[LLVMbugs] [Bug 23263] New: clang miscompiles atomics by using optimized libcalls for underaligned storage

[bugzilla-daemon at llvm.org]

https://llvm.org/bugs/show_bug.cgi?id=23263

            Bug ID: 23263
           Summary: clang miscompiles atomics by using optimized libcalls
                    for underaligned storage
           Product: clang
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: C++
          Assignee: unassignedclangbugs at nondot.org
          Reporter: richard-llvm at metafoo.co.uk
                CC: dgregor at apple.com, llvmbugs at cs.uiuc.edu
    Classification: Unclassified

Per https://gcc.gnu.org/wiki/Atomic/GCCMM/LIbrary, "All the optimized routines
expect that the object will be properly aligned for a data type of the
specified size".

However, Clang and GCC get this wrong:

  struct X { int a, b; };
  X f(X *p) { X r; __atomic_load(p, &r, 5); return r; }

... calls __atomic_load_8. And GCC's libatomic implements __atomic_load_8 on
x86_64 with a single 8-byte mov, which is only atomic for a correctly-aligned
operand.

We should instead emit a call to __atomic_load(8, p, &r, 5) because we have no
reason to think that p is 8 byte aligned.

The same happens for __atomic_store and probably all the other atomic libcalls.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20150417/36a1afa5/attachment.html>