<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 - SLP vectorizer: aligned store to unaligned address"
   href="https://bugs.llvm.org/show_bug.cgi?id=34219">34219</a>
          </td>
        </tr>

        <tr>
          <th>Summary</th>
          <td>SLP vectorizer: aligned store to unaligned address
          </td>
        </tr>

        <tr>
          <th>Product</th>
          <td>clang
          </td>
        </tr>

        <tr>
          <th>Version</th>
          <td>trunk
          </td>
        </tr>

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

        <tr>
          <th>OS</th>
          <td>Linux
          </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>C++
          </td>
        </tr>

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

        <tr>
          <th>Reporter</th>
          <td>vsna7ma4he@jorrit.de
          </td>
        </tr>

        <tr>
          <th>CC</th>
          <td>dgregor@apple.com, llvm-bugs@lists.llvm.org
          </td>
        </tr></table>
      <p>
        <div>
        <pre>The reproducer causes clang++ to generate a movdqa instruction to an address
that is not sufficiently aligned, causing a SEGFAULT on x86-64.  This happens
with -O2, but not with -O2 -fno-slp-vectorize.  This was originally observed
with

  clang version 3.8.0 (tags/RELEASE_380/final 263969)
  <a href="http://coliru.stacked-crooked.com/a/159167c4e00de3bd">http://coliru.stacked-crooked.com/a/159167c4e00de3bd</a>

it can still be seen in the assembler output of what godbolt calls clang trunk

  clang version 6.0.0 (trunk 310993)
  <a href="https://godbolt.org/g/oF9WT1">https://godbolt.org/g/oF9WT1</a>

The offending instruction is this store -- movdqa requires alignment to 16
bytes, but the address is only aligned to 8 bytes.

  movdqa %xmm0,0x200b3d(%rip) # 601058 <counters+0x8>

reproducer.cc:
====================
// compile with:
// clang++ -std=c++14 -O2 -Wall -g -pedantic -pthread reproducer.cpp
long m = 3;
struct alignas(16) Counters {
  long padding;
  long c0;
  long c1;
} counters;
int main() {
  ++counters.c0; ++counters.c1;

  if(m & 1) { ++counters.c0; ++counters.c1; }
  if(m & 2) { ++counters.c0; ++counters.c1; }
}
====================

Disassmbly (clang trunk from godbolt).  The offending instruction is the
fourth from the bottom.
====================
.plt.got:
 jmpq *0x200c12(%rip) # 600ff8 <_DYNAMIC+0x200>
 xchg %ax,%ax
main:
 movdqu 0x200b80(%rip),%xmm1 # 601058 <counters+0x8>
 pcmpeqd %xmm2,%xmm2
 movdqa %xmm1,%xmm0
 psubq %xmm2,%xmm0
 movdqu %xmm0,0x200b6c(%rip) # 601058 <counters+0x8>
 mov 0x200b3d(%rip),%rax # 601030 <m>
 test $0x1,%al
 je 400503 <main+0x33>
 paddq 0xb1(%rip),%xmm1 # 4005b0 <_IO_stdin_used+0x10>
 movdqa %xmm1,%xmm0
 test $0x3,%al
 je 40051b <main+0x4b>
 test $0x2,%al
 je 400513 <main+0x43>
 pcmpeqd %xmm1,%xmm1
 psubq %xmm1,%xmm0
 movdqa %xmm0,0x200b3d(%rip) # 601058 <counters+0x8>
 xor %eax,%eax
 retq 
 xchg %ax,%ax
====================</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>