" any accessor type meeting the right requirements.
+We considered making `aligned_accessor` "wrap" any accessor type
+that meets the right requirements.
 For example, `aligned_accessor` could take the inner accessor as a template parameter, store an instance of it, and dispatch to its member functions.
 That would give users a way to apply multiple accessor "attributes" to their data handle, such as atomic access (see P2689) and overalignment.
 
@@ -158,10 +181,13 @@ First, we would have no way to validate that the user's accessor type has the co
 We could check that their accessor's `data_handle_type` is a pointer type,
 but we could not check that their accessor's `access` function
 actually dereferences the pointer.
-For instance, `access` might instead interpret the data handle as a file handle or a key into a distributed data store.
+For instance, `access` might instead interpret the pointer as a file handle or a key into a distributed data store.
 
 Second, even if the inner accessor's `access` function actually did return the result of dereferencing the pointer, the outer `access` function might not be able to recover the effects of the inner `access` function, because `access` computes a `reference`, not a pointer.
-For example, our `aligned_accessor`'s access function can only apply `assume_aligned` to the pointer; knowledge of overalignment at compile time "lost" with the dereference at element `i`.
+In order for `aligned_accessor`'s `access` function
+to get back that pointer, it would need to reach past
+the inner accessor's public interface.
+That would defeat the purpose of generic nesting.
 
 Third, any way (not just this one) of nesting two generic accessors raises the question of order dependence.  Even if it were possible to apply the effects of both the inner and outer accessors' `access` functions in sequence, it might be unpleasantly surprising to users if the effects depended on the order of nesting.
 A similar question came up in the "properties" proposal P0900, which we quote here.
@@ -181,26 +207,30 @@ Making it `explicit` would prevent implicit conversion.
 Thus, we have decided to add this `explicit` constructor in R1.
 
 Without the `explicit` constructor,
-users must do this to convert from nonaligned to aligned `mdspan`.
+users have two options for turning a nonaligned `mdspan`
+into an aligned `mdspan`.  First, as in the following example,
+users could "take apart" the input nonaligned `mdspan`
+and use the pieces to construct an aligned `mdspan`,
+whose type they name completely.
 ```c++
 void compute_with_aligned(
-  std::mdspan<float, std::dextent<int, 2>, std::layout_left> matrix)
+  std::mdspan<float, std::dims<2>, std::layout_left> matrix)
 {
   const std::size_t byte_alignment = 4 * alignof(float);
-  using aligned_matrix_t = std::mdspan<float, std::dextent<int, 2>,
+  using aligned_matrix_t = std::mdspan<float, std::dims<2>,
     std::layout_left, std::aligned_accessor<float, byte_alignment>>;
 
   aligned_matrix_t aligned_matrix{matrix.data_handle(), matrix.mapping()};
   // ... use aligned_matrix ...
 }
 ```
-
-Another option would be to use
-constructor template argument deduction (CTAD),
-as in the following.
+Second, as in the following example,
+users could construct an `aligned_accessor` explicitly
+and use constructor template argument deduction (CTAD)
+to construct the aligned `mdspan` from its pieces.
 ```c++
 void compute_with_aligned(
-  std::mdspan<float, std::dextent<int, 2>, std::layout_left> matrix)
+  std::mdspan<float, std::dims<2>, std::layout_left> matrix)
 {
   const std::size_t byte_alignment = 4 * alignof(float);
 
@@ -209,19 +239,20 @@ void compute_with_aligned(
   // ... use aligned_matrix ...
 }
 ```
-However, `mdspan` users commonly define their own type aliases for `mdspan`,
+The first approach would likely be more common.
+This is because `mdspan` users commonly define
+their own type aliases for `mdspan`,
 with application-specific names that make code more self-documenting.
 The `aligned_matrix_t` definition above is an an example.
-Such users would not normally rely on CTAD
-for conversion from a less specific to a more specific `mdspan`.
 
 Adding an `explicit` constructor from `default_accessor`
-lets users get the same effect more concisely.
+lets users get the same effect more concisely,
+without needing to "take apart" the input `mdspan`.
 ```c++
-void compute_with_aligned(std::mdspan<float, std::dextent<int, 2>, std::layout_left> matrix)
+void compute_with_aligned(std::mdspan<float, std::dims<2, int>, std::layout_left> matrix)
 {
   const std::size_t byte_alignment = 4 * alignof(float);
-  using aligned_mdspan = std::mdspan<float, std::dextent<int, 2>,
+  using aligned_mdspan = std::mdspan<float, std::dims<2, int>,
     std::layout_left, std::aligned_accessor<float, byte_alignment>>;
 
   aligned_mdspan aligned_matrix{matrix};
@@ -251,16 +282,16 @@ We do not _object_ to such an alias, but we do not find it very useful,
 for the following reasons.
 
 1. Users of `mdspan` commonly define their own type aliases
-    whose names have meaningful names for their applications.
+    whose names are meaningful for their applications.
 
 2. It would not save much typing.
 
 Examples may define aliases to make them more concise.
-One of them in this proposal defines the following alias
+One example in this proposal defines the following alias
 for an `mdspan` of `float` with alignment `byte_alignment`.
 ```c++
 template<size_t byte_alignment>
-using aligned_mdspan = std::mdspan<float, std::dextents<int, 1>,
+using aligned_mdspan = std::mdspan<float, std::dims<1, int>,
   std::layout_right, std::aligned_accessor<float, byte_alignment>>;
 ```
 This lets the example use `aligned_mdspan<32>` and `aligned_mdspan<16>`.
@@ -268,7 +299,8 @@ This lets the example use `aligned_mdspan<32>` and `aligned_mdspan<16>`.
 The above alias is specific to a particular example.
 A _general_ version of alias would look like this.
 ```c++
-template<class ElementType, class Extents, class Layout, size_t byte_alignment>
+template<class ElementType, class Extents, class Layout,
+  size_t byte_alignment>
 using aligned_mdspan = std::mdspan<ElementType, Extents, Layout,
   std::aligned_accessor<ElementType, byte_alignment>>;
 ```
@@ -282,18 +314,18 @@ For instance, users might define the following.
 ```c++
 template<class ElementType>
 using vector_t = std::mdspan<ElementType,
-  std::dextents<int, 1>, std::layout_left>;
+  std::dims<1>, std::layout_left>;
 template<class ElementType>
 using matrix_t = std::mdspan<ElementType,
-  std::dextents<int, 2>, std::layout_left>;
+  std::dims<2>, std::layout_left>;
 
 template<class ElementType, size_t byte_alignment>
 using aligned_vector_t = std::mdspan<ElementType,
-  std::dextents<int, 1>, std::layout_left, 
+  std::dims<1>, std::layout_left, 
   std::aligned_accessor<ElementType, byte_alignment>>;
 template<class ElementType, size_t byte_alignment>
 using aligned_matrix_t = std::mdspan<ElementType,
-  std::dextents<int, 2>, std::layout_left, 
+  std::dims<2>, std::layout_left, 
   std::aligned_accessor<ElementType, byte_alignment>>;
 ```
 Such users may never type the characters "`mdspan`" again.
@@ -313,7 +345,7 @@ whether its accessor finds the caller's data handle acceptable.
 
 This is true for any accessor type, not just for `aligned_accessor`.
 It is a design feature of `mdspan` that accessors can be stateless.
-Even if they have state,
+Most of them have no state.  Even if they have state,
 they do not need to be constructed with or store the data handle.
 As a result, an accessor might not see a data handle
 until `access` or `offset` is called.
@@ -325,15 +357,17 @@ Using exceptions in the manner of `vector::at`
 could reduce performance and would also make `mdspan` unusable
 in a freestanding or no-exceptions context.
 
-Note that `aligned_accessor` does not introduce any _additional_ preconditions
+Note that `aligned_accessor` does not introduce _additional_ preconditions
 beyond those of the existing C++ Standard Library feature `assume_aligned`.
-In the words of one LEWG reviewer, `aligned_accessor` is not any more "pointy"
-than `assume_aligned`; it just passes the point through without "blunting" it.
+In the words of one LEWG reviewer,
+`aligned_accessor` is not any more "pointy" than `assume_aligned`;
+it just passes the point through without "blunting" it.
 
 Before submitting R0 of this paper,
 we considered an approach specific to `aligned_accessor`,
 that would wrap the pointer in a special data handle type.
-The data handle type would have an `explicit` constructor taking a raw pointer,
+The data handle type would have an `explicit` constructor
+that takes a raw pointer,
 with a precondition that the raw pointer have sufficient alignment.
 The constructor would be `explicit`, because it would have a precondition.
 This design would force the precondition back to `mdspan` construction time.
@@ -347,7 +381,8 @@ mdspan x{acc_type::data_handle_type{raw_pointer}, mapping, acc_type{}};
 
 We rejected this approach in favor of `is_sufficiently_aligned`
 for the following reasons.
-First, wrapping the pointer in a custom data handle class
+
+1. Wrapping the pointer in a custom data handle class
 would make every `access` or `offset` call
 need to reach through the data handle's interface,
 instead of just taking the raw pointer directly.
@@ -360,11 +395,13 @@ that may hinder compilers from optimizing.
 Performance of `aligned_accessor` matters as much or even more than
 performance of `default_accessor`, because `aligned_accessor` exists
 to communicate optimization potential.
-Second, the alignment precondition would still exist.
+
+2. The alignment precondition would still exist.
 Requiring the data handle type to throw an exception
 if the pointer is not sufficiently aligned
 would make `mdspan` unusable in a freestanding or no-exceptions context.
-Third, users should not have to pay for unneeded checks.
+
+3. Users should not have to pay for unneeded checks.
 The two examples in the wording express the two most common cases.
 If users get a pointer from a function like `aligned_alloc`,
 then they already know its alignment, because they asked for it.
@@ -374,6 +411,12 @@ then they know alignment before dispatch.
 In both cases, users already know the alignment
 before constructing the `mdspan`.
 
+4. The data handle is still a pointer,
+it's just a pointer with a constraint on its values.
+Users would reasonably expect to be able to use
+the result of `data_handle()` with existing interfaces
+that expect a raw pointer.
+
 An LEWG poll on 2023-10-10,
 "[b]lock `aligned_accessor` progressing until we have
 a way of checking alignment requirements during `mdspan` construction,"
@@ -400,10 +443,9 @@ LEWG expressed an (unpolled) interest that we explore `mdspan` safety
 in subsequent work after the fall 2023 Kona WG21 meeting.
 LEWG asked us to explore safety
 in a way that is not specific to `aligned_accessor`.
-We consider the section below
-"Generalize `is_sufficiently_aligned` for all accessors?",
-that discusses adding a member function `is_checkably_valid`
-to the accessor requirements, part of that exploration.
+Part of that exploration is in the section below
+"Generalize `is_sufficiently_aligned` for all accessors?".
+We plan further exploration of this topic elsewhere.
 
 ## `is_sufficiently_aligned` is not `constexpr`
 
@@ -443,10 +485,11 @@ More importantly for `mdspan` users,
 Standard C++ offers no way to check whether a pointer
 and a layout mapping's `required_span_size()` form a valid range.
 
-### `is_checkably_valid`: Generic validity check?
+### `detectably_invalid`: Generic validity check?
 
 During the June 2024 St. Louis WG21 meeting,
-one LEWG reviewer pointed out that code that is generic on the accessor type
+one LEWG reviewer (please see Acknowledgments below)
+pointed out that code that is generic on the accessor type
 currently has no way to check whether a given data handle is valid.
 Specifically, given a `size_t` `size`
 (e.g., the `required_span_size()` of a given layout mapping),
@@ -455,15 +498,15 @@ there is no way to check whether $[$ 0, `size` $)$ forms an accessible range
 of a given data handle and accessor.
 The reviewer suggested adding a new member function
 ```c++
-bool is_checkably_valid(data_handle_type handle, size_t size) const noexcept;
+bool detectably_invalid(data_handle_type handle, size_t size) const noexcept;
 ```
-to all `mdspan` accessors.  This would return `false`
+to all `mdspan` accessors.  This would return `true`
 if the implementation can show that $[$ 0, `size` $)$
 is *not* an accessible range for `handle` and the accessor,
 and `true` otherwise.
-The word "checkably" in the name would remind users
+The word "detectably" in the name would remind users
 that this is a "best effort" check.
-It might return `true` even if the handle is invalid
+It might return `false` even if the handle is invalid
 or if $[$ 0, `size` $)$ is not an accessible range.
 Also, it might return different values on different implementations,
 depending on their ability to check e.g., pointer range validity.
@@ -488,25 +531,28 @@ auto create_mdspan_with_check(
   LayoutMapping mapping,
   Accessor accessor)
 {
-  if (not accessor.is_checkably_valid(handle, mapping.required_span_size())) {
+  if (accessor.detectably_invalid(handle, mapping.required_span_size())) {
     throw std::out_of_range("Invalid data handle and/or size");
   }
   return mdspan{handle, mapping, accessor};
 }
 ```
 
-### Arguments against and for `is_checkably_valid`
+### Arguments against and for `detectably_invalid`
 
 We didn't include this feature in the original `mdspan` design
-because most data handle types have no way to check validity.  
+because most data handle types have no way to say with full accuracy
+whether a handle and size are valid.
 We didn't want to give users the false impression 
 that a validity check was doing anything meaningful.
 Standard C++ has no way to check a raw pointer `T*` and a size,
 though some implementations such as CHERI C++
-([Davis 2019] and [Watson 2020]) do have this feature.
-We designed `mdspan` accessors to be able to wrap libraries that implement
-a partitioned global address space (PGAS) programming model
-for accessing remote data over a network.
+([Davis 2019] and [Watson 2020])
+and run-time profiling and debugging systems such as Valgrind
+do have this feature.
+We designed `mdspan` accessors to be able to wrap libraries
+that implement a partitioned global address space (PGAS)
+programming model for accessing remote data over a network.
 (See <a href="https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2022/p0009r18.html#why-custom-accessors">P0009R18, Section 2.7, "Why custom accessors?".</a>)
 Such libraries include the one-sided communication interface in MPI
 (the Message Passing Interface for distributed-memory parallel programming)
@@ -518,11 +564,11 @@ that points to an allocation from the "symmetric heap"
 (that is accessible to all participating parallel processes).
 Such libraries generally do not have validity checks for their handles.
 
