[Bug target/80837] [7/8 regression] x86 accessing a member of a 16-byte atomic object generates terrible code: splitting/merging the bytes

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80837

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2017-05-23
   Target Milestone|---                         |7.2
            Summary|[7.1.0 regression] x86      |[7/8 regression] x86
                   |accessing a member of a     |accessing a member of a
                   |16-byte atomic object       |16-byte atomic object
                   |generates terrible code:    |generates terrible code:
                   |splitting/merging the bytes |splitting/merging the bytes
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 8 generates a __atomic_load_16 call for me while GCC 6 does

        lock cmpxchg16b (%rdi)