<html>
    <head>
      <base href="https://bugs.llvm.org/">
    </head>
    <body><table border="1" cellspacing="0" cellpadding="8">
        <tr>
          <th>Bug ID</th>
          <td><a class="bz_bug_link 
          bz_status_NEW "
   title="NEW - Implement cstring literal merging"
   href="https://bugs.llvm.org/show_bug.cgi?id=50563">50563</a>
          </td>
        </tr>

        <tr>
          <th>Summary</th>
          <td>Implement cstring literal merging
          </td>
        </tr>

        <tr>
          <th>Product</th>
          <td>lld
          </td>
        </tr>

        <tr>
          <th>Version</th>
          <td>unspecified
          </td>
        </tr>

        <tr>
          <th>Hardware</th>
          <td>PC
          </td>
        </tr>

        <tr>
          <th>OS</th>
          <td>All
          </td>
        </tr>

        <tr>
          <th>Status</th>
          <td>NEW
          </td>
        </tr>

        <tr>
          <th>Severity</th>
          <td>enhancement
          </td>
        </tr>

        <tr>
          <th>Priority</th>
          <td>P
          </td>
        </tr>

        <tr>
          <th>Component</th>
          <td>MachO
          </td>
        </tr>

        <tr>
          <th>Assignee</th>
          <td>unassignedbugs@nondot.org
          </td>
        </tr>

        <tr>
          <th>Reporter</th>
          <td>jezreel@gmail.com
          </td>
        </tr>

        <tr>
          <th>CC</th>
          <td>gkm@fb.com, jezreel@gmail.com, llvm-bugs@lists.llvm.org, smeenai@fb.com
          </td>
        </tr></table>
      <p>
        <div>
        <pre>Created <span class=""><a href="attachment.cgi?id=24903" name="attach_24903" title="ld64 cstring misalignment crash">attachment 24903</a> <a href="attachment.cgi?id=24903&action=edit" title="ld64 cstring misalignment crash">[details]</a></span>
ld64 cstring misalignment crash

Mergeable cstring literals are found under the `__TEXT,__cstring` section. In
contrast to ELF, which puts strings that need different alignments into
different sections, clang's Mach-O backend puts them all in one section.
Strings that need to be aligned have the `.p2align` directive emitted before
them, which simply translates into zero padding in the object file.

I *think* ld64 extracts the desired per-string alignment from this data by
preserving each string's offset from the last section-aligned address. I'm not
entirely certain since it doesn't seem consistent about doing this; but perhaps
this can be chalked up to cases where ld64 has to deduplicate strings with
different offset/alignment combos -- it seems to pick one of their alignments
to preserve. This doesn't seem correct in general; see the attachment, which
can induce ld64 to produce a crashing binary just by linking in an additional
object file that only contains cstrings and no code.

Moreover, this scheme seems rather inefficient: since unaligned and aligned
strings are all put in the same section, which has a single alignment value, it
doesn't seem possible to tell whether a given string doesn't have any alignment
requirements. Preserving offset+alignments for strings that don't need it is
wasteful.

In practice, the crashes seen so far seem to stem from x86_64 SIMD operations
on cstrings. X86_64 requires SIMD accesses to be 16-byte-aligned. So for now,
I'm thinking of just aligning all strings to 16 bytes on x86_64. This is indeed
wasteful, but implementation-wise it's simpler than preserving per-string
alignment+offsets. It also avoids the aforementioned crash after deduplication
of differently-aligned strings. Finally, the overhead is not huge: using
16-byte alignment (vs no alignment) is only a 0.5% size overhead when linking
chromium_framework.

This is a pretty unprincipled heuristic, but we can always revisit it later.
For now, I'd like to see how many CFStrings ICF can remove after we dedup the
cstrings.

(Another heuristic I considered: clang seems to only emit SIMD operations for
strings > 16 bytes, so I tried only aligning these longer strings, but it was a
difference of 0.1% -- didn't seem worth the additional implementation
complexity.)</pre>
        </div>
      </p>


      <hr>
      <span>You are receiving this mail because:</span>

      <ul>
          <li>You are on the CC list for the bug.</li>
      </ul>
    </body>
</html>