-On the other hand, an `is_checkably_valid` function
+On the other hand, a `detectably_invalid` function
 would let happen any checks that _could_ happen.
 For instance, a hypothetical "GPU device memory accessor"
-(not proposed for the C++ Standard,
-but existing in projects like RAPIDS RAFT)
+(not proposed for the C++ Standard, but existing in projects like
+<a href="https://github.com/rapidsai/raft">RAPIDS RAFT</a>)
 might permit access to an allocation of GPU "device" memory
 from only GPU "device" code, not from ordinary "host" code.
 A common use case for GPU allocations
@@ -533,69 +579,186 @@ in host code with that accessor.
 The accessor could use a CUDA run-time function like
 <a href="https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__UNIFIED.html">`cudaPointerGetAttributes`</a>
 to check if the pointer points to valid GPU memory.
+Even `default_accessor` could have a simple check like this.
+
+```c++
+bool detectably_invalid(data_handle_type ptr, size_t size) const {
+  return ptr == nullptr && size != 0;
+}
+```
 
 ### Nonmember customization point design
 
 C++23 defines the generic interface of accessors
-through the "accessor policy requirements" in
-<a href="https://eel.is/c++draft/mdspan.accessor.reqmts">[mdspan.accessor.reqmts]</a>.
-Adding `is_checkably_valid` to these requirements would be a breaking change.
+through the accessor policy requirements
+<a href="https://eel.is/c++draft/mdspan.accessor.reqmts">**[mdspan.accessor.reqmts]**</a>.
+Adding `detectably_invalid` to these requirements
+would be a breaking change to C++23.
 Thus, generic code that wanted to call this function
 would need to fill in default behavior
-for existing C++23 or user-defined accessors.
-The following `is_checkably_valid` nonmember function
-shows one way users could do that.
-For the full source code and tests, please see this
-<a href="https://godbolt.org/z/14YPfMoq3">Compiler Explorer link</a>
-or Appendix A below.
-
-The point of this function is to show that
-even if C++26 adds `is_checkably_valid` to the accessor requirements
-and to all the Standard accessors,
-users could still work around the lack of this function
+for both Standard accessors defined in C++23, and
+user-defined accessors that comply with the C++23 accessor requirements.
+The following `detectably_invalid` nonmember function
+(not proposed in this paper) shows one way users could do that.
+Please see Appendix A below for the full source code
+of a demonstration, along with a Compiler Explorer link.
+This demonstration shows that breaking backwards compatibility
+with C++23 is unnecessary, because users can straightforwardly
+work around the lack of a `detectably_invalid` member function
 in C++23 - compliant accessors.
-While it could make sense to standardize this function,
-users might like to fill in different default behavior,
-so we do not propose this here.
+Not standardizing this nonmember function work-around
+would also give users the freedom to fill in different default behavior.
+For example, some users may prefer to consider
+every (data handle, size) pair invalid unless proven otherwise,
+as a way to force use of custom accessors that have the ability
+to make accurate checks.
 
 ```c++
 template<class Accessor>
-concept has_is_checkably_valid = requires(Accessor acc) {
-  // Making the existence of the type alias a separate constraint
-  // gives more user-friendly error messages.
+concept has_detectably_invalid = requires(Accessor acc) {
   typename Accessor::data_handle_type;
-  { std::as_const(acc).is_checkably_valid(
+  { std::as_const(acc).detectably_invalid(
       std::declval<typename Accessor::data_handle_type>(),
       std::declval<std::size_t>()
-    ) } noexcept -> std::convertible_to<bool>;
+    ) } noexcept -> std::same_as<bool>;
 };
 
 template<class Accessor>
-bool is_checkably_valid(Accessor&& accessor,
+bool detectably_invalid(Accessor&& accessor,
   typename std::remove_cvref_t<Accessor>::data_handle_type handle,
   std::size_t size)
 {
-  if constexpr (has_is_checkably_valid<std::remove_cvref_t<Accessor>>) {
-    return std::as_const(accessor).is_checkably_valid(handle, size);
+  if constexpr (has_detectably_invalid<std::remove_cvref_t<Accessor>>) {
+    return std::as_const(accessor).detectably_invalid(handle, size);
   }
   else {
-    return true;
+    return false;
   }
 }
 ```
 
+### We need both `is_sufficiently_aligned` and `detectably_invalid`
+
+The same LEWG reviewer who suggested `detectably_invalid`
+also said that it would make `is_sufficiently_aligned` unnecessary.
+We disagree for the following reasons.
+
+1. Users will often want to check alignment separately
+    from pointer range validity.
+
+2. Checking alignment may be much less expensive
+    than checking pointer range validity.
+
+Regarding (1), we think the most common use case for
+`aligned_accessor`'s `explicit` converting constructor
+from `default_accessor` would be explicit construction
+of an `mdspan` with `aligned_accessor`
+from an `mdspan` with `default_accessor`.  The latter exists,
+so the user has already asserted that the range formed by
+its data handle and `required_span_size()` is valid.
+Thus, the only thing the user would need to check
+would be whether the data handle is sufficiently aligned.
+
+### Alternative: Nonmember `is_sufficiently_aligned`
+
+The reviewer responded to our argument above by suggesting
+that we remove `is_sufficiently_aligned` from `aligned_accessor`
+and make it a separate nonmember function.
+This is a reasonable alternative suggestion.
+We think that LWG review of `aligned_accessor` can proceed concurrently
+while LEWG considers this alternative.
+
+Into which header should this new function go?
+We think it belongs in `<bit>`,
+because it is fundamentally a bit arithmetic operation.
+There should be no reason why that could not be freestanding,
+like everything else in the `<bit>` header.
+Another option would be to put it in `<memory>`
+right after `assume_aligned`, and to mark it
+as freestanding (since `assume_aligned` is).
+
+### Do accessors need to check anything else?
+
+The only other thing an accessor's user might want to check
+besides a (data handle, size) pair
+would be converting construction from another type of accessor.
+All `mdspan` components -- `extents`, layout mappings, and accessors --
+implement conversions with preconditions via `explicit` constructors.
+(For more detail, please see the section below,
+"Explicit conversions as the model for precondition-asserting conversions.")
+Accessors do *not* store their data handles,
+so the only reason to check whether converting construction is valid
+would be if the input or result accessor has separate run-time state.
+(Otherwise, the check could be a constraint or `static_assert`.)
+It's rare for an accessor to need run-time state,
+so we don't expect to need this feature in generic code.
+It would also be a separable addition from
+the feature of checking a data handle and size.
+Nevertheless, one could consider a design.
+We would favor just overloading `detectably_invalid` for accessors,
+as there would be no risk of ambiguity.
+Converting constructors only take one argument,
+so there would be no ambiguity between calling `detectably_invalid`
+with an accessor and calling it with a data handle and size.
+
+### Naming the function
+
+1. The function describes a property:
+    "this (data handle, size) pair is not known to be invalid."
+    It's an adjective (like "`valid`" or "`is_valid`"),
+    not a verb (like "check" as in "`check_valid`").
+
+2. The function does not promise perfect accuracy.
+    In the common case, it says whether it can _detect_
+    whether the handle and size are _in_valid.
+    Whether or not they are _valid_ might be harder to say.
+
+3. As discussed above, users may also want to check
+    converting constructors from other accessor types.
+    However, there would be no risk of ambiguity between that
+    and checking a data handle and size.
+    Therefore, there's no need for the function's name to include
+    the type of the thing being checked (e.g., "range").
+
+4. Specifically, the function should not contain the word "pointer,"
+    because a data handle is not necessarily a pointer.
+    Even if `data_handle_type`​ is a pointer type,
+    a data handle might not necessarily be a pointer
+    to the elements in the Standard C++ sense.
+    For example, it might be some opaque handle
+    that a library represents as a type alias of `void*`​.
+
+These points together suggest the name `detectably_invalid`.
+
 ### Conclusions
 
-1. Adding `is_checkably_valid` to the accessor requirements and existing Standard accessors in C++26 would be a breaking change.  Nevertheless, users could fill in reasonable behavior for C++23 accessors.
+1. Adding `detectably_invalid` to the accessor requirements
+    and existing Standard accessors in C++26
+    would be a breaking change to C++23.
+    Nevertheless, users could write code
+    that fills in reasonable behavior for C++23 accessors.
+
+2. Few C++ implementations offer a way to check validity of a pointer range.
+    Thus, users would experience `detectably_invalid` as mostly not useful
+    for the common case of `default_accessor` and other accessors
+    that access a pointer range.
 
-2. Few C++ implementations offer a way to check validity of a pointer range.  Thus, users would experience `is_checkably_valid` as mostly not useful for the common case of `default_accessor` and other accessors that access a pointer range.
+3. Item (1) reduces the urgency of this suggestion for C++26.
+    Item (2) reduces its potential to improve the `mdspan` user experience
+    in a practical way.  Therefore, we do not suggest adding
+    `detectably_invalid` to the accessor requirements in this proposal.
+    However, we do not discourage further work in separate proposals.
 
-3. Item (1) reduces the urgency of this suggestion for C++26.  Item (2) reduces its potential to improve the `mdspan` user experience in a practical way.  Therefore, we do not suggest adding `is_checkably_valid` to the accessor requirements in this proposal.  However, we do not discourage further work in separate proposals.
+4. We would like LEWG to consider the alternative design that
+    removes `is_sufficiently_aligned` from `aligned_accessor` and
+    adds it to the C++ Standard Library as a separate nonmember function.
+    We think LWG review of `aligned_accessor` can proceed concurrently.
 
 ## Explicit conversions as the model for precondition-asserting conversions
 
 During the June 2024 St. Louis WG21 meeting,
-one LEWG reviewer asked about the `explicit` constructor from `default_accessor`.
+one LEWG reviewer asked about the `explicit` constructor
+from `default_accessor`.
 This constructor lets users assert that a pointer
 has sufficient alignment to be accessed by the `aligned_accessor`.
 The reviewer argued that this was an "unsafe" conversion,
@@ -721,12 +884,6 @@ auto result = use_overaligned_view(naughty_cast<
 );
 ```
 
-Requiring `naughty_cast` for `mdspan` conversions with preconditions
-is a bit like requiring `static_cast` for narrowing arithmetic conversions.
-The advantage is that one can search a large code base for `static_cast`.
-The disadvantage is that the code base will become full of `static_cast`,
-thus making potentially incorrect conversions less searchable anyway.
-
 Another argument for `naughty_cast` besides searchability
 is to make conversions with preconditions "loud," that is,
 easily seen in the code by human developers.
@@ -831,7 +988,7 @@ We have tested an implementation of this proposal with the
 ```c++
 template<size_t byte_alignment>
 using aligned_mdspan =
-  std::mdspan<float, std::dextents<int, 1>, std::layout_right, std::aligned_accessor<float, byte_alignment>>;
+  std::mdspan<float, std::dims<1, int>, std::layout_right, std::aligned_accessor<float, byte_alignment>>;
 
 // Interfaces that require 32-byte alignment,
 // because they want to do 8-wide SIMD of float.
@@ -900,6 +1057,10 @@ June 2020.
 Available online [last accessed 2024-07-05]:
 <a href="https://doi.org/10.48456/tr-947">https://doi.org/10.48456/tr-947</a>
 
+# Acknowledgments
+
+* For `detectably_invalid`, credit (with permission) to David Sankel (Adobe), `dsankel@adobe.com`
+
 # Wording
 
 > Text in blockquotes is not proposed wording, but rather instructions for generating proposed wording.
@@ -1036,14 +1197,14 @@ Otherwise, `compute` dispatches to a possibly less optimized function
 ```c++
 extern void
 compute_using_fourfold_overalignment(
-  std::mdspan<float, std::dextents<size_t, 1>, std::layout_right,
+  std::mdspan<float, std::dims<1>, std::layout_right,
     std::aligned_accessor<float, 4 * alignof(float)>> x);
 
 extern void
 compute_without_requiring_overalignment(
-  std::mdspan<float, std::dextents<size_t, 1>> x);
+  std::mdspan<float, std::dims<1>> x);
 
-void compute(std::mdspan<float, std::dextents<size_t, 1>> x)
+void compute(std::mdspan<float, std::dims<1>> x)
 {
   auto accessor = aligned_accessor<float, 4 * sizeof(float)>{};
   auto x_handle = x.data_handle();
@@ -1089,7 +1250,7 @@ requiring arrays of `float` to have 32-byte alignment.
 
 ```c++
 template<size_t byte_alignment>
-using aligned_mdspan = std::mdspan<float, std::dextents<int, 1>,
+using aligned_mdspan = std::mdspan<float, std::dims<1, int>,
   std::layout_right, std::aligned_accessor<float, byte_alignment>>;
 
 extern void vectorized_axpy(aligned_mdspan<32> y, float alpha, aligned_mdspan<32> x);
@@ -1119,11 +1280,51 @@ float user_function(size_t num_elements, float alpha)
 ```
 --*end example*]
 
-# Appendix A: `is_checkably_valid` nonmember function example
+# Wording for alternative nonmember `is_sufficiently_aligned` design
+
+> Text in blockquotes is not proposed wording, but rather instructions for generating proposed wording.
+> The instructions here are expressed as a "diff" atop P2897R2,
+> that is, as if the wording proposed in P2897R2
+> had already been applied to the Working Draft.
+> The � character is used to denote a placeholder section number which the editor shall determine.
+
+> From the `aligned_accessor` synopsis in **[mdspan.accessor.aligned.overview]**,
+> remove the member function `static bool is_sufficiently_aligned(data_handle_type p);`.
+
+> From the list of members of `aligned_accessor` in **[mdspan.accessor.aligned.members]**,
+> remove `static bool is_sufficiently_aligned(data_handle_type p);`
+> from after line 7 ("*Effects*: Equivalent to: `return p + i;`"), and
+> remove line 8 ("*Preconditions*: ...") and line 9 ("*Returns*: ...").
+> (Retain the *Example* using `is_sufficiently_aligned`
+> that immediately follows the deleted lines.)
+
+> To the Header `<bit>` synopsis **[bit.syn]**, after `enum class endian` and before the `}` that closes `namespace std`, add the following.
+
+```c++
+// [bit.aligned], is_sufficiently_aligned
+template<class ElementType, size_t byte_alignment>
+  bool is_sufficiently_aligned(ElementType*);
+```
+
+> After **[bit.endian]**, add a new section **[bit.aligned]**
+> and put in it all the material that follows.
+
+```c++
+template<class ElementType, size_t byte_alignment>
+bool is_sufficiently_aligned(ElementType* p);
+```
+
+[1]{.pnum} *Preconditions*: `p` points to an object `X` of a type similar (**[conv.qual]**) to `ElementType`.
+
+[2]{.pnum} *Returns*: `true` if `X` has alignment at least `byte_alignment`, else `false`.
+
+# Appendix A: `detectably_invalid` nonmember function example
 
 This section is nonnormative.  This is the full source code with tests
-for the `is_checkably_valid` nonmember function example above.  Please see
-this <a href="https://godbolt.org/z/hKh3vK11f">Compiler Explorer link</a>.
+for the `detectably_invalid` nonmember function example above.  Please see
+this <a href="https://godbolt.org/z/4P9MGbhdj">Compiler Explorer link</a>
+for a test with five different compilers:
+GCC 14.1, Clang 18.1.0, MSVC v19.40 (VS17.10), and nvc++ 24.5.
 
 ```c++
 #include <cassert>
@@ -1135,32 +1336,32 @@ this <a href="https://godbolt.org/z/hKh3vK11f">Compiler Explorer link</a>.
 #include <utility>
 
 template<class Accessor>
-concept has_is_checkably_valid = requires(Accessor acc) {
+concept has_detectably_invalid = requires(Accessor acc) {
   typename Accessor::data_handle_type;
-  { std::as_const(acc).is_checkably_valid(
+  { std::as_const(acc).detectably_invalid(
       std::declval<typename Accessor::data_handle_type>(),
       std::declval<std::size_t>()
     ) } noexcept -> std::convertible_to<bool>;
 };
 
 template<class Accessor>
-bool is_checkably_valid(Accessor&& accessor,
+bool detectably_invalid(Accessor&& accessor,
   typename std::remove_cvref_t<Accessor>::data_handle_type handle,
   std::size_t size)
 {
-  if constexpr (has_is_checkably_valid<std::remove_cvref_t<Accessor>>) {
-    return std::as_const(accessor).is_checkably_valid(handle, size);
+  if constexpr (has_detectably_invalid<std::remove_cvref_t<Accessor>>) {
+    return std::as_const(accessor).detectably_invalid(handle, size);
   }
   else {
-    return true;
+    return false;
   }
 }
 
 struct A {
   using data_handle_type = float*;
 
-  static bool is_checkably_valid(data_handle_type ptr, std::size_t size) noexcept {
-    return ptr != nullptr || size == 0;
+  static bool detectably_invalid(data_handle_type ptr, std::size_t size) noexcept {
+    return ptr == nullptr && size != 0;
   }
 };
 
@@ -1172,8 +1373,8 @@ struct C {
   using data_handle_type = float*;
 
   // This is nonconst, so it's not actually called.
-  bool is_checkably_valid(data_handle_type ptr, std::size_t size) {
-    throw std::runtime_error("C::is_checkably_valid: uh oh");
+  bool detectably_invalid(data_handle_type ptr, std::size_t size) {
+    throw std::runtime_error("C::detectably_invalid: uh oh");
   }
 };
 
@@ -1181,8 +1382,8 @@ struct D {
   using data_handle_type = float*;
 
   // This is const but not noexcept, so it's not actually called.
-  bool is_checkably_valid(data_handle_type ptr, std::size_t size) const {
-    throw std::runtime_error("D::is_checkably_valid: uh oh");
+  bool detectably_invalid(data_handle_type ptr, std::size_t size) const {
+    throw std::runtime_error("D::detectably_invalid: uh oh");
   }
 };
 
@@ -1191,39 +1392,39 @@ int main()
 {
   float* ptr = nullptr;
 
-  assert(is_checkably_valid(A{}, ptr, 0));
-  assert(not is_checkably_valid(A{}, ptr, 1));
+  assert(not detectably_invalid(A{}, ptr, 0));
+  assert(detectably_invalid(A{}, ptr, 1));
 
   A a{};
-  assert(is_checkably_valid(a, ptr, 0));
-  assert(not is_checkably_valid(a, ptr, 1));
+  assert(not detectably_invalid(a, ptr, 0));
+  assert(detectably_invalid(a, ptr, 1));
 
   const A a_c{};
-  assert(is_checkably_valid(a_c, ptr, 0));
-  assert(not is_checkably_valid(a_c, ptr, 1));
+  assert(not detectably_invalid(a_c, ptr, 0));
+  assert(detectably_invalid(a_c, ptr, 1));
 
-  assert(is_checkably_valid(B{}, ptr, 0));
-  assert(is_checkably_valid(B{}, ptr, 1));
+  assert(not detectably_invalid(B{}, ptr, 0));
+  assert(not detectably_invalid(B{}, ptr, 1));
 
   // B doesn't know how to check pointer validity.
 
-  assert(is_checkably_valid(B{}, ptr, 0));
-  assert(is_checkably_valid(B{}, ptr, 1));
+  assert(not detectably_invalid(B{}, ptr, 0));
+  assert(not detectably_invalid(B{}, ptr, 1));
 
   B b{};
-  assert(is_checkably_valid(b, ptr, 0));
-  assert(is_checkably_valid(b, ptr, 1));
+  assert(not detectably_invalid(b, ptr, 0));
+  assert(not detectably_invalid(b, ptr, 1));
 
   const B b_c{};
-  assert(is_checkably_valid(b_c, ptr, 0));
-  assert(is_checkably_valid(b_c, ptr, 1));
+  assert(not detectably_invalid(b_c, ptr, 0));
+  assert(not detectably_invalid(b_c, ptr, 1));
 
-  // If users make is_checkably_valid nonconst or not noexcept,
+  // If users make detectably_invalid nonconst or not noexcept,
   // the nonmember function falls back to a default implementation.
 
   try {
-    assert(is_checkably_valid(C{}, ptr, 0));
-    assert(is_checkably_valid(C{}, ptr, 1));
+    assert(not detectably_invalid(C{}, ptr, 0));
+    assert(not detectably_invalid(C{}, ptr, 1));
   }
   catch (const std::runtime_error& e) {
     std::cerr << "C{} threw runtime_error: " << e.what() << "\n";
@@ -1231,8 +1432,8 @@ int main()
 
   try {
     const C c_c{};
-    assert(is_checkably_valid(c_c, ptr, 0));
-    assert(is_checkably_valid(c_c, ptr, 1));
+    assert(not detectably_invalid(c_c, ptr, 0));
+    assert(not detectably_invalid(c_c, ptr, 1));
   }
   catch (const std::runtime_error& e) {
     std::cerr << "const C threw runtime_error: " << e.what() << "\n";
@@ -1240,16 +1441,16 @@ int main()
 
   try {
     C c{};
-    assert(is_checkably_valid(c, ptr, 0));
-    assert(is_checkably_valid(c, ptr, 1));
+    assert(not detectably_invalid(c, ptr, 0));
+    assert(not detectably_invalid(c, ptr, 1));
   }
   catch (const std::runtime_error& e) {
     std::cerr << "nonconst C threw runtime_error: " << e.what() << "\n";
   }
 
   try {
-    assert(is_checkably_valid(D{}, ptr, 0));
-    assert(is_checkably_valid(D{}, ptr, 1));
+    assert(not detectably_invalid(D{}, ptr, 0));
+    assert(not detectably_invalid(D{}, ptr, 1));
   }
   catch (const std::runtime_error& e) {
     std::cerr << "D{} threw runtime_error: " << e.what() << "\n";
@@ -1257,8 +1458,8 @@ int main()
 
   try {
     const D d_c{};
-    assert(is_checkably_valid(d_c, ptr, 0));
-    assert(is_checkably_valid(d_c, ptr, 1));
+    assert(not detectably_invalid(d_c, ptr, 0));
+    assert(not detectably_invalid(d_c, ptr, 1));
   }
   catch (const std::runtime_error& e) {
     std::cerr << "const D threw runtime_error: " << e.what() << "\n";
@@ -1266,8 +1467,8 @@ int main()
 
   try {
     D d{};
-    assert(is_checkably_valid(d, ptr, 0));
-    assert(is_checkably_valid(d, ptr, 1));
+    assert(not detectably_invalid(d, ptr, 0));
+    assert(not detectably_invalid(d, ptr, 1));
   }
   catch (const std::runtime_error& e) {
     std::cerr << "nonconst D threw runtime_error: " << e.what() << "\n";

From 9923b3f2294faf3dcce724b6341ac2190c2fb323 Mon Sep 17 00:00:00 2001
From: Mark Hoemmen <mhoemmen@nvidia.com>
Date: Wed, 10 Jul 2024 15:49:20 -0600
Subject: [PATCH 2/4] P2897R2: More non-wording

Add non-wording section explaining why `aligned_accessor`
has no `explicit` constructor from less to more alignment.
---
 aligned_accessor/P2897R2.html        | 129 +++++++++++++++++++--------
 aligned_accessor/aligned_accessor.md |  58 ++++++++++++
 2 files changed, 149 insertions(+), 38 deletions(-)

diff --git a/aligned_accessor/P2897R2.html b/aligned_accessor/P2897R2.html
index 2b6c412a..aeca3451 100644
--- a/aligned_accessor/P2897R2.html
+++ b/aligned_accessor/P2897R2.html
@@ -460,46 +460,48 @@ <h1 id="toctitle">Contents</h1>
 nestable</a></li>
 <li><a href="#explicit-constructor-from-default_accessor" id="toc-explicit-constructor-from-default_accessor"><span class="toc-section-number">5.2</span> Explicit constructor from
 <code>default_accessor</code></a></li>
-<li><a href="#we-do-not-define-an-alias-for-aligned-mdspan" id="toc-we-do-not-define-an-alias-for-aligned-mdspan"><span class="toc-section-number">5.3</span> We do not define an alias for
+<li><a href="#why-no-explicit-constructor-from-less-to-more-alignment" id="toc-why-no-explicit-constructor-from-less-to-more-alignment"><span class="toc-section-number">5.3</span> Why no explicit constructor from
+less to more alignment?</a></li>
+<li><a href="#we-do-not-define-an-alias-for-aligned-mdspan" id="toc-we-do-not-define-an-alias-for-aligned-mdspan"><span class="toc-section-number">5.4</span> We do not define an alias for
 aligned mdspan</a></li>
-<li><a href="#mdspan-construction-safety" id="toc-mdspan-construction-safety"><span class="toc-section-number">5.4</span> mdspan construction
+<li><a href="#mdspan-construction-safety" id="toc-mdspan-construction-safety"><span class="toc-section-number">5.5</span> mdspan construction
 safety</a></li>
-<li><a href="#is_sufficiently_aligned-is-not-constexpr" id="toc-is_sufficiently_aligned-is-not-constexpr"><span class="toc-section-number">5.5</span>
+<li><a href="#is_sufficiently_aligned-is-not-constexpr" id="toc-is_sufficiently_aligned-is-not-constexpr"><span class="toc-section-number">5.6</span>
 <code>is_sufficiently_aligned</code> is not
 <code>constexpr</code></a></li>
-<li><a href="#generalize-is_sufficiently_aligned-for-all-accessors" id="toc-generalize-is_sufficiently_aligned-for-all-accessors"><span class="toc-section-number">5.6</span> Generalize
+<li><a href="#generalize-is_sufficiently_aligned-for-all-accessors" id="toc-generalize-is_sufficiently_aligned-for-all-accessors"><span class="toc-section-number">5.7</span> Generalize
 <code>is_sufficiently_aligned</code> for all accessors?</a>
 <ul>
-<li><a href="#is_sufficiently_aligned-is-specific-to-aligned_accessor" id="toc-is_sufficiently_aligned-is-specific-to-aligned_accessor"><span class="toc-section-number">5.6.1</span>
+<li><a href="#is_sufficiently_aligned-is-specific-to-aligned_accessor" id="toc-is_sufficiently_aligned-is-specific-to-aligned_accessor"><span class="toc-section-number">5.7.1</span>
 <code>is_sufficiently_aligned</code> is specific to
 <code>aligned_accessor</code></a></li>
-<li><a href="#detectably_invalid-generic-validity-check" id="toc-detectably_invalid-generic-validity-check"><span class="toc-section-number">5.6.2</span> <code>detectably_invalid</code>:
+<li><a href="#detectably_invalid-generic-validity-check" id="toc-detectably_invalid-generic-validity-check"><span class="toc-section-number">5.7.2</span> <code>detectably_invalid</code>:
 Generic validity check?</a></li>
-<li><a href="#arguments-against-and-for-detectably_invalid" id="toc-arguments-against-and-for-detectably_invalid"><span class="toc-section-number">5.6.3</span> Arguments against and for
+<li><a href="#arguments-against-and-for-detectably_invalid" id="toc-arguments-against-and-for-detectably_invalid"><span class="toc-section-number">5.7.3</span> Arguments against and for
 <code>detectably_invalid</code></a></li>
-<li><a href="#nonmember-customization-point-design" id="toc-nonmember-customization-point-design"><span class="toc-section-number">5.6.4</span> Nonmember customization point
+<li><a href="#nonmember-customization-point-design" id="toc-nonmember-customization-point-design"><span class="toc-section-number">5.7.4</span> Nonmember customization point
 design</a></li>
-<li><a href="#we-need-both-is_sufficiently_aligned-and-detectably_invalid" id="toc-we-need-both-is_sufficiently_aligned-and-detectably_invalid"><span class="toc-section-number">5.6.5</span> We need both
+<li><a href="#we-need-both-is_sufficiently_aligned-and-detectably_invalid" id="toc-we-need-both-is_sufficiently_aligned-and-detectably_invalid"><span class="toc-section-number">5.7.5</span> We need both
 <code>is_sufficiently_aligned</code> and
 <code>detectably_invalid</code></a></li>
-<li><a href="#alternative-nonmember-is_sufficiently_aligned" id="toc-alternative-nonmember-is_sufficiently_aligned"><span class="toc-section-number">5.6.6</span> Alternative: Nonmember
+<li><a href="#alternative-nonmember-is_sufficiently_aligned" id="toc-alternative-nonmember-is_sufficiently_aligned"><span class="toc-section-number">5.7.6</span> Alternative: Nonmember
 <code>is_sufficiently_aligned</code></a></li>
-<li><a href="#do-accessors-need-to-check-anything-else" id="toc-do-accessors-need-to-check-anything-else"><span class="toc-section-number">5.6.7</span> Do accessors need to check
+<li><a href="#do-accessors-need-to-check-anything-else" id="toc-do-accessors-need-to-check-anything-else"><span class="toc-section-number">5.7.7</span> Do accessors need to check
 anything else?</a></li>
-<li><a href="#naming-the-function" id="toc-naming-the-function"><span class="toc-section-number">5.6.8</span> Naming the function</a></li>
-<li><a href="#conclusions" id="toc-conclusions"><span class="toc-section-number">5.6.9</span> Conclusions</a></li>
+<li><a href="#naming-the-function" id="toc-naming-the-function"><span class="toc-section-number">5.7.8</span> Naming the function</a></li>
+<li><a href="#conclusions" id="toc-conclusions"><span class="toc-section-number">5.7.9</span> Conclusions</a></li>
 </ul></li>
-<li><a href="#explicit-conversions-as-the-model-for-precondition-asserting-conversions" id="toc-explicit-conversions-as-the-model-for-precondition-asserting-conversions"><span class="toc-section-number">5.7</span> Explicit conversions as the model
+<li><a href="#explicit-conversions-as-the-model-for-precondition-asserting-conversions" id="toc-explicit-conversions-as-the-model-for-precondition-asserting-conversions"><span class="toc-section-number">5.8</span> Explicit conversions as the model
 for precondition-asserting conversions</a>
 <ul>
-<li><a href="#example-conversion-to-aligned_accessor" id="toc-example-conversion-to-aligned_accessor"><span class="toc-section-number">5.7.1</span> Example: conversion to
+<li><a href="#example-conversion-to-aligned_accessor" id="toc-example-conversion-to-aligned_accessor"><span class="toc-section-number">5.8.1</span> Example: conversion to
 <code>aligned_accessor</code></a></li>
-<li><a href="#status-quo" id="toc-status-quo"><span class="toc-section-number">5.7.2</span> Status quo</a></li>
-<li><a href="#mdspan-uses-explicit-conversions-to-assert-preconditions" id="toc-mdspan-uses-explicit-conversions-to-assert-preconditions"><span class="toc-section-number">5.7.3</span> <code>mdspan</code> uses
+<li><a href="#status-quo" id="toc-status-quo"><span class="toc-section-number">5.8.2</span> Status quo</a></li>
+<li><a href="#mdspan-uses-explicit-conversions-to-assert-preconditions" id="toc-mdspan-uses-explicit-conversions-to-assert-preconditions"><span class="toc-section-number">5.8.3</span> <code>mdspan</code> uses
 explicit conversions to assert preconditions</a></li>
-<li><a href="#alternative-explicit-cast-function-naughty_cast" id="toc-alternative-explicit-cast-function-naughty_cast"><span class="toc-section-number">5.7.4</span> Alternative: explicit cast
+<li><a href="#alternative-explicit-cast-function-naughty_cast" id="toc-alternative-explicit-cast-function-naughty_cast"><span class="toc-section-number">5.8.4</span> Alternative: explicit cast
 function <code>naughty_cast</code></a></li>
-<li><a href="#conclusion-retain-mdspans-current-design" id="toc-conclusion-retain-mdspans-current-design"><span class="toc-section-number">5.7.5</span> Conclusion: retain
+<li><a href="#conclusion-retain-mdspans-current-design" id="toc-conclusion-retain-mdspans-current-design"><span class="toc-section-number">5.8.5</span> Conclusion: retain
 <code>mdspan</code>’s current design</a></li>
 </ul></li>
 </ul></li>
@@ -588,6 +590,9 @@ <h1 data-number="2" id="revision-history"><span class="header-section-number">2<
 <li><p>P2389R2 was voted into the Working Draft at St. Louis, so replace
 use of <code>dextents</code> in examples with
 <code>dims</code>.</p></li>
+<li><p>Add non-wording section explaining why
+<code>aligned_accessor</code> has no <code>explicit</code> constructor
+from less to more alignment</p></li>
 </ul></li>
 </ul>
 <h1 data-number="3" id="purpose-of-this-paper"><span class="header-section-number">3</span> Purpose of this paper<a href="#purpose-of-this-paper" class="self-link"></a></h1>
@@ -774,7 +779,55 @@ <h2 data-number="5.2" id="explicit-constructor-from-default_accessor"><span clas
 <div class="sourceCode" id="cb4"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb4-1"><a href="#cb4-1" aria-hidden="true" tabindex="-1"></a>aligned_matrix_t aligned_matrix<span class="op">{</span>matrix<span class="op">.</span>data_handle<span class="op">()</span>, matrix<span class="op">.</span>mapping<span class="op">()}</span>;</span></code></pre></div>
 <p>Now, users can do the same thing with fewer characters.</p>
 <div class="sourceCode" id="cb5"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb5-1"><a href="#cb5-1" aria-hidden="true" tabindex="-1"></a>aligned_matrix_t aligned_matrix<span class="op">{</span>matrix<span class="op">}</span>;</span></code></pre></div>
-<h2 data-number="5.3" id="we-do-not-define-an-alias-for-aligned-mdspan"><span class="header-section-number">5.3</span> We do not define an alias for
+<h2 data-number="5.3" id="why-no-explicit-constructor-from-less-to-more-alignment"><span class="header-section-number">5.3</span> Why no explicit constructor
+from less to more alignment?<a href="#why-no-explicit-constructor-from-less-to-more-alignment" class="self-link"></a></h2>
+<p>As explained in the previous section, <code>aligned_accessor</code>
+has an <code>explicit</code> converting constructor from
+<code>default_accessor</code> so that users can assert overalignment. It
+also has an (implicit) converting constructor from another
+<code>aligned_accessor</code> with more alignment, to an
+<code>aligned_accessor</code> with less alignment. However,
+<code>aligned_accessor</code> does <em>not</em> have an
+<code>explicit</code> converting constructor from another
+<code>aligned_accessor</code> with <em>less</em> alignment, to an
+<code>aligned_accessor</code> with <em>more</em> alignment. Why not?</p>
+<p>Consider the three typical use cases for
+<code>aligned_accessor</code>.</p>
+<ol type="1">
+<li><p>User knows an allocation’s alignment at compile time.</p></li>
+<li><p>User knows an allocation’s alignment at run time, but not at
+compile time. For example, the value might depend on run-time detection
+of particular hardware features.</p></li>
+<li><p>User doesn’t know whether an allocation is overaligned. They
+might need to ask some system at run time, or check the pointer value
+themselves, in order to decide whether to call code that expects a
+particular alignment.</p></li>
+</ol>
+<p>In Case (1), users would normally declare the maximum alignment. They
+would want to preserve this information at compile time as much as
+possible, by keeping the <code>aligned_accessor</code>
+<code>mdspan</code> with maximum compile-time alignment for the entire
+scope of its use. Users would only want implicit conversions to less
+alignment or <code>default_accessor</code> when calling functions whose
+parameter types encode these requirements.</p>
+<p>Case (2) reduces to Case (3).</p>
+<p>Case (3) reduces to Case (1). This works like any conversion from
+run-time type to compile-time type, with a fixed list of possible
+compile-time types (the alignments). As soon as a user’s
+<code>mdspan</code> enters a scope where the alignment is known at
+compile time, the user would want to preserve that compile-time
+information and maximize the alignment for as large of a scope as
+possible.</p>
+<p>None of these cases involve starting with more alignment, going to
+less (but still some) alignment, and then going back to more alignment
+again. Code that does that probably does not correctly use the types of
+function parameters to express its overalignment requirements. It’s like
+code that uses <code>dynamic_cast</code> a lot. Users can still convert
+from less or more alignment by creating the result’s
+<code>aligned_accessor</code> manually. However, we don’t want to
+encourage this pattern, so we don’t offer an explicit conversion for
+it.</p>
+<h2 data-number="5.4" id="we-do-not-define-an-alias-for-aligned-mdspan"><span class="header-section-number">5.4</span> We do not define an alias for
 aligned mdspan<a href="#we-do-not-define-an-alias-for-aligned-mdspan" class="self-link"></a></h2>
 <p>In LEWG’s 2023-10-10 review of R0, participants observed that this
 proposal’s examples define an example-specific type alias for
@@ -827,7 +880,7 @@ <h2 data-number="5.3" id="we-do-not-define-an-alias-for-aligned-mdspan"><span cl
 For this reason, while we do not object to an
 <code>aligned_mdspan</code> alias, we do not find the proliferation of
 aliases particularly ergonomic.</p>
-<h2 data-number="5.4" id="mdspan-construction-safety"><span class="header-section-number">5.4</span> mdspan construction safety<a href="#mdspan-construction-safety" class="self-link"></a></h2>
+<h2 data-number="5.5" id="mdspan-construction-safety"><span class="header-section-number">5.5</span> mdspan construction safety<a href="#mdspan-construction-safety" class="self-link"></a></h2>
 <p>LEWG’s 2023-10-10 review of R0 expressed concern that
 <code>mdspan</code>’s constructor has no way to check
 <code>aligned_accessor</code>’s alignment requirements. Users can call
@@ -947,7 +1000,7 @@ <h2 data-number="5.4" id="mdspan-construction-safety"><span class="header-sectio
 in the section below “Generalize <code>is_sufficiently_aligned</code>
 for all accessors?”. We plan further exploration of this topic
 elsewhere.</p>
-<h2 data-number="5.5" id="is_sufficiently_aligned-is-not-constexpr"><span class="header-section-number">5.5</span>
+<h2 data-number="5.6" id="is_sufficiently_aligned-is-not-constexpr"><span class="header-section-number">5.6</span>
 <code>is_sufficiently_aligned</code> is not <code>constexpr</code><a href="#is_sufficiently_aligned-is-not-constexpr" class="self-link"></a></h2>
 <p>LEWG reviewed R1 of this proposal at the June 2024 St. Louis WG21
 meeting, and polled 1/10/0/0/1 (SF/F/N/A/SA) to remove
@@ -968,9 +1021,9 @@ <h2 data-number="5.5" id="is_sufficiently_aligned-is-not-constexpr"><span class=
 <code>is_sufficiently_aligned</code> to be <code>constexpr</code> and
 did not want to constrain implementations to use compiler-specific
 functionality.</p>
-<h2 data-number="5.6" id="generalize-is_sufficiently_aligned-for-all-accessors"><span class="header-section-number">5.6</span> Generalize
+<h2 data-number="5.7" id="generalize-is_sufficiently_aligned-for-all-accessors"><span class="header-section-number">5.7</span> Generalize
 <code>is_sufficiently_aligned</code> for all accessors?<a href="#generalize-is_sufficiently_aligned-for-all-accessors" class="self-link"></a></h2>
-<h3 data-number="5.6.1" id="is_sufficiently_aligned-is-specific-to-aligned_accessor"><span class="header-section-number">5.6.1</span>
+<h3 data-number="5.7.1" id="is_sufficiently_aligned-is-specific-to-aligned_accessor"><span class="header-section-number">5.7.1</span>
 <code>is_sufficiently_aligned</code> is specific to
 <code>aligned_accessor</code><a href="#is_sufficiently_aligned-is-specific-to-aligned_accessor" class="self-link"></a></h3>
 <p>The <code>is_sufficiently_aligned</code> function exists so users can
@@ -982,7 +1035,7 @@ <h3 data-number="5.6.1" id="is_sufficiently_aligned-is-specific-to-aligned_acces
 <code>mdspan</code> users, Standard C++ offers no way to check whether a
 pointer and a layout mapping’s <code>required_span_size()</code> form a
 valid range.</p>
-<h3 data-number="5.6.2" id="detectably_invalid-generic-validity-check"><span class="header-section-number">5.6.2</span>
+<h3 data-number="5.7.2" id="detectably_invalid-generic-validity-check"><span class="header-section-number">5.7.2</span>
 <code>detectably_invalid</code>: Generic validity check?<a href="#detectably_invalid-generic-validity-check" class="self-link"></a></h3>
 <p>During the June 2024 St. Louis WG21 meeting, one LEWG reviewer
 (please see Acknowledgments below) pointed out that code that is generic
@@ -1026,7 +1079,7 @@ <h3 data-number="5.6.2" id="detectably_invalid-generic-validity-check"><span cla
 <span id="cb11-9"><a href="#cb11-9" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
 <span id="cb11-10"><a href="#cb11-10" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> mdspan<span class="op">{</span>handle, mapping, accessor<span class="op">}</span>;</span>
 <span id="cb11-11"><a href="#cb11-11" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
-<h3 data-number="5.6.3" id="arguments-against-and-for-detectably_invalid"><span class="header-section-number">5.6.3</span> Arguments against and for
+<h3 data-number="5.7.3" id="arguments-against-and-for-detectably_invalid"><span class="header-section-number">5.7.3</span> Arguments against and for
 <code>detectably_invalid</code><a href="#arguments-against-and-for-detectably_invalid" class="self-link"></a></h3>
 <p>We didn’t include this feature in the original <code>mdspan</code>
 design because most data handle types have no way to say with full
@@ -1067,7 +1120,7 @@ <h3 data-number="5.6.3" id="arguments-against-and-for-detectably_invalid"><span
 <div class="sourceCode" id="cb12"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb12-1"><a href="#cb12-1" aria-hidden="true" tabindex="-1"></a><span class="dt">bool</span> detectably_invalid<span class="op">(</span>data_handle_type ptr, <span class="dt">size_t</span> size<span class="op">)</span> <span class="kw">const</span> <span class="op">{</span></span>
 <span id="cb12-2"><a href="#cb12-2" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> ptr <span class="op">==</span> <span class="kw">nullptr</span> <span class="op">&amp;&amp;</span> size <span class="op">!=</span> <span class="dv">0</span>;</span>
 <span id="cb12-3"><a href="#cb12-3" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
-<h3 data-number="5.6.4" id="nonmember-customization-point-design"><span class="header-section-number">5.6.4</span> Nonmember customization point
+<h3 data-number="5.7.4" id="nonmember-customization-point-design"><span class="header-section-number">5.7.4</span> Nonmember customization point
 design<a href="#nonmember-customization-point-design" class="self-link"></a></h3>
 <p>C++23 defines the generic interface of accessors through the accessor
 policy requirements
@@ -1110,7 +1163,7 @@ <h3 data-number="5.6.4" id="nonmember-customization-point-design"><span class="h
 <span id="cb13-19"><a href="#cb13-19" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="kw">false</span>;</span>
 <span id="cb13-20"><a href="#cb13-20" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
 <span id="cb13-21"><a href="#cb13-21" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
-<h3 data-number="5.6.5" id="we-need-both-is_sufficiently_aligned-and-detectably_invalid"><span class="header-section-number">5.6.5</span> We need both
+<h3 data-number="5.7.5" id="we-need-both-is_sufficiently_aligned-and-detectably_invalid"><span class="header-section-number">5.7.5</span> We need both
 <code>is_sufficiently_aligned</code> and
 <code>detectably_invalid</code><a href="#we-need-both-is_sufficiently_aligned-and-detectably_invalid" class="self-link"></a></h3>
 <p>The same LEWG reviewer who suggested <code>detectably_invalid</code>
@@ -1132,7 +1185,7 @@ <h3 data-number="5.6.5" id="we-need-both-is_sufficiently_aligned-and-detectably_
 <code>required_span_size()</code> is valid. Thus, the only thing the
 user would need to check would be whether the data handle is
 sufficiently aligned.</p>
-<h3 data-number="5.6.6" id="alternative-nonmember-is_sufficiently_aligned"><span class="header-section-number">5.6.6</span> Alternative: Nonmember
+<h3 data-number="5.7.6" id="alternative-nonmember-is_sufficiently_aligned"><span class="header-section-number">5.7.6</span> Alternative: Nonmember
 <code>is_sufficiently_aligned</code><a href="#alternative-nonmember-is_sufficiently_aligned" class="self-link"></a></h3>
 <p>The reviewer responded to our argument above by suggesting that we
 remove <code>is_sufficiently_aligned</code> from
@@ -1147,7 +1200,7 @@ <h3 data-number="5.6.6" id="alternative-nonmember-is_sufficiently_aligned"><span
 option would be to put it in <code>&lt;memory&gt;</code> right after
 <code>assume_aligned</code>, and to mark it as freestanding (since
 <code>assume_aligned</code> is).</p>
-<h3 data-number="5.6.7" id="do-accessors-need-to-check-anything-else"><span class="header-section-number">5.6.7</span> Do accessors need to check
+<h3 data-number="5.7.7" id="do-accessors-need-to-check-anything-else"><span class="header-section-number">5.7.7</span> Do accessors need to check
 anything else?<a href="#do-accessors-need-to-check-anything-else" class="self-link"></a></h3>
 <p>The only other thing an accessor’s user might want to check besides a
 (data handle, size) pair would be converting construction from another
@@ -1168,7 +1221,7 @@ <h3 data-number="5.6.7" id="do-accessors-need-to-check-anything-else"><span clas
 constructors only take one argument, so there would be no ambiguity
 between calling <code>detectably_invalid</code> with an accessor and
 calling it with a data handle and size.</p>
-<h3 data-number="5.6.8" id="naming-the-function"><span class="header-section-number">5.6.8</span> Naming the function<a href="#naming-the-function" class="self-link"></a></h3>
+<h3 data-number="5.7.8" id="naming-the-function"><span class="header-section-number">5.7.8</span> Naming the function<a href="#naming-the-function" class="self-link"></a></h3>
 <ol type="1">
 <li><p>The function describes a property: “this (data handle, size) pair
 is not known to be invalid.” It’s an adjective (like
@@ -1192,7 +1245,7 @@ <h3 data-number="5.6.8" id="naming-the-function"><span class="header-section-num
 </ol>
 <p>These points together suggest the name
 <code>detectably_invalid</code>.</p>
-<h3 data-number="5.6.9" id="conclusions"><span class="header-section-number">5.6.9</span> Conclusions<a href="#conclusions" class="self-link"></a></h3>
+<h3 data-number="5.7.9" id="conclusions"><span class="header-section-number">5.7.9</span> Conclusions<a href="#conclusions" class="self-link"></a></h3>
 <ol type="1">
 <li><p>Adding <code>detectably_invalid</code> to the accessor
 requirements and existing Standard accessors in C++26 would be a
@@ -1215,7 +1268,7 @@ <h3 data-number="5.6.9" id="conclusions"><span class="header-section-number">5.6
 a separate nonmember function. We think LWG review of
 <code>aligned_accessor</code> can proceed concurrently.</p></li>
 </ol>
-<h2 data-number="5.7" id="explicit-conversions-as-the-model-for-precondition-asserting-conversions"><span class="header-section-number">5.7</span> Explicit conversions as the
+<h2 data-number="5.8" id="explicit-conversions-as-the-model-for-precondition-asserting-conversions"><span class="header-section-number">5.8</span> Explicit conversions as the
 model for precondition-asserting conversions<a href="#explicit-conversions-as-the-model-for-precondition-asserting-conversions" class="self-link"></a></h2>
 <p>During the June 2024 St. Louis WG21 meeting, one LEWG reviewer asked
 about the <code>explicit</code> constructor from
@@ -1226,7 +1279,7 @@ <h2 data-number="5.7" id="explicit-conversions-as-the-model-for-precondition-ass
 more explicit than an <code>explicit</code> constructor: e.g., a new
 <code>*_cast</code> function template. We do not agree with this idea;
 this section explains why.</p>
-<h3 data-number="5.7.1" id="example-conversion-to-aligned_accessor"><span class="header-section-number">5.7.1</span> Example: conversion to
+<h3 data-number="5.8.1" id="example-conversion-to-aligned_accessor"><span class="header-section-number">5.8.1</span> Example: conversion to
 <code>aligned_accessor</code><a href="#example-conversion-to-aligned_accessor" class="self-link"></a></h3>
 <p>Suppose that some function that users can’t change returns an
 <code>mdspan</code> of <code>float</code> with
@@ -1241,7 +1294,7 @@ <h3 data-number="5.7.1" id="example-conversion-to-aligned_accessor"><span class=
 Its return type doesn’t matter for this example.</p>
 <div class="sourceCode" id="cb15"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb15-1"><a href="#cb15-1" aria-hidden="true" tabindex="-1"></a>SomeReturnType use_overaligned_view<span class="op">(</span></span>
 <span id="cb15-2"><a href="#cb15-2" aria-hidden="true" tabindex="-1"></a>  mdspan<span class="op">&lt;</span><span class="dt">float</span>, dims<span class="op">&lt;</span><span class="dv">1</span><span class="op">&gt;</span>, layout_right, aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, <span class="dv">8</span><span class="op">&gt;&gt;)</span>;</span></code></pre></div>
-<h3 data-number="5.7.2" id="status-quo"><span class="header-section-number">5.7.2</span> Status quo<a href="#status-quo" class="self-link"></a></h3>
+<h3 data-number="5.8.2" id="status-quo"><span class="header-section-number">5.8.2</span> Status quo<a href="#status-quo" class="self-link"></a></h3>
 <p>How do users call <code>use_overaligned_view</code> with the object
 returned from <code>overaligned_view</code>? The status quo offers two
 ways. Both of them rely on
@@ -1275,7 +1328,7 @@ <h3 data-number="5.7.2" id="status-quo"><span class="header-section-number">5.7.
 it’s nondefault, and it’s a class with a short definition. Users can go
 to the specification, see <code>assume_aligned</code>, and know they are
 dealing with a low-level function that has a precondition.</p>
-<h3 data-number="5.7.3" id="mdspan-uses-explicit-conversions-to-assert-preconditions"><span class="header-section-number">5.7.3</span> <code>mdspan</code> uses
+<h3 data-number="5.8.3" id="mdspan-uses-explicit-conversions-to-assert-preconditions"><span class="header-section-number">5.8.3</span> <code>mdspan</code> uses
 explicit conversions to assert preconditions<a href="#mdspan-uses-explicit-conversions-to-assert-preconditions" class="self-link"></a></h3>
 <p>The entire system of <code>mdspan</code> components was designed so
 that</p>
@@ -1307,7 +1360,7 @@ <h3 data-number="5.7.3" id="mdspan-uses-explicit-conversions-to-assert-precondit
 range <span class="math inline">[</span> <code>ptr</code>,
 <code>ptr + size</code> <span class="math inline">)</span> is
 accessible.</p>
-<h3 data-number="5.7.4" id="alternative-explicit-cast-function-naughty_cast"><span class="header-section-number">5.7.4</span> Alternative: explicit cast
+<h3 data-number="5.8.4" id="alternative-explicit-cast-function-naughty_cast"><span class="header-section-number">5.8.4</span> Alternative: explicit cast
 function <code>naughty_cast</code><a href="#alternative-explicit-cast-function-naughty_cast" class="self-link"></a></h3>
 <p>Everything we have described above is the status quo. What did the
 one LEWG reviewer want to see? They wanted all conversions with
@@ -1403,7 +1456,7 @@ <h3 data-number="5.7.4" id="alternative-explicit-cast-function-naughty_cast"><sp
 points for Senders and Receivers”) changes P2300 (std::execution) to use
 member functions instead of <code>tag_invoke</code>-based customization
 points.</p>
-<h3 data-number="5.7.5" id="conclusion-retain-mdspans-current-design"><span class="header-section-number">5.7.5</span> Conclusion: retain
+<h3 data-number="5.8.5" id="conclusion-retain-mdspans-current-design"><span class="header-section-number">5.8.5</span> Conclusion: retain
 <code>mdspan</code>’s current design<a href="#conclusion-retain-mdspans-current-design" class="self-link"></a></h3>
 <p>For all these reasons, we do not support replacing
 <code>mdspan</code>’s current “conversions with preconditions are
diff --git a/aligned_accessor/aligned_accessor.md b/aligned_accessor/aligned_accessor.md
index 10fbbbb7..f3a4a7e6 100644
--- a/aligned_accessor/aligned_accessor.md
+++ b/aligned_accessor/aligned_accessor.md
@@ -81,6 +81,9 @@ toc: true
     * P2389R2 was voted into the Working Draft at St. Louis,
         so replace use of `dextents` in examples with `dims`.
 
+    * Add non-wording section explaining why `aligned_accessor`
+        has no `explicit` constructor from less to more alignment
+
 # Purpose of this paper
 
 We propose adding `aligned_accessor` to the C++ Standard Library.
@@ -272,6 +275,61 @@ Now, users can do the same thing with fewer characters.
 aligned_matrix_t aligned_matrix{matrix};
 ```
 
+## Why no explicit constructor from less to more alignment?
+
+As explained in the previous section,
+`aligned_accessor` has an `explicit` converting constructor
+from `default_accessor` so that users can assert overalignment.
+It also has an (implicit) converting constructor
+from another `aligned_accessor` with more alignment,
+to an `aligned_accessor` with less alignment.
+However, `aligned_accessor` does *not* have
+an `explicit` converting constructor
+from another `aligned_accessor` with *less* alignment,
+to an `aligned_accessor` with *more* alignment.  Why not?
+
+Consider the three typical use cases for `aligned_accessor`.
+
+1. User knows an allocation's alignment at compile time.
+
+2. User knows an allocation's alignment at run time,
+    but not at compile time.  For example, the value might depend
+    on run-time detection of particular hardware features.
+
+3. User doesn't know whether an allocation is overaligned.
+    They might need to ask some system at run time,
+    or check the pointer value themselves, in order to decide
+    whether to call code that expects a particular alignment.
+
+In Case (1), users would normally declare the maximum alignment.
+They would want to preserve this information at compile time
+as much as possible, by keeping the `aligned_accessor` `mdspan` with
+maximum compile-time alignment for the entire scope of its use.
+Users would only want implicit conversions to less alignment
+or `default_accessor` when calling functions whose parameter types
+encode these requirements.
+
+Case (2) reduces to Case (3).
+
+Case (3) reduces to Case (1).  This works like
+any conversion from run-time type to compile-time type,
+with a fixed list of possible compile-time types (the alignments).
+As soon as a user's `mdspan` enters a scope
+where the alignment is known at compile time,
+the user would want to preserve that compile-time information
+and maximize the alignment for as large of a scope as possible.
+
+None of these cases involve starting with more alignment,
+going to less (but still some) alignment,
+and then going back to more alignment again.
+Code that does that probably does not correctly use the types
+of function parameters to express its overalignment requirements.
+It's like code that uses `dynamic_cast` a lot.
+Users can still convert from less or more alignment
+by creating the result's `aligned_accessor` manually.
+However, we don't want to encourage this pattern,
+so we don't offer an explicit conversion for it.
+
 ## We do not define an alias for aligned mdspan
 
 In LEWG's 2023-10-10 review of R0,

From b40d37e57b1cc96f1e862b624cca0edd357a4daf Mon Sep 17 00:00:00 2001
From: Mark Hoemmen <mhoemmen@nvidia.com>
Date: Thu, 11 Jul 2024 09:30:39 -0600
Subject: [PATCH 3/4] P2897R2: Add godbolt link w/ full impl and demo

---
 aligned_accessor/P2897R2.html        | 193 ++++++++++++++++++++++++++-
 aligned_accessor/aligned_accessor.md | 192 +++++++++++++++++++++++++-
 2 files changed, 382 insertions(+), 3 deletions(-)

diff --git a/aligned_accessor/P2897R2.html b/aligned_accessor/P2897R2.html
index aeca3451..15202b6e 100644
--- a/aligned_accessor/P2897R2.html
+++ b/aligned_accessor/P2897R2.html
@@ -4,7 +4,7 @@
   <meta charset="utf-8" />
   <meta name="generator" content="mpark/wg21" />
   <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes" />
-  <meta name="dcterms.date" content="2024-07-10" />
+  <meta name="dcterms.date" content="2024-07-11" />
   <title>aligned_accessor: An mdspan accessor expressing pointer overalignment</title>
   <style>
 code{white-space: pre-wrap;}
@@ -425,7 +425,7 @@ <h1 class="title" style="text-align:center"><code>aligned_accessor</code>:
   </tr>
   <tr>
     <td>Date: </td>
-    <td>2024-07-10</td>
+    <td>2024-07-11</td>
   </tr>
   <tr>
     <td style="vertical-align:top">Project: </td>
@@ -520,6 +520,8 @@ <h1 id="toctitle">Contents</h1>
 <code>is_sufficiently_aligned</code> design</a></li>
 <li><a href="#appendix-a-detectably_invalid-nonmember-function-example" id="toc-appendix-a-detectably_invalid-nonmember-function-example"><span class="toc-section-number">12</span> Appendix A:
 <code>detectably_invalid</code> nonmember function example</a></li>
+<li><a href="#appendix-b-implementation-and-demo" id="toc-appendix-b-implementation-and-demo"><span class="toc-section-number">13</span> Appendix B: Implementation and
+demo</a></li>
 </ul>
 </div>
 <h1 data-number="1" id="authors"><span class="header-section-number">1</span> Authors<a href="#authors" class="self-link"></a></h1>
@@ -593,6 +595,8 @@ <h1 data-number="2" id="revision-history"><span class="header-section-number">2<
 <li><p>Add non-wording section explaining why
 <code>aligned_accessor</code> has no <code>explicit</code> constructor
 from less to more alignment</p></li>
+<li><p>Add Compiler Explorer link with full implementation and
+demo</p></li>
 </ul></li>
 </ul>
 <h1 data-number="3" id="purpose-of-this-paper"><span class="header-section-number">3</span> Purpose of this paper<a href="#purpose-of-this-paper" class="self-link"></a></h1>
@@ -1952,6 +1956,191 @@ <h1 data-number="12" id="appendix-a-detectably_invalid-nonmember-function-exampl
 <span id="cb39-148"><a href="#cb39-148" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>cerr <span class="op">&lt;&lt;</span> <span class="st">&quot;Made it to the end</span><span class="sc">\n</span><span class="st">&quot;</span>;</span>
 <span id="cb39-149"><a href="#cb39-149" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> <span class="dv">0</span>;</span>
 <span id="cb39-150"><a href="#cb39-150" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
+<h1 data-number="13" id="appendix-b-implementation-and-demo"><span class="header-section-number">13</span> Appendix B: Implementation and
+demo<a href="#appendix-b-implementation-and-demo" class="self-link"></a></h1>
+<p><a href="https://godbolt.org/z/hjqExKaeh">This Compiler Explorer
+link</a> gives a full implementation of <code>aligned_accessor</code>
+and a demonstration. We show the full source code from that link here
+below.</p>
+<div class="sourceCode" id="cb40"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb40-1"><a href="#cb40-1" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;https://raw.githubusercontent.com/kokkos/mdspan/single-header/mdspan.hpp&gt;</span></span>
+<span id="cb40-2"><a href="#cb40-2" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;bit&gt;</span></span>
+<span id="cb40-3"><a href="#cb40-3" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;cassert&gt;</span></span>
+<span id="cb40-4"><a href="#cb40-4" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;cmath&gt;</span></span>
+<span id="cb40-5"><a href="#cb40-5" aria-hidden="true" tabindex="-1"></a><span class="pp">#if defined(_MSC_VER)</span></span>
+<span id="cb40-6"><a href="#cb40-6" aria-hidden="true" tabindex="-1"></a><span class="pp">#  include </span><span class="im">&lt;cstdlib&gt;</span><span class="pp"> </span><span class="co">// MSVC&#39;s _aligned_malloc</span></span>
+<span id="cb40-7"><a href="#cb40-7" aria-hidden="true" tabindex="-1"></a><span class="pp">#endif</span></span>
+<span id="cb40-8"><a href="#cb40-8" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;exception&gt;</span></span>
+<span id="cb40-9"><a href="#cb40-9" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;functional&gt;</span></span>
+<span id="cb40-10"><a href="#cb40-10" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;memory&gt;</span></span>
+<span id="cb40-11"><a href="#cb40-11" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;numeric&gt;</span></span>
+<span id="cb40-12"><a href="#cb40-12" aria-hidden="true" tabindex="-1"></a><span class="pp">#include </span><span class="im">&lt;type_traits&gt;</span></span>
+<span id="cb40-13"><a href="#cb40-13" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-14"><a href="#cb40-14" aria-hidden="true" tabindex="-1"></a><span class="kw">namespace</span> stdex <span class="op">=</span> std<span class="op">::</span>experimental;</span>
+<span id="cb40-15"><a href="#cb40-15" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-16"><a href="#cb40-16" aria-hidden="true" tabindex="-1"></a><span class="co">// P2389 (voted into C++ at June 2024 STL plenary)</span></span>
+<span id="cb40-17"><a href="#cb40-17" aria-hidden="true" tabindex="-1"></a><span class="kw">namespace</span> std <span class="op">{</span></span>
+<span id="cb40-18"><a href="#cb40-18" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="dt">size_t</span> Rank, <span class="kw">class</span> IndexType <span class="op">=</span> <span class="dt">size_t</span><span class="op">&gt;</span></span>
+<span id="cb40-19"><a href="#cb40-19" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> dims <span class="op">=</span> dextents<span class="op">&lt;</span>IndexType, Rank<span class="op">&gt;</span>;</span>
+<span id="cb40-20"><a href="#cb40-20" aria-hidden="true" tabindex="-1"></a><span class="op">}</span> <span class="co">// namespace std</span></span>
+<span id="cb40-21"><a href="#cb40-21" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-22"><a href="#cb40-22" aria-hidden="true" tabindex="-1"></a><span class="kw">namespace</span> <span class="op">{</span></span>
+<span id="cb40-23"><a href="#cb40-23" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-24"><a href="#cb40-24" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, std<span class="op">::</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb40-25"><a href="#cb40-25" aria-hidden="true" tabindex="-1"></a><span class="kw">class</span> aligned_accessor <span class="op">{</span></span>
+<span id="cb40-26"><a href="#cb40-26" aria-hidden="true" tabindex="-1"></a><span class="kw">public</span><span class="op">:</span></span>
+<span id="cb40-27"><a href="#cb40-27" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static_assert</span><span class="op">(</span>std<span class="op">::</span>has_single_bit<span class="op">(</span>byte_alignment<span class="op">)</span>,</span>
+<span id="cb40-28"><a href="#cb40-28" aria-hidden="true" tabindex="-1"></a>    <span class="st">&quot;byte_alignment must be a power of two.&quot;</span><span class="op">)</span>;</span>
+<span id="cb40-29"><a href="#cb40-29" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static_assert</span><span class="op">(</span>byte_alignment <span class="op">&gt;=</span> <span class="kw">alignof</span><span class="op">(</span>ElementType<span class="op">)</span>,</span>
+<span id="cb40-30"><a href="#cb40-30" aria-hidden="true" tabindex="-1"></a>    <span class="st">&quot;Insufficient byte alignment for ElementType&quot;</span><span class="op">)</span>;</span>
+<span id="cb40-31"><a href="#cb40-31" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-32"><a href="#cb40-32" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> offset_policy <span class="op">=</span> stdex<span class="op">::</span>default_accessor<span class="op">&lt;</span>ElementType<span class="op">&gt;</span>;</span>
+<span id="cb40-33"><a href="#cb40-33" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> element_type <span class="op">=</span> ElementType;</span>
+<span id="cb40-34"><a href="#cb40-34" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> reference <span class="op">=</span> ElementType<span class="op">&amp;</span>;</span>
+<span id="cb40-35"><a href="#cb40-35" aria-hidden="true" tabindex="-1"></a>  <span class="kw">using</span> data_handle_type <span class="op">=</span> ElementType<span class="op">*</span>;</span>
+<span id="cb40-36"><a href="#cb40-36" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-37"><a href="#cb40-37" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> aligned_accessor<span class="op">()</span> <span class="kw">noexcept</span> <span class="op">=</span> <span class="cf">default</span>;</span>
+<span id="cb40-38"><a href="#cb40-38" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-39"><a href="#cb40-39" aria-hidden="true" tabindex="-1"></a>  <span class="kw">template</span><span class="op">&lt;</span></span>
+<span id="cb40-40"><a href="#cb40-40" aria-hidden="true" tabindex="-1"></a>    <span class="kw">class</span> OtherElementType,</span>
+<span id="cb40-41"><a href="#cb40-41" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span><span class="dt">size_t</span> other_byte_alignment<span class="op">&gt;</span></span>
+<span id="cb40-42"><a href="#cb40-42" aria-hidden="true" tabindex="-1"></a>  <span class="kw">requires</span><span class="op">(</span>std<span class="op">::</span>is_convertible_v<span class="op">&lt;</span></span>
+<span id="cb40-43"><a href="#cb40-43" aria-hidden="true" tabindex="-1"></a>    OtherElementType<span class="op">(*)[]</span>, element_type<span class="op">(*)[]&gt;)</span></span>
+<span id="cb40-44"><a href="#cb40-44" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> aligned_accessor<span class="op">(</span></span>
+<span id="cb40-45"><a href="#cb40-45" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span>OtherElementType, other_byte_alignment<span class="op">&gt;)</span></span>
+<span id="cb40-46"><a href="#cb40-46" aria-hidden="true" tabindex="-1"></a>      <span class="kw">noexcept</span></span>
+<span id="cb40-47"><a href="#cb40-47" aria-hidden="true" tabindex="-1"></a>  <span class="op">{</span></span>
+<span id="cb40-48"><a href="#cb40-48" aria-hidden="true" tabindex="-1"></a>    <span class="kw">constexpr</span> <span class="dt">size_t</span> the_gcd <span class="op">=</span></span>
+<span id="cb40-49"><a href="#cb40-49" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>gcd<span class="op">(</span>other_byte_alignment, byte_alignment<span class="op">)</span>;</span>
+<span id="cb40-50"><a href="#cb40-50" aria-hidden="true" tabindex="-1"></a>    <span class="kw">static_assert</span><span class="op">(</span>the_gcd <span class="op">==</span> byte_alignment<span class="op">)</span>;</span>
+<span id="cb40-51"><a href="#cb40-51" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb40-52"><a href="#cb40-52" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-53"><a href="#cb40-53" aria-hidden="true" tabindex="-1"></a>  <span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> OtherElementType<span class="op">&gt;</span></span>
+<span id="cb40-54"><a href="#cb40-54" aria-hidden="true" tabindex="-1"></a>  <span class="kw">requires</span><span class="op">(</span>std<span class="op">::</span>is_convertible_v<span class="op">&lt;</span></span>
+<span id="cb40-55"><a href="#cb40-55" aria-hidden="true" tabindex="-1"></a>    OtherElementType<span class="op">(*)[]</span>, element_type<span class="op">(*)[]&gt;)</span></span>
+<span id="cb40-56"><a href="#cb40-56" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="kw">explicit</span> aligned_accessor<span class="op">(</span></span>
+<span id="cb40-57"><a href="#cb40-57" aria-hidden="true" tabindex="-1"></a>    stdex<span class="op">::</span>default_accessor<span class="op">&lt;</span>OtherElementType<span class="op">&gt;)</span> <span class="kw">noexcept</span></span>
+<span id="cb40-58"><a href="#cb40-58" aria-hidden="true" tabindex="-1"></a>  <span class="op">{}</span></span>
+<span id="cb40-59"><a href="#cb40-59" aria-hidden="true" tabindex="-1"></a> </span>
+<span id="cb40-60"><a href="#cb40-60" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span></span>
+<span id="cb40-61"><a href="#cb40-61" aria-hidden="true" tabindex="-1"></a>    <span class="kw">operator</span> stdex<span class="op">::</span>default_accessor<span class="op">&lt;</span>element_type<span class="op">&gt;()</span> <span class="kw">const</span></span>
+<span id="cb40-62"><a href="#cb40-62" aria-hidden="true" tabindex="-1"></a>  <span class="op">{</span></span>
+<span id="cb40-63"><a href="#cb40-63" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="op">{}</span>;</span>
+<span id="cb40-64"><a href="#cb40-64" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb40-65"><a href="#cb40-65" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-66"><a href="#cb40-66" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> reference</span>
+<span id="cb40-67"><a href="#cb40-67" aria-hidden="true" tabindex="-1"></a>    access<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span></span>
+<span id="cb40-68"><a href="#cb40-68" aria-hidden="true" tabindex="-1"></a>  <span class="op">{</span></span>
+<span id="cb40-69"><a href="#cb40-69" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> std<span class="op">::</span>assume_aligned<span class="op">&lt;</span>byte_alignment<span class="op">&gt;(</span>p<span class="op">)[</span>i<span class="op">]</span>;</span>
+<span id="cb40-70"><a href="#cb40-70" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb40-71"><a href="#cb40-71" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-72"><a href="#cb40-72" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="kw">typename</span> offset_policy<span class="op">::</span>data_handle_type</span>
+<span id="cb40-73"><a href="#cb40-73" aria-hidden="true" tabindex="-1"></a>  offset<span class="op">(</span>data_handle_type p, <span class="dt">size_t</span> i<span class="op">)</span> <span class="kw">const</span> <span class="kw">noexcept</span> <span class="op">{</span></span>
+<span id="cb40-74"><a href="#cb40-74" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> p <span class="op">+</span> i;</span>
+<span id="cb40-75"><a href="#cb40-75" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb40-76"><a href="#cb40-76" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-77"><a href="#cb40-77" aria-hidden="true" tabindex="-1"></a>  <span class="kw">static</span> <span class="dt">bool</span></span>
+<span id="cb40-78"><a href="#cb40-78" aria-hidden="true" tabindex="-1"></a>    is_sufficiently_aligned<span class="op">(</span>data_handle_type p<span class="op">)</span> <span class="kw">noexcept</span></span>
+<span id="cb40-79"><a href="#cb40-79" aria-hidden="true" tabindex="-1"></a>  <span class="op">{</span></span>
+<span id="cb40-80"><a href="#cb40-80" aria-hidden="true" tabindex="-1"></a>    <span class="cf">return</span> <span class="kw">alignof</span><span class="op">(</span>ElementType<span class="op">)</span> <span class="op">&gt;=</span> byte_alignment <span class="op">&amp;&amp;</span> </span>
+<span id="cb40-81"><a href="#cb40-81" aria-hidden="true" tabindex="-1"></a>      std<span class="op">::</span>bit_cast<span class="op">&lt;</span><span class="dt">uintptr_t</span><span class="op">&gt;(</span>p<span class="op">)</span> <span class="op">%</span> byte_alignment <span class="op">==</span> <span class="dv">0</span>;</span>
+<span id="cb40-82"><a href="#cb40-82" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb40-83"><a href="#cb40-83" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb40-84"><a href="#cb40-84" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-85"><a href="#cb40-85" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb40-86"><a href="#cb40-86" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span></span>
+<span id="cb40-87"><a href="#cb40-87" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>mdspan<span class="op">&lt;</span><span class="dt">float</span>, std<span class="op">::</span>dims<span class="op">&lt;</span><span class="dv">1</span>, <span class="dt">int</span><span class="op">&gt;</span>, std<span class="op">::</span>layout_right,</span>
+<span id="cb40-88"><a href="#cb40-88" aria-hidden="true" tabindex="-1"></a>    aligned_accessor<span class="op">&lt;</span><span class="dt">float</span>, byte_alignment<span class="op">&gt;&gt;</span>;</span>
+<span id="cb40-89"><a href="#cb40-89" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-90"><a href="#cb40-90" aria-hidden="true" tabindex="-1"></a><span class="co">// Interfaces that require 32-byte alignment,</span></span>
+<span id="cb40-91"><a href="#cb40-91" aria-hidden="true" tabindex="-1"></a><span class="co">// because they want to do 8-wide SIMD of float.</span></span>
+<span id="cb40-92"><a href="#cb40-92" aria-hidden="true" tabindex="-1"></a><span class="dt">void</span></span>
+<span id="cb40-93"><a href="#cb40-93" aria-hidden="true" tabindex="-1"></a>vectorized_axpby<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y,</span>
+<span id="cb40-94"><a href="#cb40-94" aria-hidden="true" tabindex="-1"></a>  <span class="dt">float</span> alpha, aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> x, <span class="dt">float</span> beta<span class="op">)</span></span>
+<span id="cb40-95"><a href="#cb40-95" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb40-96"><a href="#cb40-96" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>x<span class="op">.</span>extent<span class="op">(</span><span class="dv">0</span><span class="op">)</span> <span class="op">==</span> y<span class="op">.</span>extent<span class="op">(</span><span class="dv">0</span><span class="op">))</span>;</span>
+<span id="cb40-97"><a href="#cb40-97" aria-hidden="true" tabindex="-1"></a>  <span class="cf">for</span> <span class="op">(</span><span class="dt">int</span> k <span class="op">=</span> <span class="dv">0</span>; k <span class="op">&lt;</span> x<span class="op">.</span>extent<span class="op">(</span><span class="dv">0</span><span class="op">)</span>; <span class="op">++</span>k<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb40-98"><a href="#cb40-98" aria-hidden="true" tabindex="-1"></a>    y<span class="op">[</span>k<span class="op">]</span> <span class="op">=</span> beta <span class="op">*</span> y<span class="op">[</span>k<span class="op">]</span> <span class="op">+</span> alpha <span class="op">*</span> x<span class="op">[</span>k<span class="op">]</span>; </span>
+<span id="cb40-99"><a href="#cb40-99" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb40-100"><a href="#cb40-100" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
+<span id="cb40-101"><a href="#cb40-101" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-102"><a href="#cb40-102" aria-hidden="true" tabindex="-1"></a><span class="co">// 1-norm of the vector y</span></span>
+<span id="cb40-103"><a href="#cb40-103" aria-hidden="true" tabindex="-1"></a><span class="dt">float</span> vectorized_norm<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">32</span><span class="op">&gt;</span> y<span class="op">)</span></span>
+<span id="cb40-104"><a href="#cb40-104" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb40-105"><a href="#cb40-105" aria-hidden="true" tabindex="-1"></a>  <span class="dt">float</span> one_norm <span class="op">=</span> <span class="fl">0.0</span><span class="bu">f</span>;</span>
+<span id="cb40-106"><a href="#cb40-106" aria-hidden="true" tabindex="-1"></a>  <span class="cf">for</span> <span class="op">(</span><span class="dt">int</span> k <span class="op">=</span> <span class="dv">0</span>; k <span class="op">&lt;</span> y<span class="op">.</span>extent<span class="op">(</span><span class="dv">0</span><span class="op">)</span>; <span class="op">++</span>k<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb40-107"><a href="#cb40-107" aria-hidden="true" tabindex="-1"></a>    one_norm <span class="op">+=</span> std<span class="op">::</span>fabs<span class="op">(</span>y<span class="op">[</span>k<span class="op">])</span>; </span>
+<span id="cb40-108"><a href="#cb40-108" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb40-109"><a href="#cb40-109" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> one_norm;</span>
+<span id="cb40-110"><a href="#cb40-110" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
+<span id="cb40-111"><a href="#cb40-111" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-112"><a href="#cb40-112" aria-hidden="true" tabindex="-1"></a><span class="co">// Interfaces that require 16-byte alignment,</span></span>
+<span id="cb40-113"><a href="#cb40-113" aria-hidden="true" tabindex="-1"></a><span class="co">// because they want to do 4-wide SIMD of float.</span></span>
+<span id="cb40-114"><a href="#cb40-114" aria-hidden="true" tabindex="-1"></a><span class="dt">void</span> fill_x<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">16</span><span class="op">&gt;</span> x<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb40-115"><a href="#cb40-115" aria-hidden="true" tabindex="-1"></a>  <span class="cf">for</span> <span class="op">(</span><span class="dt">int</span> k <span class="op">=</span> <span class="dv">0</span>; k <span class="op">&lt;</span> x<span class="op">.</span>extent<span class="op">(</span><span class="dv">0</span><span class="op">)</span>; <span class="op">++</span>k<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb40-116"><a href="#cb40-116" aria-hidden="true" tabindex="-1"></a>    x<span class="op">[</span>k<span class="op">]</span> <span class="op">=</span> <span class="kw">static_cast</span><span class="op">&lt;</span><span class="dt">float</span><span class="op">&gt;(</span>k <span class="op">+</span> <span class="dv">2</span><span class="op">)</span>;</span>
+<span id="cb40-117"><a href="#cb40-117" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span>  </span>
+<span id="cb40-118"><a href="#cb40-118" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
+<span id="cb40-119"><a href="#cb40-119" aria-hidden="true" tabindex="-1"></a><span class="dt">void</span> fill_y<span class="op">(</span>aligned_mdspan<span class="op">&lt;</span><span class="dv">16</span><span class="op">&gt;</span> y<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb40-120"><a href="#cb40-120" aria-hidden="true" tabindex="-1"></a>  <span class="cf">for</span> <span class="op">(</span><span class="dt">int</span> k <span class="op">=</span> <span class="dv">0</span>; k <span class="op">&lt;</span> y<span class="op">.</span>extent<span class="op">(</span><span class="dv">0</span><span class="op">)</span>; <span class="op">++</span>k<span class="op">)</span> <span class="op">{</span></span>
+<span id="cb40-121"><a href="#cb40-121" aria-hidden="true" tabindex="-1"></a>    y<span class="op">[</span>k<span class="op">]</span> <span class="op">=</span> <span class="kw">static_cast</span><span class="op">&lt;</span><span class="dt">float</span><span class="op">&gt;(</span>k <span class="op">-</span> <span class="dv">1</span><span class="op">)</span>;</span>
+<span id="cb40-122"><a href="#cb40-122" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span>  </span>
+<span id="cb40-123"><a href="#cb40-123" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
+<span id="cb40-124"><a href="#cb40-124" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-125"><a href="#cb40-125" aria-hidden="true" tabindex="-1"></a><span class="co">// Helper functions for making overaligned array allocations.</span></span>
+<span id="cb40-126"><a href="#cb40-126" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-127"><a href="#cb40-127" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
+<span id="cb40-128"><a href="#cb40-128" aria-hidden="true" tabindex="-1"></a><span class="kw">struct</span> delete_raw <span class="op">{</span></span>
+<span id="cb40-129"><a href="#cb40-129" aria-hidden="true" tabindex="-1"></a>  <span class="dt">void</span> <span class="kw">operator</span><span class="op">()(</span>ElementType<span class="op">*</span> p<span class="op">)</span> <span class="kw">const</span> <span class="op">{</span></span>
+<span id="cb40-130"><a href="#cb40-130" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>free<span class="op">(</span>p<span class="op">)</span>;</span>
+<span id="cb40-131"><a href="#cb40-131" aria-hidden="true" tabindex="-1"></a>  <span class="op">}</span></span>
+<span id="cb40-132"><a href="#cb40-132" aria-hidden="true" tabindex="-1"></a><span class="op">}</span>;</span>
+<span id="cb40-133"><a href="#cb40-133" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-134"><a href="#cb40-134" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType<span class="op">&gt;</span></span>
+<span id="cb40-135"><a href="#cb40-135" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> allocation <span class="op">=</span></span>
+<span id="cb40-136"><a href="#cb40-136" aria-hidden="true" tabindex="-1"></a>  std<span class="op">::</span>unique_ptr<span class="op">&lt;</span>ElementType<span class="op">[]</span>, delete_raw<span class="op">&lt;</span>ElementType<span class="op">&gt;&gt;</span>;</span>
+<span id="cb40-137"><a href="#cb40-137" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-138"><a href="#cb40-138" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="kw">class</span> ElementType, std<span class="op">::</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
+<span id="cb40-139"><a href="#cb40-139" aria-hidden="true" tabindex="-1"></a>allocation<span class="op">&lt;</span>ElementType<span class="op">&gt;</span> allocate_raw<span class="op">(</span><span class="kw">const</span> std<span class="op">::</span><span class="dt">size_t</span> num_elements<span class="op">)</span></span>
+<span id="cb40-140"><a href="#cb40-140" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb40-141"><a href="#cb40-141" aria-hidden="true" tabindex="-1"></a>  <span class="kw">const</span> std<span class="op">::</span><span class="dt">size_t</span> num_bytes <span class="op">=</span> num_elements <span class="op">*</span> <span class="kw">sizeof</span><span class="op">(</span>ElementType<span class="op">)</span>;</span>
+<span id="cb40-142"><a href="#cb40-142" aria-hidden="true" tabindex="-1"></a>  <span class="dt">float</span><span class="op">*</span> ptr <span class="op">=</span> <span class="kw">reinterpret_cast</span><span class="op">&lt;</span><span class="dt">float</span><span class="op">*&gt;(</span></span>
+<span id="cb40-143"><a href="#cb40-143" aria-hidden="true" tabindex="-1"></a><span class="pp">#if defined(_MSC_VER)</span></span>
+<span id="cb40-144"><a href="#cb40-144" aria-hidden="true" tabindex="-1"></a>    _aligned_malloc<span class="op">(</span>byte_alignment, num_bytes<span class="op">)</span></span>
+<span id="cb40-145"><a href="#cb40-145" aria-hidden="true" tabindex="-1"></a><span class="pp">#else</span></span>
+<span id="cb40-146"><a href="#cb40-146" aria-hidden="true" tabindex="-1"></a>    std<span class="op">::</span>aligned_alloc<span class="op">(</span>byte_alignment, num_bytes<span class="op">)</span></span>
+<span id="cb40-147"><a href="#cb40-147" aria-hidden="true" tabindex="-1"></a><span class="pp">#endif</span></span>
+<span id="cb40-148"><a href="#cb40-148" aria-hidden="true" tabindex="-1"></a>  <span class="op">)</span>;</span>
+<span id="cb40-149"><a href="#cb40-149" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> <span class="op">{</span>ptr, delete_raw<span class="op">&lt;</span>ElementType<span class="op">&gt;{}}</span>;</span>
+<span id="cb40-150"><a href="#cb40-150" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
+<span id="cb40-151"><a href="#cb40-151" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-152"><a href="#cb40-152" aria-hidden="true" tabindex="-1"></a><span class="dt">float</span> user_function<span class="op">(</span><span class="dt">size_t</span> num_elements, <span class="dt">float</span> alpha, <span class="dt">float</span> beta<span class="op">)</span></span>
+<span id="cb40-153"><a href="#cb40-153" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb40-154"><a href="#cb40-154" aria-hidden="true" tabindex="-1"></a>  <span class="kw">constexpr</span> <span class="dt">size_t</span> max_byte_alignment <span class="op">=</span> <span class="dv">32</span>;</span>
+<span id="cb40-155"><a href="#cb40-155" aria-hidden="true" tabindex="-1"></a>  <span class="co">// unique_ptr&lt;float[], deleter_something&gt;</span></span>
+<span id="cb40-156"><a href="#cb40-156" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> x_alloc <span class="op">=</span> allocate_raw<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
+<span id="cb40-157"><a href="#cb40-157" aria-hidden="true" tabindex="-1"></a>  <span class="kw">auto</span> y_alloc <span class="op">=</span> allocate_raw<span class="op">&lt;</span><span class="dt">float</span>, max_byte_alignment<span class="op">&gt;(</span>num_elements<span class="op">)</span>;</span>
+<span id="cb40-158"><a href="#cb40-158" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-159"><a href="#cb40-159" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> x<span class="op">(</span>x_alloc<span class="op">.</span>get<span class="op">()</span>, num_elements<span class="op">)</span>;</span>
+<span id="cb40-160"><a href="#cb40-160" aria-hidden="true" tabindex="-1"></a>  aligned_mdspan<span class="op">&lt;</span>max_byte_alignment<span class="op">&gt;</span> y<span class="op">(</span>y_alloc<span class="op">.</span>get<span class="op">()</span>, num_elements<span class="op">)</span>;</span>
+<span id="cb40-161"><a href="#cb40-161" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-162"><a href="#cb40-162" aria-hidden="true" tabindex="-1"></a>  <span class="co">// expect aligned_mdspan&lt;16&gt;, get aligned_mdspan&lt;32&gt;</span></span>
+<span id="cb40-163"><a href="#cb40-163" aria-hidden="true" tabindex="-1"></a>  fill_x<span class="op">(</span>x<span class="op">)</span>; <span class="co">// automatic conversion from 32-byte aligned to 16-byte aligned</span></span>
+<span id="cb40-164"><a href="#cb40-164" aria-hidden="true" tabindex="-1"></a>  fill_y<span class="op">(</span>y<span class="op">)</span>; <span class="co">// automatic conversion again</span></span>
+<span id="cb40-165"><a href="#cb40-165" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-166"><a href="#cb40-166" aria-hidden="true" tabindex="-1"></a>  <span class="co">// expect aligned_mdspan&lt;32&gt;, get aligned_mdspan&lt;32&gt;</span></span>
+<span id="cb40-167"><a href="#cb40-167" aria-hidden="true" tabindex="-1"></a>  vectorized_axpby<span class="op">(</span>y, alpha, x, beta<span class="op">)</span>;</span>
+<span id="cb40-168"><a href="#cb40-168" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> vectorized_norm<span class="op">(</span>y<span class="op">)</span>;</span>
+<span id="cb40-169"><a href="#cb40-169" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span>
+<span id="cb40-170"><a href="#cb40-170" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-171"><a href="#cb40-171" aria-hidden="true" tabindex="-1"></a><span class="op">}</span> <span class="co">// namespace (anonymous)</span></span>
+<span id="cb40-172"><a href="#cb40-172" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb40-173"><a href="#cb40-173" aria-hidden="true" tabindex="-1"></a><span class="dt">int</span> main<span class="op">(</span><span class="dt">int</span> argc, <span class="dt">char</span><span class="op">*</span> argv<span class="op">[])</span></span>
+<span id="cb40-174"><a href="#cb40-174" aria-hidden="true" tabindex="-1"></a><span class="op">{</span></span>
+<span id="cb40-175"><a href="#cb40-175" aria-hidden="true" tabindex="-1"></a>  <span class="dt">float</span> result <span class="op">=</span> user_function<span class="op">(</span><span class="dv">10</span>, <span class="fl">1.0</span><span class="bu">f</span>, <span class="op">-</span><span class="fl">1.0</span><span class="bu">f</span><span class="op">)</span>;</span>
+<span id="cb40-176"><a href="#cb40-176" aria-hidden="true" tabindex="-1"></a>  <span class="co">// 3 + 3 + ... + 3 = 30</span></span>
+<span id="cb40-177"><a href="#cb40-177" aria-hidden="true" tabindex="-1"></a>  <span class="ot">assert</span><span class="op">(</span>result <span class="op">==</span> <span class="fl">30.0</span><span class="bu">f</span><span class="op">)</span>;</span>
+<span id="cb40-178"><a href="#cb40-178" aria-hidden="true" tabindex="-1"></a>  <span class="cf">return</span> <span class="dv">0</span>;</span>
+<span id="cb40-179"><a href="#cb40-179" aria-hidden="true" tabindex="-1"></a><span class="op">}</span></span></code></pre></div>
 </div>
 </div>
 </body>
diff --git a/aligned_accessor/aligned_accessor.md b/aligned_accessor/aligned_accessor.md
index f3a4a7e6..fe12547a 100644
--- a/aligned_accessor/aligned_accessor.md
+++ b/aligned_accessor/aligned_accessor.md
@@ -84,6 +84,8 @@ toc: true
     * Add non-wording section explaining why `aligned_accessor`
         has no `explicit` constructor from less to more alignment
 
+    * Add Compiler Explorer link with full implementation and demo
+
 # Purpose of this paper
 
 We propose adding `aligned_accessor` to the C++ Standard Library.
@@ -1535,4 +1537,192 @@ int main()
   std::cerr << "Made it to the end\n";
   return 0;
 }
-```
\ No newline at end of file
+```
+
+# Appendix B: Implementation and demo
+
+<a href="https://godbolt.org/z/hjqExKaeh">This Compiler Explorer link</a>
+gives a full implementation of `aligned_accessor` and a demonstration.
+We show the full source code from that link here below.
+
+```c++
+#include <https://raw.githubusercontent.com/kokkos/mdspan/single-header/mdspan.hpp>
+#include <bit>
+#include <cassert>
+#include <cmath>
+#if defined(_MSC_VER)
+#  include <cstdlib> // MSVC's _aligned_malloc
+#endif
+#include <exception>
+#include <functional>
+#include <memory>
+#include <numeric>
+#include <type_traits>
+
+namespace stdex = std::experimental;
+
+// P2389 (voted into C++ at June 2024 STL plenary)
+namespace std {
+template<size_t Rank, class IndexType = size_t>
+using dims = dextents<IndexType, Rank>;
+} // namespace std
+
+namespace {
+
+template<class ElementType, std::size_t byte_alignment>
+class aligned_accessor {
+public:
+  static_assert(std::has_single_bit(byte_alignment),
+    "byte_alignment must be a power of two.");
+  static_assert(byte_alignment >= alignof(ElementType),
+    "Insufficient byte alignment for ElementType");
+
+  using offset_policy = stdex::default_accessor<ElementType>;
+  using element_type = ElementType;
+  using reference = ElementType&;
+  using data_handle_type = ElementType*;
+
+  constexpr aligned_accessor() noexcept = default;
+
+  template<
+    class OtherElementType,
+    std::size_t other_byte_alignment>
+  requires(std::is_convertible_v<
+    OtherElementType(*)[], element_type(*)[]>)
+  constexpr aligned_accessor(
+    aligned_accessor<OtherElementType, other_byte_alignment>)
+      noexcept
+  {
+    constexpr size_t the_gcd =
+      std::gcd(other_byte_alignment, byte_alignment);
+    static_assert(the_gcd == byte_alignment);
+  }
+
+  template<class OtherElementType>
+  requires(std::is_convertible_v<
+    OtherElementType(*)[], element_type(*)[]>)
+  constexpr explicit aligned_accessor(
+    stdex::default_accessor<OtherElementType>) noexcept
+  {}
+ 
+  constexpr
+    operator stdex::default_accessor<element_type>() const
+  {
+    return {};
+  }
+
+  constexpr reference
+    access(data_handle_type p, size_t i) const noexcept
+  {
+    return std::assume_aligned<byte_alignment>(p)[i];
+  }
+
+  constexpr typename offset_policy::data_handle_type
+  offset(data_handle_type p, size_t i) const noexcept {
+    return p + i;
+  }
+
+  static bool
+    is_sufficiently_aligned(data_handle_type p) noexcept
+  {
+    return alignof(ElementType) >= byte_alignment && 
+      std::bit_cast<uintptr_t>(p) % byte_alignment == 0;
+  }
+};
+
+template<size_t byte_alignment>
+using aligned_mdspan =
+  std::mdspan<float, std::dims<1, int>, std::layout_right,
+    aligned_accessor<float, byte_alignment>>;
+
+// Interfaces that require 32-byte alignment,
+// because they want to do 8-wide SIMD of float.
+void
+vectorized_axpby(aligned_mdspan<32> y,
+  float alpha, aligned_mdspan<32> x, float beta)
+{
+  assert(x.extent(0) == y.extent(0));
+  for (int k = 0; k < x.extent(0); ++k) {
+    y[k] = beta * y[k] + alpha * x[k]; 
+  }
+}
+
+// 1-norm of the vector y
+float vectorized_norm(aligned_mdspan<32> y)
+{
+  float one_norm = 0.0f;
+  for (int k = 0; k < y.extent(0); ++k) {
+    one_norm += std::fabs(y[k]); 
+  }
+  return one_norm;
+}
+
+// Interfaces that require 16-byte alignment,
+// because they want to do 4-wide SIMD of float.
+void fill_x(aligned_mdspan<16> x) {
+  for (int k = 0; k < x.extent(0); ++k) {
+    x[k] = static_cast<float>(k + 2);
+  }  
+}
+void fill_y(aligned_mdspan<16> y) {
+  for (int k = 0; k < y.extent(0); ++k) {
+    y[k] = static_cast<float>(k - 1);
+  }  
+}
+
+// Helper functions for making overaligned array allocations.
+
+template<class ElementType>
+struct delete_raw {
+  void operator()(ElementType* p) const {
+    std::free(p);
+  }
+};
+
+template<class ElementType>
+using allocation =
+  std::unique_ptr<ElementType[], delete_raw<ElementType>>;
+
+template<class ElementType, std::size_t byte_alignment>
+allocation<ElementType> allocate_raw(const std::size_t num_elements)
+{
+  const std::size_t num_bytes = num_elements * sizeof(ElementType);
+  float* ptr = reinterpret_cast<float*>(
+#if defined(_MSC_VER)
+    _aligned_malloc(byte_alignment, num_bytes)
+#else
+    std::aligned_alloc(byte_alignment, num_bytes)
+#endif
+  );
+  return {ptr, delete_raw<ElementType>{}};
+}
+
+float user_function(size_t num_elements, float alpha, float beta)
+{
+  constexpr size_t max_byte_alignment = 32;
+  // unique_ptr<float[], deleter_something>
+  auto x_alloc = allocate_raw<float, max_byte_alignment>(num_elements);
+  auto y_alloc = allocate_raw<float, max_byte_alignment>(num_elements);
+
+  aligned_mdspan<max_byte_alignment> x(x_alloc.get(), num_elements);
+  aligned_mdspan<max_byte_alignment> y(y_alloc.get(), num_elements);
+
+  // expect aligned_mdspan<16>, get aligned_mdspan<32>
+  fill_x(x); // automatic conversion from 32-byte aligned to 16-byte aligned
+  fill_y(y); // automatic conversion again
+
+  // expect aligned_mdspan<32>, get aligned_mdspan<32>
+  vectorized_axpby(y, alpha, x, beta);
+  return vectorized_norm(y);
+}
+
+} // namespace (anonymous)
+
+int main(int argc, char* argv[])
+{
+  float result = user_function(10, 1.0f, -1.0f);
+  // 3 + 3 + ... + 3 = 30
+  assert(result == 30.0f);
+  return 0;
+}
+```

From fe29e0f014060541cda03def2805e4907b866e3c Mon Sep 17 00:00:00 2001
From: Mark Hoemmen <mhoemmen@nvidia.com>
Date: Thu, 11 Jul 2024 09:34:04 -0600
Subject: [PATCH 4/4] P2897R2: Add ref to Appendix B

---
 aligned_accessor/P2897R2.html        | 3 ++-
 aligned_accessor/aligned_accessor.md | 1 +
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/aligned_accessor/P2897R2.html b/aligned_accessor/P2897R2.html
index 15202b6e..e300fc54 100644
--- a/aligned_accessor/P2897R2.html
+++ b/aligned_accessor/P2897R2.html
@@ -1467,7 +1467,8 @@ <h3 data-number="5.8.5" id="conclusion-retain-mdspans-current-design"><span clas
 explicit conversions” design with a cast function design.</p>
 <h1 data-number="6" id="implementation"><span class="header-section-number">6</span> Implementation<a href="#implementation" class="self-link"></a></h1>
 <p>We have tested an implementation of this proposal with the <a href="https://github.com/kokkos/mdspan/">reference mdspan
-implementation</a>.</p>
+implementation</a>. Appendix B below lists the source code of a full
+implementation.</p>
 <h1 data-number="7" id="example"><span class="header-section-number">7</span> Example<a href="#example" class="self-link"></a></h1>
 <div class="sourceCode" id="cb24"><pre class="sourceCode cpp"><code class="sourceCode cpp"><span id="cb24-1"><a href="#cb24-1" aria-hidden="true" tabindex="-1"></a><span class="kw">template</span><span class="op">&lt;</span><span class="dt">size_t</span> byte_alignment<span class="op">&gt;</span></span>
 <span id="cb24-2"><a href="#cb24-2" aria-hidden="true" tabindex="-1"></a><span class="kw">using</span> aligned_mdspan <span class="op">=</span></span>
diff --git a/aligned_accessor/aligned_accessor.md b/aligned_accessor/aligned_accessor.md
index fe12547a..cf004f59 100644
--- a/aligned_accessor/aligned_accessor.md
+++ b/aligned_accessor/aligned_accessor.md
@@ -1042,6 +1042,7 @@ design with a cast function design.
 
 We have tested an implementation of this proposal with the
 [reference mdspan implementation](https://github.com/kokkos/mdspan/).
+Appendix B below lists the source code of a full implementation.
 
 # Example